Research And Implementation Of Road Traffic Sign Recognition Based On Driving Video

The detection and recognition of road traffic signs is an important content in the research of intelligent transportation system.Accurate and efficient recognition of traffic signs can provide guarantee for normal auxiliary driving system,and also provide necessary conditions for the realization of driverless vehicle system in the future,which has great theoretical value and application prospect.The subject of this paper is to detect and recognize road traffic signs based on the method of deep learning.The ultimate goal is to detect and recognize traffic signs in complex traffic scenes.Driving video is a good record of the real traffic scene in the driving process.Therefore,this paper uses driving video as the research material to detect and recognize the traffic signs in the driving video sequence image.The research contents are as follows:(1)An algorithm of traffic sign detection and recognition in driving video sequence image is proposed.Based on the analysis of the structure of yolo-v3 neural network,the algorithm improves part of the structure,making the network more suitable for the detection and recognition of traffic signs in driving video sequence image.The algorithm trains the improved yolo-v3 network on the traffic scene image dataset,and then uses the trained network as the detection network to detect the traffic video image frame by frame,which realizes the automatic detection and recognition of traffic signs in the traffic video.(2)A traffic sign detection and recognition algorithm based on traffic sign feature fusion of multi frame video image is proposed.The algorithm uses the yolo-v3 network to detect the first frame of traffic signs,uses the local feature matching method to search other subsequent frames of traffic signs,uses the feature extraction layer of vgg19 network to extract features and fuse them,uses the full connection layer of vgg19 neural network and the output layer of softmax for feature recognition,and the experiment shows that the algorithm can improve the recognition accuracy of traffic signs.(3)A small-scale traffic sign recognition algorithm based on super-resolution image reconstruction is proposed.The algorithm solves the problem of low recognition accuracy of smallscale traffic sign image(traffic sign image is less than 30 × 30 pixels).The algorithm first detects the first frame of traffic sign image.When the small-scale traffic sign image is identified,the fsrcnn super-resolution image reconstruction network is used to enlarge the small-scale traffic sign image,and then the traffic sign recognition network is used for recognition,which improves the recognition accuracy of the small-scale traffic sign.