Research On Traffic Sign Recognition Algorithms Based On MSER And Optimized SVM Using Genetic Algorithm

With the continuous development of the socio-economic level,the traffic problem has become increasingly serious,and the issue of traffic safety is particularly important.To this end,"smart transportation" came into being."Assisted driving" or "automated driving" in "smart traffic" is an important part,and among them,automatic detection and identification of roadside traffic signs is a key technology.For this reason,this paper will study an algorithm that can effectively identify traffic signs under natural conditions.The traffic sign recognition algorithm proposed in this paper firstly pre-processes the image to be detected,reduces the interference of light and noise on the detection and recognition stage through the operations of image equalization,sharpening,and normalization,and improves the recognition accuracy.Using the MERS maximum stable extremal region algorithm to extract and segment the region of interest from the pre-processed image,the feature region is extracted from the block HOG gradient direction histogram,and the training set in the GTSDB traffic sign image database is used to support the SVM.Vector machine training learning,access to traffic signs SVM classifier.In order to further optimize the recognition performance,the types of kernel functions that affect the performance of SVM and the corresponding penalty parameters C and kernel function parameters g are analyzed in depth.An improved genetic algorithm optimization parameter search algorithm based on adaptive crossover mutation is proposed.The approximate optimal penalty parameter C and kernel function parameter g are obtained.The traffic sign recognition algorithm of genetically optimized SVM classifier is simulated by MATLAB,and a good recognition rate is obtained.