This work mainly combines yolov5 framework and deepsort algorithm to realize the identification and tracking of vehicles and pedestrians on the road. First, the whole road is divided into up lane and down lane, and two collision lines of different colors are set in the middle of the road. The blue collision line is used to classify and count vehicles in the up lane, and the yellow collision line is used to classify and count vehicles in the down lane. According to the different label categories of vehicles and the collision between ID and collision line, the vehicle categories and quantity of different lanes are counted. Secondly, the things on the road are divided into five categories: cars, bicycles, motorcycles, buses and pedestrians, which can be trained as data sets and can be identified and classified in real time. Thirdly, this algorithm can track and count the targets, and calculate the vehicles and pedestrians flowing into the picture, so as to judge the traffic flow density at this time, and thus plan the passable area of vehicles. Finally, we also realize the calculation of centroid, return the centroid coordinates of the target to locate the target position in real time, which lays a good foundation for the realization of autonomous navigation. This project is based on the data set of traffic, and can be applied to the positioning of the robot, target tracking and autonomous navigation, so as to realize the robot in special circumstances timely and accurate positioning of the target object, planning the travel route, timely response to achieve the goal.