Reversing judgment method and system based on vehicle key point track and medium

Abstract

The invention discloses a reversing judgment method and system based on a vehicle key point track and a medium, and the method comprises the following steps: detecting continuous video frames, and outputting a target vehicle detection frame; inputting the target vehicle picture into a vehicle key point extraction model, and outputting a coarse feature map, a fine feature map and a vehicle orientation feature vector; judging the orientation of the target vehicle according to the vehicle orientation feature vector, screening out a visual vehicle key point fine feature map, and then obtaining visual vehicle key point image coordinates; and judging whether the vehicle backs up or not according to the movement track of the visible vehicle key point image coordinates in the continuous video frames and corresponding judgment rules. Vehicle reversing is judged by obtaining the key point coarse feature map, the key point fine feature map and the vehicle orientation feature vector, the precisionof the key point movement track of the target vehicle in the continuous video frames shot by the indefinite angle camera is improved, the accuracy of judging whether the vehicle is reversed or not isimproved, and the method can be widely applied to the technical field of deep learning.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a method, system and medium for reversing judgment based on vehicle key point trajectories. Background technique [0002] The reversing judgment based on the vehicle key point trajectory refers to detecting the key points of the target vehicle and its orientation in continuous video frames, so as to obtain the key point trajectory of the target vehicle, and then judge whether the target vehicle is reversing in the continuous video frames. [0003] Because the front and rear of the vehicle are extremely similar, it is difficult to judge whether the target vehicle is reversing from the continuous frames taken by the camera at an indeterminate angle. If you can judge the direction of the target vehicle and its visual key points, then you can use the movement trajectory of the visual key points Determine whether the target vehicle is reversing. This method can be applied to in...

AI Technical Summary

Problems solved by technology

However, these discriminative methods are very sensitive to the camera shooting angle and are not robust to the orientation of the target vehicle. It is difficult to judge whether the target vehicle is reversing based on the continuous frames captured by the camera at an indeterminate angle.

Images