Front vehicle line pressing behavior detection and dynamic tracking method based on vehicle-mounted video

Abstract

The invention discloses a front vehicle line pressing behavior detection and dynamic tracking method based on a vehicle-mounted video, and the method comprises the steps: carrying out the target detection of a front vehicle through employing a yolo-series deep neural network recognition algorithm, enabling a vehicle tracking frame to be correlated with a detection frame in a current frame, and fusing the vehicle tracking frame with the appearance information of a target; replacing the vehicle tracking frame with a detection frame; recognizing lane line pixel points, removing noisy points, anddividing lane lines into solid lines and dotted lines; fitting the identified lane line pixel points into a lane line linear equation; selecting two position points on the lower frame of the detectionframe, judging the relationship with the lane line linear equation, and judging whether the vehicle is in a line pressing state or not; setting judgment conditions for the start time and the end timeof rolling the lane line, and recording the vehicle id, the position bbox of the detection frame and the start time information; and comparing the starting moment information and the ending moment information of the same vehicle, combining and outputting. According to the invention, the line pressing event can be dynamically detected, and the detection precision is improved.

Description

technical field [0001] The invention relates to a vehicle-mounted video-based line-pressing behavior detection and dynamic tracking method of a front vehicle, which belongs to the technical field of computer vision detection. Background technique [0002] Violations of vehicles while driving are an important reason for frequent traffic accidents. Violations of the traffic line can easily lead to collisions with vehicles in adjacent lanes, which is more dangerous. However, the monitoring area of ​​the electronic police fixed-view camera is limited and concealed. Poor performance, which is convenient for drivers to avoid violation detection. Therefore, it is increasingly important to dynamically track and detect violations, expand traffic law enforcement areas, and improve the efficiency, mobility and accuracy of traffic administrative law enforcement inspections and violation punishments. [0003] In the existing vehicle line-pressing behavior detection, some of them use tra...

AI Technical Summary

Problems solved by technology

Although existing methods can detect vehicles and lane lines through semantic segmentation and mature convolutional neural networks, and even determine the state of vehicle line-crossing in a single picture, they are rarely able to combine the two to dynamically detect the line-breaking events of vehicles ahead. detection

Images