Research On Real-time Traffic Sign Detection In Complex Road Environment

Road traffic signs,as facilities for assisting traffic management and ensuring the safety of road participants,have important guidance and warning functions for motor vehicle drivers and pedestrians.The purpose of this paper is to extract and detect traffic signs under complex natural background conditions,so as to prepare for identifying traffic signs more effectively.Due to the influence of the camera shake,the speed of the motor vehicle,and the complex external natural environments,it is still a long way to get high accuracy while ensure the real-time detection.Based on the above background,this paper mainly studies the traffic signs from traffic sign image preprocessing,traffic sign area segmentation,traffic sign feature extraction and positioning,etc.The main work is as follows:(1)In the traffic sign preprocessing stage,based on the common color space,a color probability model for processing traffic sign color information is adopted to enhance traffic sign specific colors(such as red,blue,and yellow)and suppress traffic sign image background color.Then,an offline look up table(LUT)is calculated to give all possible probabilities,thereby reducing the algorithm's search space and detection time.Finally,the traffic signs are processed using histogram equalization and compared before and after the results.(2)The color probability model is used to convert the input color image into a traffic sign probability map,and the input color image is subjected to a histogram equalization process to obtain a gray image.Then the two treated traffic sign images are used by Maximally Stable Extremal Regions segmentation algorithm to get the rough extraction results of the traffic sign area,and then use the MSER improved algorithm to finely screen the traffic sign area,remove the interference area,mainly combined with the gray threshold of the sign and set the aspect ratio of the extraction area.Determine the final candidate traffic sign area.(3)This paper computes the Histogram of Oriented Gradient(HOG)features in probability maps to make full use of traffic signs' color and shape information.At the same time,another HOG feature is calculated on the basis of histogram equalization,and these two gradient direction histogram features are combined to form a color gradient histogram feature.For a case where there is more than one detection frame for a sign,the process of non-maximal suppression of traffic sign and the processing result are given.Finally,using the algorithm proposed in this paper to get the experimental results under the CTSD test set,and corresponding analysis is given.At the same time,the algorithm of this paper is compared with the other two algorithms,which shows that the algorithm has high accuracy and basically meets the requirement of real-time.