Traffic Signs Recognition Based On Pulse Coupled Neural Network

As one of the most important parts of Intelligent Transportation System (ITS), Traffic SignRecognition (TSR) has great application value for the actual traffic environment. The title of thispaper has important value in theory and application for on-board video identification.The research topic of this paper take pulse coupled neural network (PCNN) as the theoreticalbasis, studied the parameters choosing process during the extraction of image edge information.Based on RGB and HSV color space theory model, we classified the entropy sequences of all trafficsigns in six different color spaces from PCNN through the experiment and observe the effects withthe3-D surface chart form. Introducing the concept of image distance transform, we used thesimplified PCNN model to achieve transform results just like Euclidean distance transform, andalso build a traffic sign recognition system.With the automatic wave diffusion characteristic of PCNN, we use the distance transformimage got from the simplified PCNN module as the classification features, the minimum variance topromote the matching analysis, and choose the optimal PCNN parameters through experiments.According to the image library GB5768-1999of road traffic sign of national standard, theexperiments were conducted to analysis the results. Then we get some conclusion of betterextraction effect of the shape information in two value edge image with Euclidean distancetransformation from the simplified PCNN module than the traditional PCNN entropy sequencefeature vector method.This paper presents a changed PCNN model to extract the feature of the traffic signs, whichmodel extract the distance transform image of the edge image of the original image as a featurevector. We verified the feasibility of the scheme by experiments and prove the feature extractionmethod is effective.