Estimation of Rotational Axis and Attitude Variation of Satellite by Integrated Image Processing

Abstract

As a result of the increased number of missions in space, the number of satellites that have completed their missions or have broken down has increased, leaving a great deal of space debris in orbit. Most space debris is found in GEO or low-altitude polar orbits and more than 9,600 pieces of debris having a diameter of over 10 cm are currently in orbit. The number of pieces of debris may increase further due to break up, which increases the chance of debris colliding with other spacecraft. To solve this problem, the development of space robots to capture and eliminate space debris from orbit has been explored extensively, and areas such as attitude estimation, formation flying or rendezvous(Kojima , 2005), manipulator control(Inaba & Oda , 2000), de-orbiting of space debris using electro-dynamic tether systems(Forward et al. , 2000; Ishige et al. , 2004) have been investigated. If the target satellite is incorporative, that is, for an example, if radar communication between the target satellite and the debris eliminator satellite is not possible, then image processing will be required in order to monitor the attitude of the satellite and to capture it by means of a robot manipulator. Image processing algorithms with lower computational costs are desired, because the computer resources installed on satellites are usually fewer than those on the ground. A great number of image processing algorithms have been developed for various purposes, such as edge extraction(Harris & Stephens , 1988; Kitchen & Rosenfeld , 1982) and silhouette extraction(Tomasi & Kaneda , 1991). In the EST-VII mission(Inaba & Oda , 2000), the target markers were installed on the daughter satellite so that the mother satellite can easily monitor the attitude of the daughter satellite. However, normally it is not easy to recognize the attitude of satellites in orbit, because commercial satellites that are not equipped with target markers are usually covered by multi-layer insulator (MLI) with numerous wrinkles that randomly reflect the Sun’s light, and such random reflection makes silhouette extraction more difficult. Furthermore, satellites often overlap the Earth, and the direction of the Sun’s light relative to the satellite varies with time. To estimate the attitude of a satellite, the iterative closest point (ICP) algorithm(Besl & McKay , 1992) has been studied by JAXA(Terui et al. , 2002), but this algorithm needs a computational cost. In order to avoid the computational cost of ICP, the grid closest point (GCP) algorithm (Yamany et al. , 1998) was developed using a hash-table technique. The GCP algorithm has 22