Schlüsselwort: runway detection

The following thesis presents a board-autonomous, optical positioning system for aircraft navigation during landing. The optical positioning system processes images captured by an optical camera for detecting the runway and calculating the position relative to the runway during the phases: final approach, landing and roll-out on the runway. This system can help increase the automation level in cockpit, lower the workload for the pilot and contribute to safety improvement in aviation. The whole tool chain of image-processing and position calculation is presented in this work. It is implemented within a C++ - based software environment. Additionally open-source libraries like OpenCV, Qt among other libraries are used for image-processing, data-management and user interaction. This work focuses on individually developed algorithms for feature filtering (runway detection) and position calculation, which are not covered by the open source libraries, because of their unique, customized functionalities. Hence, the whole tool chain from source image processing to final position calculation will be covered in this work. The data processing starts with time-synchronized color images, which are merged with the aircraft state vector. The images are then reduced to grayscale images and noise is minimized. During the process of image segmentation the content of the images is separated from the background and is described by edges, areas and other features. Subsequently, the image features can be filtered in order to identify those which represent the runway. The runway features identified by computer vision algorithms can be used for position calculation. In this case, point and line correspondences between the detected runway and the actual runway on ground must be established. This is done by linear equations which will be derived in this thesis. They require additional input parameters like width and orientation of the runway on ground and Euler angles of the aircraft provided by the inertial navigation system on board. The optical positioning system is evaluated by means of flight-experiment data gathered with the research aircraft Do 128-6 of the Institute of Flight Guidance. Several landing approaches have been performed under varying visual conditions, at different approach angles and on different runways. All data were recorded and can be re-processed in the laboratory. A reference solution, which is derived from a satellite-based navigation solution with an accuracy of 2cm, is used during the evaluation. The accuracy and availability of the optical system will be investigated in this thesis. Furthermore, the challenges of operational deployment and image processing, such as outlier detection and exclusion, will be discussed. It will be shown, that during the landing, up to 200ft above runway an optical position calculation is continuously possible without the need of an external feed forward.