Research of Intelligence Classifier for Traffic Sign Recognition

Support vector machine (SVM) is a novel machine learning method based on statistical learning theory, which can avoid over-fitting and provide good generalization performance. In this research, multi-category SVMs (M-SVMs) is applied to traffic sign recognition and is compared with BP algorithm, which has been commonly used in neural network. 116 Chinese ideal signs and 23 Japanese signs are first chosen for training M-SVMs and BP intelligence classifiers. Next, noise signs, level twisted signs from Chinese and Japanese real signs are selected as testing set for the purpose of two networks testing. Experiment results indicate that, in approximated classification for traffic sign, SVM has achieved nearly 100% recognition rate and has certain advantages over BP algorithm. In fine classification, SVM shows its superiority to BP algorithm. Based on the analysis for the results, one may come to a conclusion that SVM algorithm is well worth the research effort and very promising in the area of traffic sign recognition