Vehicle and pedestrian tracking detection method

Abstract

The invention provides a vehicle and pedestrian tracking detection method, which is used on edge equipment with relatively low computing power, tracks a target in real time, and improves the accuracy of vehicle tracking, matching and detection. Comprising the following steps: (a) acquiring an image by using a wide-angle lens camera to obtain an original image of a road surface; (b) training a multi-category target detection model based on deep learning; (c) deploying the target detection model on a terminal development board, inputting a real-time image acquired by the wide-angle lens camera during application, and automatically detecting and identifying vehicles and pedestrians in the image by the terminal development board according to the deployed target detection model; and (d) obtaining the positions of the vehicle and the pedestrian according to the marked marking frames of the vehicle and the pedestrian, and tracking the vehicle and the pedestrian according to the positions.

Description

technical field [0001] The present invention relates to the technical field of vehicle pedestrian tracking, and more particularly, to a vehicle pedestrian tracking detection method. Background technique [0002] With the continuous reform of the expressway toll system, especially after the abolition of the provincial boundary, vehicles entering and leaving the expressway take the vehicle model as the main basis for judging the toll standard. Driven by two layers of policies and technologies, vehicle identification and detection equipment based on radar point cloud, microwave radar and image processing technology emerge in an endless stream. In addition to the relevant requirements, it is also particularly important to obtain evidence of the state of the vehicle when entering and exiting the high-speed. At present, the evidence collection of vehicle information mainly includes capturing images of the front, body, and rear of the vehicle. At the same time, it is especially im...

AI Technical Summary

Problems solved by technology

It can effectively make up for the target tracking in the case of insufficient video field of view, but in the stitching process, on the one hand, it is affected by the material and color of the detected target itself, and on the other hand, it requires high computing power of the algorithm and processor and the target image after tracking High distortion rate, relatively poor tracking effect

2. Radar ranging assists in overlapping areas to realize vehicle tracking. This technology can effectively compensate for adjacent target interference and splicing dislocation problems. However, the whole set of technologies involves two fields of image processing and radar ranging, and there are image distortion and computing power in splicing at the same time. Demanding higher phenomena

Images