Entropy Minimization Based Multi Object Tracking

摘要

Tracking in image sequences finds its application in many fields including that of surveillance. Object tracking in such sequences is however challenging due various reasons like presence of noise, similarity of the objects, resolution of the images etc. The paper evaluates entropy minimization based technique for tracking on various data sequences to show that it gives accurate results even in the presence of illumination changes, occlusion and drastic motion of the camera that is common in image sequences with moving camera. Multi object tracking has been implemented using this approach and tested for efficiency on various benchmark sequences.