In an autonomous space docking system using a variable geometry truss (VG-T), a docking port is controlled to a direction of a target satellite approaching with various postures and velocities. Consequently, the motion of the target has to be identified, where the target is to be measured at the position apart from the docking system. This paper is concerned with three dimensional measurement of the position and posture of the target as well as the object recognition by using a stereo vision. A hierarchical shape description graph is introduced to perform the stereo matching and model matching efficiently. The normalized fitting function is introduced to reduce an error of the model matching. If values of the fitting function for all of the database model are lower than a threshold value, the object is treated as an obstacle. The efficiency of the proposed system is demonstrated by an experiment.
Read full abstract