Smart Car Visual Perception Research

With the advance of wisdom City, intelligent transportation is an indispensablelink, the development of intelligent vehicles also affects the intelligent traffic, theyhave close relations. Along with the development of the city more and more quickly,city traffic safety problem is becoming more and more serious. In order to betterbuilding wisdom City, visual perception study technology of intelligent vehicleimminent.In this field there there is a continuous flow of target detection and recognitiontechnology, due to the traffic lights and traffic signs of the diversity and complexity ofroad environment, real-time and other reasons intelligent vehicle requirements, theexisting algorithms cannot achieve the target detection and recognition is very good.This paper presents identification algorithm for circular traffic lights, the algorithmcan satisfy the real-time requirement of intelligent vehicle, strong robustness; at thesame time puts forward the circular traffic signs detection algorithm, the algorithmcan detect more stable, real-time. The main research work is as follows:(1) Circular traffic light detection: the final choice of image processing in theRGB color space in order to meet the real-time requirements of intelligent vehiclebetter, for image segmentation using color threshold; then use thetraffic lights aroundthe existence of black rectangular box of this characteristic,the specific location ofrectangular box position; finally, through the above twosteps to determine whetherthere are traffic lights in the image, if there is traffic light position determination.(2) Circular traffic lights recognition: create traffic lights in the identificationprocess of traffic lights in the template, the sequential improved by templatematchingsimilarity detection algorithm; also presents a traffic light contrastrecognitionalgorithm, using color thresholding the image with single thresholdimage contrastrecognition, traffic lamp identified.(3) Detection of circular traffic signs: Circular traffic signs detection algorithm isproposed in this paper, the detection of traffic signs in RGB color space, is used tosegment the image by color threshold; the circular Hough transformdetectionalgorithm for shape detection to determine whether there is a circular traffic signs inthe image, if there exists a positioning circular traffic signs location; puts forward another kind of circularity detection algorithm, the detection and localization of thecircular traffic lights use this feature round degree.The experiment results show that, the proposed target detection and recognitionalgorithm with real-time and strong stability.