ANOMALY DETECTION IN

SURVEILLANCE CAMERA
Capstone Project Report
End-Semester Evaluation
Submitted by:
(101603005)ABHIJEET SINGH
(101611044) RAJAN GOYAL
(101611001)ABHISHEK GARG
(101603013)ABHISHEK MODI

BE Fourth Year, COE

CPG No: 104

Under the Mentorship of

Mr. Shatrughan Modi
Lecturer

Computer Science and Engineering Department

TIET, Patiala
Dec 2019

1
 ABSTRACT

In this project, we develop a system for detecting anomalies in surveillance footage. An

anomaly refers to a deviation from the normal routine, for example: a person abandoning a bag
at a crowded or public location, or a computer being stolen from a lab. For efficient large-scale
monitoring, it is critical to design a system based on intelligent video analysis. The system is
based on a simple mathematical model, and works efficiently at resolutions at which most
CCTV cameras function. The framework is based on a Background Subtraction technique, with
pre-processing involving a dual-time background subtraction algorithm, which maintains two
backgrounds, one updated with a relatively high frequency, and the other at a low frequency.
Results show that the system is fast, and robustsinsboth moderately and sparsely populated
areas. As the system does not use expensive filters and statistical calculations, it is less intensive
computationally, and at the same time deliver goodadetection results.

2
 DECLARATION

We hereby declare that the design principles and working prototype model of the project
entitled Anomaly detection in surveillance camera is an authentic record of our own work
carried out in the Computer Science and Engineering Department, TIET, Patiala, under the
guidance of Mr. Shatrughan Modi during 7th semester (2019).

Date: 18-Dec-2019

Roll No. Name Signature

101603005 Abhijeet Singh Chahal

101603013 Abhishek Modi

101611001 Abhishek Garg

101611044 Rajan Goyal

Counter Signed By:

Faculty Mentor:
Mr. Shatrughan Modi
Lecturer
CSED,
TIET, Patiala

3
 ACKNOWLEDGEMENT

We would like to express our thanks to our mentor Mr Shatrughan Modi. He has been of great
help in our venture, and an indispensable resource of technical knowledge. He is truly an
amazing mentor to have.
We are also thankful to Dr Maninder Singh, Head, Computer Science and Engineering
Department, entire faculty and staff of Computer Science and Engineering Department, and
also our friends who devoted their valuable time and helped us in all possible ways towards
successful completion of this project. We thank all those who have contributed either directly
or indirectly towards this project.
Lastly, we would also like to thank our families for their unyielding love and encouragement.
They always wanted the best for us and we admire their determination and sacrifice.

Date: 18-Dec-2019

Roll No. Name Signature

101603005 Abhijeet Singh Chahal

101603013 Abhishek Modi

101611001 Abhishek Garg

101611044 Rajan Goyal

4
TABLE OF CONTENTS
ABSTRACT 2
DECLARATION 3
ACKNOWLEDGEMENT 4
LIST OF TABLES 8
LIST OF FIGURES 9
LIST OF ABBREVIATIONS 11
CHAPTER 1- INTRODUCTION 12
1.1 Project Overview 12
1.1.1 Technical terminology 12
1.1.2 Problem Statement 13
1.1.3 Goal 13
1.1.4 Solution 14
1.2 Need Analysis 14
1.3 Research Gaps 15
1.4 Problem Definition and Scope 17
1.5 Assumptions and Constraints 17
1.6 Approved Objectives 18
1.7 Methodology Used 18
1.8 Project Outcomes and Deliverables 18
1.9 Novelty 19

CHAPTER 2- REQUIREMENT ANALYSIS 20

2.1 Literature Survey 20
2.1.1 Theory Associated With Problem Area 20
2.1.2 Existing System and Solutions 20
2.1.3 Research Findings for Existing Literature 21
2.1.4 Problem Identified 21
2.1.5 Survey of Tools and Technologies Used 22
2.2 Standards 24
2.3 Software Requirement Specification 24
2.3.1 Introduction 24
2.3.1.1 Purpose 24
2.3.1.2 Intended Audience and Reading Suggestions 25
2.3.1.3 Project Scope 25
2.3.2 Overall Description 25
2.3.2.1 Product Perspective 25
2.3.2.2 Product Features 25

5
 2.3.3 External Interface Requirements 25
2.3.3.1 User Interfaces 25
2.3.3.2 Hardware Interfaces 25
2.3.3.3 Software Interfaces 26
2.3.4 Other Non-functional Requirements 26
2.3.4.1 Performance requirements 26
2.3.4.2 Safety Requirements 26
2.3.4.3 Security Requirements 26
2.4 Cost Analysis 27
2.5 Risk Analysis 27
2.5.1 Risks to consider 27
2.5.2 Impact of the risk 27
2.5.3 Unpredictable Factors 27
2.5.4 Predictable Factors 27
CHAPTER 3- METHODOLOGY ADOPTED 28
3.1 Investigative Techniques 28
3.2 Proposed Solution 29
3.3 Work Breakdown Structure 29
3.4 Tools and Technologies Used 29

CHAPTER 4- DESIGN SPECIFICATIONS 30

4.1 System Architecture 30
4.1.1 Component Diagram 30
4.1.2 MVC Architecture 31
4.1.3 Tier Architecture 31
4.2 Design Level diagrams 32
4.2.1 Use case diagram 32
4.2.2 Use Case template 33
4.2.3 Data Design diagram 35
4.2.4 Class diagram 36
4.2.5 Context diagram 36
4.2.6 Sequence diagram 37
4.2.7 State chart diagram 38
4.2.8 Activity diagram 38
4.3 User Interface diagrams 39
4.4 System Screenshots 39

CHAPTER 5- IMPLEMENTATION & EXPERIMENTAL RESULTS 41

5.1 Experimental setup 41
5.2 Experimental Analysis 42
5.2.1 Data 42
5.2.2 Performance Parameter 42

6
 5.3 Working of the Project 42
5.3.1 Procedural workflow 42
5.3.2 Algorithmic Approaches Used 43
5.3.3 System Screenshots 44
5.4 Testing Processes 44
5.4.1 Test Plan 44
5.4.1.1 Features to be tested 44
5.4.1.2 Test strategies 45
5.4.1.3 Test techniques 45
5.4.2 Test cases 46
5.4.3 Test results 47
5.5 Result & Discussions 48
5.6 Inferences drawn 48
5.7 Validation of Objectives 48
CHAPTER 6- CONCLUSIONS AND FUTURE DIRECTIONS 49
6.1 Conclusions 49
6.2 Environmental & Social Benefits 49
6.3 Reflections 50
6.4 Future Work 50

CHAPTER-7 PROJECT METRICS 51

7.1 Challenges Faced 51
7.2 Relevant Subjects 51
7.3 Interdisciplinary Knowledge Sharing 51
7.4 Peer Assessment Matrix 52
7.5 Role Playing and Work schedule 52
7.6 Student Outcome Description & Performance indicators 54
7.7 Brief Analytical Assessment 54

APPENDIX-A: REFERENCES 56

APPENDIX-B: PLAGIARISM REPORT 57

7
LIST OF TABLES

Table No. Table Caption Page No.

Table 1.1 Assumptions and Constraints 17

Table 2.1 Research Findings for Existing Literature 21

Table 2.2 Cost Analysis 27

Table 2.3 Impact of risk 27

Table 3.1 Investigative Techniques 28

Table 4.1 Use Case Template 33

Table 4.2 Use Case Template 34

Table 5.1 Validations of Objectives 48

Table 7.1 Subject codes & Subject names 51

Table 7.2 Peer assessment matrix 52

Table 7.3 Student outcome description & performance Indicators 54

8
LIST OF FIGURES

Figure No. Figure Caption Page No.

Figure 3.1 Work Breakdown Structure 29

Figure 4.1 Component Diagram 30

Figure 4.2 MVC Architecture 31

Figure 4.3 Tier Architecture 31

Figure 4.4 Use Case Diagram 32

Figure 4.5 Data Design Diagram 35

Figure 4.6 Class Diagram 36

Figure 4.7 Context Diagram 36

Figure 4.8 Sequence Diagram 37

Figure 4.9 State Chart Diagram 38

Figure 4.10 Activity Diagram 38

Figure 4.11 Interface Design 39

9
Figure 4.12 GUI pop-up 39

Figure 4.13 Video selection pop-up 40

Figure 4.14 Video popup that is trained 40

Figure 4.15 Graph showed anomaly at peak point 40

Figure 5.1 GUI interface after code processed 41

Figure 5.2 Camera implementation 41

Figure 5.3 Flow Chart 43

Figure 5.4 GUI interface in front of backend 44

Figure 5.5 No anomaly 46

Figure 5.6 Anomaly detected 47

Figure 5.7 End result 47

Figure 7.1 Gantt Chart 52

Figure 7.2 Gantt Chart 53

Figure 7.3 Gantt Chart 53

Figure 7.4 Gantt Chart 53

10
LIST OF ABBREVIATION

MIL Machine Instance Learning

QVGA Quarter Video Graphics Array

CCTV Closed-Circuit Television

HOG Histogram of Oriented Gradients

MDT Mixture of Dynamic textures

KL Kullback–Leibler

Wi-Fi Wireless-Fidelity

GPU Graphical Processing Unit

HMM Hidden Markov Model

PF Particle Filter

GUI Graphical User Interface

11
 CHAPTER 1-INTRODUCTION

1.1 Project Overview

In this undertaking, we build up a framework for location of irregularities in observation

film. Envision a framework that takes feed from reconnaissance cameras, comprehends
when something anomalous happens and transfers this data to the client that is pertinent.
This would free up the HR that are presently being utilized to screen feed from these
cameras, capable cameras that are not being utilized right now to participate without
additional cost and increase significantly more than a post occurrence logical
instrument. In this undertaking the present arrangements accessible for strange
recognition will be audited, a technique for anomaly discovery will be planned,
executed and tried. Results show that the framework is quick, and vigorous in both
modestly and inadequately populated regions. As the framework doesn't utilize costly
channels and measurable estimations, it is less dynamic computationally, and
simultaneously convey great recognition results.

1.1.1 Technical terminology

Surveillance recordings can get an assortment of anomalies. In this project, we propose
a framework to learn abnormalities by using both ordinary and atypical feed recordings.
To abstain from remarking the strange portions or clasps in preparing recordings, which
is very tedious, we have proposed to adapt peculiarity through multiple instance ranking
structure by utilizing weakly named preparing videos. In our methodology, we think
about ordinary and peculiar recordings as packs and video sections as instances in
multiple instance learning (MIL), and consequently get familiar with an irregularity
positioning model that predicts high inconsistency scores for strange video outlines.

12
1.1.2 Problem Statement
The main requirement of anomaly detection through a surveillance camera is
recording the real-time data with high accuracy and create alert during anomaly.
Sometimes, when the camera is not working properly, it may send wrong values
which will generate a false alert. The issue tackled by this project is the location of
abandoned objects in a single colour camera’s video feed quickly, efficiently, and
reliably.

The system must subscribe to certain specifications. The system should work online in
real time, must stay active all the time, and thus should work with varying light
conditions.

If these specifications are not met, then it might be too late to take proper action, as a
bomb may explode, or a shoplifter can escape. The algorithm used by us runs in real
time on modern inexpensive computer systems. It should have no problem running on
old hardware, if the video resolution is QVGA (320*240 pixel resolution), as is the case
with most security CCTV cameras

To summarize, the main problems in anomaly detection through a camera are:

● Accuracy of dataset

● Fast & real-time response

● Secure and functional network

● Working on modern inexpensive systems

● Varying light Conditions

1.1.3 Goal
The goal is to build an application that can detect the anomaly with high accuracy and
to achieve the fastest possible response time of the Anomaly detection algorithm.

13
1.1.4 Solution

We are planning to use CCTV cameras connected with computer. Footage from CCTV
with be sent to application. After that, the data will be sent as input parameters of anomaly
detection algorithm. Algorithm will compute the degree of anomaly and will then classify
the anomaly. If the result is beyond threshold limit, an alert will be generated on
application to check if it is false alert or not.

1.2 Need Analysis

There have been an increasing number of terrorist attacks in crowded public places
such as train stations, subways, airports, malls, etc. Security staff and police forces
rely on video surveillance systems; however, it is labour intensive, and there are limits
of human effectiveness in monitoring and analysing crowded scenes, particularly
when multiple cameras are involved. It is rather unfortunate if there is no one to
monitor the feed, or if negligence leads to an attack. Automated surveillance systems
have been used for offline video analysis, much research is underway on online
analysis and detection of events. We can detect anomalies and suspicious behaviours
in the video feed online, abandoned objects particularly. Many algorithms proposed
to deal with the challenge of abandoned object detection depend on complex
probabilistic mathematics, resulting in the need for large amount of resources, and
their failure to work fast in real time. Other difficulties in detecting abandoned objects
are occlusion by people in a crowd, and the background subtraction lacking
dynamically updating features or being computationally intensive. Successive frames,
the system keeps a lookout for the bag positions, whose presence in anomaly in
footage is not limited to detection of static objects. It can be extended to detecting a
fight, a person jumping a fence or an accident.

When an abandoned bag is detected, the system analyses its database to determine the
most likely bag position(s), where the position is defined as the location where the
bag into the scene is left unattended. Thro or disappearance from the scene defines
the status of the bag, and decides the appropriate course of action. These types of
systems can assist security personnel by providing excellent situational awareness,
enabling them to respond to critical situations more efficiently.

14
1.3 Research Gaps

Loss of consciousness is one of the leading problems faced by the user who monitor
surveillance camera’s feed. To prevent this from happening, a constant supervision of
the application is needed. Hence an alternate surveillance method is required. Anomaly
require immediate attention and the only way one can establish that is by designing a
sensing and alerting method.

According to a survey done by Authors [1], concentrated on circumstances where

following and article recognition is troublesome because of the objects of enthusiasm
being little, the casing rate very low (time-pass like film) and picture quality
additionally low. In this manner the information is treated as a flood of isolated pictures
concentrating on the spatial data in each picture. The fundamental thought is that
everything not saw in the past is bizarre. They separate between uncommon scenes
(never observed) and uncommon scenes (comparable scene seen previously, however
not all the time). A model is found out utilizing the past watched information. At the
point when another picture is caught a closeness measure is determined that contrasts
it and the scholarly model. Diverse comparability measures can be utilized, they utilize
standardized cross relationship. Since the portrayal of the picture has a high
dimensionality most focuses are similarly far off to a question point. Along these lines
idea of helpful closest neighbours is utilized. To do this they store dissemination of the
similitudes. On the off chance that it is improbable that the closeness score of another
picture is drawn from this circulation, at that point the picture is evaluated as irregular.
No transient data is spoken to with this technique. There may be an issue with the testing
being a lot of episodic nature. The outcomes are not altogether reported with ROC or
perplexity grid.
Authors [2] focused around low casing rate information where optical stream or
different strategies can't be utilized. They use the HOG descriptor to portray every area
and a database order framework where every section is a Hoard descriptor. At the point
when a descriptor is brought into the classifier the good ways from the info descriptor
to the closest section in the database is estimated. On the off chance that the closest
separation is more prominent than the inward model separation of the database
classifier the two closest sections are consolidated and another is made with the most

15
up to date input. There is a maximum breaking point to what number of sections there
can be in the database, to keep the computational necessities to low. This is managed
by a steady which can be set in an unexpected way. The strategy accomplish 20 fps.
The strategy works very well for distinguishing strange articles and runs continuously.
There is a solo learning calculation which is ideal and it is information driven so no
physically limits must be set.
Authors [3] in its research, Optical stream is figured for every pixel in every locale
utilizing the Lucas-Kanade strategies. The movement in every area is quantised into
four headings and the movement in the following edges are summarized to make a four-
container histogram containing the movement of 50 edges. This will at that point be
changed over to a codebook of 16 potential blends extending from no critical movement
in either of the 4 headings to huge movement in the entirety of the four bearings. The
classifier is a naive Bayes classifier and a limit is utilized for deciding whether a casing
is irregular or not.
Authors [4] Developed strategy joins both spatial and fleeting anomaly recognition.
There is achieving and a testing stage, so they don't use on-line refreshing of the
ordinary model. For the transient irregularities an MDT is processed on each casing in
the preparation set and in the running stage the registered MDTs are analyzed by
utilizing log probability for the group that is nearest to the MDT. This will create the
fleeting oddity map. For the spatial oddities saliency technique is utilized. This is a
proportion of how a lot of the district stands apart of its environment. A MDT is
determined utilizing a guess, since it would somehow or another be excessively
computationally overwhelming. The spatial irregularity map is registered by utilizing
KL disparity between the MDT of the preparation set and the info outline. The two
abnormality maps are consolidated by including them and the inconsistency discovery
is finished utilizing a straightforward limit. In the testing stage it takes 25s for each
casing to be handled, which makes this technique a long way from constant.

All the above research papers provided us an insight of the techniques and the
algorithms being currently being used in this research area. We, hence, deduced the
need of a portable, low cost and environment independent system which can alert the
user as soon as the anomaly is detected to prevent.

16
1.4 Problem Definition and Scope

The main problem in detection of anomaly is the generation of false alarm. It can have
various reasons like the application is not working properly, the application crashes off
from the user accidentally and some more.
A CCTV camera will be attached to computer through wires or Wi-Fi. Footages from
the camera will be sent to application for the detection of anomaly. This will be done
with the help of Anomaly detection algorithm which will classify the results obtained
after computing the recorded data into normal or anomalous. An alert will be generated
on the application with a 10 seconds window to respond, failing which the security will
be alerted through alarm system.

1.5 Assumptions and Constraints

TABLE 1.1: Assumptions and Constraints

S. No. Assumptions and Constraints

1 The surveillance system requires a fraction of seconds to detect an anomaly, so any anomaly taking
less time than the threshold time will not be detected.

2 A backup power supply should be maintained 24×7 to keep a check on the operational efficiency
of the surveillance system

3 The camera must be super sensitive to low light conditions too.

4 The sensor used may or may not have a higher surviving capacity, thus, malfunctioning of it can
occur, so we need to keep a check on the sensor condition timely.

5 The hardware must contain GPU that efficiently process the algorithms and image processing.

6 For the alarm to get to the active state, a proper trained algorithm must be used to measure the
level of anomaly. If the level is greater than the threshold value, then only the alarm will go the
active state.

17
1.6 Approved Objectives
1. To successfully detect the anomaly
2. To successfully alert the user
3. To ensure that device is recording valid values and in every possible light
condition on modern inexpensive systems.
4. To optimize the time complexity of our anomaly detection algorithm
5. To classify the normal and anomalous videos

1.7 Methodology Used

The proposed methodology uses a machine learning approach for the anomaly
detection.
1. Data Collection
Data is recorded by using surveillance cameras, attached with detection computer
system. Live data is then transmitted to the server by the hardware using the wires
or Wi-Fi.
2. Data Processing
Data is processed through the anomaly detection algorithm which uses machine
learning to classify the data and based on the results, it generates an alert when
required.
3. Anomaly Detection and Alert
If an anomaly is detected, an alert will be sent to the concerned person, i.e., first to
the user, to check if it’s the false alarm or not. In case of true alarm, alert will be
sent to security personnel.

1.8 Project Outcomes and Deliverables

● Anomaly Detection algorithm.
● A software that classify normal and anomalous videos
● Alert generation after anomaly detection on application.
● Window based application for the user and its security system.

18
1.9 Novelty
This project is the one of few area of research and application. We can detect anomalies
and suspicious behaviours in the video feed online, abandoned objects particularly.
Many algorithms proposed to deal with the challenge of abandoned object detection
depend on complex probabilistic mathematics, resulting in the need for large amount
of resources, and their failure to work fast in real time. This leads to extensively save
of man power to detect anomaly on camera as previous systems work manually.

19
 CHAPTER 2- REQUIREMENT ANALYSIS

2.1 Literature Survey

2.1.1 Theory Associated With Problem Area

Literature and research on related work has given us several techniques useable for
anomaly detection within and outside of the domain of this research. Most of the
techniques require knowledge about the characteristics or are specifically made to
detect one anomaly type. This research does not focus on one single anomaly type but
rather tries to find a method to detect any given event in the dataset. A prototype capable
of detecting different anomaly types using the techniques mentioned earlier is designed
for testing and evaluation purposes, but can also be used as anomaly detection tool
outside the scope of this research.

2.1.2 Existing System and Solutions

Computerized video content investigation of a packed scene has been a functioning

examination region in the field of PC vision over the most recent couple of years. This
solid intrigue is driven by the expanded interest for open security at swarmed spaces,
for example, air terminals, train stations, shopping centres, arenas, and so on. In such
scenes, customary PC vision strategies for video observation can't be legitimately
applied in the jam-packed scene because of enormous varieties of group densities,
complex group elements and serious impediments in the scene. Calculations for
individual identification, following and action investigation which think about a person
in detachment (i.e., singular article division and following) frequently face troublesome
circumstances,. Hence, numerous papers think about the group as a solitary element
and investigate its elements. Be that as it may, a packed condition can likewise be
unstructured where people on foot are moderately allowed to move in numerous ways
rather than an organized group where every individual moves rationally one basic way.
In an unstructured swarmed scene, considering the group as one element will neglect
to recognize strange occasions which emerge because of an unseemly activity of a
person in a group.

20
 2.1.3 Research Findings for Existing Literature
TABLE 2.1: Research Findings for Existing Literature
S. Roll Name Paper Title Tools/ Findings Citation
No. Number Technology

1 101603013 Abhishek Hunting Nessie – Real- CCTV-camera Use of real time CCTV [1]
Modi Time Abnormality camera with algorithm in
Detection from Webcams real-time

2 101603005 Abhijeet Real-time detection of CCTV, Reduction of [2]

Singh unusual regions in image microcontroller. misclassification for large
streams non-anomaly movements

3 101611001 Abhishek Stream-based active CCTV HMM has high accuracy [3]
Garg unusual event detection compared to other models

4 101611044 Rajan Goyal Anomaly Detection in CCTV Accuracy of the Anomaly [4]
Crowded Scenes detection

2.1.4 Problem Identified

Many paper deals with demonstrating swarmed scenes to distinguish distinctive group
occasions as well as to recognize irregular occasions. In any case, the meaning of
unusual occasion or occasion of intrigue has been causing a lot of perplexity in the
writing because of its abstract nature. A few analysts consider an uncommon and
extraordinary occasion as anomalous while some consider occasions that have not been
watched are unusual. The issue turns out to be additionally testing as the thickness of
individuals increments. Thus, more PC vision calculations are being investigated as of
late. In spite of the extraordinary intrigue and an enormous number of techniques
created, there is an absence of a far reaching survey on swarm video examination. The
review paper by Zhan et, as far as we could possibly know, is the just one concentrating
on swarm video investigation. Zhan et al. explored some group thickness estimators
and group demonstrating systems, concentrating on walker discoveries, and following
in a jumbled scene. In any case, they didn't examine the point of group conduct
comprehension and variation from the norm recognition which is shrouded in this

21
 overview. We additionally present a few advances on swarm movement demonstrating
and multi-target following in a packed scene which are not shrouded in the past
overview. To accomplish this, we have partitioned the examination on swarm video
investigation into three general classifications: naturally visible demonstrating, tiny
displaying and group occasion discovery.

2.1.5 Survey of Tools and Technologies Used

The methods related to each task are further divided into sub-categories and a
comprehensive description of representative methods is provided.

 Macroscopic Modelling

So as to get familiar with the run of the mill movement designs in a jam-packed scene,
naturally visible perception based techniques use all-encompassing properties of the
scene, for example, movements in nearby spatial-fleeting cuboid or quick movement
are used. It is additionally the favoured strategy in following and examining the conduct
of both meagre and thick group utilizing the accompanying properties, for example,
thickness, speed and stream.

 Optical Flow Feature

Optical stream is a thick field of immediate speeds figured between two successive
casings usually utilized in removing movement highlights. The removed stream vectors
may contain clamour and excess data. So as to diminish the computational expense and
expel clamour, scientists use solo dimensional decrease procedures. In this manner, the
following stage is to locate the agent movement examples of the scene by consolidating
stream vectors from all video outlines.

 Spatial-Temporal Gradient Feature

Notwithstanding optical stream data, different highlights, for example, spatial-worldly

slope are likewise used to show the customary development of a group. In the coupled
Gee is prepared dependent on the circulation of spatial-worldly movements to identify
restricted anomalies in thickly jam-packed scenes. The typicality of a packed scene is

22
 found out utilizing a blend of dynamic surfaces. At that point, worldly and spatial
anomalies are independently identified by discovering deviations from the ordinary
example.

 Microscopic Modelling

Tiny investigation and demonstrating relies upon the examination of video directions
of moving substances. This methodology, all in all, contains the accompanying
advances: 1. identification of the moving targets present in the scene; 2. following of
the distinguished targets; and 3. examination of the directions to identify predominant
streams, and to show run of the mill movement designs. Scientists have utilized diverse
discovery and following calculations to produce solid directions. Following individuals
in groups can be either utilized as a way to improve swarm elements investigation,
utilizing the tracks and mining patterns out of these (base up way to deal with swarm
examination); or, on the other hand, following strategies can utilize signals acquired
from the investigation of group elements, so as to improve precision (top-down
approach). The multifaceted nature of following calculations relies upon the unique
situation and condition where the following is performed.

 The Particle Filter (PF) Framework

The most well-known methodology for following is the Molecule Channel based
system. Molecule sifting system was first presented for visual following by Blake. At
first, molecule channel approaches were just founded on shading prompts, and could
just track one single objective.

 Improving Tracking Using Crowd-Level Cues

Following techniques can utilize signs got from the examination of group elements, so
as to improve their exactness, in a top-down approach. These more significant level
signals can be either logical or originating from the social cooperations among the
individuals in the group. Following in Groups from Different Perspectives Specialists
have likewise investigated the utilization of numerous cameras for following
individuals under extreme impediment in an unpredictable domain. Numerous camera

23
 following strategies expect to grow the observed region and give total data about
fascinating people by social affair confirmations from various camera views.

 Event Detection in Crowds

Recognizing abnormalities or extraordinary events in swarms has moved a great deal

of research endeavours. Programmed frameworks would permit diminishing the weight
of manual video supervision, which makes is infeasible as a rule, given the tremendous
measures of information. Location of oddities in jam-packed scenes can be viewed as
an arrangement issue where just two classes are characterized (i.e., "typical" versus
"atypical"). The overview examinations the works in the writing crosswise over five
viewpoints: 1. the objective/s of intrigue (an individual, a group); 2. The meanings of
what is odd, and the presumptions taken; 3. The kinds of sensors included, and the
highlights utilized; 4. The learning techniques; and 5. The demonstrating calculations.

2.2 Standards

1. Hardware Device Standards

a) A 2.4 MP Bullet HD Surveillance Camera.

b) Connecting Wires.

c) PC System.

2. Software Application Interface Standards

a) Platform Used: Anaconda 3.6

b) Language Used: Python, MATLAB

c) Accessibility: Free/Open Source

2.3 Software Requirement Specification

2.3.1 Introduction
2.3.1.1 Purpose

24
 The purpose of this document is to provide a detailed overview of the software product,
its parameters, and goals. This document describes the project and its target audience
and its user interface, hardware and software requirements. It defines how our client,
team, and the audience see the product and its functionality.

2.3.1.2 Intended Audience and Reading Suggestions

This report is planned for people finding an answer forever supervision of the
inconsistencies in reality. Readers primarily intrigued by a concise presentation of the
item should concentrate on the remainder of Section 1 (Introduction), just as Section 2
of the archive (overall description), which give a primary diagram of each purpose of
the undertaking all in all. Readers who wish to investigate our task in more detail should
peruse on to Section 3 (system feature), which develops the data spread out in the
principle outline. Section 4 (External interface requirement) offers other specialized
subtleties, remembering data for the client level interface just as the equipment and
programming stages on which the application will run. Readers keen on the non-
specialized parts of the undertaking, which covers execution, well-being, security and
different properties can likewise allude of section 2 for clearness. Readers who have
not discovered the data they are searching for should check different necessities, which
incorporates any extra data which doesn't fit coherently into different segments.

2.3.2 Overall Description

2.3.2.1 Product Features
The product is designed to detect any kind of anomaly in real time and alarm the
concerned authorities side by side. It contains an Infra sensitive sensor that could detect
the anomaly in the dark also. Along with it, it is much capable of recording the real time
event and store it in its database.

2.3.3 External Interface Requirements

2.3.3.1 User Interfaces
● The GUI should be user friendly.
● The GUI should be highly responsive.
● User’s data should be secured.

25
2.3.3.2 Hardware Interfaces
● Accurate data is recorded by the hardware device.
● The hardware device and the software should be connected to the internet 24*7.
2.3.3.3 Software Interfaces
● The application should generate alarm in real-time.
● Anomaly detected should be accurate.

2.3.4 Other Non-functional Requirements

2.3.4.1 Performance requirements
● The software will be connected to the hardware device as soon as it is turned on by
the user.
● A user-interface will be displayed on the software so that it can show any possible
warning.
● The data will be transferred to the cloud system for real time calculations and
predictions.
● The user will be given an option to cancel the alarm in case the software has made
false assumptions.

2.3.4.2 Safety Requirements

The data being collected and the alert being generated is available only to the authorized
user. Moreover, the software requires login id and a password to be assessed. Hence,
safety of the data is being taken care of.

2.3.4.3 Security Requirements

The location and anomaly information that is being recorded by the application is secure
with our interface. The data won’t be shared with anyone at all. All the data will be
used exclusively for the predictions from our machine learning model.

2.4 Cost Analysis

26
Table 2.2: Cost Analysis

Product Name Quantity Price/Piece (INR) Total Price (INR)

SP005 IP Wireless Camera 1 2275 2275

AC 1900 USB 2.0 Interface 1 795 795
ABS Electronic Buffer 1 450 450
TOTAL 3520

2.5 Risk Analysis

2.5.1 Risks to Consider

● False alarm generated.

● Too much delay in response.

● Inaccurate values by the sensors.

● Data lost or corrupted during transmission.

2.5.2 Impact of Risks

Table 2.3: Impact of Risks

Risk Probability Impact

False alarm generated Medium High
Too much delay in response Low Medium
Inaccurate values by sensors Medium Medium
Data lost during transmission Low Medium

2.5.3 Predictable Factors

● Sensors not working.

● Low connectivity.

● Model not trained properly

27
 CHAPTER 3- METHODOLOGY ADOPTED

3.1 Investigative Techniques

TABLE 3.1: Investigative Techniques

S. Investigative Project Investigative Techniques Investigative Projects Examples

No. Techniques Description

1 Descriptive An investigation in which scientific questions Projects based on designing completely

are investigated and observations of new system models, concepts, algorithms
phenomenon are recorded and catalogued. etc.

2 Comparative Investigations where observations are made Comparison Based Projects (Algorithm
that compare two objects or phenomenon. based, System based etc.)

3 Experimental An organized investigation that includes a Image Processing based Projects etc.
control group and is designed to test the
hypothesis, includes independent and
dependent variables

4 Analytical An investigation in which a proposed method Image Processing based Projects etc.
is analysed to calculate its effectiveness.

5 Case Study An investigation in which the problem Software Engineering projects, quality
statement is studied in real time Assurance projects, etc.

6 Hypothesis An investigation in which a hypothesis is Image processing and mathematical

proposed to solve the problem statement and projects.
then proved

7 Questionnaire Survey An investigation in which a survey is Comparison Based Projects (Algorithm

circulated to compare results based on based, System based etc.)
feedback either from a targeted audience or
from a variable audience.

8 Literature Review An investigation in which a comprehensive All types of projects.

study of the existing projects is done.

28
3.2 Proposed Solution

Research on Anomaly detection is going on from decades and many fascinating new
technologies have been introduced in the said research, our model will represent an
working model of automated implemented with methodology. Now most of the CCTV
need manual service to run it consumes their precious time. By efficient system number
of incidents can also be reduced. Our proposed model build with deep learning would
run on real track made by us following the same rules of a camera running.

3.3 Work Breakdown Structure

FIGURE 3.1: Work Breakdown Structure

3.4 Tools and Technologies Used

 CCTV camera
 Deep learning
 Sequential model

29
 CHAPTER 4- DESIGN SPECIFICATIONS

4.1 System Architecture

4.1.1 Component Diagram

FIGURE 4.1: Component Diagram

30
4.1.2 MVC Architecture

FIGURE 4.2: MVC Architecture

4.1.3 Tier Architecture

FIGURE 4.3: Tier Architecture

31
4.2 Design Level diagrams

4.2.1 Use case diagram

FIGURE 4.4: Use Case Diagram

32
4.2.2 Use Case template
TABLE 4.1: Use Case Template

Name Login

This use-case depicts that if you need to use our software you need proper user-
Description id and password to use it for security reasons.

1. Software is uploaded on hardware.

Assumptions 2. Software should be configured properly.

3. User and Caregivers are comfortable with using software.

Trigger Users is provided user id and password for login.

Primary Actor User

Secondary Actor -

Interests This use-case is created to ensure their safety.

1. Software is built on hardware.

Preconditions 2. User have registered and have their login id and password.

1. User starts the software and enters the correct user-id and password to
Normal Scenario monitor screen displays.

Registration of user is done during software uploading.

2.

Extensions 1. If user forgets password then they can verify themselves by forgot
password option.

Alternate Scenario -

Success end conditions:

1. Login id and password verified successfully.

2. If user forgets password they can verify themselves by forgot password
option.

Post-conditions Failure end conditions:

1. Login id and password is not verified.

2. User is not verified by forgot password option.

33
 Minimal Guarantee:
1. User is verified by customer care.

Special Requirements -

Frequency of Occurrence Login use-case is used only one time while using the software to open homepage.

Issues -

TABLE 4.2: Use Case Template

Name Anomaly Acknowledged

The Anomaly acknowledgement in diagram depicts that when our device tracks
Description the movements of people and if any event which seem to be odd it alerts the user
so they can take action accordingly.

1. Hardware should be ON all the time.

2. Sufficient environment condition for camera.
Assumptions 3. Hardware must be compatible with software.
4. User have little knowledge to use software.

Trigger It is trigger when any unexpected event occurs.

Primary Actor User (Security head/Local Retailer)

Secondary Actor -

Stakeholders and Interests People owning shops, gyms, malls etc. who want security in their system.

Precondition 1. All camera/cameras must be connected to hardware.

2. The user should be logged in.
3. Hardware contain GPU.

1. On occurrence of Anomaly in frame software detects it and alert is given to

Normal Scenario user.
2. On false detection user can ignore alert.

34
 1. On true anomaly detection user informs the security handler/ attendant.
Extensions
2. Attendant takes the action according to the situation.

Alternate Scenario -

Success end conditions:

1. Anomaly detected successfully.
2. Right action taken by the security
Post-conditions Failure end conditions:
1. False anomaly alert message.

Minimal Guarantee:
1. Data is recorded correctly and is then classified to
Detect the anomaly.

Performance:
1. The anomaly should be detected in real time.
Special Requirements 2. Power consumption should be less
User Interface:
1. It should be very simple for the user.
2. It should take correct decision.

Frequency of Occurrence When anomaly is detected at that time it will give alert message.

Issues False anomaly generated by the device.

4.2.3 Data Design diagram

FIGURE 4.5: Data Design Diagram

35
4.2.4 Class diagram

FIGURE 4.6: Class Diagram

4.2.5 Context diagram

FIGURE 4.7: Context Diagram

36
4.2.6 Sequence diagram

FIGURE 4.8: Sequence Diagram

37
4.2.7 State chart diagram

FIGURE 4.9: State Chart Diagram

4.2.8 Activity diagram

FIGURE 4.10: Activity Diagram

38
4.3 User Interface diagrams

FIGURE 4.11: Interface Design

4.4 System Screenshots

FIGURE 4.12: GUI pop-up after running the code

39
FIGURE 4.13: Interface for selecting the video that will be trained

FIGURE 4.14: Video pop-up that is trained

FIGURE 4.15: Graph showed anomaly at peak point

40
 CHAPTER 5-IMPLEMENTATON & EXPERIMENTAL
RESULTS

5.1 Experimental Setup

FIGURE 5.1: GUI interface after processed code

FIGURE 5.2: CAMERA Implementation

In the project the camera is connected to the USB port of the laptop which records the video
signals which is then tested by the anomaly detection model and the final result is shown by
the graph.

41
5.2 Experimental Analysis

5.2.1 Data
For our model we have collected the data from the different CCTV camera all around the
s world in different cities which was easily available on the Internet. In our model we have
d trained our model both on negative data and positive data to increase the accuracy.

5.2.2 Performance parameter

 User satisfaction
We are trying to provide more security to the user by detecting anomaly by artificial
a Intelligence, so it’s beneficial initiative from user’s end.
 Time
Completion of the project is 11 months starting from February 2019 to November
xds 2019. In the given time frame all the deliverables have been achieved.
 Cost
It is a cost efficient prototype till now, if we implement it on a large scale the
effectiveness would be good but we will be requiring the dataset of different type of
anomaly which can be easily available through the CCTV cameras.
 Quality
This projects success would definitely increase the quality of current research going
O in this field as we are using deep learning. Development of this project would lead to
m increase in security and reduce the number of crimes. Hence proving quality to nation
x as well by reducing the crimes.

5.3 Working of project

5.3.1 Procedural Workflow

This device consists of camera which records the real time videos which are constantly,
and C3D features are extracted from the recorded whole video and divide the video into
32 segments. These features are then send to the created CNN (convolutional neural
network) which is trained with training videos .Model is created in Json language and
weights are set for the neural network. The data is displayed in frame vs. probability
graph. When-ever there is uncertainty in video probability in graph rises which shows

42
 that there is possibility of anomaly at that frame number. Model can also detect anomaly
in previously recorded videos.

5.3.2 Algorithmic Approaches Used

Step 1: Algorithm is loaded and camera is turned on.

Step 2: Camera start recording the video and features of the model are passed in CNN model.

Step 3: weights and json model is ready for any conflict detection.

Step 4: Anomaly occurs.

Step 5: Graph raises for the frame at which anomaly occurs.

Step 6: Even after anomaly there is continuous formation of graph. When there is low

probability of anomaly graph remains down, when there is high probability graph rises.

Figure 5.3: Flow chart

43
5.3.3 System Screenshots

Figure 5.4: GUI interface in front of backend

GUI will come which will have button to start the Uploading of the video.

5.4 Testing Process

5.4.1 Test Plan

A test plan for software project is defined as a document that defines the objective, scope,
emphasis and approach on a software testing effort. Components of Test plan include: Test
plan id, features to be tested, test techniques, testing tasks, features pass or fail criteria, test
deliverables, responsibilities, and schedule.

5.4.1.1 Features to be tested

 Receiving data from the camera into the anomaly detection algorithm.

 Video data conversion into text file for c3d feature extraction.

 Anomaly Detection Algorithm.

44
5.4.1.2 Test Strategy

Test strategy is a set of guidelines that explains test design and determines how testing
needs to be done.

So here we are going to test different modules of the algorithm and check if there are
any errors. If any errors are encountered, we will try to correct them with minimal
changes that we can do in the project, making sure that these changes won't create any
new errors.

5.4.1.3 Test Techniques

Unit Testing: Unit testing is a level of software testing where singular module of the
product are verified. The intention is to approve that every unit of the product proceeds
as structured. A unit is the littlest testable piece of any product. It as a rule has one or a
couple of sources of inputs and normally a singular yield.

Modules to be tested are:

Hardware

Adjusting the camera from how far it can detect.

Software

 3d feature extraction module - Checking if feature is being extracted properly.

 Jason model setting module - Taking care of the weights are being adjusted properly.

 Graph formation module - If graph is constantly showing its results.

 Detecting module - if the algorithm is detecting the anomaly

Integration Testing: Integration Testing is a level of software testing where singular

units are consolidated and verified as a gathering. The motivation behind this degree of
testing is to uncover errors in the association between incorporated units. Test pilots
and test stubs are utilized to aid Integration Testing.

 Model Formation - 3d feature extraction and json model for adjusting the weight is
tested combined.

45
  Detecting and Graph formation - Check if a graph is being built constantly and if
anomaly occurs the graph is changing accordingly.

 Whole Model - Checking if the model is working as whole.

Acceptance Testing: Acceptance Testing is a level of software testing where a system

is tested for acceptability. The purpose of this test is to evaluate the system’s compliance
with the business requirements and assess whether it is acceptable for delivery. All
modules are to be tested with real time user values.

Regression Testing: Regression Testing is a kind of software testing that expects to

guarantee that changes (upgrades or imperfection fixes) to the product have not
unfavorably influenced it. The probability of any code change affecting functionalities
that are not legitimately connected with the code is consistently there and it is basic that
regression testing is directed to ensure that fixing one thing has not broken something
else. During this testing, new experiments are not made however recently made
experiments are re-executed.

5.4.2 Test Case

We tested the anomaly detection accuracy of the system with the help of the fighting
videos and we got this results

Figure 5.5: No Anomaly

This picture has no anomaly as there is no fight scene and graph is also on normal side.

46
 Figure 5.6: Anomaly Detected

This figures show the graph before and after fighting scene as it can be seen that the graph
goes up and it is detected by the graph. There were many videos which were used for test
cases and above fig is one of it.

Though the test case we found had somewhat delay in the detecting which can be further be
improved.

5.4.3 Test Results

Figure 5.7: End Result

We will get this type of graph after processing the real time feed . Peak of the graph will
show the presence of anomaly in the video feed which we are getting through the camera.

47
5.5 Results and Discussions

Following results are achieved from the project:

● The device is successfully generating alert in the case of anomaly as an alarm, in the real
time anomaly scenario and surveillance centre too.

● In case of a false alarm, system is resetting successfully on pressing the signal button
provided with the help of manual supervision.

● Device is continuously notifying the supervisors and nearby people in case of anomaly site,
till the problem is not solved and normalcy is not retained.

5.6 Inferences Drawn

Following inferences can be drawn by the working of this project:

● The hardware and the algorithm needs to be connected every time for anomaly detection.

● Hundreds & thousands of types of anomalies can be detected through this system.

● The alarm for the anomaly detection is being generated in an appropriate time frames.

● Based on the expected outcomes, the system is working well.

5.7 Validation of Objectives

Table 5.1: Validation of Objectives

S. No. Objective Status

1 To successfully alert about the anomaly and its surroundings. Done

2 To ensure that the algorithm is recording optimal values. Done

3 To optimize the time complexity of the anomaly detection algorithm. Done

48
 CHAPTER 6- CONCLUSIONS AND FUTURE DIRECTIONS

6.1 Conclusion

In order to detect real world irregularities in tracking images, we suggest a deep

research approach. The data collected from real-world applications are becoming larger
and larger in size and dimension. The information gathered from real world applications
are increasing and bigger in size and measurement. As the dimensionality expands, the
information objects become scanty, bringing about recognizing oddities being more
challenging. In addition, the traditional peculiarity identification techniques can't work
adequately and proficiently. We have talked about ordinary issues of oddity location
related with the high-dimensional and blended sort information and quickly checked on
the methods of anomaly recognition. To offer a superior comprehension of the
peculiarity recognition strategies for experts, we led a trial on openly accessible
datasets. In spite of the fact that the advances of abnormality identification for the high-
dimensional and blended sort information have been accomplished somewhat, there are
additionally a few issues, for example, irregularities are moderately uncommon and the
ground truth is frequently inaccessible in genuine situations, how to adequately and
thoroughly assess the location execution is likewise a difficult issue.

6.2 Environmental & Social Benefits

The product does not harm the environment in any sense. The project ‘Anomaly
Detection in Surveillance Camera’ to help the detection of abnormal behaviour of
anyone. There have been an increasing number of terrorist attacks in crowded public
places such as train stations, subways, airports, malls, etc. Security staff and police
forces rely on video surveillance systems; however, it is labour intensive, and there are
limits of human effectiveness in monitoring and analysing crowded scenes, particularly
when multiple cameras are involved. It is rather unfortunate if there is no one to monitor
the feed, or if negligence leads to an attack. Automated surveillance systems have been
used for mostly offline video analysis after an event has occurred, much research is

49
 underway on online analysis and prediction of events. We can detect anomalies and
suspicious behaviours in the video feed online, abandoned objects particularly.

6.3 Reflections
Various skills were enhanced by working on this project as team work, time
management, decision making, multitasking etc. All these reflections from the project
helped us to develop personally as well as on academic grounds.

6.4 Future Work

Due to time constraints, only a limited amount of implementation has been done till
now. The next step in the project could be to train more dataset to detect anomaly in
every scenario. In addition to this, System could be trained to automatically inform the
security personnel. Also, we could classify the type of anomaly by analysing the
algorithm readings. This will convert the simple classification machine learning
algorithm to a multi-class classification algorithm. Other sensors can be used to record
other necessary data.

50
 CHAPTER 7-PROJECT METRICS

7.1 Challenges Faced

Choosing the hardware parts: We faced various problems while choosing hardware camera
as it had to be a High Definition camera with various type of defined parameters.

Selection of the anomaly algorithm: The prime problem we faced during the project was to
use what kind of algorithm, that is, it should be clearly application focused and not
generalised at the same optimal level.

Implementation and Data Collection: The main problem was to train the model with the data
set which was very difficult to find and was over 100 of GB’s and required a very powerful
UI to run.

7.2 Relevant Subjects

Table 7.1: Subject Code and Subject Names

Subject Code Subject name Description

It helped us to implement the concept of Deep Learning
UCS742 Deep Learning
Models like CNN.
To analyse the real time data frames we are getting through the
UCS522 Computer Vision
algorithm.
UML602 Machine Learning We were taught about various class definers like SVM etc.
It helped us in creating a dataset that is far more advanced
UCS633 Data Analytics and Visualization
when it comes to implementation.

7.3 Interdisciplinary Knowledge Sharing

This project does not have much of an interdisciplinary approach as many objects and
processes of the relevant subjects such as deep learning, computer vision belong to the same
discipline i.e. computer engineering.

51
 7.4 Peer Assessment Matrix
Table 7.2: Peer assessment matrix

EVALUATION OF

Abhishek Modi Abhishek Garg Rajan Goyal Abhijeet Singh

Abhishek Modi 5 5 5 5

EVALUATION Abhishek Garg 5 5 5 5

BY Rajan Goyal 5 5 5 5

Abhijeet Singh 5 5 5 5

7.5 Role Playing and Work Schedule

1. Abhijeet Singh: Project analysis and design, hardware analysis, poster, dataset creation and
report work.

2. Rajan Goyal: Hardware design, project analysis and design, dataset creation, video, report
work and research findings.

3. Abhishek Garg: Project idea and design, hardware analysis, report work, research findings
and document formatting.

4. Abhishek Modi: Hardware analysis and design, project design, document formatting,
dataset creation, report work and research findings.

Figure 7.1-Gantt chart of Abhijeet Singh

52
 Figure 7.2 Gantt chart of Rajan Goyal

Figure 7.3- Gantt chart of Abhishek Garg

Figure 7.4- Gantt chart of Abhishek Modi

53
 7.6 Student Outcomes Description and Performance Indicators
Table 7.3: Student Outcomes Description and Performance Indicators

SO DESCRIPTION OUTCOME

Applying basic principles of science towards solving

A2. Used a camera to predict the anomaly.
real life problems.
Identify the constraints, restrictions and models for the Various restrictions were identified by us which could lead to
B1
problems. false alerts.
Use tested methods, tools and techniques for data Data frames from the algorithm were stored as .txt files while
B2
collection. performing various reforms.
Analyse and interpret results with respect to False alarm button was introduced in case a false alarm was
B3
assumptions, restrictions and theory. ringing.
All the team members worked on both, the working of
D2 Can play different roles as a team player.
algorithm as well as the report documentation.
Use analytical and computational methods to obtain Used the readings from each and every dimension to analyse
E3
solutions. each and every aspect of the anomaly.
Showcase professional responsibility while interacting For the data collection, we surveyed a lot of malls and
F1
with peers and professional communities. crowded places to learn and define the real time problems.
Produce a variety of documents such as laboratory or All the team members worked on documentation process and
G1
project reports using appropriate formats. ensured the value of time is maintained.
The team delivered PPTs on various occasions to show the
G2 Deliver a clear and effective oral presentation.
progress to our mentor and the panel.
Various codes and conversions were used while writing the
K2. Apply various codes and algorithmic processes.
code.

7.7 Brief Analytic Assessment

Q1. What sources of information did your team explored to arrive at the list of possible
Project Problems?
Ans: We studied various Research papers, visited different websites like sih.gov.in,
kaggle and consulted our mentor to arrive at the list of possible project problems.

Q2. What analytical, computational and/or experimental methods did your project team
use to obtain solutions to the problems in the project?
Ans: We looked at different resources like research papers, journals, websites etc. and
consulted our mentor to obtain solutions to the problems in the project.

54
Q3. Did the project demand demonstration of knowledge of fundamentals, scientific
and/or engineering principles? If yes, how did you apply?
Ans: Yes, our object demanded demonstration of various principles. It includes rigorous
training of algorithm using dataset that has to convert into txt files.

Q4. How did your team shares responsibility and communicate the information of
schedule with others in team to coordinate design and manufacturing dependencies?
Ans: All the work was divided equally among the team members. As all the members were
from the same class group and had already worked together previously on various projects,
there was no communication gap between them.

Q5. What resources did you use to learn new materials not taught in class for the course
of the project?
Ans: Our mentor helped us a lot during the project. Whenever we reached a dead end, she
was always there to rescue us. Various research papers were also used to learn new materials
that were not taught previously in the class.

Q6. Does the project make you appreciate the need to solve problems in real life using
engineering and could the project development make you proficient with software
development tools and environments?
Ans: Yes, the project makes us appreciate the need to solve problems in real life. We are
hopeful that one day our project will surely help old people or those people who cannot take
care of themselves and also, the project development made us proficient with software
development and environments.

55
REFERENCES

[1] Michael D. Breitenstein, Helmut Grabner and Luc Van Gool. “Hunting Nessie – Real-Time
Abnormality Detection from Webcams." IEEE 12th International Conference on Computer
Vision Workshops (ICCV Workshops) 118.18 (2009): 2537-2543.

[2] R. Schuster, R. M¨orzinger, W. Haas, H. Grabner and Luc Van Gool, etc. al " Real-time
detection of unusual regions in image streams." Proceedings of the international conference
on Multimedia 39.10 (2010): 105002.

[3] Chen Change Loy, Tao Xiang and Shaogang Gong, et al. “Stream-based active unusual
event detection” Procedia Computer Science 141 (2008): 358-365.

[4] Vijay Mahadevan, Weixin Li, Viral Bhalodia and Nuno Vasconcelos. “Anomaly Detection
in Crowded Scenes." Instrumentation Science & Technology 44.4 (2010): 333-342.

56
PLAGIARISM REPORT

57
58
59