Correlation tracking algorithm based on adaptive template update
Zhuojin Pan, Xiujuan Wang · 2010 3rd International Congress on Image and Signal Processing · 2010
Aiming at the common problem of target changing and partial occlusion in the process of infrared dynamic tracking, a weighted function based on the distance of each pixel to the template center is put forward. And the traditional mean absolute difference (MAD) algorithm is improved to solve the interference. For template update we introduce an adaptive template update strategy which can adaptively correct template through the analysis of current frame matching coefficient and the gray scale information of the template edge. The experimental results show that our algorithm not only maintains the low computational complexity of MAD algorithm but also has higher stability and accuracy during continuous tracking.