Traffic sign detection algorithm based on lightweight SSD
[Objective]Real-time and accurate traffic sign detection is critical for autonomous driving and intelligent transportation.Existing intelligent detection algorithms exhibits low detection speeds and cannot be efficiently adapted to embedded-terminal devices for detecting complex real-world road scenes.To address this problem,a traffic sign detection algorithm based on lightweight SSD is proposed.[Methods]The algorithm replaces the VGG16 network with the MobileNetV3_large network to reduce model parameters and improve detection speed.Furthermore,it uses an inverse residual structure B-neck with an additional SE module to replace the corresponding standard convolution,thereby enhancing the semantic information of low-level feature layers.The improved RFB network is designed to enhance the detection capability of small traffic signs,and the size of the preset prior boxes is reset to improve the detection capability of the model for specific datasets.[Results]The experimental results show that the improved SSD algorithm achieves an mAP value of 89.04％on the Chinese traffic sign detection dataset CCTSDB,which is 5.26％higher than that of the MobileNet-SSD algorithm.FPS reaches 60 frames/s,which is 23 frames/s higher than that of the SSD algorithm.[Conclusions]The proposed algorithm exhibits high real-time performance and detection accuracy and is robust in complex traffic environments.
万方数据
维普
