Experimental Evaluation of a Machine Vision Based Pose Estimation System for Autonomous Capture of Satellites with Interface Rings

Abstract

This paper presents an experimental evaluation and description of a machine visionbased system for estimating the pose (position and orientation) of a replicated satellite featuring an interface ring of 1.22m diameter. Images of the satellite are obtained using a camera with a wide angle lens (125 deg) in order to have the entire ring in view. The sequences of images are then processed by a modified ellipse extractor. Next, an ellipse detector selects the extracted ellipse that best matches the ring on the image. The normal vector to the ring plane and the three coordinates of the ring center with respect to the camera are found analytically using the geometric parameters of the detected ellipse. The test setup features two 6DOF robotic arms. One arm holds a satellite mockup while the second arm is equipped with the camera system. Pose estimation using a laser tracker with active markers was used as the source of ground truth. Experiments have shown that the system is robust to camera/target relative position and motions, light conditions, and to some degree of occlusions. Visual servoing for tracking the mockup and approaching to grapple position was performed successfully.