Adjustable Tracking Algorithm with Adaptive Template Matching

This paper presents an adjustable algorithm for object tracking in a sequence of images. It is a method designed to be effective in applications where some of the information on the object tracked is known. To establish the object position and size in a frame, adaptive template matching, location prediction through known trajectory, border information in the image, and data acquired in previous frames are used. This algorithm behaves nicely with sequences of images where the tracked object size and features vary considerably. It is also robust to some levels of occlusion, and it can be adapted to simpler applications to compete in speed with a basic correlation algorithm.