Vehicle Tracking and Video Surveillance

Abstract

This project aims at creating a full edged Vehicle Tracking and Vehicle Identification System. The project presents a thorough literature survey of the work done in this field. Usually there are many hurdles in the process of tracking viz. illumination changes, occlu- sion, shadows etc. In this paper a model formulating solutions to overcome illumination variations is discussed. This work compares various background subtraction techniques and analyses gingerly the advantages and drawbacks of the existing Background Subtrac- tion Algorithms. This project discusses various approaches for tracking moving objects, and compared these methods. Our objective is to detect moving vehicles and classify them in different classes such as cars and bikes. We have approached the classification problem in two ways namely using SVM classifier and using Deep Neural Networks. We have used ORB and HOG features for training the SVM classifier.