Vehicle and Vehicle Distance Detection Based on Improved YOLOv5

Vehicle detection is one of the important applications of target detection in the field of autonomous driving. In response to the shortcomings of vehicle model recognition and low detection accuracy in existing vehicle target detection methods, an improved vehicle and distance detection method based on YOLOv5. By designing the adjacent feature layer correlation module in the feature pyramid structure, the advantages of adjacent feature layers with different receptive fields are used to provide more accurate spatial position signals, thereby improving the detection accuracy. To solve the problem of non-fixed output feature map size in the SPP module, a spatial pyramid pooling module with average pooling and residual edges is designed to comprehensively improve the overall performance. Finally, the vehicle distance detection module is designed by combining the RealSense D415 depth camera to add distance measurement display to the detection header, achieving vehicle detection and distance measurement simultaneously. Through experimental horizontal comparison with other algorithms, vertical comparison before and after improvement, the results show that the precision, recall rate, and average precision have increased by 5.17%, 7.92% and 4.25% respectively after improvement, fully verifying the feasibility of the algorithm improvement. The corresponding camera distance measurement error experiment is designed, and the distance measurement error results are all within 3%, indicating practical value.