AUTOMATIC LIDAR CALIBRATION BASED ON CROSS VALIDATION FOR AUTONOMOUS DRIVING

摘要

In one embodiment, a set of LIDAR images representing LIDAR point cloud data captured by a LIDAR device of an ADV at different points in time is received. For each of the LIDAR images, a perception method is utilized to determine a location of an obstacle captured in the LIDAR image in a local coordinate system. The LIDAR image is transformed using a coordinate converter (e.g., a LIDAR to GPS coordinate conversion logic or function) from the local coordinate system to a global coordinate system. The coordinate converter is optimized based on the transformed LIDAR images by adjusting one or more parameters of the coordinate converter and the above operations are iteratively performed to obtain a set of optimal parameters. The optimized coordinate converter can then be utilized to process subsequent LIDAR images during autonomous driving at real-time.