Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR2690

Identification of Missing Person using CNN

J. Dhanalakshmi1 Ashok Kumar M2 ; Shalini J3 ; Soundharya Devi M4
Assistant professor/CSE SSM Institute of Engineering and Technology,
SSM Institute of Engineering and Technology Tamil Nadu, India

Abstract:-Our project aims to leverage Convolutional to match them against a database of known individuals. This
Neural Networks (CNNs) for the identification of missing technology can help law enforcement agencies, non-profit
persons. CNNs, a class of deep learning algorithms organizations, and rescue teams to locate missing persons
widely used in image recognition tasks, offer promising more efficiently and accurately.
potential in automating and enhancing the identification
process. The project aims to develop a robust system In this context, CNNs offer several key advantages.
using CNN models to match unidentified individuals They can analyze large volumes of data quickly, making
with missing person databases, improving identification them ideal for processing surveillance footage or large-scale
accuracy and providing closure. image databases. Additionally, CNNs can identify subtle
features and patterns that might be overlooked by human
The proposed approach demonstrates potential in observers, enhancing the accuracy of identification.
assisting law enforcement agencies and missing persons
organizations by providing a reliable and efficient means However, implementing CNN-based systems for
of cross-referencing images from various sources, such identifying missing persons involves several challenges.
as surveillance footage, social media, and public records. These include ensuring data privacy, addressing bias in
Additionally, the flexibility of CNNs allows for the training datasets, and maintaining the reliability of the
integration of other biometric markers, including models in various scenarios. Moreover, integrating CNNs
fingerprints and voice recognition, to increase the into existing workflows and ensuring interoperability with
accuracy and reliability of identifications. other systems are crucial considerations.

This research underscores the importance of using II. LITERATURE REVIEW

artificial intelligence and machine learning in social good
applications, highlighting the potential for technology to TITLE : Identification of Missing Person Using
play a transformative role in reuniting families and Convolutional Neural Networks
bringing closure to unresolved cases. Future work will
focus on refining the model, enhancing privacy AUTHORS: P.D Nharsha Sai; V.V Sai Kiran, K. Rohith;
protections, and ensuring ethical use in real-world D.RajeswaraRao
applications.
The image of person is matched using the Deep
Keywords:- Automated Identification, Convolutional Neural learning model, VGG16. This CNN model can be used for
Networks (CNNs), Deep Learning, Facial Recognition, Face recognition in images. VGG-16 is a pre-trained model
Missing Person Identification, Pattern Recognition, Social which is run on millions of images. Those weights are used
Impact. in our proposed work. The NumPy module is used to
represent mathematical and logical operations on arrays.
I. INTRODUCTION This Predictive model probably identifies the missing person
if the image matches and is found in the database. This
The identification of missing persons is a critical issue methodology uses several steps which are data
in law enforcement, humanitarian work, and disaster preprocessing, data augmentation, VGG-16, training and
response. Recent advancements in artificial intelligence (AI) testing the model. Using this model, the accuracy for the
and deep learning have opened new possibilities for training data is 90.01%. On testing, the accuracy obtained is
enhancing the accuracy and efficiency of identifying 85.21%.
missing individuals. Convolutional Neural Networks
(CNNs), a type of deep learning algorithm, have gained TITLE : Facial Matching and Reconstruction Techniques in
significant popularity due to their effectiveness in image Identification of Missing Person Using Deep Learning,
analysis and pattern recognition.
AUTHORS: R Annamalai; Sudharson S; Kolla Gnapika
CNNs can process visual data and learn complex Sindhu
features, making them well-suited for facial recognition,
object detection, and other tasks related to visual The number of missing person cases has dramatically
identification. When applied to the problem of identifying increased nowadays, leaving loved ones with a lot of
missing persons, CNNs can analyze images or video footage unanswered questions. Police inquiries and public

IJISRT24APR2690 www.ijisrt.com 3381

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR2690

announcements are two regularly used traditional methods the images of missing individuals, thereby improving the
for locating missing persons, although they frequently fall accuracy and speed of identification and making the process
short, especially over time. Artificial intelligence (AI) is more reliable and efficient. To provide a comprehensive
gaining popularity and could be used to enhance the search solution, we have developed a user-friendly webbased
process. This study offers a revolutionary approach for application system that facilitates the search for missing
solving the unsolved cases of missing individuals by using persons. The application efficiently collects and stores
AI-based facial matching and face reconstruction information about missing individuals in a centralized
approaches. The proposed method successfully uses the database. Whenever a missing person is identified in a
ORL (Olivetti Research Laboratory) Dataset's Support CCTV video stream, our application actively tracks their
Vector Machine (SVM) classifier to reach an outstanding location. Once the missing person is successfully identified
accuracy of 93% by combining face landmarks and machine within the video stream, the application promptly sends
learning algorithms. Additionally, a 3D face reconstruction location details via email to the person’s relative.
method based on Convolutional Neural Networks (CNN)
trained on the varied 300-W dataset achieves a high TITLE: AI BASED – ASSISTED SEARCH FOR MISSING
accuracy of 90%. These results demonstrate the potential of PERSON
AI and deep learning models for improving missing person
identification. The proposed approach offers a viable option AUTHORS: Hemadharshini S, Bheena Dhevi V, Bama Devi
that aids in providing closure to the impacted families, YEAR: 2023
making a significant contribution to the field and reducing
crimes in the future. Face recognition is a biometric-based technology that
mathematically maps a particular person’s or individual’s
TITLE: Drone-based Autonomous Human Identification for facial features and stores all that data as a face print. By
Search and Rescue Missions in Real-time. using this technique, the information on the face of a person
is saved mathematically or in the format of graphs in the
AUTHORS: K. Jayalath, S.R. Munasinghe database, which is used for detecting that particular face.
The face recognition model in our system will find a match
A drone-based human identification system to make of that person in the database. If a match is found, it will be
search and rescue missions more effective. The drone, once notified to the police and the guardian of that person. In this
detects signs of human presence in real-time aerial videos, paper, we will use the ideas of the AWS facial recognition
autonomously navigates towards the suspicious location to algorithm which is based on Artificial Intelligence (AI) and
get a better view to verify human presence. It processes will detect faces with the maximum accuracy to find the
images and sends selected frames for the operator for missing person.
verification and issuance of instructions for follow-up
actions. A custom built Tensorflow neural network is used III. PROBLEM STATEMENT
as the object detector, which processes images in real-time
and report human objects if detected on the ground. The objective is to use convolutional neural networks
(CNN) to precisely identify missing people by using their
TITLE: Efficient Face Recognition in Real Time for visual recognition skills for age progression analysis and
Locating Missing Person face feature detection.

AUTHOR: Nimmy Pailochan, Ms. R. Keerthana, Mr. N. Large data sets, unique face patterns, and a dependable
Santhosh, Dr. J. Sreerambabu YEAR:2023 tool for law enforcement and humanitarian agencies are the
goals of this model.
This paper presents a web-based application system
that utilizes deep learning as the core technology of a Face IV. KEY POINTS WITHIN THE PROBLEM
Recognition System, with the primary objective of STATEMENT
enhancing the accuracy and efficiency of locating missing
individuals. Every day, a significant number of individuals  Certainly, here are key points to consider within the
go missing due to various factors such as old age, mental problem statement for the project of identifying missing
health issues, or conditions like Alzheimer’s. The persons by comparing live video with dataset images:
conventional methods employed for searching for missing
persons are typically slow, expensive, and involve  Identification of Missing Persons: The primary objective
protracted physical searches lasting weeks or even months. is to develop a system capable of identifying missing
In contrast, deep learning-based technologies offer a persons through comparison of live video footage with a
promising solution by rapidly analyzing substantial volumes dataset of known images of missing individuals.
of data within minutes or hours. By leveraging facial  Real-Time Video Analysis: The system must be capable
recognition technology, which is an application of deep of analyzing live video streams in real-time, ensuring
learning, our proposed system aims to compare images and swift identification and response to potential matches.
videos obtained from surveillance cameras with pictures of
missing persons to identify potential matches. Specifically,
we employ the Resnet deep learning algorithm to examine

IJISRT24APR2690 www.ijisrt.com 3382

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR2690

 Integration of Datasets: The project involves integrating  Data Collection:

a comprehensive dataset containing images and relevant
information of missing persons with the system. This  Gather a comprehensive dataset of images of missing
dataset serves as the reference for comparison during the persons from various sources such as law enforcement
identification process. databases, NGOs, social media, and family submissions.
 Machine Learning Algorithms: Utilization of advanced  Ensure that the dataset is diverse, including images of
machine learning algorithms, such as convolutional people of different ages, genders, ethnicities, and
neural networks (CNNs), for accurate and efficient appearances.
matching between live video frames and dataset images.
 Feature Extraction and Comparison: Implementation of  Preprocessing:
techniques for feature extraction from both live video
frames and dataset images, enabling meaningful  Standardize the images in the dataset to a common
comparison to identify potential matches. format and resolution.
 Accuracy and Reliability: Ensuring the system delivers  Apply preprocessing techniques such as normalization,
high accuracy and reliability in identifying missing resizing, and noise reduction to improve consistency and
persons, minimizing false positives and false negatives. quality.
 Privacy and Ethical Considerations: Incorporating
mechanisms to protect the privacy of individuals  Feature Extraction:
captured in live video feeds and adhere to ethical
guidelines regarding the use of sensitive personal data.  Utilize computer vision techniques to extract features
 Scalability and Deployment: Designing the system to be from both the live video feed and the dataset images.
scalable, allowing it to handle large volumes of live  Features could include facial landmarks, color
video feeds and dataset images efficiently. Additionally, histograms, texture descriptors, or deep learning
ensuring ease of deployment across various locations embeddings.
and platforms.
 Collaboration with Law Enforcement and Authorities:  Model Training:
Collaboration with law enforcement agencies and
relevant authorities to facilitate the integration and  Train a machine learning or deep learning model to learn
utilization of the system within existing missing persons the relationship between the features extracted from the
identification frameworks. live video and dataset images.
 User Interface and Accessibility: Developing a user-  Consider using convolutional neural networks (CNNs)
friendly interface for operators to interact with the for feature extraction and classification tasks due to their
system, facilitating ease of use and accessibility in effectiveness in image analysis.
various operational scenarios.
 Evaluation Metrics: Establishing clear evaluation metrics  Model Integration:
to assess the performance of the system, including
metrics such as precision, recall, and F1-score, to  Integrate the trained model into a real-time video
measure its effectiveness in identifying missing persons processing pipeline capable of analyzing live video
accurately. streams.
 Continuous Improvement and Updates: Implementing  Ensure that the system can handle the computational
mechanisms for continuous improvement and updates to requirements for processing video in real-time.
the system, including regular updates to the dataset,
optimization of algorithms, and incorporation of  Real-time Detection and Matching:
feedback from users and stakeholders.
 Implement algorithms to detect faces or individuals in
V. PROPOSED METHODOLOGY the live video feed.
 Use the trained model to compare the detected faces with
 Define Project Scope and Objectives: the dataset of missing persons.
 Employ techniques such as similarity metrics,
 Clearly define the purpose of the project, whether it's to thresholding, or machine learning classifiers to match
aid law enforcement, humanitarian efforts, or another live faces with the dataset.
cause.
 Set specific objectives such as improving the efficiency  Verification and Validation:
of identifying missing persons, reducing search time, or
increasing accuracy.  Validate the performance of the system using a separate
validation dataset and metrics such as accuracy,
precision, recall, and F1-score.
 Conduct thorough testing to ensure the system's
reliability and robustness in various scenarios and
conditions.

IJISRT24APR2690 www.ijisrt.com 3383

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR2690

 Deployment and Monitoring:  Matching Algorithm: Implement a matching algorithm to

identify potential matches between the live video and
 Deploy the system in relevant environments such as law images in the dataset. This algorithm should consider
enforcement agencies, public spaces, or humanitarian factors such as the similarity threshold for considering a
organizations. match and any additional constraints, such as age,
 Monitor the system's performance in real-world usage gender, or ethnicity, to improve the accuracy of the
and collect feedback for continuous improvement. identification.
 Verification and Notification: Once a potential match is
identified, the system should verify the match through
 Privacy and Ethical Considerations: additional checks, such as human review or additional
biometric verification methods. If the match is
 Implement privacy measures to protect the identities and confirmed, notify the relevant authorities or individuals
data of individuals, ensuring compliance with responsible for locating the missing person.
regulations such as GDPR or HIPAA.  Feedback Loop: Incorporate a feedback loop to
 Address ethical concerns related to surveillance, consent, continuously improve the accuracy of the system. This
bias, and potential misuse of the technology. could involve updating the dataset with new images of
missing persons, fine-tuning the matching algorithm
 Documentation and Reporting: based on performance feedback, and incorporating
advances in computer vision and facial recognition
 Document the entire methodology, including data technology.
sources, preprocessing steps, model architecture, training
process, and evaluation results. VII. RESULTS & DISCUSSIONS
 Provide clear and comprehensive reports on the project
outcomes, including successes, limitations, and  Dataset
recommendations for future enhancements. This dataset contains different images of persons. The
image size is width=640, height=480 Each person of the
 Continuous Improvement: image will be stored with their name in a separate folder.
Each image in jpeg file or jpg file format
 Iterate on the system based on feedback, technological
advancements, and evolving needs to continually
enhance its effectiveness and usability.  Training And Testing
Training:Images of known faces are loaded using the
This methodology provides a structured approach to load_encoding_images method of the SimpleFacerec class.
leveraging live video footage for identifying missing Each image is converted into a face encoding using the
persons, integrating techniques from computer vision, face_recognition.face_encodings function. The resulting
machine learning, and real- time processing. face encodings and their corresponding names are stored in
lists (known_face_encodings and known_face_names) for
VI. TECHNIQUE OF THE SOLUTION recognition.

 One technical solution for identifying missing persons by Testing: The ‘start_video’method initializes the
comparing live video with a dataset of images involves webcam and continuously reads frames. Faces are detected
using computer vision and facial recognition algorithms. using the ‘detect_known_faces’ method, which compares
Here's a high-level overview of the process: the detected face encodings with known faces. If a match is
found, the recognized person's name is displayed on the
 Face Detection: Use a face detection algorithm to locate frame. The ‘capture_imag’e method allows capturing an
and extract faces from the live video stream in real-time. image when the "Capture" button is clicked, saving it with
This step ensures that only the relevant portions of the the detected faces and showing a success message.
video containing faces are processed further.
 Feature Extraction: Extract facial features from the  Library Used
detected faces in the live video. These features could A Python library built on top of dlib and OpenCV
include the relative positions of key facial landmarks, Provides easy-to-use interfaces for face detection, face
such as eyes, nose, and mouth, as well as more complex encoding, and face comparison. Offers pre-trained deep
features derived from the face's texture and structure. learning models for accurate face recognition tasks. Enables
 Database Query: Compare the extracted features from developers to integrate face recognition capabilities into
their applications efficiently. Convolutional Neural Network
the live video with the dataset of images containing faces
of missing persons. This comparison could involve a : CNN is a type of deep learning architecture commonly
similarity measure, such as cosine similarity or used for image classification and recognition tasks. Also
Euclidean distance, to determine how closely the used to extract features from human images to identify
different humans.
features match.

IJISRT24APR2690 www.ijisrt.com 3384

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR2690

 Integration REFERENCES
The face recognition module is seamlessly integrated
into the larger application, enabling real-time face detection, [1]. Artyani,I.(2019,0910).http://repository.uinjkt.ac.id/ds
identification, and image capture functionalities It leverages pace/bitstream/123456789/47930/1/ISMA%20ARTYA
the capabilities of the face recognition library to provide N I-FST.pdf. Diambil kembali dari UINJKT:
accurate and reliable face recognition services within the http://repository.uinjkt.ac.id/dspace/bitstream/1234567
graphical user interface. 89/47930/1/ ISMA%20ARTYANI-FST.pdf
[2]. Astuti, D. L. (2019, 08 01). klasifikasi ekspresi wajah
VIII. CONCLUSION menggunakan metode principal component analysis
(pca) dan convolutional neural network (cnn).
The project aiming to identify missing persons by Diambilkembali dari RepositoryUnsri:http://repository.
comparing live video with a dataset of images represents a unsri.ac.id/6479/3/RA MA_55101_09042621721004_
significant advancement in leveraging technology for 0004027101_0023027804_01_FRONT_REF.pdf
humanitarian purposes. [3]. Fahmi, K., Santosa , S., & Fanani , A. Z.
(2015).optimasi parameter artificial neural network
By harnessing the power of computer vision and dengan menggunakan algoritma genetika untuk
machine learning algorithms, this project offers a promising memprediksi nilai tukar rupiah. Jurnal Teknologi
solution to expedite the process of locating and reuniting Informasi, 196-206.
missing individuals with their families and loved ones. [4]. Fermansah , D. (2018, 01 20). Machine Learning.
Diambil kembali dari Universitas Siliwang i:
Through rigorous testing and refinement, it has the http://repositori.unsil.ac.id/233/6/bab%202.pdf
potential to become an invaluable tool for law enforcement [5]. Jiang, A., Yan, N., Wang, F., Huang, H., Zhu, H., &
agencies, humanitarian organizations, and communities Wei, B. (2019). Visible Image Recognition of Power
worldwide, offering hope and closure to those affected by Transformer Equipment Based on Mask R-CNN.
the tragedy of disappearance. Sustainable Power and Energy Conference (iSPEC)
(hal . 222-299). Beijing, China, China: IEEE.
As we continue to refine and develop this technology,
we move closer to a future where no one is left behind, and
every missing person is given the opportunity to be found.

FUTURE SCOPE

In the realm of future work for identifying missing

persons, advancements in computer vision and artificial
intelligence hold significant promise. One avenue of
exploration could involve enhancing the accuracy and
efficiency of matching live video footage with a
comprehensive dataset of images. This could be achieved
through the refinement of deep learning algorithms capable
of recognizing facial features, gauging age progression, and
accounting for variations in lighting, angles, and resolution.

Additionally, integrating real-time monitoring

capabilities could bolster the system's responsiveness,
enabling swift identification and location tracking.
Collaborations with law enforcement agencies, forensic
experts, and technologists could facilitate the development
of a robust framework that respects privacy concerns while
maximizing the potential for reuniting missing individuals
with their loved ones.

Furthermore, exploring the integration of biometric

data beyond facial recognition, such as gait analysis or voice
recognition, could provide further layers of identification,
enhancing the system's overall effectiveness and reliability.

IJISRT24APR2690 www.ijisrt.com 3385