Integrating depth and intensity information for vision-based head tracking
Head tracking has been an area of active research in computer vision for several years. Until recently, systems that produce real-time, stereo depth data have been unavailable in the commercial market. As a result, most approaches to this problem relied exclusively on intensity images, using color c...
Saved in:
| Main Author: | |
|---|---|
| Format: | Dissertation |
| Language: | English |
| Published: |
ProQuest Dissertations & Theses
01-01-2008
|
| Subjects: | |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Head tracking has been an area of active research in computer vision for several years. Until recently, systems that produce real-time, stereo depth data have been unavailable in the commercial market. As a result, most approaches to this problem relied exclusively on intensity images, using color cues and intensity edges for head detection and tracking. In this project we present an algorithm for head tracking using depth information provided by stereo camera system. This algorithm works better than existing head tracking algorithms that use intensity information which suffer due to background clutter as well as lighting variations. In order to overcome the above mentioned problems, our method uses stereo depth information along with intensity information (where the depth information is not available) to perform effective segmentation of the foreground. Once the segmentation is done, the lean of the body (the angle made with the vertical axis) is found. Local edge detectors are used to find the neck point. Using the neck point location, we look for the head in the region above the neck point and fit the head model in the image. |
|---|---|
| ISBN: | 0549696229 9780549696223 |