Research on pedestrian target trajectory detection of mobile robot in a complex scene

In the complex road environment, given the decline in accuracy and poor real-time performance of pedestrian detection and recognition under the conditions of illumination intensity and occlusion, this paper proposes a multi-target walking tracking method using YOLOV5 and deep sort, which combines the two modules of walking identification and standard walking tracking under different conditions, and improves the Kalman filter on this basis, to realize the function of walking trajectory prediction. First, the YOLOV5 method is used to mark the pedestrians in the video through the recorded video number, pedestrian number, and other location data, to obtain the correct location of all pedestrians in the video; Then, pedestrian tracking is realized by the deep sort method through information.