Please use this identifier to cite or link to this item:
Title: Pedestrian and car traffic analysis by computer vision – classification development
Authors: Chan, Hoi Man
Department: Department of Electronic Engineering
Issue Date: 2008
Supervisor: Supervisor: Dr. Yuen, Kelvin S Y.; Assessor: Dr. So, H C
Abstract: This project presents a technique for classifying the regions for moving cars and walking pedestrians on a road. A static camera is required to acquire video sequences. The data obtained by this classification algorithm is useful for traffic analysis and behaviors monitor of cars and pedestrians. In this project, optical flow is used to find the velocity vector field which indicates the moving directions of the pixels. Then, Gaussian mixture models (GMMs) are obtained by applying the regularized Expectation-Maximization (EM) algorithm. The GMM represents a continuous probability density function of the direction of the vector for a pixel. Finally, delta rule training algorithm is used to train the GMM in order to classify it. For the classification performance, the classification algorithm developed can successfully classify the region for moving cars and walking pedestrians and the background can be extracted.
Appears in Collections:Electronic Engineering - Undergraduate Final Year Projects

Files in This Item:
File SizeFormat 
fulltext.html146 BHTMLView/Open

Items in Digital CityU Collections are protected by copyright, with all rights reserved, unless otherwise indicated.