Multi-target tracking algorithm by corner feature for video stream

In the field of computer vision research, moving target tracking of sequence images is not well solved, and is an old, important and challenging task. Fast and accurate completion of moving target tracking is very an important and difficult problem. In a dynamic environment, moving target tracking is often influenced by various factors, such as weather changes, shadows, occlusion, background confusion. This paper presents a multi-target tracking algorithm by corner feature for video stream, which can solve the problems of video multi-objective motion tracking. The improved Harris operator can extract some even and stable corners, and the matching optimization reduces the amount of false matching points, so they improve the accuracy of target tracking. Experimental results show that the algorithm can complete a stable matching under a variety of complex conditions, and achieve a steady target tracking under a small part of the block state.