Airport Obstacle Extraction by Aerial Photograh Stereo Matching

Terrain and obstacle detection is very important for the safety of aircraft operation. The International Civil Aviation Organization (ICAO) issued the requirements for electronic Terrain and Obstacle Data (eTOD) in 2004. The requirements request all countries to finish the survey of eTOD around large airports by 2010. Nevertheless, no traditional surveying approaches satisfy the required precision and resolution for the obstacle collection in any cost effective manner. In this paper, RealScape/Airport, a novel airport obstacle extraction system, is introduced. The RealScape/Airport system implements airport obstacle collection based on aerial photograph analysis. The system first creates the Digital Surface Model of the area covering an airport and its surrounding region from the aerial photographs of the area by a unique pixel-by-pixel stereo processing method. Then, the system compares the elevation of every pixel with that of the Obstacle Limitation Surface (OLS) on this location. The OLSs are 1.2% inclined surfaces starting from the endpoint of the runway all around until 10 km away, and from that point stretching outside horizontally as plane surfaces. An object that is higher than the OLSs is defined as an obstacle in the eTOD regulation. Based on this regulation, the system extracts the pixels over the OLSs as potential obstacles. Finally, the potential obstacle pixels are inspected visually under stereoscopic view to avoid missing projecting objects like lightning rods. From the final result, we find that the RealScape/Airport system achieves 50 cm vertical and 70 cm horizontal accuracy.