In this report we consider the problem of 3D object recognition, and the role that perceptual grouping processes must play. In particular, we argue that a single level of perceptual grouping is inadequate, and that reliance on a single level of grouping is responsible for the specific weaknesses of several well-known recognition techniques. Instead, we argue that recognition must utilize a hierarchy of perceptual grouping processes, and describe an appearance-based system that uses four distinct levels of perceptual grouping, the upper two novel, to represent 3-D objects in a form that not only allows recognition, but reasoning about 3D manipulation of a sort that has been supported in the past only by 3D geometric models.

We introduce a novel view-based object representation, called the saliency map graph (SMG), which captures the salient regions of an object view at multiple scales using a wavelet transform. This compact representation is highly invariant to translation, rotation (image and depth), and scaling, and offers the locality of representation required for occluded object recognition. To compare two saliency map graphs, we introduce two graph similarity algorithms. The first computes the topological similarity between two SMG's, providing a coarse-level matching of two graphs. The second computes the geometrical similarity between two SMG's, providing a fine-level matching of two graphs. We test and compare these two algorithms on a large database of model object views. Keywords: View-Based Object Recognition, Shape Representation and Recovery, Graph Matching. 1 Introduction The view-based approach to 3-D object recognition represents an object as a collection of 2-D views, sometimes called...

We describe and analyze an appearance-based 3-D object recognition system that avoids some of the problems of previous appearance-based schemes. We describe various large-scale performance tests and report good performance for full-sphere/hemisphere recognition of up to 24 complex, curved objects, robustness against clutter and occlusion, and some intriguing generic recognition behavior. We also establish a protocol that permits performance in the presence of quantifiable amounts of clutter and occlusion to be predicted on the basis of simple score statistics derived from clean test images and pure clutter images. Key Words: Object recognition, Appearance-based representations, Visual learning. 1 Introduction Object recognition has been an important and much-researched problem in the study of both machine and human vision. Until recently, the most successful computational work on object recognition has used model-based approaches in which the image is matched against explicitly repre...

paper analyses the improvements that can be gained in the generalized Hough transform method for recognizing objects through the use of imperfect perceptual grouping techniques. In particular, we consider simple grouping techniques that determine pairs of points that are likely to belong to the same object using a criterion based on connectedness in the image edge map. It is shown that such imperfect grouping techniques can considerably improve both the efficiency and accuracy of object recognition. Experiments are described that demonstrate the

The role of perceptual organization in motion analysis has heretofore been minimal. In this work we present a simple but powerful computational model and associated algorithms based on the use of perceptual organizational principles, such as temporal coherence (or common fate) and spatial proximity, for motion segmentation. The computational model does not use the traditional frame by frame motion analysis; rather it treats an image sequence as a single 3D spatio-temporal volume. It endeavors to find organizations in this volume of data over three levels—signal, primitive, and structural. The signal level is concerned with detecting individual image pixels that are probably part of a moving object. The primitive level groups these individual pixels into planar patches, which we call the temporal envelopes. Compositions of these temporal envelopes describe the spatio-temporal surfaces that result from object motion. At the structural level, we detect these compositions of temporal envelopes by utilizing the structure and organization among them. The algorithms employed to realize the computational model include 3D edge detection, Hough transformation, and graph based methods to group the temporal envelopes

Polyhedral descriptions of objects are needed in applications like vision-based robotics, e.g. to carry out grasping and assembling tasks. This work presents a novel methodology for the subtask of localizing a three-dimensional target object in the image and extracting the two-dimensional depiction of the boundary. By eliciting the general principles underlying the process of image formation we exhaustively make use of general, qualitative assumptions, and thus reduce the role of object-specific knowledge for boundary extraction. Geometric/photometric consensus principles are involved in a Hough transformation based approach for extracting line segments. The perceptual organization of line segments into polygons or arrangements of polygons, which originate from the silhouette or the shape of approximate polyhedral objects, is based on shape regularities and quasi-invariants of projective transformation. An affiliated saliency measure combines evaluations of geometric/photometric consen...

This paper presents an k-partition, graph theoretic approach to perceptual organization. Principal results include a generalization of the bi-partition normalized cut to a k-partition measure, and a derivation of a sub-optimal, polynomial time solution to the NP-hard k-partition problem. The solution is obtained by first relaxing to an eigenvalue problem, followed by a heuristic procedure to enforce feasible solutions. This approach is a departure from the standard k-partitioning graph literature in that the partition measure used is non-quadratic, and is a departure from image segmentation literature in that k-partitioning is used in place of a recursive bi-partition. We apply this approach to image segmentation of infra-red (IR) images, and show representative segmentation results. Initial results show promise for further investigation.

This paper introduces an intermediate grouping phase whereby structures, meant to correspond to generic 3D junctions of planar and curved surfaces, are extracted from imperfect basic constant curva-ture contour primitives. This research work is part of a more generic project for detecting and describing generic high-level structures corresponding to 3D ob-jects and/or parts of 3D objects in a single illuminance image of a cluttered scene. MAGNO (Multi-level Ac-cess to Generic Notable Objects), a computer vision system exploiting generic constraints at each of its hier-archical processing levels is at the heart of the project. 1

Robots are now employed to carry out well-defined tasks in customized static environments, often at a high speed and an astonishing level of precision. However, these robots usually depend totally in their actions on a detailed control scheme developed in advance during an off-line planning phase. Due to recent progress in electronics and computing power, in control and agent technology, and in computer vision and machine learning, the development of autonomous robots capable of solving high-level deliberate tasks in natural environments can now be approached seriously. This article provides essential vision- and learning-related aspects for developing autonomous camera-equipped robot systems. 1

Abstract not found