Research On Localization At Intersection Based On Traffic Lights Map

Recent years,the Intelligent Transportation System(ITS)has been developed greatly and the autonomous vehicle technology has drawn more and more attention.The navigation and localization are the basis as well as the difficult part in the whole autonomous vehicle system.For this difficult problem,this paper studies the intersection localization method based on traffic lights map.Traditional vehicle localization methods use information from GPS,inertial navigation,and odometer,but localization errors accumulate over time.Aiming at the above problems,this paper presents a method of intersection localization method based on traffic lights map.In particular,traffic lights with significant visual characteristics in urban environment are used as positioning markers,and the localization accuracy of autonomous vehicle at intersections is improved by combining the location information of traffic lights provided by high-precision map.Firstly,the whole scheme of positioning system is designed.The sensors used in this system are determined,and the coordinates of sensors are established respectively.The transformation and rotation are calibrated,then realization details of each module in the system are explained respectively.Accurate detection of traffic lights is the basis of the localization method proposed by this paper.However,under complicated driving conditions,the quality of images captured by vehicle camera is often poor due to the bad illumination,motion blur and fog.For the problem of illumination,we pre-processing the image with the color consistency algorithm.In order to solve the problem of motion blur in the image,we propose a motion-blur kernel estimation method based on IMU data,which can remove the motion blur in the image.In order to solve the problem of image blurring in foggy day,an improved method based on dark channel theory is proposed,which can remove the image blur caused by fog.Based on the image quality enhancement,this paper uses the visual selective attention model to realize the fast localization of traffic lights’ candidate regions,and detect the exact position of traffic lights using the support vector machine(SVM)and histogram of gradient(HOG).The method of multi-dimensional histogram feature statistic is proposed to complete the recognition of the signal type,and the current detection and recognition result is corrected by the historical detection and recognition result for the purpose of get stable and reliable result.Finally,the state model and measurement model of the vehicle vision system are established.Combined with the location and height information of traffic lights and traffic signs provided by the high-precision map,the Extended Kalman Filter(EKF)is used to fuse the vision detection results of the traffic lights with the inertial measurement unit information.The experimental results show that the method proposed in this paper improves the lateral localization accuracy and the accuracy of vehicle’s yaw angle.In summary,the intersection localization method based on traffic lights map proposed in this paper improves the localization accuracy of autonomous vehicle at intersections,especially the lateral positioning accuracy and yaw angle accuracy.