Real Time Estimation of 3 D Environmentwith

This paper deals with the problem of 3D structure estimation of a set of geometrical primitives in an active vision context. Our method is based on the structure from controlled motion approach which consists in constraining the camera motion in order to obtain an optimal estimation of the 3D structure of a geometrical primitive. Since this approach involves to focus on the considered primitive, we present a method for connecting up many estimations. It is centered on visual 2D information and rough estimation of the structure of the studied primitives. It allows us to have an optimal estimation of each primitive of a scene composed of cylinders, segments and polygons. Our method has been implemented with an automata network using the real time Signal language. These automata are able to connect up the diierent stages of the reconstruction process: selection, focusing, optimal estimation of the selected primitive and, simultaneously, rough estimation of the other ones. Real time experiments on a robotic cell are presented and experimental results are given. 1 Problem statement Many applications in robotics involve a good knowledge of the robot environment. For such applications, the aim of this paper is to obtain a complete and precise description of a scene using the visual data provided by a camera mounted on the end eeector of a robot arm. The observability of the camera motion which is necessary for the 3D structure estimation characterizes a domain of research called dynamic vision 7]. Furthermore, when the camera motion is controlled using vision data, dynamic vision becomes active vision 1]]13], which generally provides more precise results. Our method is based on the structure from controlled motion approach, which consists in constraining the camera motion in order to obtain a precise and robust estimation of geometrical primitives such as points, straight lines and cylinders 3]]6]. This related work deals with the reconstruction of only one primitive, which is insuucient to consider real scenes containing several objects. In this paper, the reconstruction problem is handled at two levels: { local aspect: Since the method proposed in 6] involves to focus on the considered primitive, it has to be successively performed for each primitive of the scene. This local aspect of the reconstruction is divided into four steps: focusing, recognition, optimal estimation and spatial localization of the primitive ; { global aspect: Developing perception strategies to get the spatial organization of complex scenes is …