Multi-target vehicle track real-time construction method based on traffic monitoring video

Abstract

The invention discloses a multi-target vehicle track real-time construction method based on a traffic monitoring video, and belongs to the technical field of computer vision. The method comprises thefollowing steps: firstly, acquiring and marking traffic data at the angle of a traffic monitoring camera; respectively extracting convolution features and HOG features of the traffic video images by using the trained deep learning model and an HOG feature extraction tool; fusion, based on the fusion features, on one hand, a license plate number detection model based on yo-v3 is trained, on the other hand, real-time accurate tracking of multiple vehicle targets is completed by adopting a correlation filtering method based on space-time regularization, and finally, license plate information anda driving track of each vehicle are recorded in combination with a tracking result and a license plate detection result. According to the method, the cost for additionally installing and maintaining equipment is reduced, and the practicability is better; and the method has higher accuracy, and particularly, the track construction effect under the conditions of shielding and deformation is remarkably improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a real-time construction method of multi-target vehicle trajectories based on traffic monitoring video. Background technique [0002] Traffic video surveillance system has been increasingly used in daily highway operation safety management. In the intelligent research of traffic monitoring system, the construction of vehicle trajectory based on video is a key step for vehicle behavior recognition. The vehicle trajectory refers to the sequence of feature points formed at different consecutive frame times when the vehicle passes through the traffic video surveillance scene. The construction of vehicle trajectories in traffic monitoring scenarios involves multi-target tracking, and the occlusion and trajectory deformation between targets have become the focus and difficulty of accurately constructing vehicle trajectories. The result accuracy of traditional vide...

AI Technical Summary

Problems solved by technology

[0003] GPS-based trajectory tracking can obtain valuable information from GPS location data to help construct vehicle trajectories. subject to great limitations

[0004] Vehicle tracking methods based on the construction of video-based vehicle trajectories include optical flow method based on target detection, frame difference method, background difference method, etc.; these traditional video-based detection methods are mainly based on background modeling to complete traffic flow parameters These methods are susceptible to misjudgment due to interference from environmental conditions, and the accuracy is not high enough

Images