Vehicle and human tracking adapting different analysis techniques

• Main Author: Mok, Serene Yin Leng
• Other Authors: Wang Gang
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/68023
• Institution: Nanyang Technological University

Object tracking is one of the main problems countered in the process of perfecting computer vision today. Object tracking is very useful in a variety of applications such as driver assistant program and camera surveillance. Tracking is a process whereby an object is being located and followed in either a real-time or pre-recorded series of image sequences. This has been a very challenging problem as most of the present methods are unable to handle different wide varieties of complex images. Therefore it is very important for a tracking system to be equipped with a versatile adaptive filter, also often known as kernel or a classifier, to be able to differentiate the target’s characteristics from its background for efficient and precise tracking. Currently there are many proposed methods and techniques of object tracking, all having their own strengths and weakness. However it is important to know which proposed solution would be best for tracking human and vehicles, therefore comes the purpose of this paper. Using MATLAB algorithms, this paper will analyze the basis of 3 different object tracking (DSST, Color Visual and KCF) and how it is applied to human and vehicle tracking. It will also explore and evaluate the different types of image analysis techniques and filters used for effective and robust tracking.