Qualitative Understanding of Scene Dynamics for Mobile Robots

In this paper, we present a new approach to analyzing motion sequences as they are observed from a mobile robot operating in a dynamic environment. In particular, we address the problems of (a) estimating the robot's egomotion, (b) recon structing the 3D scene structure, and (c) evaluating the mo tion of individual objects from a sequence of monocular images. Our approach consists of a two-stage process starting from given sets of displacement vectors between distinct image features in successive frames. First, the robot's egomo tion is computed in terms of rotations and the direction of translation. To cope with the problems of noise, we have extended the concept of the Focus of Expansion (FOE) by computing a Fuzzy FOE, which defines an image region rather than a single point. In the second stage, a 3D scene model is constructed by analyzing the movements and posi tions of image features relative to each other and relative to the Fuzzy FOE. Using a mainly qualitative strategy of rea soning and modeling, multiple scene interpretations are pursued simultaneously. This second stage allows the deter mination of moving objects in the scene. Results of this ap proach applied to a real image sequence are presented.