Abstract-This paper introduces an analytical framework for studying some properties of model acquisition and recognition techniques based on indexing. The goal is to demonstrate that several problems previously associated with the approach can be attributed to the low dimensionality of invariants used. These include limited index selectivity, excessive accumulation of votes in the look-up table buckets, and excessive sensitivity to quantization parameters. Theoretical results demonstrate that using high-dimensional, highly descriptive global invariants produces better results in terms of accuracy, false positive suppression, and computation time. A practical example of high-dimensional global invariants is introduced and used to implement a 2-D shape acquisitionhecognition system. The acquisitiodrecognition system is based on a two-step table look-up mechanism. First, local curve descriptors are obtained by correlating image contour information at short range. Then, seven-dimensional global invariants are computed by correlating triplets of local curve descriptors at longer range. This experimental system is meant to illustrate the behavior of a high-dimensional indexing scheme. Indeed, its performance shows good agreement with the analytical model with respect to database size, fault tolerance, and recognition speed. Model acquisition time is linear to cubic in the number of object features. Object recognition time is constant to linear in the number of models in the database and linear to cubic in the number of features in the image. The system has been tested extensively, with more than 250 arbitrary shapes in the database. Unsupervised shape and subpart acquisition is demonstrated. I.

... In this paper we present a more rigorous approach in which the conditions under which to accept a match are derived based on fundamental grounds. We obtain an expression that relates the probability of a match occurring at random to the fraction of model features that are accounted for by the match. This expression is a function of the number of model features, the number of image features, and a bound on the degree of sensor noise. One

We survey the main ideas behind recent research in model-based object recognition. The survey covers representations for models and images and the methods used to match them. Perceptual organization, the use of invariants, indexing schemes, and match verification are also reviewed. We conclude that there is still much room for improvement in the scope, robustness, and efficiency of object recognition methods. We identify what we believe are the ways improvements will be achieved. ii Contents 1. Introduction .................................................................................................................................... 1 2. Representation ................................................................................................................................ 3 2.1 What makes a good shape representation? ............................................................................ 3 2.2 The choice of coordinate system ..........................................

The recognition community has long avoided bridging the representational gap between traditional, low-level image features and generic models. Instead, the gap has been artificially eliminated by either bringing the image closer to the models, using simple scenes containing idealized, textureless objects, or by bringing the models closer to the images, using 3-D CAD model templates or 2-D appearance model templates. In this paper, we attempt to bridge the representational gap for the domain of model acquisition. Specifically, we address the problem of automatically acquiring a generic 2-D view-based class model from a set of images, each containing an exemplar object belonging to that class. We introduce a novel graph-theoretical formulation of the problem, and demonstrate the approach on real imagery.

We describe a model based recognition system, called LEWIS, for the identification of planar objects based on a projectively invariant representation of shape. The advantages of this shape description include simple model acquisition (direct from images), no need for camera calibration or object pose computation, and the use of index functions. We describe the feature construction and recognition algorithms in detail and provide an analysis of the combinatorial advantages of using index functions. Index functions are used to select models from a model base and are constructed from projective invariants based on algebraic curves and a canonical projective coordinate frame. Examples are given of object recognition from images of real scenes, with extensive object libraries. Successful recognition is demonstrated despite partial occlusion by unmodelled objects, and realistic lighting conditions. 1 Introduction 1.1 Overview In the context of this paper, recognition is defined as the prob...

A fast simulated annealing algorithm is developed for automatic object recognition. The object recognition problem is addressed as the problem of best describing a match between a hypothesized object and an image. The normalized correlation coeficient is used as a measure of the match. Templates are generated on-line during the search by transforming model images. Simulated annealing reduces the search time by orders of magnitude with respect to an exhaustive search. The algorithm is applied to the problem of how landmarks, e.g., trafic signs, can be recognized by a navigating robot. We illustrate the performance of our algorithm with real-world images of complicated scenes with traffic signs. False positive matches occur only for templates with very small information content. To avoid false positive matches, we propose a method to select model images for robust object recognition by measuring the information content of the model images. The algorithm works well in noisy images for model images with high information content.

The systems and concepts described in this paper document the evolution of the geometric invariance approach to object recognition over the last five years. Invariance overcomes one of the fundamental difficulties in recognising objects from images: that the appearance of an object depends on viewpoint. This problem is entirely avoided if the geometric description is unaffected by the imaging transformation. Such invariant descriptions can be measured from images without any prior knowledge of the position, orientation and calibration of the camera. These invariant measurements can be used to index a library of object models for recognition and provide a principled basis for the other stages of the recognition process such as feature grouping and hypothesis verification. Object models can be acquired directly from images, allowing efficient construction of model libraries without manual intervention. A significant part of the paper is a summary of recent results on the construction of ...

This paper proposes a novel approach to constructing a hierarchical representation of visual input that aims to enable recognition and detection of a large number of object categories. Inspired by the principles of efficient indexing (bottom-up), robust matching (top-down), and ideas of compositionality, our approach learns a hierarchy of spatially flexible compositions, i.e. parts, in an unsupervised, statistics-driven manner. Starting with simple, frequent features, we learn the statistically most significant compositions (parts composed of parts), which consequently define the next layer. Parts are learned sequentially, layer after layer, optimally adjusting to the visual data. Lower layers are learned in a category-independent way to obtain complex, yet sharable visual building blocks, which is a crucial step towards a scalable representation. Higher layers of the hierarchy, on the other hand, are constructed by using specific categories, achieving a category representation with a small number of highly generalizable parts that gained their structural flexibility through composition within the hierarchy. Built in this way, new categories can be efficiently and continuously added to the system by adding a small number of parts only in the higher layers. The approach is demonstrated on a large collection of images and a variety of object categories. Detection results confirm the effectiveness and robustness of the learned parts. 1.

We address the problem of recognition of multiple flat objects in a cluttered environment from an arbitrary view-point (weak perspective). The models are acquired au-tomatically and initially approximated by polygons with multiple line tolerances for robustness. Groups of con-secutive segments (super segments) are then Gray coded and entered into a hash table. This provides the essen-tial mechanism for indexing and fast retrieval. Once the data base of all models is built, the recognition proceeds by segmenting the scene into a polygonal approximation; the Gray code for each super segment retrieves model hy-potheses from the hash table. Hypotheses are clustered if they are mutually consistent, and represent the instance of a model. Finally, the estimate of the transformation is refined. This methodology allows us to recognize models in the presence of noise, occlusion, scale, rotation, trans-lation and weak perspective. Unlike most of the current systems, its complexity grows as O(kN) when N is the number of models, and k << 1.

this paper, an approach is proposed for describing take into account the particular measurements to be made objects for the purposes of recognition. We deal with two from the smoothed data, specifically, the multi-scale mea- issues: building general-purpose multiscale descriptions of surement of curvature. The curvature-tuned smoothing curved objects, and extracting a concise set of attributes that can be used for recognition. Typical examples of the method we have developed and present here allows us to types of curve we are able to describe as both qualitatively obtain measurements which have not been subjected to similar, yet discriminably different, are shown in Fig. 1. an unnatural "flattening" or distortion as a result of the The method is based on the use of approximating curves smoothing [13]. Our technique is based on multiscale with simple curvature properties. In contrast, alternative smoothing with a specialized form of "regularization." We methods based on line segments or curvature extrema 1