Zebra Crossing Detection And Recognition Algorithm

Road zebra crossing is an important traffic sign.Based on image recognition system seeing-eye auxiliary equipment, the research is to be able to efficiently identify road zebra crossing algorithm, guiding the blind people cross the street along the zebra crossing,has important application value and social value.In this paper, based on image processing seeing-eye auxiliary equipment question about road zebra crossing detection and recognition algorithm. For the algorithm research,mainly from the zebra crossing texture characteristics,namely the zebra crossing interval characteristics such as the width.So using the feature of the zebra crossing detection and recognition algorithm to design.And Studying on the algorithm of this paper is mainly composed of three parts:Firstly : zebra crossing edge detection. There are a lot of classical edge detection operator in image processing,such as Prewitt and Sobel operator,each one has its own advantages and disadvantages, but the operator via analysis indicates that the method is not adjust to be used for edge detection of a zebra crossing,and considering the timeliness,the image edge detection of early not to deal with the noise,so based on the above two points,using the fast algorithm is established and the anti-jamming to deal with the noise of the template operator,namely,Prewitt operator based on edge detection to complete a zebra crossing.The second:zebra crossing the extraction of characteristic.Due to the zebra crossing with strong texture characteristics that the location of the zebra crossing is black and white colour alternative with larger gradient values, therefore to exert maximum gradient projection transformation of the zebra crossing and extract the transformation peak point on the curve as the characteristic value series.In order to improve the recognition rate, peak point of the need to establish a curve sifting principles.This paper choses to calculate the frequency value method,effectively guarantee the validity of the selected peak point.The third:Identify zebra crossing.Based on the size of the peak point value and direction characteristics to establish a discriminant function of minimum error rate. This algorithm combines the idea of pattern recognition, stablishing consistency measure value R using characteristic Angle,and by a large number of training sample images to determine the R value range.If calculated by the target image, within the scope of the value of R, the final judgment zebra crossing is contained in the figure,or does not exist.The experimental results show that the algorithm has higher reliability andeffectiveness. According to use the algorithm, a large number of testing and validation to be done, and analysis the result data that zebra crossing detection accuracy can reach more than 95%,and testing time within 90 ms.so it is Better satisfied the real-time and accuracy requirements guide equipment.