An Extrinsic Calibration Method for LiDAR and Camera Based on Feature Matching

Abstract

The onboard sensor information of intelligent driving systems primarily fuses LiDAR and camera data. Accurate and stable extrinsic parameter calibrations offer the basis of effective multi-source information fusion. In order to improve the robustness of the perception system, this paper proposes an extrinsic calibration method for LiDAR and cameras based on feature matching. Firstly, a point cloud data sphere center algorithm and image data ellipse algorithm were proposed to extract point cloud three-dimensional coordinates and pixel two-dimensional coordinates of feature points. Next, the constraint of feature point pairs was established in the LiDAR and camera coordinate systems to construct a nonlinear optimization algorithm. Finally, this nonlinear optimization algorithm was used to optimize the extrinsic parameters of the LiDAR and cameras. Evaluation of the projected LiDAR point cloud onto the image based on the optimized extrinsic parameters yielded transverse and vertical average errors of 3.06 pixels and 1.19 pixels, respectively, in the LiDAR and camera joint calibration. The method proposed in the article reduces the average projection error by 40.8% and the error variance by 56.4% compared to the livox_camera_lidar_calibration method. Its accuracy and robustness are significantly better than the latter.

Keywords

• light detection and ranging(lidar)
• camera
• extrinsic parameter calibration
• feature matching