Prohibition Traffic Sign Recognition Research In Vehicle Auxiliary System

With the rapid development of China’s social and economic, urban transport has an unprecedented rapid growth, especially in the past10years, ownership of various types of vehicles is growing. Cities in our country, especially in big cities such as Guangzhou and Shenzhen, the contradictions between the existing road traffic facilities and the increasingly prominent is increasingly prominent. The traffic congestion and frequent traffic accidents has constituted a great threat to people’s life and property safety, and the nation’s economic development. Therefore, how to establish a smart and stable vehicle auxiliary system to improve the safety of the vehicle, reduce road traffic accidents, has very important social significance and research significance.Road traffic sign recognition system is an important part of the vehicle auxiliary systems. The purpose of this paper is to study the prohibition traffic sign detection and recognition under the urban traffic environment. After survey of domestic and overseas developments, and according to the characteristics of the ban traffic signs, we have designed and completed a good real-time, high recognition rate and robust ban traffic sign detection and recognition system. The concrete research contents are as follows:(1) Color segmentation of prohibition traffic sign. First, in order to weaken or eliminate the effect of light on the subsequent image segmentation of color images of natural scenes, we do the light uniformity of the color image in the YCbCr color space, and convert the processed image to the RGB color space. Then segment the color image after preprocessing directly with the improved segmentation algorithm in the RGB color space based on the color characteristics of the ban traffic signs, and compare this segmentation results with the segmentation results in other color spaces.(2) Detection of the prohibition traffic signs based on shape characteristic. First use the combination of some morphological image processing methods such as dilation, erosion, opening and closing for the binary image obtained by coarse color segmentation. Then detect the shape of the area of interest with the improved round-degree method and improved parametric method of triangle, mark and cut out the circular areas and inverted equilateral triangle areas these meet the requirements for subsequent identification. (3) The application of deep learning model MPCNN in prohibition traffic signs recognition. The deep learning theories and several deep learning models that commonly used are introduced in detail. Make full use of the advantage of deep learning method in image recognition that deep learning network can automatically learn smoke like characteristics from the training sets without any manual work for extracting features, in this paper we use the MPCNN, an improved model of Convolutional Neural Network(CNN), for prohibition traffic signs recognition. We give out the details of the network structure and the training method of MPCNN model, and test the performance of the network by using the sample data.(4) A simple and easy system for ban traffic signs detection and recognition is designed and coded, the system interface is given out and introduced in detail, and the comprehensive performance is evaluated.