E-YOLO: A traffic sign detection method in complex environments

Aiming at the problems of low accuracy and slow detection speed in traffic sign detection in complex scenes, an improved E-YOLO (Enhanced YOLO) traffic sign algorithm based on YOLOv3 is proposed. Firstly, merge the batch normalization layer into the convolutional layer to improve the forward inference speed of the model; Secondly, the binary K-means clustering algorithm is used to determine the prior boxes suitable for traffic signs; Then, a spatial pyramid pooling module is introduced to extract deep features from the feature map; Finally, the CIoU regression loss function is introduced to improve the accuracy of model detection. The experimental results show that under the reconstructed CTSDB traffic sign dataset, the proposed algorithm has improved the average accuracy and detection speed by 4.26% and 15.19% respectively compared to YOLOv3. At the same time, it has better accuracy and speed in traffic sign recognition compared to other algorithms, and has good robustness, meeting the requirements of efficient real-time detection in complex scenarios.