2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE)

Implementation of Machine Learning and KNN

Algorithm for Finding Missing Person
2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE) | 978-1-6654-3789-9/22/$31.00 ©2022 IEEE | DOI: 10.1109/ICACITE53722.2022.9823710

Mayank Kumar Singh Pooja Verma Ajay Shanker Singh

School of Computing Science & School of Computing Science & School of Computing Science &
Engineering Engineering Engineering
Galgotias University Galgotias University Galgotias University
Greater Noida . India Greater Noida, India Greater Noida, India
[email protected] [email protected] ajay.shankersingh@galgotiasuniversity
.edu.in
Anandhan.K
School of Computing Science & Engineering
Galgotias University
Greater Noida, India
[email protected]

Abstract- Nowadays, finding a missing person is a This application is used to look for people who have gone
difficult task despite the fact that we are all kept up to missing. It offers functions such as registering missing
date by social media. It requires a great deal of paper individuals and maintaining a database of people who have
work and is a time-consuming process additionally, even been located. However, there is no tool that allows you to
after all of this hard labour there are minimal aspects of match the faces of missing people. There is no form validation
obtaining a satisfactory result. This project provides a to ensure that the information submitted by the user is correct.
mechanism that benefits both the police and the normal
It declares faces matched if at least 70-75 percent of points
public by speeding up the procedure. Here we used face
are matched. It's a method of uploading photographs into an
recognition as a search process. If a person goes missing,
application using an Android phone, and the results are based
the individual's guardian can upload the photograph
on facial recognition. For comparing the faces of saved faces
which will then be stored in the system. The facial
and missing people, we use Python's Dlib and Face
recognition model in our system will then search the
Recognition packages.
database for a match of that individual. If there is a
match is found, the police and the guardian will be A. Face Detection
alerted. The process of locating a face in an image without
identifying is known as face detection. Although numerous
Keywords— KNN classifier, Face Recognition, Face face detection algorithms existed prior to 2001, the Viola-
Modeling, Face Detection, Dlib . Jones paper "Rapid Object Detection using a Boosted Cascade
of Simple Features" [3]. Unlike prior face detection methods
I. INTRODUCTION
that depended on pixel analysis, this new method does not rely
Since we knew, we certainly utilise our faces to recognise on pixel analysis.
people these days, and technology is improving at rapid rate
in terms of facial detection. Face detection is a unique Viola-Jones created the "Haar-classifier" algorithm, which
technology that exploded in popularity in recent years. The relies on Haar-like features. The Haar classifier is a machine
goal is to create a project to locate lost individuals that will be learning method that has been trained on a large number of
useful to both ordinary people and police departments, positive and negative examples.To recognise things in photos,
allowing them to locate their loved ones in a timely manner. samples are used [4]. The size of the image in the classifier
In our project, we create a full storage of missing people so must be large enough for it to perform properly.The size of the
that the police can track them down or their guardians may training set must match the size of the object detection input
submit a report on them. People can also upload a photograph image.
of the lost person . On the basis of comparing faces of missing people, the
According to data on lost persons in crime rates in India, a algorithm guarantees a 70-80% success rate. To achieve the
total of roughly 2.9 lakh people were reported missing in required goals, we need three modules: User, Police, and
2016, 3 lakh people in mid-2017, and 3.48 lakh people in Admin. Data from solved cases is automatically updated and
2018-19 with the number continuously rising [1]. According deleted from the database.
to an article "Lost in 1995 Jyotsna Found Her Family in 2016" The proposed approach assists in locating a missing
by Manabi Katoch, published on April 30 2016. The report's individual in the shortest amount of time and with the least
headlines said, "A huge thank you to Chandrapur Police" [2]. amount of effort. The programme interface is simple to use
After 21 years, a person is discovered. It is a national and free, making it worthwhile to demonstrate to others. It also
emergency in the country to take harsh action against eliminates the laborious way of recording missing person
criminals and to employ technology to respond quickly. situations results in saving time [5].

978-1-6654-3789-9/22/$31.00 ©2022 IEEE 1879

Authorized licensed use limited to: Zhejiang University. Downloaded on November 16,2024 at 23:45:22 UTC from IEEE Xplore. Restrictions apply.
 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE)

We make the entire procedure very simple for you. So, Second, in 2016, Rohit Satle and his colleagues published
here's what we came up with. a study on a face detection system based on Principal
Component Analysis (PCA) method [7].
1) Created a graphical user interface (GUI) programme
in Python that police may use to create a new case. The image Thirdly Swarna Bai Arniker and her colleagues presented
of the missing person is analysed in the backend, and face key a research that identifies missing persons using RFID
points are collected and saved in the database, along with technology [8] .
additional information such as name, father's name, age, city, Fourth, Birari Hetal and his colleagues prepared a work in
and so on. which they compared faces using SWF-SIFT. SIFT, on the
2) Created an Android application that is accessible to other hand, is computationally intensive and hence time
the general public (people like us). We may use this app to consuming [9] .
submit a photo of someone we think is questionable. We have Thomas M. Omweri and Andrew M. Kahonge published a
the option of submitting the image anonymously or with little paper in 2015 in which they proposed a method for searching
information. This image, along with its location, is saved in for missing people using a mobile-based online service [10].
the database.
Prof. Sumeet Pate and colleagues published a study in
3) To match the images provided by the police and the 2016 that demonstrated the application of the Line Edge
images submitted by the user, the GUI application employs a Method (LEM) for facial detection in the search for missing
Machine Learning Algorithm. If a match is found, we can individuals [11].
display it together with the last known location of the missing
Finally, in 2018, Zimbabwean Peace Muyambo proposed
individual.
a face detection system to locate lost individuals in Zimbabwe,
II. LITERATURE REVIEW which employed the LBPH approach to identify faces [12].
We conducted extensive research and compiled the III. PROBLEM STATEMENT
following findings from our literature review. So initially,
According to a study released by the National Criminal
Pournami S. Chandran and his team from the Centre for
Bureau in October 2019, there was a 77 percent to 80 percent
Development of Advanced Computing distributed a paper
increase in the number of crime cases in 2017 when compared
with the same goal. With the aid of SVM, they created Facial
to 2016 . The evidence is sufficient to take severe action in the
features extraction using Deep Learning (CNN) [6].
country's crime-solving procedure.
This section examines and contrasts the manual and
Furthermore, the existing situation for locating missing
automated methods of discovering individuals, as well as their
persons is based on a manual approach, which is ineffective in
associated studies. Every year over 1 lakh persons go missing,
dealing with the continuously rising number of occurrences of
with just a small percentage of them being located quickly.
missing persons. Finding a lost individual by hand is still a
However, the majority of instances take considerably longer,
lengthy and time-consuming operation. It takes a long time to
and in certain circumstances the person may die as a result. As
write a FIR (First Information Report), register it in the
a result, manually locating the individual as quickly as feasible
department book, and then distribute it to the other stations. It
is a challenging process. It takes time to submit a case and get
squanders both manpower and paper.
it to the point of execution under the manual system.
Furthermore, personnel is not well-trained to successfully As detailed in this Nature.com article, there are a slew of
search for the missing. So, while certain initiatives have been ethical issues surrounding the development of AI facial
done to work in the field of missing persons and to reduce recognition. For example, researchers at Harrisburg
crime rates, these applications are insufficient to address the University in Pennsylvania created AI face recognition
current situation. software that, according to them could predict if someone
would become a criminal with an accuracy of 80% [13]. After
Features such as adding and eliminating complaints,
a barrage of critical feedback, Harrisburg pulled their press
displaying them and searching by qualities were included in
release on the subject and decided not to publish the piece.
Crime Reporter and Missing Person Finder, however facial
recognition was not. Similarly, the task of Missing Person Another thorny issue is data harvesting without consent.
Finder papers is limited to inputting and updating information. Until the early 2000s, AI researchers usually relied on
The majority of the papers are identical to the ones listed volunteers to provide training data. However, however, the
above. As a result, automation is required to automate the job vast majority of facial photos are gathered without consent.
of locating the missing individual. The server has a method for
storing the photographs and information about the missing For example, in 2016, researchers from the University of
individual. Washington in Seattle published a database of 3.3 million
photographs of faces taken from Flickr without permission
We can get the details of all the missing people at that time if [14]. There are currently no clear legal safeguards in place for
we do it this way. In this context, we created the Missing the collection of facial recognition training data, yet Facebook
Person programme, which includes features such as face recently paid a $650 million settlement for acquiring face data.
recognition, case storing to a database, and matching related Google, for example, has publicly said that it is taking a
photographs from the database. more responsible approach to face-related technology. The
Many programmes provide the option of storing following are some of the requirements:
information to a database, however they are ineffective at
giving accurate information and take longer to deliver results. x Protecting privacy by giving an optimal degree of
The details of the missing individual can be stored on servers control and transparency.
such as AWS, Alibaba, Docker and others.

1880
Authorized licensed use limited to: Zhejiang University. Downloaded on November 16,2024 at 23:45:22 UTC from IEEE Xplore. Restrictions apply.
 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE)

x Not magnifying or reinforcing existing prejudices. used to train the classifier. Assume that KNN distributes the
facial points in the graph with RED belongs to Face 1,
x Not employing these technologies in ways that GREEN corresponding to Face 2 ,and BLUE corresponding
contravene internationally-accepted ethical principles. to Face 3. [Fig.2]
IV. SYSTEM OVERVIEW
A. Software Requirement
x Visual Studio Code
x Android Studio
x AVD for testing
x SDK for API 8 / API 29
x JDK, Java 7+ Fig. 2. KNN Classifier's Operation[15]
x EC2 instance from amazon web services
After that, to match any individual's face, we just need to
x Python 3.6 or above create face landmarks for person and then use it to forecast
using the KNN classifier that we implemented. If the
x Django 2+ similarity is above 60 then will be assumed that those two
x Dlib 19+, Face recognition, etc. as required people are the same .

x 64 bit-Operating System Now, if a face is found matched by using the KNN model,
one will obtain the encoded label as an output. This person's
x Window 7 or above. information will be encoded on this label.
x HTML "Missing Person" is the name of the programme. To begin,
download and install the programme. The Android OS will
B. Hardware Requirement then check the application's compatibility with the current
x Processor of 1.7GHz Processor Pentium 4 or above version while installing it. When we launch the app, the first
page that shows on the screen is a button that says "Get
x RAM: 1GB, 100GB of hard disk Started."
x Android Mobile with camera
V. IMPLEMENTATION
The suggested system, which we created follows the
technique shown below[Fig.1]. We're looking for the lost
person's facial features. For a face, the dlib landmark yields
nearly 68 different points.These generated points are float
values with an 8 decimal place precision as shown in below
figure.

Fig. 3. Registering New Cases

A new activity appears when you click the button. There

are two alternatives on this page: either search for a person by
submitting an image and matching the face, or provide
information about a new missing person. If we choose to add
a missing person, we had to provide information such as name,
age, address, phone number, image of the individual,
description, and so on [Fig.3].

Fig. 1. Landmarks on Face[15]

Dlib creates 136 * 3 face landmark points while the

desktop programme detects roughly three cases. The count is
136 since each point's x, y coordinate is created there are 68
points, and there are three situations. These points are then

1881
Authorized licensed use limited to: Zhejiang University. Downloaded on November 16,2024 at 23:45:22 UTC from IEEE Xplore. Restrictions apply.
 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE)

The more intricate and clever facial recognition grows, the

more difficult it becomes to comprehend how it works. The
logic of a neural network is embedded in the action of
thousands of "neurons" that are organised into hundreds of
linked layers.
In the future years, the United States will have to make
difficult decisions concerning AI: experts such as Stephen
Hawking and Elon Musk have expressed reservations about
adopting AI, claiming that it may lead to humanity's
extinction.
Fig. 4. Uploading image for searching Nonetheless, several nations are forging ahead in the field
of AI facial recognition, with China now leading the pack.
After filling out the form, the information is entered into China's objective is to build industrial standards now so that it
the database, and a new missing person is saved in our admin may influence the creation and implementation of global
panel. If we wish to look for a missing individual using an standards in the future. As the technology arms race between
image, we just upload the image and click the search button the United States and China heats up, we should expect to see
[Fig.4]. Our face matching process begins when you click the an influx of AI solutions and standards.
search button. It first counts how many photographs have been
stored, then loops over each image and compares the uploaded VII. CONCLUSION
image to the images in the database. If any of the picture faces If some of the earlier programmes used the function of
match, the loop is broken and the message "The image is preserving details, the feature of facial recognition was
matched by name XYZ" appears [Fig.5]. It indicates that the lacking, and vice versa. Face recognition with significantly
match has been located, and we may obtain information on the higher accuracy is being worked on and this technology is
missing individual by looking up the image name in the admin proving to be highly useful in locating missing people.The
panel's database. programme uploads, retrieves, deletes, and updates the
missing person's information. A search tool is provided to
match the faces and identify the missing person's details. This
programme is for ordinary people to submit and search details
but it also includes tools for police to verify the details of
missing individuals and an admin panel to alter and save the
details in the database. The application's server is also
managed by the administrator. Furthermore, the procedure for
locating missing persons is expedited in this manner. Our
technology replaced the human way of examining a database
of single images with a well-planned face recognition
algorithm that completes the task in a short amount of time.
The process of locating missing persons is speed up in this
Fig. 5. Result after searching
way. Our solution substitutes the time-consuming manual way
If the image does not match the database images, the of scanning through databases for each photo to ensure a
screen displays the message "this is not matched". match with an efficient facial recognition method that
completes the task in no time. Though our method has a minor
We made it available only to those who are permitted since flaw, the accuracy of our system decreases when the age of the
the details of the individual are confidential. People may individual is between the ages of 0 and 10. This is owing to
search the photographs, but only allowed people can view the the fact that face characteristics are still developing at that age.
information through the admin panel. It only informs them if In the future, we hope to be able to overcome this constraint.
a match has been found.
VIII. REFERENCES
VI. FUTURE SCOPE
The goal of our next work is to calculate the application's [1] [1] Crime In India(2019) Statistics Volume-1 by Nation Crime Records
working tendency so that if a problem arises, we can address Bureau.
it as quickly as possible and demonstrate that our suggested [2] “Lost in 1995 Jyotsna Found Her Family in 2016" by Manabi Katoch,
system is superior to previous applications. We also intended The Better India April 30 , 2016
to add the following features to the programme. When a [3] P. Viola and M. Jones, "Rapid object detection using a boosted cascade
of simple features," Proceedings of the 2001 IEEE Computer Society
person goes missing, an automatic alert is sent out. An Conference on Computer Vision and Pattern Recognition.
automatic notification mechanism is also being developed to [4] “Comparison of Viola-Jones Haar Cascade Classifier and Histogram of
improve the application's functionality. oriented gradients (hog) for face detection” January 2020 IOP
Conference Series Materials Science and Engineering 732(1):012038.
We intend to expand this technology in the future by
[5] Finding Missing Person using AI – Gagan Manku.
linking it to public cameras and detecting faces in real time.
The frames will be continuously transferred from the public [6] P. S. Chandran, N. B. Byju, R. U. Deepak, K. N. Nishakumari, P.
Devanand and P. M. Sasi, "Missing Child Identification System Using
cameras to our system, which will monitor the frames in real Deep Learning and Multiclass SVM," 2018 IEEE Recent Advances in
time. When a missing individual is found in one of the frames, Intelligent Computational Systems (RAICS), 2018.
the appropriate authorities will be contacted. [7] “Missing Child Identification Using Face Recognition System” by
Rohit Satle, Vishnuprasad Poojary, John Abraham, Mrs. Shilpa

1882
Authorized licensed use limited to: Zhejiang University. Downloaded on November 16,2024 at 23:45:22 UTC from IEEE Xplore. Restrictions apply.
 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE)

Wakode at International Journal of Advanced Engineering and

Innovative Technology Volume-3 Issue-1 July-Aug 2016.
[8] Arniker, Swarna & Rao, K. & Kalyani, G. & Meena, D. & Lalitha, M.
& Shirisha, K. & Seetasaikiran,. (2014). RFID based missing person
identification system. 1-4. 10.1109/ICIEV.2014.7135998.
[9] Birari Hetal, Sanyashiv Rakesh, Porje Rohan, Salveharish “Android
Based Application - Missing Person Finder” IRE Journals, Volume 1,
Issue 12, June 2018.
[10] Omweri, Thomas & Kahonge, Andrew. (2015). Using a Mobile Based
Web Service to Search for Missing People – A Case Study Of Kenya.
International Journal of Computer Applications Technology and
Research. 4. 507-511. 10.7753/IJCATR0407.1004.
[11] Robust Face Recognition System for E-Crime Alert by Prof. Sumeet
Pate, Deepak G. Maurya , Vinit J. Mhatre, Parmesh K. Yadav at
International Journal for Research in Engineering Application &
Management (IJREAM) Special Issue-01, MAR 2016
[12] Muyambo, Peace. (2018). An Investigation on the Use of LBPH
Algorithm for Face Recognition to Find Missing People in Zimbabwe.
International Journal of Engineering Research and. V7.
10.17577/IJERTV7IS070045.
[13] “Harrisburg University in Pennsylvania develop face recognition
software to predict criminality” Article by Kara Urland , May 6 2020.
[14] “The ethical questions that haunt facial-recognition research” Article
by Richard Van Noorden on Nature 18 November 2020.
[15] “Finding Missing Person Using AI” by Shefali Patil , Pratiksha Gaikar
, Divya Kare , Sanjay Pawar at International Journal of Progressive
Research in Science and Engineering, Vol.2, No.6, June 2021.

1883
Authorized licensed use limited to: Zhejiang University. Downloaded on November 16,2024 at 23:45:22 UTC from IEEE Xplore. Restrictions apply.