Intelligent stereo vision system for robotic applications

Abstract

This paper describes an intelligent vision method, which is capable of reconstructing the robot environment. Algorithms and their implementation are presented for localising known objects in the scene and determine 3D Euclidean transformation (the relative position and orientation) between recognised objects. This can be done by reconstructing the projective 3D world of the scene and involve the metrical constraints with an object recognition method. The resulting displacement information can be used as the input of the intelligent robot control system and the calibrated virtual reality.