| LiDAR and camera,as sensors that directly obtain information about the surrounding environment of the vehicle,are the basis for the interaction between the vehicle and the outside world.In order to overcome the limitations of a single sensor and increase the robustness of the sensing system,the LiDAR is usually fused with sensors such as the camera,and the premise of the fusion is to obtain the internal parameters of the camera and the external parameters between the LiDAR and the camera.How to accurately and quickly calibrate to obtain internal and external parameters has become a core issue that needs to be resolved.This paper has conducted in-depth research on the calibration method of LiDAR and camera data.The research content is as follows:(1)For the camera calibration,improvements are made on the basis of the classic Zhang’s calibration method,a complete distortion model is introduced,the original calibration model is improved,and the camera parameter calibration accuracy is improved.(2)Use the calibrated camera to perform algorithm feasibility experiments,point shift experiments,algorithm comparison experiments,and target measurement accuracy analysis experiments based on monocular cameras.The experimental results calibrated the algorithm in this paper has good practicability and high accuracy.The follow-up checkerboard joint calibration method provided support.(3)The LiDAR and camera joint calibration experiment was carried out using the checkerboard calibration method.After the camera was calibrated,the camera internal and external parameters were used to obtain the pose information in the checkerboard camera coordinate system,and the checkerboard was calculated by the point cloud information of the calibration board obtained by the LiDAR.According to the pose information in the laser coordinate system,the external parameter information between the two is obtained,and then the LM algorithm is used to optimize the optimal parameters.(4)The LiDAR and camera joint calibration experiment was carried out using the feature point calibration method.The special three-plane calibration board designed in this paper was used to fit the spatial equations of each plane through the initial point clouds of different planes,and the coordinates of the feature point radar coordinate system were obtained through the spatial relationship.Extract the feature point Pixel coordinates from the image to establish an equation,and use the least square method to calculate the calibration parameters.(5)This article has carried out a series of experiments to verify the effect of the above algorithm.The experiment proves that the two algorithms have their own advantages.The accuracy of the checkerboard method is higher than that of the feature point method,but the computational complexity is lower than that of the feature point method;the feature point method is more efficient,but its corners The point coordinate accuracy does not reach the sub-pixel level,and the initial calculation data is greatly affected by the measurement error;experiments have shown that both have strong practicability and good calibration accuracy,providing for the fusion of LiDAR and camera data in autonomous driving in different scenarios A favorable reference. |
