Validation of advanced driver assistance systems by airborne optical imagery

Abstract

Combinations of advanced driver assistance systems (ADAS) enabling semi-autonomous driving are now being brought to market. Recently developed ADAS are amongst others the adaptive break control, the lane keeping assist and the lane change support. The complexity in the development of these systems increases, as fully autonomous vehicles should become a reality within the next decade. The use of airborne optical sensors can be an independent tool to validate ADAS operating at test sites or in real traffic. The German Aerospace Center (DLR) has developed an optical camera system which is installed on a helicopter for a variety of traffic monitoring applications. With this system, it is possible to track a single test vehicle and to monitor the vehicle with the surrounding traffic situation. The camera system can acquire high resolution images with 54Mpix and a frame rate below 12fps or 4k (Ultra-HD) videos with 25fps. In combination with a high-end GNSS/IMU system, the images and videos can be timely registered within milliseconds precision and georeferenced with an absolute accuracy better than one meter, i.e. the instantaneous position, timestamp and speed of the test and surrounding vehicles can be derived very accurate. Additionally, the positions and speed of front and rear traffic can be determined automatically, which are necessary for the validation of ADAS. In combination with automatic image analysis tools, the road markings can be extracted, which allows the relative positioning of vehicles with respect to the road markings and the surrounding traffic within few centimetres. In this contribution, the potential of airborne optical imagery for the validation of ADAS is presented and demonstrated on example imagery and videos taken from motorways in Germany.