Vision System
A good visual system is hard to implement on an underwater autonomous vehicle underwater
visibility is extremely poor due to the optical properties of light propagation, namely absorption
and scattering which are important in the image formation process. To do any sort of image
recognition, the images taken underwater need to be enhanced and for navigation the enhanced
images are needed in real time. That is why the researches decided to explore directly the use of
perceptually uniform color spaces, in particular the l∝β color space.
A color space is a mathematical model through which various color combinations are
represented. The researchers selected the l∝β color space because it can simplify the analysis of
data from the underwater images. In underwater images, the background color(water tint) is
usually blue or green, these colors correspond to the limits of α and β channels respectively and
therefore identifying objects with contrasting colors to these two colors becomes much easier.
The mapping between RGB and the l∝β color space is expressed as a linear transformation:
0.3475 0.8265 0.5559 R
[] [ ][ ]
[] =[0.2162 0.4267 −0.6411][G] ,
[] [0.1304 −0.1033 −0.0269][B]
where l∈ [0.0,1.73] is the achromatic channel which determines the luminance value, α∈
[−0.6411,0.6429]is the yellow-blue opposite channel, and β∈ [−0.1304,0.1304], is the red-cyan
with a significant influence of green. The data in these channels include a wide variety of colors;
however, the information in aquatic images is contained in a very narrow interval.
(input image) (a)
(b) (c)
These figures show an underwater image and the frequency histogram for each channel of the
l∝β color space. In these images, the data of the objects are concentrated in a small interval. So,
in order to make the identification process more reliable, new limits for each channel need to be
set. These values help to increase the clarity between identifiable objects & the background of
the image. The new limits are calculated using the frequency histogram for each of the channels
and this way the extreme values in the histogram with a higher frequency than the threshold
value is calculated. Using the frequency histogram method eliminates any outliers in the data.
Finally, a data normalization process is performed using the new interval in each channel of the
color space. Then, it becomes possible to get a clear segmentation of the objects with colors at
the end values of the channels.
Visual Navigation System
The autonomous underwater vehicle shown below is called Mexibot and is created on the RHex
platform. This robot is an amphibious robot, meaning it can operate in both land and water
environments. The robot has two pairs of computers embedded into it, one is used for the visual
system, registration of data etc. and the other is used for low level control. However, both
computers are connected through Ethernet so they are able to interact with each other. QNX
operating system is installed in the control computer since the control loop of the robot runs on a
real-time constraint. The other computer that deals with the vision system has Ubuntu 12.04
installed. On this computer, advanced applications are used which use the Robot Operating
System (ROS). The vehicle also has an Inertial measurement unit (IMU), which provides the
orientation of the robot and angular velocity of the vehicle. A pressure is used to estimate the
depth of the robot under water and a set of three cameras are used for generating visual
information and navigation purposes.
