Research Of Traffic Sign Detection And Recognition Based On Deep Learning

Traffic sign detection and recognition is an important function in intelligent traffic.Its purpose is to assist drivers to more fully understand the situation in front of the road during driving,but also to enable the realization of driverless driving in the future,to liberate drivers,and thus change our lives.The core of traffic sign detection and recognition is the research on detection and recognition algorithms.However,due to the diversity of traffic signs,the surrounding environment in which it is located is also variegated and complex.Such characteristics,coupled with the contrast of daytime and night light intensity,there are problems such as misdetection,missed detection,etc.At the same time,there is a high requirement for real-time and accuracy in the recognition process,and the existing recognition algorithm is complicated.The network structure still has some room for improvement in the training model and traffic sign recognition process.Aiming at the problem of traffic sign detection brought about by different illumination in real scenes as well as occlusion,insult and deformation.Based on previous studies,this paper proposes a traffic sign detection algorithm based on color,shape features and machine learning.This algorithm was used to verify the GTSDB public traffic data set in Germany.The experimental results show that the detection accuracy rate reaches 92.31%.Aiming at the problems such as the complex network structure of the existing recognition algorithm model,network training and the time-consuming process of sign recognition,this paper proposes a traffic sign recognition algorithm based on MCCNN.On the open-source Caffe framework,this algorithm is used to train the training set in GTSRB,the data set of the German traffic sign recognition standard.The key features in these traffic signs are extracted through the deep neural network and identified by the trained recognition model.Classified traffic sign.After the experiment,the experimental results show that the recoginize accuracy rate reaches 93.024%,The results will be compared with those of the predecessors.Finally,a prototype system for the detection and identification of traffic signs was designed,and the existing data sets were augmented with data.The additional traffic sign experimental data was collected from both the school and the surrounding streets.The proposed detection and recognition algorithm is tested on the prototype to obtain experimental results and analyze.