3D dynamic scene analysis. A stereo based approach (Q1320405)

The problem of analyzing sequences of images to extract three-dimensional motion and structure has been at the heart of the research in computer vision for many years. It is very important since its success or failure will determine whether or not vision can be used as a sensory process in reactive systems. The considerable research interest in this field has been motivated at least by the following two points: 1) The redundancy of information contained in time-varying images can overcome several difficulties encountered in interpreting a single image. 2) There are a lot of important applications including automatic vehicle driving, traffic control, aerial surveillance, medical inspection and global model construction. However, there are many new problems which should be solved: how to efficiently process the abundant information contained in time-varying images, how to model the change between images, how to model the uncertainty inherently associated with the imaging system and how to solve inverse problems which are generally ill-posed. There are of course many possibilities for attacking these problems and many more remain to be explored. We discuss a few of them in this book based on work carried out during the last five years in the Computer Vision and Robotics Group at INRIA (Institut National de Recherche en Informatique et en Automatique). In fact, image sequence analysis is a rather vague term and can be used with several meanings. Our definition is that, given one or several sequences of images from one (monocular sequences) or several cameras whose relative positions are known (stereo sequences) acquired while the cameras are moving in an unknown environment containing a number of mobile rigid objects, we must determine the various relative motions (cameras and objects) and the structure of the scene. There has been a tremendous amount of work on the analysis of monocular sequences of images during the past decade. Much less has been done on the analysis of sequences of stereo images. This book describes our work in this domain. As the relative positions of the cameras are known, we can obtain a set of three-dimensional representations of the environment sequentially reconstructed by a stereo system. Therefore, we estimate three- dimensional motion from three-dimensional structure, hence the title of the book. An important feature of the work described in this book is that uncertainty is modeled as early as in the original 2D images and manipulated during all subsequent stages of processing. This distinguishes our approach from most of the approaches reported in the literature, which usually do not deal with this issue. Because of its power and simplicity, we have chosen a probabilistic framework to represent and process this uncertainty.