Mutual Absolute Calibration of Lidar Mounting Positions on Vehicles using CAD Models.

Extrinsic calibration of a sensor mounting position on a vehicle is essential for a truthful mapping of the sensor data with respect to the vehicle. However, extrinsic calibration is usually cumbersome due to complex measurement setups or requirements on the environment, e.g., locally planar surfaces. We propose to use mutual detections of sensors mounted on different vehicles for calibrating their mounting positions. To maximize the accuracy of the detected vehicle poses, our developed method uses CAD models of the involved vehicles. Based on the accuracy of each detected vehicle pose, we further derive a formulation to estimate the calibration accuracy. The benefits of our methods are demonstrated using Monte Carlo simulations and real-world experiments with two BMW test vehicles.