Research On Detection And Recognition Algorithm Of Road Traffic Sign In Natural

Traffic sign detection and recognition system is an important area of intelligent transportation system.With the popularization of vehicles,the traffic safety problem is becoming more and more prominent,and the research and development of intelligent transportation system is hot.The traffic sign detection and recognition system can improve the traffic safety problem and greatly reduce the traffic accident rate.This paper mainly studies the traffic mark detection and recognition algorithm,uses GTSRB(German Traffic Sign Recognition Benchmark)database to verify,and applies the algorithm to some local speed limit detection and recognition.In this paper,we use histogram equalization and Gaussian filtering to preprocess the original image,enhance the image and eliminate the Gaussian noise.Based on understanding the advantages and disadvantages of many color models,we propose a method based on HSV(Hue,Saturation and Value).The adaptive threshold segmentation algorithm is use to realize the preliminary detection of traffic signs.Compared with the adaptive threshold segmentation and ITS adaptive threshold segmentation of OTSU,which saves nearly half the time of processing time and the binary image obtained after processing.Clear,contains less noise.Then the binary image is morphologically processes to strengthen the edge detail and improve the image quality.Using a small area based on the contour elimination method,remove the image of non-target small area.Then the method based on edge information and contour moments is use to subdivide the target area.For the identification of traffic signs,the target area is divide into three categories:circular sign,rectangle and triangular mark,based on the method of contour moment and circularity.Then all the candidate images are normalize and the image is normalize,and all the processed images are divide into training samples and test samples.The LBP(Local)of the training model is extract.(SVM)classifier,the SVM classifier,and the SVM classifier,respectively,and then SVM reclassification based on the LBP feature of the test set candidate image.By comparing the results of the two-stage identification with the original SVM to identify the GTSRB database,the improved recognition algorithm significantly reduces the classification time and the accuracy rate is 98.5%.Finally,the traffic mark detection and identification system is applied to some local speed limit signs to detect and identify,the accuracy rate of 90.1%,the experiment proved that the proposed traffic sign detection and recognition algorithm has a certain applicability.