A Novel Method for LiDAR Camera Calibration by Plane Fitting

With the development of sensor technology, many different kinds of sensors are now utilized in various application fields. Sensors can make machines smarter and multiple-sensor systems can acquire more information to work more stably. When multiple sensors are integrated into one system, calibration, either in time domain or space domain, is very important to merge data from different sensors. In this paper, we propose a novel method to calibrate a LiDAR and a camera using 3d-3d corresponding features using a cube with ArUco Markers. In the LiDAR frame, point data on the three surfaces are selected to fit the plane's equation independently. In this way, the vertex's coordinate in 3d space and the normal vector of each plane can be obtained. In the camera frame, the corresponding point's coordinates and normal vectors of each plane can be obtained by the camera's full 6d pose estimated using ArUco Markers. In this way, we get a set of point cloud per sensor using the data above. Then a rigid body transformation can be computed by Kabsch algorithm[1]. Experiments show that our method can obtain more stable calibration results than the existing method (Ankit's method) without loss of precision.