改进的YOLOv3交通标志识别算法

Improved YOLOv3 Traffic Sign Recognition Algorithm

针对复杂场景下交通标志检测存在精度低、检测速度慢等问题,提出一种基于YOLOv3改进的S-YOLO(stronger-YOLO)交通标志算法。首先,合并批归一化层到卷积层,以提升模型前向推理速度;其次,采用二分K-means聚类算法,确定适合交通标志的先验框;然后引入空间金字塔池化模块,提取特征图深度特征;最后引入完整-交并比(complete-IoU,CIoU)回归损失函数,提升模型检测精度。实验结果表明,在重制的中国交通标志数据集(Chinese traffic sign dataset,CTSDB)下,所提算法与YOLOv3相比,平均准确率和检测速度分别提升了4.26%和15.19%,同时相较YOLOv4以及其他算法对交通标志识别有更优的精度和速度,具有良好的鲁棒性,满足复杂场景高效实时检测。

Aiming at the problems of low accuracy and slow detection speed in traffic sign detection in complex scenes, an improved S-YOLO traffic sign algorithm based on YOLOv3 was proposed. Firstly, the batch normalization layer was merged into the convolution layer to improve the forward reasoning speed of the model. Secondly, binary K-means clustering algorithm was used to determine the a priori frame suitable for traffic signs. Then the spatial pyramid pooling module was introduced to extract the depth features of the feature map. Finally, complete-IoU regression loss function was introduced to improve the detection accuracy of the model. The experimental results show that under the reproduced Chinese traffic sign dataset traffic sign dataset, the mAP and FPS of the proposed algorithm are improved by 4.26% and 15.19% respectively compared with YOLOv3. At the same time, compared with YOLOv4 and other algorithms, the proposed algorithm has better accuracy and speed for traffic sign recognition, has good robustness, and meets the efficient real-time detection of complex scenes.