With the development of automobile technology, intelligent vehicle and automatic driving technology will make due contributions to reducing traffic accidents. This paper aims to improve the dynamic identification and tracking technology in the current intelligent vehicle and automatic driving. First, it is improved based on the MobileNet V2 backbone network, and then a new tracking model framework is designed combining with the SiamRPN single target tracker. Secondly, it integrates space-time tracking clues to improve the stability and robustness of the algorithm. Finally, it constructs a pedestrian dynamic identification algorithm based on the dynamic pedestrian factors in the driving process. Through the training of data sets and video tracking experiments, the performance of the algorithm in this paper is proved quantitatively and qualitatively.