Traffic Sign Recognition Based On Pulse Coupled Neural Network And Barcode

Researching of the intelligent transportation systems(ITS) is imminent with the increasing traffic pressure caused by gradually accelerating process of urbanization.Traffic sign recognition(TSR) is an important branch of ITS,which can not only be used in the automatic driving system to assist the driver,but also can be used in the unmanned.The research of this paper proposed an effective traffic sign method based on pulse coupled neural network(PCNN) and it has an important theoretical research value and practical significance.This TSR method based on the one of the automatic identification technology—bar code. Experiments were carried out based on traffic signs images of GB5768-1999.Every image was generated to a bar code thus there was a bar code library to be used.An unknown sign can be recognized by matching its bar code and the bar code library.In this paper, the specific research are as follows:(1)Generation and application of one-dimensional barcode.A time sequence can be generated by a PCNN model as the feature sequence,and it can be converted into a gray barcode by gray coding.Then a one-dimensional barcode will transform from the gray barcode by another PCNN model. Furthermore,the input of PCNN can only be gray image and when the processing image is color image, three channels of image in RGB color space were processed respectively.(2)Generating and application of two-dimensional barcode combined with position information of firing pixel.The mirror symmetry traffic signs of different significance are indistinguishable because of the rotating\mirror invariance of PCNN model. Since the misclassifacation caused by the one-dimensional barcode,then position information of firing pixel is induced to generate two-dimensional binary barcode.(3)Exploring the performance of traffic sign recognition based on barcode technology. Non-laboratory environment is simulated by adding noise to traffic signs and experiments are carried out in different conditions.Statistical of matching rate can help analysis in which conditions the algorithm has the best performance.