Tracking people in a video. I will supply videos, and the first stage of the project is to track the people in the video and output a "bounding-frame" around them. I rather this to be done in c++, but matlab is also an option.
Additional Project Description:
06/17/2013 at 22:06 IST
1. Track walking people in video.
2. The people might walk in different speeds, they might also stop and stand still.
3. Other moving objects might appear in the video (such as vehicles), which are needed to be avoided.
4. People might occlude each other, so still wanting to at least track upper body parts.
5. People tracked might differ by age, (children should be tracked also).
6. Videos of walking people might be under different light settings. Either indoors artificial light, or be taken outdoors
in morning light, or in the evening (with artificial light).
7. Tracking should be done as fast as possible, as this should be tested on-line, with a web camera.
Input: Video stream from a web camera (online streams).
Outputs: Array of PeopleObj, with the following data:
The bounding boxes and images are obtained from the first appearance of the person.
Array of Trajectories, with the following data:
PersonTrajectory.id (the object unique id).
Example video: http://www.youtube.com/watch?v=yZu2bMO-U2M