Determining unit position is very important for indoor autonomous aerial robots. In prior research, we used externally installed cameras to detect natural unit features to reduce the burden of measuring position, focusing on the elliptical trajectory of the rotating rotors while the unit was in flight as the natural feature of interest. Ellipse detection within images allows calculation of unit position. An outstanding problem is that ellipse detection takes a considerable amount of time, and in some environments it is difficult to distinguish between the rotor and the background. In this study, we investigate methods for addressing these problems, proposing a novel algorithm for fast ellipse detection. Furthermore, we record and visualize change in the rotating rotor pattern over time to enable detection against previously problematic backgrounds. For verification, we fly a general-purpose unit in a variety of environments and measure unit position. The results show that the proposed method reduces processing times for ellipse detection and that position can be measured without depending on the composition of flight environment, and thus that unit position detection is improved through the use of infrastructure cameras.
ASJC Scopus subject areas