Study of face detection and tracking

Since the development of face detection in the 1990’s, the research on its potential applications such as automated face recognition, surveillance and security system, human-computer interaction, etc. has been very active. Together with the development of the instant messenging and teleconferencing,...

Full description

Saved in:
Bibliographic Details
Main Author: Khew, Zong Jie.
Other Authors: Zhu Ce
Format: Final Year Project
Language:English
Published: 2009
Subjects:
Online Access:http://hdl.handle.net/10356/17946
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Since the development of face detection in the 1990’s, the research on its potential applications such as automated face recognition, surveillance and security system, human-computer interaction, etc. has been very active. Together with the development of the instant messenging and teleconferencing, there is a huge increase the usage of these technologies. Integration between these two technologies has been sought after. During the last decades, there had been numerous proposition and research on methods for face detection and face tracking. In this project, it is to study in details a few of the more popular and common existing methods available. The background and introduction of these methods will be shown, together with how the algorithms worked and how they performed. The methods that will be discussed are namely, automatic human face detection and recognition under non-uniform illumination face detection using discriminating feature analysis and Support Vector Machine (SVM), face tracking in Model-based Coding (MBC), multi-expert approach for face detection and multi-view face and eye detection using discriminant features. Lastly, a more enhanced algorithm, Kalman filter algorithm, is displayed and discussed on its functions and effectiveness on multiple face tracking. The various algorithms and detection methods are studied in details such as how they are derived from and how they detect or track faces. Experimental results showing their performance and efficiency are also displayed to justify why they are the more common and popular existing methods being used in many applications nowadays. This project also provides detailed information and insights that can be used as a foundation, to fulfill the possibility of integrating the more enhanced algorithm with other applications such as real-time video streaming for future developments.