Road Traffic Signs Segmentation Method Research In Natural Scenes

As one of the most important parts of Traffic Sign Recognition (TSR), the real-timeand accuracy of Road Traffic Sign Segmentation has great influence on the actual trafficsign recognition. The title of this paper has important value in theory and application foron-board video identification.The research topic of this paper take pulse coupled neural network (PCNN) and RGBmodel as the theoretical basis, studied the color traffic signs segmentation in natural scenes.Based on the similar neurons firing synchronously of Pulse Coupled NeuralNetwork(PCNN), the firing time of each neuron is calculated in RGB color spacerespectively, then the traffic signs region are segmented by combination results.Experiments are carried out based on images of standard traffic signs database and imagesof natural scene.Analyzing the experimental results which using traditional PCNN model for trafficsign image library GB5768-1999, the optimal iterative numbers is acquired by calculatingthe maximum entropy in each color space, determine the PCNN visual system using analogprocessing methods. The salt and pepper noise add to image of traffic sign library, stillusing the method described above to process, analyzed anti-noise performance of PCNN.For the traffic sign in natural scene, the traffic signs region can well be segmented withmodified PCNN model.Pulse Coupled Neural Network(PCNN) is used for traffic signs segmentation, themethod analysis color features of image based on the similar neurons firing synchronouslyof PCNN, We verified the feasibility of the traffic sign segmentation by experiments andprove this method is effective.