Research On Traffic Sign Detection Algorithm In Complicated Environment

In recent years,more and more researchers pay attention to the Intelligent Transportation System(ITS).Traffic sign detection is an important part of intelligent transportation system.It is the premise of ITS,which has very important theoretical significance and practical application.Because of the high complexity of the real scene,the traffic sign detection is not yet mature.Therefore,it is necessary to continue to study a kind of detection scheme which is suitable for complicated environment.In this paper,we study the traffic sign detection in complicated environment.The purpose of this paper is to put forward a kind of detection scheme which can be applied to various situations.The main work of this paper is as follows:Firstly,comparative analysis of several kinds of traffic signs segmentation method based on color space,an adaptive threshold segmentation method based on RGB color space is proposed to overcome the limitation of fixed threshold color segmentation method.The proposed method can effectively overcome the defects of the fixed threshold segmentation method,which is easy to be affected by the factors such as illumination,fading,shadow and so on.The effectiveness of the RGBAT method is verified by comparing the experimental results with several methods based on color space.Secondly,according to the application of traffic sign detection,the gradient histogram feature and quantization histogram features based on HSV color space are simplified,which can reduce the feature dimension while ensuring the classification ability of the feature.Thirdly,this paper presents a new universal detection scheme for traffic sign images in complicated environment.Traffic signs are divided into three categories: high brightness,normal brightness and low brightness.This paper design the appropriate detection method for each category.Traffic signs in the candidate region is divided into three categories: high brightness,normal brightness and low brightness.The corresponding SVM classifier is trained for different classes.The SVM classifier is High_SVM,Medium_SVM and Low_SVM.The classification processing of complicated environment can improve the adaptability of detection methods to some extent.