Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

You can use Optical Flow to track. You will have the motion vectors for each features. You could average the different motions to get the average displacement of the bounding box.

You can get an exmple of Optical Flow here.

Some ideas. You could have a list of face objects. Haar cascades would return the list of detected faces in the current image. The tracking part would be used to predict the location of each face object and merged with the newly detected faces. If one face object doesn't receive any new detection in a certain time, you could drop this face object (false detection).