Vehicle converse driving detection algorithm based on centroid tracking and Yolov5

In order to solve the technical problems of missing detection, high misjudgment rate and great external influence in vehicle converse driving detection, the invention provides a vehicle converse driving detection algorithm based on centroid tracking and Yolov5. On the basis of Yolov5, a Shuffle Attention attention mechanism module is introduced, space and channel attention mechanisms are used at the same time, the detection performance is improved, the vector angle between expectation regression is considered, an SIoU loss function is adopted, and the missed detection rate and the misjudgment rate of vehicles are effectively reduced. And detecting the converse running of the vehicle in combination with a centroid tracking algorithm. According to the invention, the influence caused by large external influence and high omission ratio can be effectively solved in vehicle converse driving detection. 本发明所解决的技术问题是车辆逆行检测中漏检、误判率高以及受外界影响大的问题，提出一种基于质心跟踪和Yolov5的车辆逆行检测算法。本发明在Yolov5的基础上，引入了Shuffle Attention注意力机制模块，同时使用空间和