Research On Traffic Sign Detection Algorithm Based On Improved YOLOv5

Traffic sign detection technology is one of the key technologies for driverless driving.The traffic sign detection algorithm with excellent performance can not only alleviate road traffic congestion,but also reduce the incidence of traffic accidents and ensure the safety of drivers to the greatest extent.However,in the real road scene,due to the interference of various factors,such as the target is blocked,the detection target is too small,the light is too strong,etc.,the real-time performance and accuracy of the traffic sign detection algorithm are greatly affected.In order to further improve the performance of traffic sign detection,this paper conducts the following research on traffic sign detection technology based on the YOLOv5 algorithm:In order to improve the detection accuracy of traffic signs,this paper improves the YOLOv5 s model,uses the Swin Transformer model to focus on the characteristics of global information,embeds Swin Transformer blocks into the backbone network of YOLOv5 s,and replaces the path aggregation network of the Neck layer with feature extraction capabilities the stronger Bi FPN structure finally uses the SIo U loss function with angle features as its bounding box loss function to obtain the S-YOLO network model proposed in this paper.On the TT100 k data set,this paper compares and verifies the S-YOLO and the original YOLOv5 s model.The experiment shows that the detection accuracy of the S-YOLO model is4.4% higher than that of the YOLOv5 s model,while the detection speed is only slightly slower than that of YOLOv5 s,fully indicating that S-YOLO has stronger comprehensive performance in traffic sign detection than YOLOv5 s.In order to reduce the calculation amount of the model and improve the speed of traffic sign detection,the GS-YOLO lightweight model is designed and a traffic sign detection platform based on Py Qt5 is implemented.Through the theoretical analysis of depth separable convolution,the GSConv and Vo VGSCSP modules are introduced,and the backbone network and Neck layer network structure are redesigned.In the design of the loss function,SIo U is still used as the bounding box loss function.Finally,before the network output The first layer uses Shuffle Attention to retain the key information of the feature map,so as to balance the detection accuracy and detection speed of the model.Through experiments,the average detection time of the GS-YOLO model is calculated to be 0.012 s,and the detection speed is14.9% faster than that of the YOLOv5 s model.In order to visually display the detection effect of the GS-YOLO model,a traffic sign detection platform based on Py Qt5 was designed and deployed in the Windows environment to verify the validity and real-time performance of the GS-YOLO model.