Position Identification Using Image Processing for UAV Flights in Martian Atmosphere

抄録

This paper presents a method for the position identification of an unmanned aerial vehicle (UAV) in the Martian atmosphere in the future. It uses the image processing of craters captured via an onboard camera of the UAV and database images. The method is composed of two processes: individual crater detection using a cascade object detector and position identification using the recognition Taguchi (RT)-method. In crater detection, objects with shapes that resemble craters are detected regardless of their positions, and the positions of multiple detected craters are identified using the criterion variable D^*, which is a normalized Mahalanobis distance. D^* is calculated from several feature variables expressing the area ratios and relative positions of the detected craters in the RT-method.