On the real-time modelling of a robotic scene perception and estimation system

In this paper a real-time machine vision system for robotic scene perception and estimation is proposed. Its goal is to determine the 3D structure of the imaged scene together with the relative position and orientation of the robot with respect to the environment. Mainly, the vision system is composed of two basic elements: a 3D camera-scene depth computation method and a camera egomotion estimator. The image processing operations have been implemented into a sequential machine vision pipeline. A performance evaluation of the proposed approach is given through experimental results within an indoor environment.