Research On Traffic Identification And Recognition Algorithm In Complex Environment

The increase of vehicles,the congestion of road conditions and the irregular operation of driving will cause frequent traffic accidents.Intelligent traffic system can regulate the driver’s safe driving behavior and reduce the frequency of frequent traffic accidents,so it is very important to study it,and very profound significance.The detection and recognition of traffic signs play a dominant role in the intelligent transportation system,so the research in this paper is of great significance.Because the actual road background is very complex,it is difficult to study the detection and recognition algorithm of traffic signs in complex environment.This paper mainly from the traffic sign detection and recognition of the two aspects of the study.Detection refers to the area containing traffic signs detected from the original image,and identification is to distinguish the detected traffic signs.The main contents of this paper are:(1)Detection of traffic signs in complex environments.An improved threshold segmentation and geometric feature screening method for YCb Cr color space is proposed as the detection algorithm in this paper.The detection accuracy of this method is 89%.The image of traffic sign detection data set in this paper contains too many interference factors,such as weather factors(rain,fog),light factors(shade,backlight)and camera factors(blur,distortion),etc.The general methods need histogram equalization and median filtering,but the proposed method only needs binary processing,compared with other methods,the detection method in this paper is better.The processed image is first segmented by YCb Cr color threshold to obtain the initial target area,but some of them also have background interference area;the road traffic signs are further separated from the background by morphological processing,area screening and geometric feature screening;finally,the location of the target area is obtained and marked on the original image to achieve target detection.The experiment proves that the improved detection method is better than other detection methods based on color threshold segmentation and MSER segmentation.(2)Recognition of traffic signs in complex environments.In this paper,a Gist+PHOG parallel fusion method is proposed to extract the features of traffic signs.The fusion feature,serial fusion feature,Gist feature and PHOG feature are trained and tested by the classifier,respectively.The experimental results show that the parallel fusion features proposed in this paper are more conducive to improving the recognition performance of the classifier.The fusion features were identified by the optimized support vector machine,the limit learning machine and the random forest classification model.the recognition accuracy was 93%,78.19%,87.23%,and the time used for testing and training was 0.081 s,0.416 s,1.23 s,respectively.Therefore,this paper uses the SVM identification model of feature fusion andPSO optimization to identify.