A target detection method in dynamic scene based on Harris algorithm with sub-block threshold

In this paper, an improved Harris algorithm for the target detection in dynamic scene due to the camera motion is presented. First, a sub-block thresholding method is proposed to solve the problem of uneven distribution of corners detected by Harris algorithm. Then, the improved Harris algorithm is used to extract feature points, which are used to estimate the parameters of global motion with the random-max consistency algorithm and the least-square method. Compensated by the result parameters, the reference image together with the current image are used to detect the target with the frame difference method. Experiment results show that the algorithm can detect the moving target more accurately in dynamic scene.