Vehicle Tracking Method Based on Multilayer Detection Model and Crowd Behavior Model

Abstract

The invention relates to a vehicle tracking method based on a multilayer detection model and a group behavior model. The vehicle tracking method comprises the steps of filtering a detected image through applying a low-threshold deformable parts model (DPM) and obtaining an object candidate in a detection period; and secondly, screening the candidate by means of image segmentation based on a prior shape, thereby obtaining a final detection result. In a tracking period, modeling is performed on the group behavior according to the detection result based on a distance between objects; and afterwards tracking by means of Kalman filtering, and restraining group behavior in tracking, thereby preventing drift between objects. The vehicle tracking method has an anticipated technical effect that the vehicle can be accurately detected in a complicated crossing environment; detection is not affected by objects such as bicycles and motorcycles; and more than 80% of vehicles can be detected when many vehicles exist at the crossing. In tracking, not only is tracking for each vehicle realized, but also object drifting in meeting of two vehicles is prevented.

Description

technical field [0001] The invention belongs to the technical field of computer vision, image and video processing, and in particular relates to a vehicle tracking method based on a multi-layer detection model and a group behavior model in a traffic monitoring video. Background technique [0002] Video-based vehicle detection and tracking technology is one of the most basic technologies for realizing intelligent transportation systems. Accurate and efficient vehicle detection and tracking algorithms can help traffic departments obtain real-time traffic flow and statistical traffic status. Compared with general detection and tracking, the detection and tracking of traffic intersections has the following characteristics: first, the movement of vehicles is very complicated (going straight, turning, accelerating, emergency stop, etc.); second, in the surveillance video of traffic intersections, vehicles The mutual occlusion between them is particularly serious; third, there are ...

AI Technical Summary

Problems solved by technology

Designed for the characteristics of multiple targets at traffic intersections, complex vehicle movement, and easy confusion, it mainly solves the problems of occlusion between vehicles and easy drifting of targets during tracking.

Images