Research On Urban Road Traffic Sign Detection Method Based On Deep Learning

In recent years,with the rapid development of deep learning in the field of intelligent transportation systems,it has brought more effective solutions to road traffic problems.Traffic sign detection is an important part of intelligent transportation system,which can provide early warning for drivers in automatic driving and assisted driving,thereby reducing the occurrence of traffic accidents.In the actual driving environment,the complexity and uncertainty of natural scenes determine that the real-time and accuracy of traffic sign detection is the top priority of research.In order to solve the problem that the existing target detection methods have low realtime performance and poor detection effect on traffic signs,it is difficult to deal with various complex scenes.This paper takes the target detection algorithm SSD in deep learning as the basic framework,and takes the traffic signs in urban road traffic scenarios as the research object,and improves the basic network,auxiliary network and receptive field of SSD.The main contents of this paper are as follows:(1)In order to meet the real-time requirements of traffic sign detection in urban road traffic scenarios,an improved SSD-based traffic sign detection algorithm(MV3-SSD)is proposed,which replaces the basis of SSD with the Mobile Net V3 network based on depthwise separable convolution.The network,and the inverse residual structure is used to improve the auxiliary network,which greatly reduces the amount of parameters of the network;in order to better adapt the network to the detection task of traffic signs,the size,aspect ratio and number of the default box are redesigned,and the obtained Adapting the default frame of the traffic sign detection task improves the detection efficiency of the algorithm for traffic signs.Through data supplementation and enhancement,a dataset ICCTSDB with rich traffic scenes is obtained,and training and testing are carried out on this dataset.The experimental results show that compared with SSD and Movbile Net-SSD,MV3-SSD achieves detection speed and accuracy Efficient balance between.(2)In order to improve the feature extraction ability of traffic sign detection in urban road traffic scenes,a traffic sign detection algorithm based on improved multi-receptive field(IR-MV3-SSD)is proposed.RFB-s two network structures are embedded in the MV3-SSD network,which reduces the number of parameters of the multi-receptive field module,enriches the semantic information of the shallow feature layer of the network,and enhances the feature extraction ability of the network;in the model In the training phase,the cosine decay warm-up strategy is used to make the learning rate decrease more smoothly and evenly,and achieve better training effect.The experimental results show that the m AP of IR-MV3-SSD on the ICCTSDB test set reaches 93.16%,and the detection speed is 57 frames per second.Compared with SSD,Yolov3 and Faster-RCNN,it has higher detection accuracy and still has high real-time performance.Finally,the Raspberry pi 4B is used as a small embedded hardware platform to complete the model deployment.The experimental results prove the effectiveness of the method proposed in this paper.