Optimization of YOLOv8 algorithm for traffic sign object detection based on MBCA and C2f AKConv module

This study improves the YOLOv8 model by introducing MBCA module and C2f AKConv module, significantly enhancing the performance of traffic sign detection. On a test set containing 638 images and 770 instances, the optimized model demonstrated excellent comprehensive detection capability: the accuracy (P) reached 88.8%, the recall (R) was 85.2%, and the mAP50 was as high as 90.7%, indicating its excellent recognition accuracy for the vast majority of traffic signs. The model's detection performance for standard traffic signs (such as "Stop" signs with a recall rate of 100% and mAP50 of 99.5%) and various speed limit signs (with mAP50 generally exceeding 85%, and high category recall rates such as "Speed Limit 100/120" exceeding 97%) is particularly outstanding, verifying its strong learning ability for identifying regular shapes and clear patterns. Comparative experiments further confirmed the advantages of the model: compared to the original YOLOv8 model, the accuracy of our model improved by 2.6%, and mAP50-95 improved by 3.7%; Compared with YOLOv5, the accuracy improvement is higher, reaching 5.6%. Therefore, this improved model provides an effective solution for precise and robust recognition of traffic signs, which has a positive significance for enhancing the environmental perception ability of intelligent driving systems.