Thesis Topic Details

Topic ID:
3131
Title:
Determine face locations in video by computer vision
Supervisor:
Tim Moors
Research Area:
Computer Vision, Networks
Associated Staff
Assessor:
Mahbub Hassan
Topic Details
Status:
Active
Type:
R & D
Programs:
CS CE SE
Group Suitable:
Yes
Industrial:
No
Pre-requisites:
--
Description:
Computer vision is widely used on manufacturing floors, and due to advances in computational power, is becoming viable for real-time analysis of complex scenes such as people in meeting rooms. This project will use computer vision techniques (e.g. starting with the OpenCV library and its implementation of a Haar classifier (see Ch. 13 of this book)) to detect and locate faces in video. Challenges might include exploiting the temporal redundancy in video (and detecting scene changes), in trade-offs between location delay, accuracy and processing cost, and in relating data from multiple cameras. This will give you experience in working with video that is recorded or streamed over a network, and with computer vision techniques, and gesture interfaces that are being developed for computers (e.g. Microsoft's Kinect gaming interface). This will be used to locate people in videoconferencing rooms and to help steer cameras so that faces are properly positioned on screen. Face detection is also useful so that backgrounds can be removed from video, either to create uniform backgrounds behind all participants (and so eliminate clutter), or for privacy so that participants can show their face but not their environment (e.g. home).
Comments:
http://www.eet.unsw.edu.au/~timm/

http://opencv.willowgarage.com/wiki/
http://www.xbox.com/kinect

Past Student Reports
 
No Reports Available. Contact the supervisor for more information.

Check out all available reports in the CSE Thesis Report Library.

NOTE: only current CSE students can login to view and select reports to download.