This work has been performed at the CVLab at the École Polytechnique Fédérale de Lausanne.

Currently, automatic landing systems heavily rely on beacons and systems on the ground to determine the path towards a landing spot. In the myCopter project we are developing vision algorithms that can perform robust and real-time automated landing place assessment in man-made environments. With our approach, landing sites are assessed as featureless, but regularly shaped, areas in the image that typically characterise man-made landing structures. Examples include prepared landing surfaces such as runways and landing pads, as well as unprepared ones, including grass fields, dirt strips and building rooftops. This video shows a novel image segmentation and shape regularity measure that we have developed.