Research On Shape Analysis And Recognition Of Trafifc Signs

Now the main transport of the cities at home and abroad is the ground traffic. With therapid development of economic, the traffic congestion phenomenon and the growing numberof traffic accidents caused wide public concern in the international community. Intelligenttransportation system(ITS) becomes a hot research. The key of ITS is traffic sign recognition,which is not only timely provides instructions and warnings but also controls traffic in realtime. These play a major role in solving the traffic congestion and preventing traffic accidents.The key research content of traffic sign recognition are feature selection, feature extractionand identification.This paper analyses the shape of traffic signs firstly, and then proposes two methodsabout contour and area of the traffic sign on the basis of original algorithm. The shape contourdescription method based on chain code and fast Fourier transform uses the fast Fouriertransform to reduce the computation time and improve the efficiency of the contour extractionalgorithm. The area feature extraction algorithm combines Principal Component Analysis(PCA) with the invariant wavelet moment make the high-dimensional feature space to thelow-dimensional feature space, which describes the regional information about traffic signwith little features as accurately as possible. At last, the experimental results prove that thetwo algorithms have good resistance to RSS invariance and robust to noise, also theapplicability is very well.Because of the Support Vector Machine (SVM) theory shows advantages and highapplicability in traffic sign recognition, this paper chooses SVM theory as the recognitionalgorithm. We solve problems of the nonlinear, multi-classification and classification decisionbased on the original algorithm. Further more, the genetic algorithm is applied to the decisiontree generation to optimize the algorithm and generate the optimal decision tree at last. Thetraining data of characteristics are used to get the classification, then identify the treatmenttraffic sign experimentally. The experimental results show that the identification is correctentirely and the efficiency of the algorithm is very high. This design of traffic sign recognitionsystem has high efficiency, accuracy and robustness.