Research On Intelligent Road Sign Recognition System Based On Improved Convolutional Neural Network

As an important component of traffic information,traffic signs,including lane speed limit signs,lane direction signs,lane warning signs and other traffic information traffic signs,have become an important part of driverless technology.The public pay attention to the recognition of traffic signs.Traffic signs can help drivers reduce the pressure in the driving process,and reducing the probability of traffic accidents is also the core element of the auxiliary driving system.With the progress and development of science and technology,higher accuracy and efficiency requirements are proposed for the identification of traffic road signs.In this paper,the problem of road sign recognition is studied based on deep learning,and an automatic recognition system for traffic signs is designed and implemented.Aiming at the problem of accuracy of existing image recognition,this paper proposes a traffic sign image detection method based on Faster R-CNN network.This method uses an efficient Faster R-CNN network to reduce the detection time in the detection process.This method shares the convolution layer computation through the area generation network and the fast convolution network,so that the area generation network can achieve almost no computational cost.It realizes a unified,near real-time depth learning object detection system,and road sign detection based on improved convolutional neural network.The paper collects the data set used in the test,and gives the measurement and evaluation indicators of target detection.The preprocessing of road sign image is analyzed,including the image enhancement algorithm and the analysis of experimental results.The framework of Faster R-CNN signpost detection is given,and the signpost detection model of Faster R-CNN is improved.Aiming at the problem of low recognition rate caused by the fact that the traffic sign image can not fully reflect the original image features when it is artificially designed to extract image features,this paper presents a traffic sign image recognition method based on improved Le Net-5 network.Finally,the experiment proves that the improved method has good performance in recall rate,accuracy rate,F-score and other evaluation indicators.The improved traffic sign image recognition method is applied to the intelligent road sign recognition system.Based on the idea of software engineering,the intelligent road sign recognition system is designed and implemented.First,the system is analyzed from two aspects: functional requirements and non functional requirements.On this basis,the system is designed and implemented.Finally,the functions of the iterative system are optimized by testing the system,Ensure that the system can operate stably after going online.