Abstraction

Object tracking is a crucial task in image processing that involves identifying

and following the movement of objects within a sequence of images or
videos. This report aims to provide an overview of object tracking
techniques, their applications, and challenges. It also presents a practical
implementation of object tracking using a popular algorithm and discusses
its strengths and limitations. The report concludes by highlighting future
directions and potential advancements in object tracking.
The tracking process can be thought of as a combination of two models: the
motion model and the appearance model. The motion model tracks the
speed and direction of the object's movement, which allows it to predict a
new position of the object based on the received data. At the same time,
the appearance model is responsible for determining if the object we have
selected is inside the frame. In the case of using pre-trained classifiers, the
coordinates of the bounding box containing the object would be
determined automatically, whereas by using “online” training. we specify
the bounding box manually and the classifier does not have training data,
except for those that it can receive while tracking the object. It is worth
noting that tracking algorithms can be divided into two groups: single-
object tracking and multi-object tracking algorithms, we will consider the
former.
Simply put, locating an object in successive frames of a video is called
tracking. Usually tracking algorithms are faster than detection algorithms.
The reason is simple. When you are tracking an object that was detected in
the previous frame, you know a lot about the appearance of the object.

1
 INDEX

Chapter No Chapter Title Page No

01 Introduction 01-02

02 Literature Review 03-04

03 Proposed System Design 05-08

04 Result Analysis 09-11

05 Conclusion 12-13

References 14

2
 Chapter 1
Introduction

3
1.1 Introduction
Image processing is a form of processing with input as image such as
photograph or video frame and output can be characteristics or parameters
related to image. Computer vision is an area that consists of methods for
incorporating, analyzing and visualizing images. Surveillance stands for
monitoring the behaviour, activities, and other changing information,
usually of people for the purpose of influencing, directing and protecting
them. The process of locating moving object using a camera is video
tracking. In simple terms, tracking means associate target objects in
consecutive video frames. Difficulties arise especially when objects are
moving rapidly as compared to frame rate or when the tracked object
changes direction over time. A sequential flow of object detection, object
tracking, object identification and its behavior completes the process
framework of tracking.
1.2 Objective
The primary objective of object tracking is to follow the spatial and
temporal changes of a target object as it moves through a video or image
sequence. This involves tasks such as detecting the initial location of the
object, estimating its motion, and predicting its future trajectory. Object
tracking is a challenging problem due to factors such as object appearance
variations, occlusions, illumination changes, cluttered backgrounds, and
complex object dynamics.

1.3 Feature

⦁ Real-Time Tracking
⦁ Motion Estimation
⦁ Target Identification
⦁ Motion detection
4
⦁ Tracking framework
⦁ Data Association

5
 Chapter-2
Literature Review

6
2.1 Introduction
Object tracking is an important topic in computer vision that has numerous
applications in various fields. In recent years, there has been extensive
research on object tracking, and several techniques have been proposed to
achieve accurate and efficient tracking. In this literature review, we will
explore some of the recent works related to object tracking. Deep learning-
based methods, multi-object tracking, and online object tracking are some
of the popular research directions in this field.
2.2 Related Works
In recent years, several researchers have proposed various methods for
object tracking using OpenCV in Python.one of them. "Deep SORT: Simple
Online and Realtime Tracking with a Deep Association Metric" by Nikolay
Wojke et al. (2017): This paper introduces Deep SORT, a deep learning-
based object tracking method that combines a convolutional neural
network (CNN) for feature extraction and the Hungarian algorithm for data
association. It achieves real-time performance and demonstrates high
accuracy on multiple object tracking tasks.
Another popular technique for object tracking "SiameseFC: Tracking as a
Pairwise Ranking Problem" by Luca Bertinetto et al. (2016): This paper
proposes the SiameseFC framework for object tracking. It uses a siamese
network architecture to learn a similarity function that compares the target
object with candidate regions in subsequent frames. The method achieves
state-of-the-art performance on various tracking benchmarks.
Online object tracking is another important area of research, where the
goal is to track the object in real-time. Several techniques have been
proposed for online object tracking, including correlation filters, Siamese
networks, and Kalman filters.

7
 Chapter-3
Proposal System design

8
3.1 Introduction
After analyzing the proposed system of different sources it is concluded
that the existing systems are less effective and less efficient enough for the
satisfactions and demand on improve object tracking system. So, we
proposed a user-friendly, more efficient, more effective and fast system to
develop object tracking system.

3.2 Methodology
The project will use the OpenCV library in Python, which is a powerful tool
for image processing and computer vision tasks. The following steps will be
taken to implement the project:
Outline the methodology and steps involved in designing and implementing
the object tracking system. Include the following components:
Data Collection: Specify the data sources, such as video sequences or live
camera feeds, and discuss the requirements for data acquisition, including
resolution, frame rate, and lighting conditions.
Preprocessing: Describe the preprocessing steps, such as image
enhancement, noise reduction, and frame alignment, to improve the
quality of input data for more accurate tracking.
Object Detection/Initialization: Explain the approach for detecting or
initializing the target object in the first frame. Discuss the use of object
detection algorithms, region proposal methods, or manual initialization
techniques.
Feature Extraction: Discuss the choice of features, such as appearance-
based features, shape-based features, or deep learning-based features, and
their extraction methodologies.
Motion Estimation: Describe the methods for estimating the object's
motion between consecutive frames. Discuss techniques like optical flow,
feature tracking.

9
Data Association: Explain the algorithms used for associating the target
object across frames.
Real-Time Implementation: Describe the implementation details of the
tracking system to achieve real-time performance.

.3.3 Tools & Technology

To develop a system tools and technologies plays the most important role.
The tools denote the components that are used to create the techniques
environment for the programmers. Technology is a collection of skills,
methods and processes used in the production of goods or services or in
the accomplishment of the objectives, such as scientific investigation.
Technology is the knowledge of techniques, processes etc. or it is the
embedded in machines, computers, devices and factories, which can be
operated by individuals without detailed knowledge of the workings of such
things The technology describes the way to code a particular project.

3.4 Hardware Requirements

⦁ Desktop PC or a Laptop

⦁ Operating System – Windows 10

⦁ Keyboard and Mouse

⦁ Ram 4 GB

⦁ Webcam

⦁ Computer Vision Robot Arm

3.5 Software Requirements

⦁ PyCharm IDE

10
⦁ Python 3.10

⦁ OpenCV

⦁ SVZone

⦁ Media Pipe

⦁ NumPy

⦁ Matplotlib

11
3.6 ER Diagram

12
 Chapter-4
Result analysis

13
4.1 Result & Discussion
We were able to successfully implement the background removal system
using OpenCV in Python. Several experiments had been done to evaluate
the tracking algorithms. These sequences used in experiments consist of
indoors and outdoors testing environments.so that the proposed scheme
can be fully evaluated. First, target object of interest is defined from the
first some frames. Then SIFT features are obtained from the target object. A
trajectory is the path that a moving object follows through space as a
function of time. A trajectory can be described mathematical either by the
geometry of the path or as the position of the object over time. It will store
the actual path of object of interest information of target in consecutive
frames. We will get the all information about target object that in which
direction it moves and what is the speed of target. The features are stored
by their key points descriptors. The results were tested on various sample
images and videos, and the following observations were made:
⦁ The algorithm accurately identifies the foreground object tracking
with high precision.

⦁ The computational efficiency of the algorithm is high, allowing real-

time processing of video streams with minimal latency.

⦁ The algorithm performs well under different lighting conditions,

including low-light and high-contrast scenarios.

⦁ The algorithm is robust to noise and interference, producing accurate

results even in the presence of image artifacts and visual clutter.

⦁ The algorithm can be easily integrated with existing video

conferencing applications, such as Zoom, to provide a seamless and
user-friendly experience.

14
4.2 Screenshot & Output
1. The result of Human Face Tracking in image sequences computed by the
method here is shown in the following figures:

A. Motion in Left Direction B. Motion in right Direction

15
C. Motion in Down Direction D. Motion in Up Direction

16
Chapter-5
Conclusion

17
5.1 Conclusion
In conclusion, this project demonstrates the use of OpenCV in Python to
implement object tracking. The work introduces an efficient method for
motion detection of a moving object utilizing frame differencing, median
filtering and morphological operations. The experiment method has good
performance and efficiency and provides accurate results.

5.2 Limitations

However, there are some limitations to this approach. One of the main
challenges When objects are partially or fully occluded by other objects or
obstacles, it becomes challenging for the tracking system to maintain
continuous tracking. The tracker may lose track of the object or mistakenly
track the occluding object instead.

18
 Reference
[1] Murtaza's Workshop - Robotics and AI YouTube Channel

(https://www.youtube.com/watch?v=1FJWXOO1SRI)

[2] Learning Python & Different Type of Package

(https://www.stayahead.com/training-courses/Python-Programming-
training-courses/StayAhead-Python-Training-Course-List.html?gad=1
&gclid=Cj0KCQjw0tKiBhC6ARIsAAOXutlf1iNmTclotTrSZnpauxm9SYWlwZmk
VUKaqke68o2J-cHZknOWc4saAqrIEALw_wcB)

[3] Karan Gupta, Anjali V. Kulkarni, “Implementation of an Automated

Single Camera Object Tracking System using Frame differencing and
Dynamic Template Matching”, International Joint Conferences on Advances
in Computer, Information, System Sciences and Engg. Sponsored by IEEE,
Bridgeport,pp. 245-250, 2007.

[4] G. Mallikarjuna Rao and Dr.Ch.Satyanarayana, “Object Tracking System

Using Approximate Median Filter, Kalman Filter and Dynamic Template
Matching”, I.J. Intelligent Systems and Applications,Vol.6, No.5, pp. 83-89,
2014.

19