哈尔滨商业大学学报（自然科学版）2025，Vol.41Issue(5)：533-544,12.
基于改进YOLOv8的交通标志检测算法
Traffic sign recognition algorithm based on improved YOLOv8
摘要
Abstract
Aiming at the current problems of low recognition accuracy,missed detection and inaccurate detection of small objects in traffic sign detection,a high-precision model based on YOLOv8s was proposed.The model adopted RepGhost Bottleneck instead of the Bottleneck module in the traditional C2f to form the C2fRepGhost structure and reduced the model computation.The backbone network incorporated the CBAM attention module,which improved the integrated representation of semantic and location information of features to enhance the capture of traffic sign features.For the small target recognition problem,a special small target detection layer was added to improve the detection accuracy of small targets.In addition,the convergence rate and regression accuracy were optimized by replacing the standard loss function with the MPDIoU loss function.Experiments on the TT100K dataset showed that the optimized network model achieved 4.8%,9.7%and 6.1%improvements in accuracy,recall,and mAP metrics,respectively,while the computational effort decreased by 0.5 GFLOPS,which confirmed a significant improvement in detection performance with almost no increase in network complexity.关键词
交通标志检测/RepGhost/C2fRepGhost/CBAM/小目标检测层/MPDIoUKey words
traffic sign inspection/RepGhost/C2fRepGhost/CBAM/small target detection layer/MPDIoU分类
信息技术与安全科学引用本文复制引用
王坤相,王正刚,王世康..基于改进YOLOv8的交通标志检测算法[J].哈尔滨商业大学学报（自然科学版）,2025,41(5):533-544,12.基金项目
国家自然科学基金(U22A2079) （U22A2079）
安徽省高校科学研究重点项目(2022AH050977) （2022AH050977）
安徽高校自然科学研究重大项目(J2021ZD0116) （J2021ZD0116）
