Traffic Signs Recognition Method Based On Mixed Forecasting Model

With the rapid development of Advanced Driver Assistance Systems(ADAS)and unmanned technology,traffic sign recognition has become an important research direction.In order to assist the drivers,achieve safe driving and reduce the possibility of traffic accidents,this paper combines Convolutional Neural Network(CNN)and Ada Boost-SVM classification algorithm to construct a Mixed Forecasting Model(MFM).In MFM,the CNN is used as a trained feature extractor and Ada Boost-SVM is used as a recognizer,which can effectively identify traffic signs.The traffic sign recognition method based on MFM includes image preprocessing,CNN feature extraction and Ada Boost-SVM classification.Firstly,a simple and effective Region Of Interesting(ROI)enhancement preprocessing method is proposed according to the characteristics of the German Traffic Sign Recognition Benchmark(GTSRB).The method includes image cropping,graying,image enhancement,traffic sign shape correction and size normalization.Through experiment contrast found after ROI stronger single image time-consuming preprocessing of image recognition and processing is better than that of other preprocessing methods.Secondly,this paper proposes to combine the CNN and Ada Boost-SVM algorithm to extract the features of the pre-processed image and predict the classification.Two sets of hidden layer structures are constructed by using the convolution layer and the downsampled layer in the CNN network.The image is taken as the input of CNN network,and the CNN is trained by Back Propagation(BP)until it converges or stabilizes.Finally,the high dimensional features of the test set are extracted and classified by Ada Boost-SVM classifier.Finally,the simulation results show that compared with PCA-SIF,K-mean and traditional CNN algorithm,the proposed algorithm not only retains the advantages of traditional CNN feature extraction,but also improves the performance of the classifier.Experiments show that the MFM has a high recognition rate and robustness for traffic sign recognition,and the recognition rate and convergence time are superior to other traditional algorithms,which improves the safety of assisted driving and driverless driving.