Research on the Helmet Wearing Detection Algorithm of Improved YOLOV5
In view of the current problems of low target detection accuracy,poor detection speed and low efficiency in the hel-met detection of electric bicycles,this article proposes an improved electric bicycle helmet detection algorithm based on YOLOV5.Firstly,aiming at the different sizes and styles of helmets,the K-means++algorithm is introduced to select the initial anchor boxes for helmets,which increases the network convergence speed and solves the problem of slow model training caused by improper initial anchor point selection.At the same time,the spatial channel attention mechanism(CBAM)is introduced to take into account both channel and spatial dimensions,improving network feature learning ability.In the Neck part,the bidirectional feature pyramid net-work structure(BiFPN)is used instead of the original feature extraction structure.Finally,the modified GIoU loss function is used as the loss function to improve model detection accuracy.Experiments show that compared with the original YOLOV5 model,the improved YOLOV5 algorithm model has increased precision by3.7％,recall by5.9％,mAPby3.1％,and meets the requirement of helmet detection accuracy,indirectly reducing the traffic accident rate to some extent.
万方数据
维普
