Traffic Sign Ranging For Autonomous Driving

With the rapid development of the global economy,people's living standards have also improved,and the base of private vehicles has grown rapidly.The growth of the base of vehicles has not only made traffic more and more congested,but also threatened people's lives and property.High-tech autonomous vehicles came into being.At the same time,as the performance of computers and embedded systems improves,large data such as images can be processed quickly,providing a good hardware environment for the development of autonomous vehicles.Therefore,it was decided to use the principle of binocular vision ranging in computer vision to study the real-time ranging problem of traffic signs in autonomous driving.After in-depth research on the principle of binocular vision ranging,through the introduction of correction maps and other methods in binocular vision ranging,to meet the real-time requirements of traffic sign ranging in automatic driving applications.The traditional binocular vision ranging principle will perform stereo correction on the binocular image,and then perform stereo matching,calculate the distance based on the obtained parallax,and all processes are online calculation processes.In order to improve the detection speed of the algorithm,two main improvements have been made:(1)Abandon the step of stereo correction,directly perform template matching on the distorted image,crop the template according to the position coordinates of the traffic sign in the left view,and view in the right In the template matching,the position coordinates of the right side traffic sign are obtained;(2)It is proposed to construct a correction mapping table,through which the corrected coordinates corresponding to the distortion coordinate of the traffic sign can be directly read,so as to obtain the parallax and calculate the distance.In this algorithm,template matching is an online calculation process,and correction map calculation is an offline process.At the same time,a faster pyramid template matching algorithm is used.Both methods effectively improve the ranging speed.Finally,in order to improve the robustness of the algorithm,corresponding image enhancement methods are proposed for the problems of overexposure and underexposure in the vehicle-mounted environment.The algorithm was tested in the laboratory environment and the vehicle environment.In the laboratory environment: within the range of 20m-60 m,the maximum relative error of the algorithm is 1.23%,and the maximum absolute error is less than 0.6 meters.Each group of images The ranging time is within 0.007ms;in the vehicle environment: the maximum relative error of the algorithm is 1.51%,the maximum absolute error is less than 0.64 meters,and the ranging time of each group of images is within 20 ms.Experiments show that the algorithm in this paper can meet the requirements of automatic driving on traffic Identify the requirements for real-time ranging.