Abstract: This paper proposes a tracking algorithm based on Kalman filtering, image perceptual hashing, and template matching to solve the problem that traditional template matching tracking algorithm has high computational complexity and is prone to tracking drift in the case of target deformation, illumination variation, and occlusion. This proposed algorithm uses the hash algorithm to extract the target hash value as a template to match, and the Kalman filtering is used to predict the search region of candidate objects in the next frame, then, a new matching rule is designed to match the templates hash value and target position change information. Finally, according to the matching results, the adaptive template updating strategy and the tracking strategy are used to track the target.The result shows that this proposed algorithm has some robustness in this case of target deformation, illumination variation, and occlusion, and can meet the requirements of real-time tracking.

Key words: target tracking, template matching, Kalman filtering, image perceptual hashing