Introduction - If you have any usage issues, please Google them yourself
The application uses the approach introduced in paper Covariance Tracking using Model Update Based on Means on Riemannian Manifolds , F.Porikli, O.Tuzel, P.Meer.
The tracking is based on:
1) initializing the target region
2) constructing the Feature Vectors for each pixel in the target region (first frame)
3) forming the Covariance Matrix using feature vectors generated in step 1
4) determining the candidate regions for the following frames and constructing the covariance matrices of these regions
5) finding the minimum covariance-distanced region from these candidate region
6) assign this region as the estimated region