PROJECT TITLE :

Patch-based Scale Calculation for Real-time Visual Tracking

ABSTRACT:

Strong scale calculation is a challenging drawback in visual tracking. Most existing trackers fail to handle large scale variations in complex videos. To address this issue, we tend to propose a sturdy and economical scale calculation methodology in tracking-by-detection framework, that divides the target into four patches and computes the dimensions factor by finding the most response position of every patch via color attributes kernelized correlation filter. In specific, we have a tendency to use the weighting coefficients to remove the abnormal matching points and transform the specified training output of the standard classifier to solve the location ambiguity drawback. Experiments are performed on several difficult color sequences with scale variations within the recent benchmark evaluation. And also the results show that our technique outperforms state-of-the-art tracking strategies whereas operating in real-time.