LiDAR–camera calibration method based on ranging statistical characteristics and improved RANSAC algorithm

Abstract

For sensory data fusion, a calibration method between 3D light detection and ranging (LiDAR) and color camera based on ranging statistical characteristics and improved RANSAC algorithm is proposed. The multi-frame LiDAR point cloud data of the calibration triangular board are recorded. The scanned points with close angles are defined a cluster within same degrees. Furthermore, accurate points are preserved using statistical filtering based on Gaussian distribution. Afterwards, the plane and edge parameters of the triangular board are estimated by the reserved point cloud using improved the random sample consensus (RANSAC) algorithm to obtain the 3D locations of the vertices. Meanwhile, corner points in the image can be extracted manually. Finally, the projection matrix between the camera and the LiDAR is estimated by using the 2D–3D​ correspondences in different positions. The projection errors of different frames and corresponding points are calculated. The results demonstrate that the average error with 300 frames is reduced by 23.05% compared to 1 frame. Moreover, the standard deviation diminishes with the increasing of corresponding points. The reliability and advantage of the method are verified compared with other state-of-art methods. It provides theoretical and technical support for low resolution LiDAR.