Research On Traffic Sign Recognition Based On Deep Learning

With the rapid economic development,the number of cars in China continues to increase,and the risks of traffic accidents and road congestion also increase.Automatic detection and identification of traffic signs can effectively alleviate these problems.For safety reasons,traffic sign detection and recognition systems have extremely high requirements for accuracy and real-time.Therefore,how to ensure that the detection and recognition algorithms of traffic signs have high accuracy and real-time is always the focus and difficulty of research.This article uses deep learning-based methods to study traffic sign detection and recognition to solve the difficulties and challenges faced by traffic signs in natural scenes.The research contents are as follows:1)Aiming at the problem of missed detection of traffic signs that are far away in natural scenes,this paper proposes a new algorithm for traffic sign detection based on deep learning.First,by adjusting the lightweight network structure and designing the feature extraction module of the traffic sign detection algorithm.Secondly,the extracted features are fused using multi-scale feature pyramids,and the size of the labeled frame in the data set is aggregated using the K-means clustering algorithm.Experimental results show that the traffic sign detection algorithm in this paper achieves an accuracy rate of 94.5% and a recall rate of 92.4%,and the image detection speed reaches 40 FPS.2)Aiming at the difficulty of recognizing certain traffic signs in natural scenes,this paper proposes a traffic sign recognition algorithm based on feature fusion.The shallow features in the detection model are merged with the features of the traffic sign image extracted by convolution.And adopt a variety of methods to improve overfitting to improve the robustness of the algorithm.Experimental results show that the recognition accuracy of the traffic sign recognition algorithm in this paper has reached 99.53%.3)Design and implement a multi-scale feature fusion traffic sign detection and recognition system.The test results show that the system is not only easy to operate and easy to use,but also has a high accuracy rate of traffic sign detection and recognition.Aiming at complex natural scenes,this paper proposes a new algorithm for traffic sign detection and recognition.This method has good application prospects in assisting driving,reducing traffic accidents and alleviating road congestion.