Automated and Targetless Multi-LiDAR Calibration using RANSAC-ICP

In recent years, 3D LiDAR has become a crucial sensor in various applications such as autonomous vehicles, robotics, object detection, forest mapping. However, certain LiDAR sensors, such as Velodyne VLP16, exhibit some drawbacks, such as a limited field of view and sparse data density, making them insufficient for certain applications. Consequently, the calibration and fusion of multiple sensors have gained importance in acquiring optimal coverage, reducing blind spots, and enhancing data density. Hence, this study mainly focused on calibrating two Velodyne VLP16 LiDAR to optimise sensing coverage range and measurement density. To perform the multi-LiDAR calibration and fusion, calibration parameters such as rotation and translation values are estimated using automatic registration algorithms as point-to-plane Iterative closest approach (ICP), which eliminates the requirement for manual intervention along with the Random sample consensus (RANSAC) algorithm. The algorithm helped in retrieving calibration parameters and achieved a registration accuracy of up to four centimetres.