<title>Vision-based robotic system for space applications</title>

Abstract

This paper describes a robotic system which accepts motion and control commands which can be generated autonomously. The system developed has been designed to perform an autonomous grapple based on guidance control feedback provided by an image from a single camera mounted on the slave robot's end effector. The vision system consists of three parts. The first signature based, trained on an arbitrary grapple interface (i.e., no special targets are required for guidance), and provides estimates for the 3D attitude by interpolating sampled signature correlations. These signatures are essentially the distribution of line orientations obtained by radial integration of the FFT of a pre-processed edge image. The second part estimates the range and bearing of the interface based on the first and second moments of the preprocessed edge image of the interface. And the third stage of the algorithm verifies the results.