A Traffic Sign Recognition Algorithm Based On Deep Learning

With the continuous improvement of people’s living standards,the number of cars on the road is also increasing.Intelligent transportation system is favored by many scholars.The detection and recognition of traffic signs is the key link of intelligent transportation system.In the specific detection and recognition process,due to the increasingly complex road environment,it is easy to be interfered by many factors,such as light,weather,etc.,the traditional methods can not meet the current increasingly complex road traffic situation.Based on this,this paper designs a traffic sign recognition algorithm based on deep learning,in order to further improve the robustness and precision of traffic sign detection Accuracy.This paper first summarizes the classification of traffic signs and the commonly used data sets.According to a series of problems existing in the current traffic environment,such as poor robustness,cumbersome calculation and low detection efficiency,a traffic sign detection algorithm based on fast Faster R-CNN is proposed.By preprocessing the original data of the image,the effect of the training model is further improved,and the Faster R-CNN algorithm is optimized and improved.ZF is selected as the pre training model.Through the test experiments in different environments and the result analysis,the effectiveness of the proposed method is proved.In order to further optimize the traffic sign data set,the fusion of image clustering algorithm and Faster R-CNN model is proposed,which not only improves the overall quality of training samples to a certain extent,but also further improves the detection and recognition effect of the final model.The experimental results verify the actual performance of the algorithm,which has the advantages of faster response speed and higher detection and recognition accuracy Advantages.According to the demand of traffic sign detection and recognition technology,the traffic road detection and recognition system is designed,and its software and hardware,system environment and functions are designed.Through the establishment of new data sets for the detection system and recognition system,the system performance is tested.The most popular test results show that the system has the advantages of easy operation and simple process.The final test results also verify that the system designed in this paper has better robustness and detection accuracy than the traditional system,in order to provide some reference for the further development of traffic sign recognition.