In order to enable the testing platform AI For Mobility (AFM) to perceive its environment, it is equipped with multiple cameras. The camera images are forwarded to several neural networks, which have been trained to detect objects, traffic participants as well as road boundaries. However, the cameras need to be calibrated extrinsically to allow a correct localization of all detected objects in the vehicle’s environment.
Therefore, several calibration targets are mounted on and around the vehicle which are observed by the cameras on the AFM. This setup is then captured by an external camera and forwarded to the calibration software which compares the arrangement of the calibration targets with the targets each camera perceives. Based on this comparison a spatial mapping of the setup can be constructed which gives information on where exactly each camera on the vehicle is located.