Research Of Extrinsic Parameter Calibration Of Lidar And Camera Based On Dual Echo

With the rapid development of intelligent fields such as autonomous driving and robotics,sensors of varying functionality are increasingly being applied to these fields to achieve environmental sensing.To overcome the shortcomings of individual sensors in environmental sensing,fusion of data from multiple sensors is often chosen to obtain richer and more robust information.In the fusion of multi-sensor data,the extrinsic parameters calibration is one of the first and most critical steps,and its calibration accuracy directly determines the performance of downstream multi-sensor fusion tasks,so this topic has become a current research hotspot.Lidar and camera are often fused and widely used in autonomous driving,scene reconstruction,etc.because of their strong information complementarity.Lidar can be classified into mechanical lidar and solid-state lidar according to the scanning mode.This thesis will focus on the extrinsic parameters calibration between these two types of lidars and camera.Firstly,this thesis conducts an in-depth study on the dual-echo,analyzes in detail the correlation between the dual-echo and the object edge,and proposes an edge feature extraction algorithm based on the dual-echo.Based on this,the dual-echo is incorporated into the proposed mechanical lidar and camera external reference calibration algorithm based on the edge correlation point cloud and the solid-state lidar and camera extrinsic parameters calibration algorithm based on the internal point cloud to improve the extrinsic parameters calibration accuracy.Second,this thesis proposes an edge-correlation point cloud-based mechanical lidar and camera extrinsic parameters calibration method to address the low resolution of the mechanical lidar point cloud and the edge local expansion problem.The method extracts the edge-associated point cloud of the calibration plate using dual-echo and proposes a nonlinear optimization method to extract vertex features from the edge-associated point cloud that are consistent with the actual calibration plate size.After extracting the vertex features,the point cloud is matched with the vertices in the image,and the optimal extrinsic parameters between the mechanical lidar and the camera is solved by the bundle ajustment method.The experimental results show that the reprojection error of the method is 1.602 px,which is lower than that of similar comparison methods,verifying the effectiveness and accuracy of the method.Finally,the solid-state lidar has problems such as many edge point cloud noises and uneven spatial distribution of point clouds,which can affect the accuracy of point cloud feature extraction and thus reduce the accuracy of extrinsic parameters calibration.Therefore,this thesis proposes a solid-state lidar and camera extrinsic parameters calibration method based on the internal point cloud.The method first removes the edge point cloud containing noise from the original calibration plate point cloud using dual-echo,and defines the remaining calibration plate point cloud as the inner point cloud.Then,based on the inner point cloud characteristics,corner point features are extracted by the optimization method.Finally,the constraint equation is established by the correspondence between the point cloud and the corner points in the image,and the extrinsic parameters are solved optimally.The experimental results show that the dual-echo can effectively reduce the impact of edge noise on the calibration accuracy of the extrinsic parameters.Compared with similar methods,the reprojection error of this method is as low as 1.432 px,which is better than the existing methods,and the data fusion results also visually verify the accuracy of the extrinsic parameters calibration results of this method.