Solution Guide II-D

1D Measuring

8.0
MVTec Software GmbH Building Vision for Business
Overview

With HALCON, you can perform highly accurate measurements in 1D (i.e., measure distances and
angles along lines and arcs), 2D (i.e., measure features like the size and orientation of objects), and
3D (i.e., measure in 3D world coordinates). This Solution Guide describes 1D measuring, a fast and
easy-to-use tool that enables you to measure the dimensions of objects with high accuracy.
A first example and a short characterization of the various methods is given in section 1 on page 1. A
description of how to create a measure object as well as the basic concepts of 1D measuring can be found
in section 2 on page 2. Then, the methods to perform 1D measuring are described in detail.
For more complex measurement tasks, it might be necessary to control the selection of the edges. To
do this, the measure object can be extended to a fuzzy measure object (section 4 on page 22) where the
selection of the edges is controlled by fuzzy membership functions.
Unless specified otherwise, the HDevelop example programs can be found in the subdirectory
1d_measuring of the directory %HALCONROOT%\examples\solution_guide.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in
any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior written
permission of the publisher.

Edition 1 July 2005 (HALCON 7.1)

Edition 1a December 2006 (HALCON 7.1.2)
Edition 2 June 2007 (HALCON 8.0)

Copyright
c 2005-2008 by MVTec Software GmbH, München, Germany MVTec Software GmbH

Microsoft, Windows, Windows NT, Windows 2000, and Windows XP are either trademarks or registered trademarks
of Microsoft Corporation.
All other nationally and internationally recognized trademarks and tradenames are hereby recognized.

More information about HALCON can be found at:

http://www.halcon.com/
Contents

1 First Example and Overview 1

2 The Basics of Measure Objects 2

2.1 The Process of 1D Edge Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Creating a Measure Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Using the Measure Object 8

3.1 The Position of Edges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 The Position of Edge Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 The Position of Points With a Particular Gray Value . . . . . . . . . . . . . . . . . . . . 13
3.4 The Gray Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.5 Measuring in World Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.6 Compensate for Radial Image Distortions . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.7 Changing the Position and Orientation of the Measure Object . . . . . . . . . . . . . . . 18
3.8 Tips and Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.9 Destroying the Measure Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4 Advanced Measuring: The Fuzzy Measure Object 22

4.1 A First Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 Controlling the Selection of Edges and Edge Pairs . . . . . . . . . . . . . . . . . . . . . 25
4.3 Creating the Fuzzy Measure Object . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4 Using the Standard Measure Object Functionality . . . . . . . . . . . . . . . . . . . . . 34

A Slanted Edges 35

B Examples for Fuzzy Membership Functions 40

B.1 Set Type contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
B.2 Set Type position . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
B.3 Set Type position_pair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
B.4 Set Type size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
B.5 Set Type gray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
B.6 Set Type position_pair combined with size . . . . . . . . . . . . . . . . . . . . . . . . . 59

C Definition of the Geometric Mean 60

 1 First Example and Overview D-1

1 First Example and Overview

First Example
The example program measure_switch.dev is a simple application where the width of and the distance
between the pins of a switch (see figure 1a) must be measured. This task can easily be solved by 1D
measuring because positions and distances are measured along a line.

(a) (b)

Figure 1: Image of a switch: (a) The width of the pins and their distance from each other must be mea-
sured; (b) the rectangular ROI within which the edges are determined.

First, a measure object is created by specifying a rectangle that lies over the pins as depicted in figure 1b.
The operator gen_measure_rectangle2 returns a handle to the created measure object.

gen_measure_rectangle2 (Row, Column, Phi, Length1, Length2, Width, Height,

Interpolation, MeasureHandle)

To perform the measurement, you pass this handle to the measuring operators. To inspect the switch,
we use measure_pairs, which extracts the edge pairs corresponding to the pins and returns their width
(distance between the edges of a pair, IntraDistance) and their distance (distance between edges of
subsequent pairs, InterDistance).

measure_pairs (Image, MeasureHandle, Sigma, Threshold, Transition, Select,

RowEdgeFirst, ColumnEdgeFirst, AmplitudeFirst,
RowEdgeSecond, ColumnEdgeSecond, AmplitudeSecond,
IntraDistance, InterDistance)

The resulting edges are displayed in figure 2 together with the measured width of the pins and the distance
between them.
The following sections take a closer look at these steps and explain how to solve more complex tasks:
Section 2 describes the basics of measuring, i.e., how the edges are determined (section 2.1 on page 3)
and how to create measure objects for rectangular and circular ROIs (section 2.2 on page 6).
Section 3 on page 8 explains how to use the measure object to extract edges (section 3.1 on page 8), edge
pairs (section 3.2 on page 10), points that have a particular gray value (section 3.3 on page 13), and the
gray value profile (section 3.4 on page 13).
D-2 2 The Basics of Measure Objects

Figure 2: The result of measuring the width of and the distance between the pins of a switch.

Often, it is necessary to determine the dimensions of an object in world coordinates, e.g., in µm. The
transformation of the results of a measure object into world coordinates is described in section 3.5 on
page 17.
Section 3.7 on page 18 describes how to change the position and orientation of measure objects. Some
tips and tricks for the use of measure objects are given in section 3.8 on page 19.
Section 4 on page 22 shows how to solve more complex measurement tasks, e.g., if the object shows
different gray levels or if parts of the image are disturbed by reflections or by shadows that are cast on
the object. For this purpose, it is possible to introduce special weighting functions to suppress unwanted
edges. This extension is called fuzzy measure object.

2 The Basics of Measure Objects

The main idea behind the measure object is to extract edges that run approximately perpendicular to
the line or arc of measurement. This process is described in section 2.1. It is important to understand
this process because you influence it with parameters both when creating the measure object and when
applying it. Section 2.2 on page 6 shows how to specify the shape of the measuring ROI. Please note that
in contrast to other HALCON methods, measuring ROIs are not created using reduce_domain, but by
specifying their dimensions when creating a measure object. A measure object can be used to determine
the position of straight edges that run approximately perpendicular to the given ROI.
 2.1 The Process of 1D Edge Extraction D-3

2.1 The Process of 1D Edge Extraction

HALCON determines the position of 1D edges by the following approach: First, equidistant lines of
projection are constructed perpendicular to the line or arc of measurement (also called profile line) with
a length equal to the width of the ROI (figure 3).

ROI

End 1P
ROI ixe
l
Profile
Line

Basics
End
e
Lin
ile
Prof

Start
Start

l
ixe
1P

(a) (b)

Figure 3: Lines of projection for a (a) rectangular ROI and (b) circular ROI.

Then, the mean gray value along each line of projection is calculated. The sequence of these mean val-
ues is called the profile. If the lines of projection are not oriented horizontally or vertically, the pixel
values along them must be interpolated. You can select the interpolation mode with the parameter In-
terpolation of the operators for the creation of measure objects (e.g., gen_measure_rectangle2,
see section 1 on page 1). If Interpolation = ’nearest_neighbor’, the gray values in the measurement
are obtained from the gray values of the closest pixel, i.e., by constant interpolation. This is the fastest
method. However, the geometric accuracy is slightly lower in this mode. For Interpolation = ’bilin-
ear’, bilinear interpolation is used, while for Interpolation = ’bicubic’, bicubic interpolation is used.
The bicubic interpolation yields the most accurate results. However, it is the slowest method.
The length of the lines of projection, i.e., the width of the ROI, defines the degree of averaging in the
direction perpendicular to the profile line. The thick lines in figure 4c and d show the profiles for the two
ROIs displayed in figure 4a and b. It can be seen that the profile that corresponds to the wider ROI is
less noisy. Therefore, as long as the edges run approximately perpendicular to the profile line, the ROI
should be chosen as wide as possible. If the edges do not run perpendicular to the profile line, the width
of the ROI must be chosen smaller to allow the detection of edges. Note that in this case, the profile will
contain more noise. Consequently, the edges will be determined less accurate (see appendix A on page
35).
The profile is then smoothed with a Gaussian smoothing filter (see the thick lines in figure 4e and f),
whose standard deviation is specified with the parameter Sigma of the measuring operators (e.g., mea-
sure_pairs). Now, the first derivative of the smoothed profile√ is calculated (thin lines in figure 4e-f).
Note that the amplitude of the derivative is scaled with a factor 2π · Sigma. The subpixel precise po-
sitions of all local extrema of the first derivative are edge candidates. In figure 5a on page 5, these edge
candidates are indicated by vectors pointing from the respective position of the derivative (thin line) to
the smoothed gray value profile (thick line). Only those edge candidates that have an absolute value of
D-4 2 The Basics of Measure Objects

(a) ROI of Width=2 (b) ROI of Width=20

250 250

200 200
gray value

gray value
150 150

100 100

50 50

0 0
0 20 40 60 80 100 120 0 20 40 60 80 100 120
profile line profile line

(c) Width=2, no smoothing (d) Width=20, no smoothing

250 250

200 200

150 150

100 100
gray value

gray value

50 50

0 0

-50 -50

-100 -100

-150 -150
0 20 40 60 80 100 120 0 20 40 60 80 100 120
profile line profile line

(e) Width=2, Sigma=0.9 (f) Width=20, Sigma=0.9

Figure 4: Profiles (thick lines) and derivatives (thin lines) of the profiles for two ROIs with different width
with and without smoothing the profiles (the orientation of the ROIs is from top left down to the
right).

the first derivative larger than a given Threshold (another parameter of the measuring operators) are
considered as edges in the image (see figure 5b on page 5).
For each edge, the position of its intersection with the profile line is returned. Figure 6 shows the detected
edges, displayed by straight lines with a length similar to the width of the ROI.
 2.1 The Process of 1D Edge Extraction D-5

250 250

200 200

150 150

100 100
gray value

gray value
50 50

0 0

-50 -50

-100 -100

-150 -150
0 20 40 60 80 100 120 0 20 40 60 80 100 120

Basics
profile line profile line

(a) (b)

Figure 5: Positions of (a) edge candidates and (b) edges along the profile. The profile was derived from
the 20 pixel wide ROI, it was smoothed with Sigma = 0.9.

Note that the above described approach for the extraction of 1D edges from a profile differs significantly
from the 2D edge extraction performed, e.g., with the operator edges_sub_pix (compare figure 7 on
page 6). In the case of the 1D edge extraction, only the positions of the intersections of the edges with
the profile line are determined as described above whereas the 2D edge extraction returns the complete
edges as contours. If the edges are straight and run approximately perpendicular to the profile line, the
results of the 1D edge extraction are similar to the intersection of the 2D edges with the profile line. The
more the edges are curved or not perpendicular to the profile line, the more different the results of the
two approaches will be. An example where the results of the two approaches are completely different is
given in section 3.1 on page 8.
Keep in mind that measure objects are designed to measure the position of straight edges that run approx-
imately perpendicular to the profile line. This means that measure objects are not suited to measure the
position of, e.g., curved edges. In those cases, the measure position will typically not coincide with the
intersection of the gray value edge with the profile line. An example for such a case is given in figure 8.

Figure 6: Edges detected from the 20 pixel wide ROI. The profile was smoothed with Sigma = 0.9 and
thresholded with a Threshold of 12.
D-6 2 The Basics of Measure Objects

Figure 7: Edges detected in 2D with the operator edges_sub_pix .

Figure 8: An example for an incorrect use of the measure object. The ROI is oriented from left to right.

Here, the edges are curved, which leads to wrong measurement results. This effect will be reduced if the
width of the ROI is chosen smaller.

2.2 Creating a Measure Object

Measure objects can be created with two different shapes: Rotated rectangles and circular arcs. They are
created with the operators gen_measure_rectangle2 or gen_measure_arc.
To create a rectangular measure object, you use the operator gen_measure_rectangle2:

gen_measure_rectangle2 (Row, Column, Phi, Length1, Length2, Width, Height,

Interpolation, MeasureHandle)

A rotated rectangle is defined by its center (Row, Column), its orientation Phi, and its half edge lengths
Length1 and Length2 (see figure 9). The parameter Phi is given in radians in mathematically positive
sense. All the other parameters are given in pixels.
 2.2 Creating a Measure Object D-7

ROI

End

Phi

e Center (Row,Column)
Lin
ile
P rof

Start
Le

1
th
ng

ng
Le
th
2

Measuring
Figure 9: Rectangular region of interest (ROI).

To create a circular measure object, you use the operator gen_measure_arc:

gen_measure_arc (CenterRow, CenterCol, Radius, AngleStart,

AngleExtent, AnnulusRadius, Width, Height,
Interpolation, MeasureHandle)

The circular arc is defined by the following parameters (see figure 10): The position and size are defined
by the center (CenterRow, CenterColumn) and the radius (Radius). The segment of the circle is defined
by its start angle (AngleStart) and the angular extent (AngleExtent). The width of the circular ROI
is defined by the annulus radius (AnnulusRadius). Again, all parameters are given in pixels, except
for the angles (AngleStart and AngleExtent), which are given in radians in mathematically positive
sense.
The first parameters of both operators for the creation of measure objects define the position, size, and
orientation of the respective ROI. The parameters Width and Height are the size of the images, on which
the measure can be applied. The parameter Interpolation sets the interpolation mode (see section 2.1
on page 3). Note that when using interpolation, not only the ROI but additionally the margin around the
ROI must fit into the image. The width of the margin (in all four directions) has to be at least 1 pixel
for a bilinear interpolation and two pixels for a bicubic interpolation. For lines of projection that do not
fulfill this condition, no result is obtained.
The result for both operators is a handle for the newly created measure object (MeasureHandle), which
can then be used to specify the measure object, e.g., in calls to the operator measure_pos. Note that if
you use HALCON’s COM or C++ interface and call the operator via the classes HMeasureX or HMea-
sure, no handle is returned because the instance of the class itself acts as your handle.
D-8 3 Using the Measure Object

ROI
Profile
Line

End

An di
Ra
nu us AngleExtent
lu
s Start

Ra
di
us
AngleStart

Center (CenterRow,CenterColumn)

Figure 10: Circular region of interest (ROI).

3 Using the Measure Object

For different tasks, depending on your application, you can measure the position of individual edges and
the distance between them (see section 3.1). If the object to inspect has two edges (like the edge pairs in
section 1 on page 1), you can let HALCON group the edges to edge pairs (see section 3.2 on page 10).
Furthermore, the position of points on the profile that have a particular gray value can be determined
(section 3.3 on page 13). It is also possible to access the gray value profile itself (section 3.4 on page
13).
Often, it is necessary to determine the dimensions of an object in world coordinates, e.g., in µm. The
transformation of the results of a measure object into world coordinates is described in section 3.5 on
page 17.
Section 3.7 on page 18 describes how to change the position and orientation of measure objects. Some
tips and tricks for the use of measure objects are given in section 3.8 on page 19.

3.1 The Position of Edges

You measure the position of the individual edges with the operator measure_pos. The edges are deter-
mined with the approach described in section 2.1 on page 3. With this approach, it is also possible to
extract edges that appear only if the image is smoothed in a direction perpendicular to the ROI.
In the example program measure_ic_leads.dev, we use this operator to measure the length of the
leads of the IC that is displayed in figure 11a. At a first glance, it is not clear how to solve this task because
the individual leads are not enclosed by two edges. However, we can exploit the averaging perpendicular
to the profile line: If the ROI encloses multiple leads, as depicted in figure 11a, the averaged gray value of
the leads is darker than the background, but lighter than the body of the IC (see figure 11b and figure 11c).
 3.1 The Position of Edges D-9

250

200

150

100

gray value
50

-50

-100

-150
0 20 40 60 80 100
profile line

(b)
250

200

150

100

gray value

Measuring
50

-50

-100

-150
(a) 0 20 40 60 80 100
profile line

(c)

Figure 11: (a) Image of an IC with the ROIs that are used for the determination of the lead length; (b) and
(c) show the profile (thick line) and its derivation (thin line) for (b) the ROI at the top and (c) the
ROI at the bottom of the image in (a). The ROIs are oriented from the top to the bottom of the
image.

Let’s take a look at the corresponding code. First, a measure object is created for the ROI at the top of
the image. The ROI must have a width such that it contains several leads.

gen_measure_rectangle2 (Row, Column, Phi, Length1, Length2, Width, Height,

Interpolation, MeasureHandle)

Now, all edges that are perpendicular to the ROI are extracted.

measure_pos (Image, MeasureHandle, Sigma, Threshold, Transition, Select,

RowEdge, ColumnEdge, Amplitude, Distance)

The resulting edges enclose the leads (see figure 12). Their distance is returned in the parameter Dis-
tance. In the example, only two edges are returned. Therefore, the parameter Distance contains only
one element, which is the length of the leads at the top of the image. To determine the length of the leads
at the bottom of the image, the same measurements are carried out within the second ROI.
D-10 3 Using the Measure Object

Figure 12: The result of the determination of the lead length.

The parameter Sigma determines the degree of smoothing of the profile along the profile line. The
parameter Threshold defines the threshold that is applied to the derivative of the profile for the selection
of edges. Both parameters are described in section 2.1 on page 3.
The parameter Transition can be used to select edges with a particular transition. If Transition is
set to ’negative’, only edges at which the gray values change from light to dark will be returned. In
the current example, the two edges in the ROI at the top of the image have a negative transition. If
Transition is set to ’positive’, only edges at which the gray values change from dark to light (e.g., the
two edges in the ROI at the bottom of the image) will be returned. To determine all edges, Transition
must be set to ’all’ in this application.
The parameter Select can be used to restrict the result to the ’first’ or ’last’ edge. In the current example,
Select must be set to ’all’.
Besides the distance of the edges, measure_pos also returns their position in the output parameters
RowEdge and ColumnEdge as well as their amplitude (Amplitude). In the example, the position is used
to visualize the edges:

disp_line (WindowHandle, RowEdge, ColumnEdge-Length2, RowEdge,

ColumnEdge+Length2)

3.2 The Position of Edge Pairs

As already noted in section 1 on page 1, whenever the object to measure is enclosed by two edges, you
can use the operator measure_pairs, which groups edges to edge pairs. In the following example, we
use this operator to determine the width of the leads from the image displayed in figure 11a on page 9.
 3.2 The Position of Edge Pairs D-11

Measuring
(a) (b)

Figure 13: (a) The ROI for and (b) a part of the result of the determination of the lead width.

First, a suitable ROI must be defined. Because we want to determine the edges of the individual leads,
the ROI must be oriented perpendicular to the leads (see figure 13a). Here, it is oriented from the left to
the right. With this ROI, a measure object is created as described in section 2 on page 2.
Now, all edge pairs that enclose dark image regions are extracted (measure_ic_leads.dev):

measure_pairs (Image, MeasureHandle, Sigma, Threshold, Transition, Select,

RowEdgeFirst, ColumnEdgeFirst, AmplitudeFirst,
RowEdgeSecond, ColumnEdgeSecond, AmplitudeSecond,
IntraDistance, InterDistance)

The resulting edges are displayed in figure 13b, which shows only a part of the image.
The parameters of measure_pairs are partly identical to the ones of measure_pos (see section 3.1 on
page 8): Sigma determines the degree of smoothing of the profile. The parameter Threshold defines
the threshold that is applied to the derivative of the profile for the selection of edges. Both parameters
are described in section 2.1 on page 3.
The parameter Transition can be used to select edge pairs with a particular transition. If Transition
is set to ’negative’, as in the example, only edge pairs where the first edge has a negative transition
and the second edge has a positive transition, i.e., edge pairs that enclose a dark image region will be
returned. If Transition is set to ’positive’, the behavior is exactly opposite, i.e., edge pairs that enclose
a bright image region will be returned. If Transition is set to ’all’, the transition is determined by the
first edge. I.e., dependent on the positioning of the measure object, edge pairs with a light-dark-light
transition or edge pairs with a dark-light-dark transition are returned. This is suited, e.g., to measure
objects with different brightness relative to the background.
If more than one consecutive edge with the same transition is found, only the first one will be used
as a pair element. This behavior may cause problems in applications in which the Threshold cannot
D-12 3 Using the Measure Object

be selected high enough to suppress consecutive edges of the same transition. This problem can be
demonstrated, e.g., by defining the ROI in the example program measure_switch.dev (see section 1
on page 1) the other way round, i.e., from bottom right to top left. Then, the first edge with a negative
transition corresponds to the shadow of the pin and the edge that corresponds to the pin itself is not used
as a pair element (see figure 14a).
For these applications, an advanced pairing mode exists that only selects the strongest edges of a se-
quence of consecutive edges of the same transition. This mode is selected by appending ’_strongest’
to any of the above modes for Transition, e.g., ’negative_strongest’. Figure 14b shows the result of
setting Transition to ’negative_strongest’.

(a) (b)

Figure 14: The results of measuring the width of and the distance between the pins of a switch if the
ROI is defined from bottom right to top left: (a) Transition = ’negative’, (b) Transition =
’negative_strongest’.

The parameter Select can be used to restrict the result to the ’first’ or ’last’ edge pair. To determine all
edge pairs, Select must be set to ’all’.
The operator measure_pairs returns the coordinates of the edge positions separately for the first and
the second edge of each edge pair. The output parameters RowEdgeFirst and ColumnEdgeFirst
contain the coordinates of the first edges of the edge pairs and the parameters RowEdgeSecond and
ColumnEdgeSecond contain the coordinates of the second edges. With this, the center of the pairs can
easily be determined by calculating the mean of the respective first and second edge positions:

RowPairCenter := (RowEdgeFirst+RowEdgeSecond)/2.0
ColumnPairCenter := (ColumnEdgeFirst+ColumnEdgeSecond)/2.0

The output parameters IntraDistance and InterDistance contain the distance between the two
edges of the edge pairs and the distance between the edge pairs.
 3.3 The Position of Points With a Particular Gray Value D-13

3.3 The Position of Points With a Particular Gray Value

With the operator measure_thresh, it is also possible to determine the subpixel precise positions where
the gray value profile has a given value. This 1D thresholding works like the 2D subpixel precise thresh-
olding operator threshold_sub_pix.
This operator can be useful to extract particular edges in applications where the lighting conditions can
be controlled very accurately. Note that the measured position depends heavily on the brightness of the
image, i.e., the measured position will vary if the brightness level of the image changes, e.g., due to
illumination changes.

3.4 The Gray Projection

For applications that require the determination of special 1D features, it is possible to access the gray
value profile directly via the operator measure_projection. It returns the original, i.e., unsmoothed
gray value profile as a tuple in the output parameter GrayValues. On the profile, points that satisfy a

Measuring
particular condition can be determined. Then, the pixel coordinates of these points can be calculated.
In the example program measure_caliper.dev, the distances between the pitch lines of a caliper are
determined. The used image (figure 15a) shows the pitch lines as approximately one pixel wide lines
(see, e.g., figure 15b). To determine the line positions in the gray value profile, a 1D line extraction is
implemented based on the gray value profile.

(a) (b)

Figure 15: (a) The image of a caliper; (b) detail of (a).

First, the measure object is created. It is positioned such that only the longest pitch lines, i.e., those
which indicate full centimeters run through the measuring ROI.
D-14 3 Using the Measure Object

gen_measure_rectangle2 (Row, Column, Phi, Length1, Length2, Width, Height,

’bilinear’, MeasureHandle)

Then, the profile is determined with the operator measure_projection, which returns the profile in the
output parameter GrayValues:

measure_projection (Image, MeasureHandle, GrayValues)

To reduce the noise, the profile must be smoothed because the operator measure_projection returns
the original, i.e., unsmoothed profile.

Sigma := 0.3
create_funct_1d_array (GrayValues, Function)
smooth_funct_1d_gauss (Function, Sigma, SmoothedFunction)

The smoothed profile is displayed in figure 16.

250

200
gray value

150

100

50

0
0 50 100 150 200 250 300 350 400
profile line

Figure 16: The smoothed profile.

To detect line points on the profile, it is sufficient to determine the points where the first derivative of the
smoothed profile vanishes. The first derivative of the smoothed profile is determined with the operator
derivate_funct_1d with the control parameter Mode set to ’first’; the points where the first derivative
vanishes are determined with the operator zero_crossings_funct_1d:

derivate_funct_1d (SmoothedFunction, ’first’, FirstDerivative)

zero_crossings_funct_1d (FirstDerivative, ZeroCrossings)

The first derivative of the smoothed profile is displayed in figure 17a.

Because of noise, the first derivative shows many more zero crossings than just in the middle of each line,
i.e., between the large negative and positive spikes in figure 17a. We can select the desired zero crossings
by looking at the magnitude of the second derivative: Bright lines on a dark background will have a
negative second derivative while dark lines on a bright background will have a positive second derivative.
The second derivative of the smoothed profile is determined with the operator derivate_funct_1d with
the control parameter Mode set to ’second’:
 3.4 The Gray Projection D-15

15 30

10 20

second derivative
5 10
first derivative

0 0

-5 -10

-10 -20

-15 -30
0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350 400
profile line profile line

(a) (b)

Figure 17: The (a) first and (b) second derivative of the smoothed profile.

Measuring
derivate_funct_1d (SmoothedFunction, ’second’, SecondDerivative)

The second derivative of the smoothed profile is displayed in figure 17b.

In our case, the pitch lines appear as dark lines on a bright background. Therefore, we have to select
those line positions, where the second derivative has a large positive value: For each line position, i.e.,
zero crossing of the first derivative, the value of the second derivative is determined with the operator
get_y_value_funct_1d. If this value is larger than a user defined threshold, the line position will be
stored in a tuple that holds the positions of all salient lines.

MinimumMagnitudeOfSecondDerivative := 8
PositionOfSalientLine := []
for i := 0 to |ZeroCrossings|-1 by 1
get_y_value_funct_1d (SecondDerivative, ZeroCrossings[i], ’constant’,
Y)
if (Y > MinimumMagnitudeOfSecondDerivative)
PositionOfSalientLine := [PositionOfSalientLine,ZeroCrossings[i]]
endif
endfor

Finally, the positions of the lines on the profile must be transformed into pixel coordinates because
measure_projection returns just the gray value profile along the profile line. Note that internally, the
coordinates of the center of the measuring ROI are rounded and the length of the measuring ROI is set
to the largest integer value that is smaller or equal to the given value.
The pixel coordinates of the start point of the profile line can be determined as follows (see figure 18):

RowStart = b(Row + 0.5)c + bLength1c · sin(Phi)

ColStart = b(Column + 0.5)c − bLength1c · cos(Phi)

This can be realized by the following lines of code:

D-16 3 Using the Measure Object

RowStart := floor(Row+0.5)+floor(Length1)*sin(Phi)
ColStart := floor(Column+0.5)-floor(Length1)*cos(Phi)

ROI

End

e
Lin
ile
Prof
Phi

n Center (Row,Column)
io
s it ne
Po l i
of

Start

1
th
ng
Le

Figure 18: The measuring ROI with the position of a point on the profile.

Now, the pixel coordinates of the positions of the salient lines on the profile can be determined as follows:
RowLine = RowStart − PositionOfSalientLine · sin(Phi)
ColLine = ColStart + PositionOfSalientLine · cos(Phi)

In HDevelop, this can be expressed by the following lines of code:

RowLine := RowStart-PositionOfSalientLine*sin(Phi)
ColLine := ColStart+PositionOfSalientLine*cos(Phi)

Figure 19 shows the positions of the determined lines.

In the following section, the pixel coordinates of the pitch lines and the distance between them are
transformed into world coordinates.
If you are using a circular measure object, which has been created with the operator gen_measure_arc,
the following formulae must be used to determine the pixel coordinates RowPoint and ColPoint of a
point on the profile:
RowPoint = b(CenterRow + 0.5)c − Radius · sin(AngleStart + Position)
ColPoint = b(CenterColumn + 0.5)c + Radius · cos(AngleStart + Position)

Here, Position is the position of the point on the profile and CenterRow, CenterColumn, Radius,
and AngleStart are the parameters used for the creation of the measure object with the operator
gen_measure_arc.
 3.5 Measuring in World Coordinates D-17

Measuring
(a) (b)

Figure 19: (a) The determined lines; (b) detail of (a).

3.5 Measuring in World Coordinates

In many applications, the dimensions of the measured object must be determined in world coordinates,
e.g., in µm. This can either be achieved by transforming the original measurement results into world
coordinates (see Solution Guide II-F, section 3.2 on page 40) or, by first rectifying the image and applying
the measurement in the transformed image (see Solution Guide II-F, section 3.3 on page 45). As a
prerequisite, the camera must be calibrated (Solution Guide II-F, section 3.1 on page 25).
In the example program measure_caliper.dev, first the positions of the pitch lines of a caliper are
determined (see section 3.4 on page 13). These positions are now transformed into world coordinates
with the operator image_points_to_world_plane:

image_points_to_world_plane (CamParam, WorldPose, RowLine, ColLine, ’mm’,

X, Y)

The parameter CamParam contains the camera parameters and the parameter WorldPose the position and
orientation of the world plane in which the measurement takes place. Both parameters can be determined
during the camera calibration (Solution Guide II-F, section 3.1 on page 25). In the current example
program, they are assumed to be known.
Now, the distance between the pitch lines can be calculated with the operator distance_pp:
D-18 3 Using the Measure Object

Num := |X|
distance_pp (X[0:Num-2], Y[0:Num-2], X[1:Num-1], Y[1:Num-1], Distance)

The resulting distances are displayed in figure 20.

Figure 20: The distances between the pitch lines in world coordinates.

3.6 Compensate for Radial Image Distortions

If the image suffers from heavy radial distortions, a correction of the image coordinates or a rectification
of the image will be necessary.
A detailed description of how to correct the image coordinates can be found in Solution Guide II-F,
section 3.2.6 on page 44. The description of how to rectify images can be found in the Solution Guide
II-F, section 3.3.2 on page 52.

3.7 Changing the Position and Orientation of the Measure Object

Often, the position of the objects to be measured varies from image to image. In this case, the ROI of the
measure must be adapted accordingly. This process is also called alignment.
The position and orientation of the object to be measured can be determined, e.g., with shape-based
matching. For a detailed description of shape-based matching, please refer to the Solution Guide II-B.
 3.8 Tips and Tricks D-19

This manual also contains an example for aligning a measure ROI (see Solution Guide II-B, section 4.3.4
on page 33).
If only the position of the measure ROI must be changed, the operator translate_measure can be
used. If also the orientation must be adapted as in the above mentioned example in the Solution Guide
II-B, it is necessary to transform the parameters that define the measure ROI and to create a new measure
object with these parameters.

3.8 Tips and Tricks

3.8.1 Distances for Circular ROIs

There are multiple representations for the distance between two edges that are determined with a measure
object that is based on a circular ROI (see figure 21):

• arc length,

Measuring
• linear distance, and

• angular distance.

ROI
Second edge Arc
len
Lin gth
ear
dis
tan
End ce
First edge
Ang
ula
r d
ist
a nce Start
Ra
di
us

Center (CenterRow,CenterColumn)

Figure 21: Distances for circular ROIs.

Measure objects based on a circular ROI return the distances as arc lengths.
The example program measure_ring.dev shows that it is very easy to transform the measurements
into another representation. In this example, the size of the cogs of a cogwheel must be measured (see
figure 22).
First, a measure object is created based on a circular ROI:
D-20 3 Using the Measure Object

Figure 22: The size of the cogs of the ring in the upper left corner of the image given in different repre-
sentations.

gen_measure_arc (CenterRow, CenterCol, Radius, AngleStart,

AngleExtent, AnnulusRadius, Width, Height,
Interpolation, MeasureHandle)

Then, the edges of the cogs are determined with the operator measure_pairs:

measure_pairs (Image, MeasureHandle, Sigma, Threshold, Transition,

Select, RowEdgeFirst, ColumnEdgeFirst,
AmplitudeFirst, RowEdgeSecond, ColumnEdgeSecond,
AmplitudeSecond, IntraDistance, InterDistance)

The size of the cogs, i.e., the distance between the edges of the edge pairs is returned in the output
parameter ’IntraDistance’ as arc length.
The linear distance can be calculated from the coordinates of the edges with the operator distance_pp:

distance_pp (RowEdgeFirst, ColumnEdgeFirst, RowEdgeSecond,

ColumnEdgeSecond, LinearDistance)

The angular distance can be derived from the arc lengths as follows:
Arc length
AngularDistance =
Radius
 3.8.2 ROIs that lie partially outside the image D-21

which is implemented in the HDevelop example program as follows:

AngularDistance := deg(IntraDistance/Radius)

Note that here the angular distance is determined in degrees.

3.8.2 ROIs that lie partially outside the image

Gray value information exists only for positions on the profile line where at least a part of the respective
line of projection (see section 2.1 on page 3) lies inside the image. If parts of an ROI lie outside the
image, the edges inside the image and the distances between them will still be determined correctly.
The example presented in figure 23 demonstrates such a case. Here, the edges of a Siemens star must be
determined with a measure object centered at the center of the Siemens star. In figure 23a, the image as
well as the outline of the ROI are shown. The upper part of the ROI lies outside the image. The distances
between the edges as returned by the measure object are displayed in figure 23b.

Measuring
(a) (b)

Figure 23: Image of a Siemens star: (a) The circular ROI lies partially outside the image; (b) the measured
distances between the edges.

Note that at positions where no gray value information is available, the profile contains zeros. Figure 24
shows the gray value profile determined with the operator measure_projection from the measure
object that was created with the above described circular ROI (see figure 23a).
D-22 4 Advanced Measuring: The Fuzzy Measure Object

250

200

gray value
150

100

50

0
0 100 200 300 400 500 600 700
profile line

Figure 24: The gray value profile from the circular ROI that lies partially outside the image.

3.8.3 Accuracy vs. Speed

By default, the averaging of gray values (see section 2.1 on page 3) is performed using fixed point
arithmetic, to achieve maximum speed. You can trade some of the speed against accuracy by setting the
system parameter ’int_zooming’ to ’false’ with the operator set_system. Then the internal calculations
are performed using floating point arithmetic. Note that the gain in accuracy is minimal.
See also the description of the parameter Interpolation at section 2.1 on page 3.

3.8.4 Image Type

Measure objects can only be applied to images of type ’byte’ and ’uint2’.
Note that if a multi-channel image is passed to the measuring operators (e.g., measure_pos), only the
first channel of the image will be used.

3.9 Destroying the Measure Object

If the measure object is not longer needed in the program, it should be destroyed with the operator
close_measure:

close_measure (MeasureHandle)

This deletes the measure object and frees all of the associated memory.

4 Advanced Measuring: The Fuzzy Measure Object

Some applications require that the selection of the edges or edge pairs can be controlled in more detail.
With the standard measure object (see above), edges or edge pairs can be selected only based on their
 4.1 A First Example D-23

contrast and transition. With a fuzzy measure object it is possible to select edges also based on their
contrast and position. The selection of edge pairs can be additionally controlled based on their position
and size, as well as based on the gray value between the two edges of the pair. The name fuzzy measure
object does not mean that the measurements are “fuzzy”; it comes from the so-called “fuzzy membership
functions” that control the selection of edges (see section 4.2.2 on page 27).
In the following section, a first example demonstrates how to create and use a fuzzy measure object. The
features that can be used to control the selection of edges and edge pairs are described in section 4.2 on
page 25 as well as the fuzzy membership functions, which are used to formalize the desired constraints.
Finally, section 4.3 on page 29 shows how to create a fuzzy measure object from a standard measure
object and one or more fuzzy membership functions.

4.1 A First Example

The fuzzy measure object is an extension to the standard measure object. It allows to control the selection
of edges by specifying weighting functions.
The example program fuzzy_measure_switch.dev shows a possible usage of the fuzzy measure ob-
ject. Again, the width of and the distance between the pins of a switch must be measured. In this case,
there are reflections on the middle pin (see figure 25), which will lead to wrong results if the standard
measure object is used (figure 26a).

Fuzzy Measure
Figure 25: Image of a switch with reflections on the middle pin.

The selection of edges can be improved if additional information is available. In the example, the pins
are approximately 9 pixels wide. This information can be translated into a fuzzy membership function.
Figure 27 shows this fuzzy membership function. It returns 1.0 for the desired pair size and 0.0 if the
size of the pairs deviates more than two pixels from the desired pair size. The intermediate values are
interpolated linearly (see figure 27). In HALCON, this fuzzy membership function is created with the
operator create_funct_1d_pairs:

create_funct_1d_pairs ([7,9,11], [0.0,1.0,0.0], SizeFunction)

D-24 4 Advanced Measuring: The Fuzzy Measure Object

(a) (b)

Figure 26: The results of measuring the width of and the distance between the pins of the switch with (a)
a standard measure object and (b) a fuzzy measure object .

0.8
membership value

0.6

0.4

0.2

5 6 7 8 9 10 11 12 13
feature value

Figure 27: The fuzzy membership function that is used in the current example to describe the constraint
on the pair width.

The standard measure object is transformed into a fuzzy measure object with the operator
set_fuzzy_measure. As parameters, you pass the handle of the standard measure object, the fuzzy
membership function, and the name of the feature to which the fuzzy membership function applies.

SetType := ’size’
set_fuzzy_measure (MeasureHandle, SetType, SizeFunction)

The resulting fuzzy measure object can be used to extract edge pairs that are approximately 9 pixels wide,
with the term “approximately” being defined by the fuzzy membership function.
Then, the fuzzy measure object is applied to extract the desired edge pairs:
 4.2 Controlling the Selection of Edges and Edge Pairs D-25

Sigma := 0.9
AmpThresh := 12
FuzzyThresh := 0.5
Transition := ’negative’
Select := ’all’
fuzzy_measure_pairs (Image, MeasureHandle, Sigma, AmpThresh, FuzzyThresh,
Transition, RowEdgeFirst, ColumnEdgeFirst,
AmplitudeFirst, RowEdgeSecond, ColumnEdgeSecond,
AmplitudeSecond, RowEdgeCenter, ColumnEdgeCenter,
FuzzyScore, IntraDistance, InterDistance)

Internally, the specified fuzzy membership function is used to weight all the possible edge pairs. The
width of each possible edge pair is determined and transformed into a fuzzy value. If this fuzzy value
lies above the user defined threshold FuzzyThresh, the edge pair is accepted, otherwise it is rejected.
The result is displayed in figure 26b on page 24. With this, the width of all pins is measured correctly.

4.2 Controlling the Selection of Edges and Edge Pairs

The selection of edges and edge pairs can be controlled based on various features. Section 4.2.1 gives
an overview over the possible features. The constraints on the feature values are defined with fuzzy
membership functions. A short introduction into the concept of fuzzy logic and how fuzzy membership
functions are created in HALCON is given in section 4.2.2 on page 27.

Fuzzy Measure
4.2.1 Features that Can Be Used to Control the Selection of Edges and Edge Pairs

In the example program fuzzy_measure_switch.dev described above, the selection of edge pairs has
been controlled by the approximately known pair size. There are further features according to which
the edges or edge pairs can be selected. These features are grouped into so-called set types (parameter
SetType of the operators set_fuzzy_measure set_fuzzy_measure_norm_pair).
The set types that describe positions or sizes can be specified either with fuzzy membership functions
that directly describe the constraint on the feature value like the one displayed in figure 27 on page 24
(see section 4.2.2 on page 27) or with normalized fuzzy membership functions, i.e., fuzzy membership
functions that are defined relative to the desired pair width (section 4.3.2 on page 30). The first method
uses the operator set_fuzzy_measure to specify the fuzzy membership function. The second method
uses the operator set_fuzzy_measure_norm_pair, which enables a generalized usage of the fuzzy
membership function because the pair size is set independently of the fuzzy membership function.
The set types that refer to features of edge pairs will only take effect if edge pairs are determined, i.e., if
the operators fuzzy_measure_pairs or fuzzy_measure_pairing are used.
In the following, a list of the possible set types is given. It is possible to use multiple set types. Details
can be found in section 4.3.3 on page 32. Note that the subtypes of each set are mutually exclusive, i.e.,
only one subtype of each set can be specified for a fuzzy measure object.
D-26 4 Advanced Measuring: The Fuzzy Measure Object

Set Type Subtype Function Type Short Description

Direct Norm.
contrast ’contrast’ • - Evaluates the amplitude of the
edges.
position ’position’ • • Evaluates the signed distance of
the edges to the reference point
of the fuzzy measure object.
The reference point is located at
the start of the profile line (see
figure 9 on page 7 and figure 10
on page 8).
’position_center’ • • Like ’position’ with the refer-
ence point located at the center
of the profile line.
’position_end’ • • Like ’position’ with the refer-
ence point located at the end of
the profile line.
’position_first_edge’ • • Like ’position’ with the refer-
ence point located at the first
edge.
’position_last_edge’ • • Like ’position’ with the refer-
ence point located at the last
edge.
position_pair ’position_pair’ • • Evaluates the signed distance of
(for edge pairs the edge pairs to the reference
only) point of the fuzzy measure ob-
ject. The position of an edge
pair is defined as the center of
the two edges. The reference
point is located at the start of the
profile line (see figure 9 on page
7 and figure 10 on page 8).
’position_pair_center’ • • Like ’position_pair’ with the
reference point located at the
center of the profile line.
’position_pair_end’ • • Like ’position_pair’ with the
reference point located at the
end of the profile line.
’position_first_pair’ • • Like ’position_pair’ with the
reference point located at the
position of the first edge pair.
’position_last_pair’ • • Like ’position_pair’ with the
reference point located at the
position of the last edge pair.
size (for edge ’size’ • • Evaluates the size of the edge
pairs only) pairs, i.e., the distance between
the two edges.
 4.2.2 Defining the Constraints on the Feature Values with Fuzzy Membership Functions D-27

Set Type Subtype Function Type Short Description

Direct Norm.
’size_diff’ - • Evaluates the signed difference
between the desired PairSize
and the actual size of the edge
pairs.
’size_abs_diff’ - • Evaluates the absolute differ-
ence between the desired Pair-
Size and the actual size of the
edge pairs.
gray (for edge ’gray’ • - Evaluates the mean gray value
pairs only) between the two edges of edge
pairs.

Note that for the extraction of edge pairs all possible pairs of edges are internally analyzed according
to the specified fuzzy membership functions. If only partial knowledge about the edge pairs to be de-
termined is specified with the fuzzy membership functions, this will possibly lead to unintuitive results.
For example, if the position of the edge pairs is specified but not their size, some small edge pairs will
possibly be omitted because of a large edge pair that consists of the first edge of the first small edge pair
and the second edge of the last small edge pair and that has a membership value equal to or higher than
that of the small edge pairs (see appendix B.3.1 on page 48). Therefore, all the available knowledge
about the edge pairs to be extracted should be specified with fuzzy membership functions. In the above

Fuzzy Measure
example, the definition of the desired edge pairs can be made more clear by also specifying the size of
the edge pairs (see appendix B.6 on page 59).

4.2.2 Defining the Constraints on the Feature Values with Fuzzy Membership Functions

The constraints for the selection of edges and edge pairs are defined with fuzzy membership functions.
These functions transform the feature values into membership values. The decision whether an edge or
an edge pair is accepted is made based on these membership values. For this so-called defuzzyfication,
the membership values are compared with a given threshold.
This section describes the concept of fuzzy membership functions and how they are created in HALCON
as well as the defuzzyfication.

Fuzzy Membership Functions

Fuzzy logic handles the concept of partial truth, i.e., truth values between “completely true” and “com-
pletely false”. A fuzzy set is a set whose elements have degrees of membership. The degree of mem-
bership is defined by the feature value of the element and by the fuzzy membership function, which
transforms the feature value into a membership value, i.e., the degree of membership.
With this, fuzzy logic provides a framework for the representation of knowledge that is afflicted with
uncertainties or can only be described vaguely. This applies often to knowledge that is expressed in
natural language, e.g., “the distance between the two edges of a pair is approximately 9 pixel,” or “the
edges are strong and they are located at the beginning of the profile line.”
D-28 4 Advanced Measuring: The Fuzzy Measure Object

For example, a fuzzy membership function for the requirement that the edges must be strong could look
like the one displayed in figure 28a. Here, any edge that has an amplitude of less than 50 will not be part
of the fuzzy set of strong edges. Edges with an amplitude higher than 100 are full members of the fuzzy
set, i.e, they have a membership value of 1.0. Edges with amplitudes between 50 and 100 are partial
members of the fuzzy set, e.g., an edge with an amplitude of 75 will have a membership value of 0.5.

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 50 100 150 200 250 0 10 20 30 40 50 60 70 80

feature value for ’contrast’ feature value for ’position’

(a) (b)

Figure 28: Examples for fuzzy membership functions. They can, e.g., be used to select (a) strong edges
or (b) edges at the beginning of the profile line. The dots indicate the points to pass to
create_funct_1d_pairs .

In general, a fuzzy membership function is a mathematical function that defines the degree of an ele-
ment’s membership in a fuzzy set.
In HALCON, fuzzy membership functions are defined as piecewise linear functions by at least two
pairs of values, sorted in an ascending order by their x value. They are created with the operator cre-
ate_funct_1d_pairs:

create_funct_1d_pairs (XValues, YValues, FuzzyMembershipFunction)

The x values represent the feature of the edges or edge pairs and must lie within the parameter space of
the set type, i.e., in case of ’contrast’ and ’gray’ features and, e.g., byte images within the range 0.0 ≤ x
≤ 255.0. For ’size’, x has to satisfy 0.0 ≤ x, whereas for ’position’ and ’position_pair’, x can be any real
number. The y values of the fuzzy membership function represent the resulting membership value for the
corresponding feature value and have to satisfy 0.0 ≤ y ≤ 1.0. Outside of the function’s interval, defined
by the smallest and the largest specified x value, the y values of the interval borders are continued with
constant values equal to the closest defined y value.
Therefore, the fuzzy membership function displayed in figure 28a is defined by the following code,
which specifies the coordinates of the two control points of the fuzzy membership function.

create_funct_1d_pairs ([50,100], [0,1], FuzzyMembershipFunction1)

A fuzzy membership function describing the condition that the edge must be at the beginning of the
profile line could look like the one displayed in figure 28b (assuming that the profile line has a length of
80 pixels). It can be defined as follows:
 4.3 Creating the Fuzzy Measure Object D-29

Length := 80
create_funct_1d_pairs ([Length/8.0, Length/2.0], [1,0],
FuzzyMembershipFunction2)

Further examples for fuzzy membership functions can be found in the appendix B on page 40.

Defuzzyfication
Finally, to decide wether an edge or an edge pair meets the constraint, the membership value must be
“defuzzyfied”. This is done by comparing it with a user-defined fuzzy threshold. If the membership
value lies below this threshold, the edge (pair) will be rejected. Otherwise, it will be accepted.
If, e.g., a fuzzy threshold of 0.5 is assumed as in figure 29, all edges that have a membership value of at
least 0.5 will be accepted. For the fuzzy membership function displayed in figure 29a, these are all edges
with a feature value of at least 75; for the fuzzy membership function of figure 29b, edges will only be
accepted if their feature value is at most 25.

1 1

0.8 0.8
membership value

membership value

0.6 0.6

0.4 0.4

Fuzzy Measure
0.2 0.2

0 0

0 50 100 150 200 250 0 10 20 30 40 50 60 70 80

feature value for ’contrast’ feature value for ’position’

(a) (b)

Figure 29: Examples for fuzzy membership functions (thick lines) together with a fuzzy threshold at 0.5
(thin lines).

The specification of the fuzzy membership function and the definition of the fuzzy threshold are sep-
arated for fuzzy measure objects. The fuzzy membership function is specified with the operator
set_fuzzy_measure or set_fuzzy_measure_norm_pair, respectively, i.e., during the creation of
the fuzzy measure object (see section 4.3). In contrast, the fuzzy threshold is defined when the fuzzy
measure object is used with one of the operators fuzzy_measure_pos, fuzzy_measure_pairs, or
fuzzy_measure_pairing. This separation between the specification of the fuzzy membership func-
tion and the definition of the fuzzy threshold allows a flexible use of the fuzzy measure object: Often, it
will be sufficient to adjust only the fuzzy threshold to varying imaging conditions.

4.3 Creating the Fuzzy Measure Object

This section describes how to create a fuzzy measure object from an existing measure object. Fuzzy
measure objects can be created by specifying a fuzzy membership function as described in section 4.3.1.
D-30 4 Advanced Measuring: The Fuzzy Measure Object

Fuzzy measure objects can also be created with normalized fuzzy membership functions, i.e., fuzzy
membership functions that are defined relative to the desired pair width (see section 4.3.2). Section 4.3.3
on page 32 shows that it is even possible to specify multiple fuzzy membership functions for different
set types of one fuzzy measure object.

4.3.1 Specifying a Fuzzy Membership Function

To transform a standard measure object into a fuzzy measure object, you specify a fuzzy membership
function with the operator set_fuzzy_measure.
For example, you can create a fuzzy measure object for the extraction of edge pairs that have a size of
approximately 10 pixels with the following lines of code:

create_funct_1d_pairs ([5,10,15], [0,1,0],

FuzzyMembershipFunctionPairSize10)
set_fuzzy_measure (MeasureHandle, ’size’,
FuzzyMembershipFunctionPairSize10)

The fuzzy membership function used in this example is displayed in figure 30.

0.8
membership value

0.6

0.4

0.2

0 5 10 15 20
feature value

Figure 30: Fuzzy membership function, which can be used to extract edge pairs that are approximately
10 pixels wide.

4.3.2 Specifying a Normalized Fuzzy Membership Function

Fuzzy measure objects can also be created with normalized fuzzy membership functions, i.e., fuzzy
membership functions that are defined relative to the desired pair width. For this, the operator
set_fuzzy_measure_norm_pair must be used.
To create a fuzzy measure object that is identical with the one created in the above example (see sec-
tion 4.3.1), the following code can be used:
 4.3.2 Specifying a Normalized Fuzzy Membership Function D-31

create_funct_1d_pairs ([0.5,1,1.5], [0,1,0],

FuzzyMembershipFunctionPairSizeNormalized)
PairSize := 10
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, ’size’,
FuzzyMembershipFunctionPairSizeNormalized)

Here, the normalized fuzzy membership function is used, which is displayed in figure 31. Internally, the
operator set_fuzzy_measure_norm_pair multiplies the x values of the normalized fuzzy membership
function with the given pair size.

0.8
membership value

0.6

0.4

0.2

0 0.5 1 1.5 2
normalized feature value

Fuzzy Measure
Figure 31: Normalized fuzzy membership function, which can be used for the extraction of edge pairs of
various sizes, according to the defined pair size.

One major advantage of using the normalized form of fuzzy membership functions is that the application
can very easily be adapted to varying sizes of the object to be measured. If, e.g., larger objects must be
measured or if a camera with a higher resolution is used, it will suffice to adapt the pair size.
D-32 4 Advanced Measuring: The Fuzzy Measure Object

If, e.g., the above example for the extraction of 10 pixel wide edge pairs must be adapted to the extraction
of 20 pixel wide edge pairs, simply the definition of the pair size must be changed:

PairSize := 20
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, ’size’,
FuzzyMembershipFunctionPairSizeNormalized)

This creates a fuzzy measure object that uses the fuzzy membership function displayed in figure 32,
internally.

0.8
membership value

0.6

0.4

0.2

0 5 10 15 20 25 30 35 40
feature value

Figure 32: Fuzzy membership function, which can be used for the extraction of approximately 20 pixel
wide edge pairs.

To achieve such a fuzzy measure object without normalized fuzzy membership functions, the definition
of the fuzzy membership function itself must be adapted:

create_funct_1d_pairs ([10,20,30], [0,1,0],

FuzzyMembershipFunctionPairSize20)
set_fuzzy_measure (MeasureHandle, ’size’,
FuzzyMembershipFunctionPairSize20)

4.3.3 Specifying Multiple Fuzzy Membership Functions

You can also specify fuzzy membership functions for different set types by repeatedly calling the operator
set_fuzzy_measure or set_fuzzy_measure_norm_pair, respectively:

set_fuzzy_measure (MeasureHandle, ’contrast’, FuzzyMembershipFunction1)

set_fuzzy_measure (MeasureHandle, ’position’, FuzzyMembershipFunction2)

If more than one set is defined in this way, the individual membership values will be aggregated with a
fuzzy and operator: The overall membership value is derived by the geometric mean (see appendix C on
page 60) of the individual membership values of each set.
 4.3.3 Specifying Multiple Fuzzy Membership Functions D-33

For example, the two set types ’contrast’ and ’position’ are used to achieve the selection of strong
edges at the beginning of the profile line. For this, the two fuzzy membership functions as displayed in
figure 28a and b on page 28 are used.
Assuming an edge at the position 20 along the profile line that has an amplitude of 80, the membership
value is determined as follows. First, the membership values of the individual fuzzy sets are determined:
membership valueposition=20 = 0.67
membership valuecontrast=80 = 0.60

Then, the geometric mean is calculated:

q
membership value20/80 = membership valueposition=20 · membership valuecontrast=80
= 0.63

This value is compared to the fuzzy threshold. If it lies above this threshold, the respective edge will be
accepted, otherwise it will be rejected.
Figure 33a shows a plot of the membership value over the two feature values for ’contrast’ and ’position’.
Assuming a fuzzy threshold of 0.5, the thick line in figure 33b marks the boundary of the domain of
the feature values that lead to an acceptance of the respective edge. All edges with position/amplitude
combinations that lie in this domain in the upper left corner will be accepted. The thin line marks the
domain that results if both constraints must be satisfied individually. As can be seen, the domain that
results from the combination of several set types is a bit larger. In particular, if one feature value meets

Fuzzy Measure
the requirement very well, the other value may be slightly worse than allowed by the respective individual
constraint.

250

membership value 200

edge amplitude

1
0.8 150
0.6
0.4
0.2 100
0
250
200 50
0 10 150
20 30 100
40 50 50 edge amplitude
edge position 60 70 80 0 0
0 10 20 30 40 50 60 70 80
edge position

(a) (b)

Figure 33: Membership value as a function of two feature values: (a) The membership value; (b) the
boundary of the domain where edges will be accepted (thick line) compared to the boundary
of the domain if the two constraints must be satisfied individually.

To illustrate this, an edge at the position 10 having an amplitude of 65 is assumed. The individual
membership values are:
membership valueposition=10 = 1.00
membership valuecontrast=65 = 0.30
D-34 4 Advanced Measuring: The Fuzzy Measure Object

With this, the membership value can be calculated:

q
membership value10/65 = membership valueposition=10 · membership valuecontrast=65
= 0.54

The overall membership value lies above 0.5 although the individual membership value for the contrast
of the edge lies well below this value.
Note that it is not possible to specify multiple fuzzy membership functions for one and the same set
type. Setting a further fuzzy membership function to a set discards the previously defined function and
replaces it by the new one.

4.3.4 Changing and Resetting a Fuzzy Measure Object

Fuzzy membership functions that have been specified for a particular set type can be changed and reset.
To change a fuzzy membership function for a set type, simply specify the new fuzzy membership function
for this set type with one of the operators set_fuzzy_measure and set_fuzzy_measure_norm_pair,
as described in section 4.3.1 on page 30.
A fuzzy measure object can be reset with the operator reset_fuzzy_measure. This discards the fuzzy
membership function of the specified fuzzy set SetType. For example, the fuzzy membership function
of the set type ’contrast’ is discarded with the following lines of code:

SetType := ’contrast’
reset_fuzzy_measure (MeasureHandle, SetType)

4.4 Using the Standard Measure Object Functionality

All operators that provide functionality for the standard measure object can also be applied to fuzzy mea-
sure objects. This includes especially the operators measure_projection, measure_thresh, trans-
late_measure, and close_measure. The operators for the standard measure object ignore the speci-
fied fuzzy membership functions and return the same results as if they were applied to a standard measure
object.
 A Slanted Edges D-35

Appendix

A Slanted Edges

This section shows the effect of the width of the ROI and the angle between the profile line and the edges.
In the figures on the next pages, the images on the left hand show the size and orientation of the ROI and
the diagrams on the right hand show the resulting profile (thick line) and its derivative (thin line, Sigma
= 0.9).
As long as the profile line runs approximately perpendicular to the edges, the ROI should be chosen as
wide as possible because then the profile is less noisy (compare, e.g., figure 35a and figure 35c).
If the profile line does not run perpendicular to the edges (see figure 36 on page 37 and figure 37 on page
38), the width of the ROI must be chosen smaller to allow the detection of edges. Note that in this case,
the profile will contain more noise. Consequently, the edges will be determined less accurately. What is
more, the distance between the determined edges does not represent the perpendicular, i.e., the shortest
distance between the respective image edges (see figure 34). If the angle δ between the profile line and
the perpendicular of the edges is known, the perpendicular distance can be determined from the distance
between the determined edges (Distance) as follows:

perpendicular distance = cos(δ) · Distance

ROI

ce
an
st
Di
δ

Appendix
e Perpendicular distance
Lin between the edges
le
ofi
Pr

Figure 34: Relationship between the determined Distance and the perpendicular distance between
edges.

If the profile line is heavily slanted with respect to the edges, it may become impossible to determine the
edges reliably, even with a very narrow ROI (see, e.g., figure 38a on page 39).
D-36 A Slanted Edges

250

200

150

100

gray value
50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(a) Width of the ROI: 6 pixels.

250

200

150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(b) Width of the ROI: 30 pixels.

250

200

150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(c) Width of the ROI: 60 pixels.

Figure 35: Angle between the profile line and the perpendicular of the edges: 0◦ .
 A Slanted Edges D-37

250

200

150

100

gray value
50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(a) Width of the ROI: 6 pixels.

250

200

150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(b) Width of the ROI: 30 pixels.

250

200

Appendix
150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(c) Width of the ROI: 60 pixels.

Figure 36: Angle between the profile line and the perpendicular of the edges: 15◦ .
D-38 A Slanted Edges

250

200

150

100

gray value
50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(a) Width of the ROI: 6 pixels.

250

200

150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(b) Width of the ROI: 30 pixels.

250

200

150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(c) Width of the ROI: 60 pixels.

Figure 37: Angle between the profile line and the perpendicular of the edges: 30◦ .
 A Slanted Edges D-39

250

200

150

100

gray value
50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(a) Width of the ROI: 6 pixels.

250

200

150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(b) Width of the ROI: 30 pixels.

250

200

Appendix
150

100
gray value

50

-50

-100

-150
0 20 40 60 80 100 120 140
profile line

(c) Width of the ROI: 60 pixels.

Figure 38: Angle between the profile line and the perpendicular of the edges: 45◦ .
D-40 B Examples for Fuzzy Membership Functions

B Examples for Fuzzy Membership Functions

In this section, examples for fuzzy membership functions are given. For each set type (see section 4.2.1
on page 25), the typical range of the x values of the fuzzy membership functions as well as one fuzzy
membership function is given. If the set type allows the creation of the fuzzy measure object with fuzzy
membership functions and normalized fuzzy membership functions, both fuzzy membership functions
are displayed. In this case, the normalized fuzzy membership function is defined such that, together with
the stated PairSize, it creates the same fuzzy measure object as the direct fuzzy membership function.
Then, mostly a PairSize of 10.0 is used to make it easy to compare the two fuzzy membership functions.
The effect of the specification of the fuzzy membership function is shown by means of the results of the
fuzzy measure object applied to the test image shown in figure 39.

Figure 39: The test image that is used to show the effect of the different fuzzy membership functions .

The results of the standard measure object applied to this test image are shown in figure 40.

(a) (b)

Figure 40: Result of the standard measure object applied to the test image: (a) Edges determined with
the operator measure_pos, (b) edge pairs determined with the operator measure_pairs .

Figure 40a shows the edges determined with the operator measure_pos and figure 40b shows the edge
pairs determined with the operator measure_pairs. Both operators were applied with the following
parameter settings:

Parameter Value
Sigma 1.0
Threshold 10.0
Transition ’all’
Select ’all’
FuzzyThresh 0.5

The following examples shall give an impression of the effect of different fuzzy membership functions
specified for different set types. They are in no way complete. Note that all the edge pairs are extracted
with the operator fuzzy_measure_pairs, not with the operator fuzzy_measure_pairing.
 B.1 Set Type contrast D-41

B.1 Set Type contrast

B.1.1 Subtype ’contrast’

Evaluates the amplitude of the edges.

The x values of fuzzy membership functions for the sub set type ’contrast’ must lie within the range of
gray values, i.e.,

0.0 ≤ x ≤ 255.0 for ’byte’ images and

0.0 ≤ x ≤ 65535.0 for ’uint2’ images.

Figure 41 shows a fuzzy membership function that can be used to extract strong edges in ’byte’ images.

0.8
membership value

0.6

0.4
The sub set type ’contrast’ cannot be specified
0.2 with a normalized fuzzy membership function.
0

0 50 100 150 200 250

feature value

(a) (b)

Figure 41: A fuzzy membership function that can be used to extract strong edges in ’byte’ images (SetType
= ’contrast’).

Figure 42 shows the positions of the edges that will be extracted if the the fuzzy measure object is created
with the fuzzy membership function displayed in figure 41. With this fuzzy measure object, only those
edges are extracted that have a contrast between 75 and 125.

Appendix
Figure 42: The extracted edges.

The fuzzy membership function displayed in figure 41a can be created with the following HDevelop
code:

SetType := ’contrast’
create_funct_1d_pairs ([50,100,150], [0,1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)
D-42 B Examples for Fuzzy Membership Functions

B.2 Set Type position

B.2.1 Subtype ’position’

Evaluates the signed distance of the edges to the reference point of the fuzzy measure object. The
reference point is located at the start of the profile line (see figure 9 on page 7 and figure 10 on page 8).
The x values of fuzzy membership functions for the sub set type ’position’ must lie within the range

0.0 ≤ x ≤ length of the ROI.

Figure 43 shows fuzzy membership functions that can be used to extract edges at the beginning of the
profile line.

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 100 200 300 400 500 0 10 20 30 40 50

feature value normalized feature value

(a) (b)

Figure 43: Fuzzy membership functions that can be used to extract edges at the beginning of the profile
line (SetType = ’position’, PairSize = 10).

Figure 44 shows the positions of the edges that will be extracted if the the fuzzy measure object is created
with the fuzzy membership function displayed in figure 43. Only the first four edges are returned.

Figure 44: The extracted edges.

The fuzzy membership function displayed in figure 43a can be created with the following HDevelop
code:

SetType := ’position’
create_funct_1d_pairs ([50,130], [1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)
 B.2.2 Subtype ’position_center’ D-43

The equivalent normalized fuzzy membership function, which is displayed in figure 43b can be created
with the following HDevelop code:

SetType := ’position’
PairSize := 10
create_funct_1d_pairs ([5,13], [1,0], NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.2.2 Subtype ’position_center’

’position_center’ behaves like ’position’ with the reference point located at the center of the profile line.
The x values of fuzzy membership functions for the sub set type ’position_center’ must lie within the
range

length of the ROI length of the ROI

− 2 ≤x≤ 2 .

Figure 45 shows fuzzy membership functions that can be used to extract edges in the center of the profile
line.

1 1

0.8 0.8
membership value

membership value

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-200 -100 0 100 200 -25 -20 -15 -10 -5 0 5 10 15 20 25

feature value normalized feature value

(a) (b)

Figure 45: Fuzzy membership functions that can be used to extract edges in the center of the profile line
(SetType = ’position_center’, PairSize = 10).

Appendix
Figure 46 shows the positions of the edges that will be extracted if the the fuzzy measure object is created
with the fuzzy membership function displayed in figure 45. Only the edges in the center of the profile
line are returned.
The fuzzy membership function displayed in figure 45a can be created with the following HDevelop
code:
SetType := ’position_center’
create_funct_1d_pairs ([-70,-20,20,70], [0,1,1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)
D-44 B Examples for Fuzzy Membership Functions

Figure 46: The extracted edges.

The equivalent normalized fuzzy membership function, which is displayed in figure 45b can be created
with the following HDevelop code:

SetType := ’position_center’
PairSize := 10
create_funct_1d_pairs ([-7,-2,2,7], [0,1,1,0],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.2.3 Subtype ’position_end’

’position_end’ behaves like ’position’ with the reference point located at the end of the profile line.
The x values of fuzzy membership functions for the sub set type ’position_end’ must lie within the range

−length of the ROI ≤ x ≤ 0.0.

Figure 47 shows fuzzy membership functions that can be used to extract edges at the end of the profile
line.

1 1

0.8 0.8
membership value

membership value

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-500 -400 -300 -200 -100 0 -50 -40 -30 -20 -10 0
feature value normalized feature value

(a) (b)

Figure 47: Fuzzy membership functions that can be used to extract edges at the end of the profile line
(SetType = ’position_end’, PairSize = 10).

Figure 48 shows the positions of the edges that will be extracted if the the fuzzy measure object is created
with the fuzzy membership function displayed in figure 47. Only the edges at the end of the profile line
are returned.
 B.2.4 Subtype ’position_first_edge’ D-45

Figure 48: The extracted edges.

The fuzzy membership function displayed in figure 47a can be created with the following HDevelop
code:

SetType := ’position_end’
create_funct_1d_pairs ([-250,0], [0,1], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 47b on page 44 can
be created with the following HDevelop code:

SetType := ’position_end’
PairSize := 10
create_funct_1d_pairs ([-25,0], [0,1], NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.2.4 Subtype ’position_first_edge’

’position_first_edge’ behaves like ’position’ with the reference point located at the first edge.
The x values of fuzzy membership functions for the sub set type ’position_first_edge’ must lie within the
range

0.0 ≤ x ≤ length of the ROI.

Figure 49 shows fuzzy membership functions that can be used to extract edges close to the first edge.
Figure 50 shows the positions of the edges that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 49. Only the edges at the beginning of
the profile line are returned. Note that even though the same fuzzy membership function has been used
as in figure 43 on page 42, one more edge has been extracted. This is because the reference point is
located at the first edge instead of the start of the profile line. Appendix

The fuzzy membership function displayed in figure 49a can be created with the following HDevelop
code:

SetType := ’position_first_edge’
create_funct_1d_pairs ([50,130], [1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)
D-46 B Examples for Fuzzy Membership Functions

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 100 200 300 400 500 0 10 20 30 40 50

feature value normalized feature value

(a) (b)

Figure 49: Fuzzy membership functions that can be used to extract edges close to the first edge (SetType
= ’position_first_edge’, PairSize = 10).

Figure 50: The extracted edges.

The equivalent normalized fuzzy membership function, which is displayed in figure 49b can be created
with the following HDevelop code:

SetType := ’position_first_edge’
PairSize := 10
create_funct_1d_pairs ([5,13], [1,0], NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.2.5 Subtype ’position_last_edge’

’position_last_edge’ behaves like ’position’ with the reference point located at the last edge.
The x values of fuzzy membership functions for the sub set type ’position_last_edge’ must lie within the
range

−length of the ROI ≤ x ≤ 0.0.

Figure 51 shows fuzzy membership functions that can be used to extract edges close to the last edge.
Figure 52 shows the positions of the edges that will be extracted if the the fuzzy measure object is created
with the fuzzy membership function displayed in figure 51. Only the edges at the end of the profile line
are returned. Note that even though the same fuzzy membership function has been used as in figure 47
 B.3 Set Type position_pair D-47

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

-500 -400 -300 -200 -100 0 -50 -40 -30 -20 -10 0
feature value normalized feature value

(a) (b)

Figure 51: Fuzzy membership functions that can be used to extract edges close to the last edge (SetType
= ’position_last_edge’, PairSize = 10).

Figure 52: The extracted edges.

on page 44, more edges have been extracted. This is because the reference point is located at the last
edge instead of the end of the profile line.
The fuzzy membership function displayed in figure 51a can be created with the following HDevelop
code:
SetType := ’position_last_edge’
create_funct_1d_pairs ([-250,0], [0,1], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 51b can be created
with the following HDevelop code:

SetType := ’position_last_edge’
PairSize := 10
create_funct_1d_pairs ([-25,0], [0,1], NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType, Appendix
NormalizedFuzzyMembershipFunction)

B.3 Set Type position_pair

Note that for the extraction of edge pairs all possible pairs of edges are internally analyzed according
to the specified fuzzy membership functions. If only partial knowledge about the edge pairs to be de-
termined is specified with the fuzzy membership functions, this will possibly lead to unintuitive results.
D-48 B Examples for Fuzzy Membership Functions

For example, if the position of the edge pairs is specified but not their size, some small edge pairs will
possibly be omitted because of a large edge pair that consists of the first edge of the first small edge pair
and the second edge of the last small edge pair and that has a membership value equal to or higher than
that of the small edge pairs (see appendix B.3.1). Therefore, all the available knowledge about the edge
pairs to be extracted should be specified with fuzzy membership functions. In the above example, the
definition of the desired edge pairs can be made more clear by also specifying the size of the edge pairs
(see appendix B.6 on page 59).

B.3.1 Subtype ’position_pair’

Evaluates the signed distance of the edge pairs to the reference point of the fuzzy measure object. The
position of an edge pair is defined as the center of the two edges. The reference point is located at the
start of the profile line (see figure 9 on page 7 and figure 10 on page 8).
The x values of fuzzy membership functions for the sub set type ’position_pair’ must lie within the range

0.0 ≤ x ≤ length of the ROI.

Figure 53 shows fuzzy membership functions that can be used to extract edge pairs in approximately the
first half of the profile line.

1 1

0.8 0.8
membership value

membership value

0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 100 200 300 400 500 0 10 20 30 40 50

feature value normalized feature value

(a) (b)

Figure 53: Fuzzy membership functions that can be used to extract edge pairs in approximately the first
half of the profile line (SetType = ’position_pair’, PairSize = 10).

Figure 54 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 53. Only edge pairs in approximately
the first half of the profile line are returned. Note that one large edge pair is extracted instead of multiple
small ones. This happens because internally, all possible pairs of edges are analyzed according to the
specified fuzzy membership functions and no condition for the size of the edge pairs was given. See
appendix B.6 on page 59 for an example where in addition a fuzzy membership function for the size of
the edge pairs is specified.
The fuzzy membership function displayed in figure 53a can be created with the following HDevelop
code:
 B.3.2 Subtype ’position_pair_center’ D-49

Figure 54: The extracted edge pairs.

SetType := ’position_pair’
create_funct_1d_pairs ([0,50,250,300], [0,1,1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 53b on page 48 can
be created with the following HDevelop code:

SetType := ’position_pair’
PairSize := 10
create_funct_1d_pairs ([0,5,25,30], [0,1,1,0],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.3.2 Subtype ’position_pair_center’

’position_pair_center’ behaves like ’position_pair’ with the reference point located at the center of the
profile line.
The x values of fuzzy membership functions for the sub set type ’position_pair_center’ must lie within
the range

length of the ROI length of the ROI

− 2 ≤x≤ 2 .

Figure 55 shows fuzzy membership functions that can be used to extract edge pairs in the middle of the
profile line.
Figure 56 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 55. Only one edge pair in the middle of
the profile line is returned. Note that only the position of the edge pair, i.e., the position of the center
Appendix
between the two edges of the edge pair is used to restrict the extraction of edge pairs. The position of the
individual edges of the edge pair is not restricted. In this case, one large edge pair with its center in the
middle of the profile line is returned.
The fuzzy membership function displayed in figure 55a can be created with the following HDevelop
code:
SetType := ’position_pair_center’
create_funct_1d_pairs ([-25,0,25], [0,1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)
D-50 B Examples for Fuzzy Membership Functions

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

-200 -100 0 100 200 -25 -20 -15 -10 -5 0 5 10 15 20 25

feature value normalized feature value

(a) (b)

Figure 55: Fuzzy membership functions that can be used to extract edge pairs in the middle of the profile
line (SetType = ’position_pair_center’, PairSize = 10).

Figure 56: The extracted edge pairs.

The equivalent normalized fuzzy membership function, which is displayed in figure 55b can be created
with the following HDevelop code:

SetType := ’position_pair_center’
PairSize := 10
create_funct_1d_pairs ([-2.5,0,2.5], [0,1,0],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.3.3 Subtype ’position_pair_end’

’position_pair_end’ behaves like ’position_pair’ with the reference point located at the end of the profile
line.
The x values of fuzzy membership functions for the sub set type ’position_pair_end’ must lie within the
range

−length of the ROI ≤ x ≤ 0.0.

Figure 57 shows fuzzy membership functions that can be used to extract edge pairs at the end of the
profile line.
 B.3.3 Subtype ’position_pair_end’ D-51

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

-500 -400 -300 -200 -100 0 -50 -40 -30 -20 -10 0
feature value normalized feature value

(a) (b)

Figure 57: Fuzzy membership functions that can be used to extract edge pairs at the end of the profile
line (SetType = ’position_pair_end’, PairSize = 10).

Figure 58 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 57. Only edge pairs at the end of the
profile line are returned.

Figure 58: The extracted edge pairs.

The fuzzy membership function displayed in figure 57a can be created with the following HDevelop
code:

SetType := ’position_pair_end’
create_funct_1d_pairs ([-300,-200,0], [0,0.9,1], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 57b can be created
with the following HDevelop code:

SetType := ’position_pair_end’
PairSize := 10 Appendix
create_funct_1d_pairs ([-30,-20,0], [0,0.9,1],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)
D-52 B Examples for Fuzzy Membership Functions

B.3.4 Subtype ’position_first_pair’

’position_first_pair’ behaves like ’position_pair’ with the reference point located at the position of the
first edge pair.
The x values of fuzzy membership functions for the sub set type ’position_first_pair’ must lie within the
range

0.0 ≤ x ≤ length of the ROI.

Figure 59 shows fuzzy membership functions that can be used to extract edge pairs in approximately the
first half of the profile line.

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 100 200 300 400 500 0 10 20 30 40 50

feature value normalized feature value

(a) (b)

Figure 59: Fuzzy membership functions that can be used to extract edge pairs in approximately the first
half of the profile line (SetType = ’position_first_pair’, PairSize = 10).

Figure 60 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 59. Only one edge pair in the first half
of the profile line is returned. Note that a large edge pair is extracted instead of multiple small ones.
This happens because internally, all possible pairs of edges are analyzed according to the specified fuzzy
membership functions and no condition for the size of the edge pairs was given. See appendix B.6 on
page 59 for an example where in addition to a fuzzy membership function for the position of the edge
pairs a fuzzy membership function for their size is specified. Note also that even though the same fuzzy
membership function has been used as in appendix B.3.1 on page 48 one more edge pair is extracted
because the reference point is located at the first edge pair instead of the start of the profile line.

Figure 60: The extracted edge pairs.

The fuzzy membership function displayed in figure 59a can be created with the following HDevelop
code:
 B.3.5 Subtype ’position_last_pair’ D-53

SetType := ’position_first_pair’
create_funct_1d_pairs ([0,50,250,300], [0,1,1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 59b on page 52 can
be created with the following HDevelop code:

SetType := ’position_first_pair’
PairSize := 10
create_funct_1d_pairs ([0,5,25,30], [0,1,1,0],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.3.5 Subtype ’position_last_pair’

’position_last_pair’ behaves like ’position_pair’ with the reference point located at the position of the
last edge pair.
The x values of fuzzy membership functions for the sub set type ’position_last_pair’ must lie within the
range

−length of the ROI ≤ x ≤ 0.0.

Figure 61 shows fuzzy membership functions that can be used to extract edge pairs close to the last edge
pair.

1 1

0.8 0.8
membership value

membership value

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-500 -400 -300 -200 -100 0 -50 -40 -30 -20 -10 0
feature value normalized feature value
Appendix
(a) (b)

Figure 61: Fuzzy membership functions that can be used to extract edge pairs close to the last edge pair
(SetType = ’position_last_pair’, PairSize = 10).

Figure 62 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 61. Only edge pairs in the second half of
the profile line are returned. Note that even though the same fuzzy membership function has been used
D-54 B Examples for Fuzzy Membership Functions

as in appendix B.3.3 on page 50, more edge pairs are extracted. This is because the reference point is
located at the last edge pair instead of the end of the profile line.

Figure 62: The extracted edge pairs.

The fuzzy membership function displayed in figure 61a on page 53 can be created with the following
HDevelop code:

SetType := ’position_last_pair’
create_funct_1d_pairs ([-300,-200,0], [0,0.9,1], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 61b on page 53 can
be created with the following HDevelop code:

SetType := ’position_last_pair’
PairSize := 10
create_funct_1d_pairs ([-30,-20,0], [0,0.9,1],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

B.4 Set Type size

B.4.1 Subtype ’size’

Evaluates the size of the edge pairs, i.e., the distance between the two edges.
The x values of fuzzy membership functions for the sub set type ’size’ must lie within the range

x ≥ 0.0.

Figure 63 shows fuzzy membership functions that can be used to extract edge pairs that are approximately
50 pixels wide.
Figure 64 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 63. Only edge pairs that are approxi-
mately 50 pixels wide are returned.
The fuzzy membership function displayed in figure 63a can be created with the following HDevelop
code:
 B.4.2 Subtype ’size_diff’ D-55

1 1

0.8 0.8
membership value

membership value
0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 100 200 300 400 500 0 0.5 1 1.5 2

feature value normalized feature value

(a) (b)

Figure 63: Fuzzy membership functions that can be used to extract edge pairs that are approximately
50 pixels wide (SetType = ’size’, PairSize = 50).

Figure 64: The extracted edge pairs.

SetType := ’size’
create_funct_1d_pairs ([20,50,80], [0,1,0], FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

The equivalent normalized fuzzy membership function, which is displayed in figure 63b can be created
with the following HDevelop code:

SetType := ’size’
PairSize := 50
create_funct_1d_pairs ([0.4,1,1.6], [0,1,0],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

Appendix
B.4.2 Subtype ’size_diff’

Evaluates the signed difference between the desired PairSize and the actual size of the edge pairs.The
signed difference is defined by:
PairSize − actual size of a pair
x =
PairSize

The x values of fuzzy membership functions for the sub set type ’size_diff’ must lie within the range
D-56 B Examples for Fuzzy Membership Functions

x ≤ 1.0.

Figure 65 shows a fuzzy membership function that can be used to extract edge pairs that have the speci-
fied PairSize or are a little bit smaller.

0.8

membership value
0.6
The sub set type ’size_diff’ can only be speci-
0.4
fied with a normalized fuzzy membership func-
tion. 0.2

-1 -0.5 0 0.5 1
normalized feature value

(a) (b)

Figure 65: A fuzzy membership function that can be used to extract edge pairs that have the specified
PairSize or are a little bit smaller (SetType = ’size_diff’, PairSize = 15).

Figure 66 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 65. Only edge pairs that are between
11 pixels and 15 pixels wide are returned.

Figure 66: The extracted edge pairs.

The normalized fuzzy membership function displayed in figure 65b can be created with the following
HDevelop code:

SetType := ’size_diff’
PairSize := 15
create_funct_1d_pairs ([-0.01,0,0.5], [0,1,0],
NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

Note that with the subtype ’size_diff’ asymmetrical fuzzy membership functions can be specified for
the difference between the PairSize and the actual size of an edge pair. This is not possible with the
subtype ’size_abs_diff’ (see appendix B.4.3).
 B.4.3 Subtype ’size_abs_diff’ D-57

B.4.3 Subtype ’size_abs_diff’

Evaluates the absolute difference between the desired PairSize and the actual size of the edge pairs.The
absolute difference is defined by:

|PairSize − actual size of a pair|

x =
PairSize

The x values of fuzzy membership functions for the sub set type ’size_abs_diff’ must lie within the range

x ≥ 0.0.

Figure 67 shows a fuzzy membership function that can be used to extract edge pairs that are approxi-
mately as wide as the specified PairSize.

membership value 0.8

0.6
The sub set type ’size_abs_diff’ can only be
0.4
specified with a normalized fuzzy membership
function. 0.2

0 0.2 0.4 0.6 0.8 1

normalized feature value

(a) (b)

Figure 67: A fuzzy membership function that can be used to extract edge pairs that are approximately as
wide as the specified PairSize (SetType = ’size_abs_diff’, PairSize = 15).

Figure 68 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership function displayed in figure 67. Only edge pairs that are between
11 pixels and 19 pixels wide are returned. Note that in addition to the edge pairs that are extracted in
appendix B.4.2 on page 55 also larger edge pairs are returned.

Appendix

Figure 68: The extracted edge pairs.

The normalized fuzzy membership function displayed in figure 67b can be created with the following
HDevelop code:
D-58 B Examples for Fuzzy Membership Functions

SetType := ’size_abs_diff’
PairSize := 15
create_funct_1d_pairs ([0,0.5], [1,0], NormalizedFuzzyMembershipFunction)
set_fuzzy_measure_norm_pair (MeasureHandle, PairSize, SetType,
NormalizedFuzzyMembershipFunction)

Note that with the subtype ’size_abs_diff’ only symmetrical fuzzy membership functions can be specified
for the difference between the PairSize and the actual size of an edge pair. To specify asymetrical fuzzy
membership functions, use the subtype ’size_abs_diff’ (see appendix B.4.2 on page 55).

B.5 Set Type gray

B.5.1 Subtype gray

Evaluates the mean gray value between the two edges of edge pairs.
The x values of fuzzy membership functions for the sub set type ’gray’ must lie within the range of gray
values, i.e.,

0.0 ≤ x ≤ 255.0 for ’byte’ images and

0.0 ≤ x ≤ 65535.0 for ’uint2’ images.

Figure 69 shows a fuzzy membership function that can be used to extract edge pairs that enclose areas
with a gray value of approximately 50 or 150.

0.8
membership value

0.6

0.4
The sub set type ’gray’ cannot be specified
0.2 with a normalized fuzzy membership function.
0

0 50 100 150 200 250

feature value

(a) (b)

Figure 69: A fuzzy membership function that can be used to extract edge pairs that enclose areas with a
gray value of approximately 50 or 150 (SetType = ’gray’).

Figure 70 shows the positions of the edges that will be extracted if the the fuzzy measure object is created
with the fuzzy membership function displayed in figure 69. With this fuzzy measure object, only those
edge pairs are returned that enclose areas of the specified gray values.
The fuzzy membership function displayed in figure 69a can be created with the following HDevelop
code:
 B.6 Set Type position_pair combined with size D-59

Figure 70: The extracted edges.

SetType := ’gray’
create_funct_1d_pairs ([20,40,60,80,120,140,160,180], [0,1,1,0,0,1,1,0],
FuzzyMembershipFunction)
set_fuzzy_measure (MeasureHandle, SetType, FuzzyMembershipFunction)

B.6 Set Type position_pair combined with size

Figure 71 shows two fuzzy membership functions that can be used to determine edge pairs in approxi-
mately the first half of the profile line. In addition to only specifying the position of the edge pairs (see
appendix B.3.1 on page 48), also a fuzzy membership function for the size of the edge pairs is specified
(see appendix B.4.1 on page 54). With this, it is possible to define precisely which kind of edge pairs
must be returned.

1 1

0.8 0.8
membership value

membership value

0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 100 200 300 400 500 0 100 200 300 400 500
feature value feature value

(a) (b)

Figure 71: Fuzzy membership functions that can be used to extract edge pairs in approximately the first
half of the profile line: (a) The restriction for the position of the pairs (SetType = ’position_pair’)
and (b) a fuzzy membership function that favors small edge pairs (SetType = ’size’).
Appendix
Figure 72 shows the positions of the edge pairs that will be extracted if the the fuzzy measure object is
created with the fuzzy membership functions displayed in figure 71. Only edge pairs in approximately
the first half of the profile line are returned. Note that smaller edge pairs than in the example in ap-
pendix B.3.1 on page 48 are extracted. This is because of the constraint on the size of the edge pairs,
which is introduced by the fuzzy membership function for the size of the edge pairs (figure 71b).
The fuzzy membership function displayed in figure 71a can be created with the following HDevelop
code:
D-60 C Definition of the Geometric Mean

Figure 72: The extracted edge pairs.

create_funct_1d_pairs ([0,50,250,300], [0,1,1,0],

FuzzyMembershipFunctionPositionPair)

The fuzzy membership function displayed in figure 71b on page 59 can be created with the following
HDevelop code:

create_funct_1d_pairs ([0,50], [1,0.5], FuzzyMembershipFunctionSize)

Then, the fuzzy measure object can be created:

set_fuzzy_measure (MeasureHandle, ’position_pair’,

FuzzyMembershipFunctionPositionPair)
set_fuzzy_measure (MeasureHandle, ’size’, FuzzyMembershipFunctionSize)

C Definition of the Geometric Mean

The geometric mean of a sequence {ai }ni=1 is defined by

n
!1/n
Y
G(a1 , . . . , an ) ≡ ai
i=1

Thus,
p
G(a1 , a2 ) = (a1 a2 )
G(a1 , a2 , a3 ) = (a1 a2 a3 )1/3

and so on.