Showing papers on "3D single-object recognition published in 1989"

Part segmentation for object recognition

Abstract: Visual object recognition is a difficult problem that has been solved by biological visual systems. An approach to object recognition is described in which the image is segmented into parts using two simple, biologically-plausible mechanisms: a filtering operation to produce a large set of potential object “parts,” followed by a new type of network that searches among these part hypotheses to produce the simplest, most likely description of the image's part structure.

Distortion invariant object recognition by matching hierarchically labeled graphs

Abstract: A graph-matching process of object recognition is proposed. It is applied to face recognition. Gray-level images are represented by a resolution hierarchy of local Gabor components, which are all scaled and rotated versions of each other. The components centered on one image point form a Gabor jet. A single jet provides a distortion-insensitive local representation of part of an image. Object recognition is achieved by matching image point jets to jets in stored prototype patterns. For a selected image jet the best matches are determined, under a constraint preserving spatial arrangement. The procedure amounts to labeled graph matching, with Gabor jets forming labels to nodes and topology determining links. A contrast-insensitive similarity measure provides for invariance with respect to lighting conditions. The authors have formulated the matching procedure as an optimization task solved by diffusion of match points. This diffusion is controlled by a potential determined by jet similarity and the topology-preserving constraint. The algorithm implements a neural network architecture. >

CAD-based computer vision: from CAD models to relational graphs

Abstract: The authors outline their approach for automatic translation of geometric entities produced by a CAD system into a relational graph structure. They have developed a system which uses 3-D object descriptions created on a commercial CAD system and expressed in the industry-standard IGES form, and performs geometric inferencing to object a relational graph representation of the object which can be stored in a database of models of object recognition. Details of the IGES standard, the geometric inference engine, and some formal properties of 3-D models are discussed. In addition to the process of translation from one data format to another, the interference engine extracts higher-level information from the CAD model and stores it explicitly in the new data structure. The higher-level features will allow the search space explored during the object recognition stage to be pruned early. >

Fundamentals of pattern recognition

Abstract: The fundamental problem of pattern recognition images, fibre images and shapes - the topological framework objects or images structures or patterns, classification and recognition - the structural or syntactic framework general formalization of the pattern recognition problem - the categorical framework

From Surfaces to Objects: Computer Vision and Three Dimensional Scene Analysis

Abstract: An introduction to recognition using surfaces object recognition from surface information surface data as input for recognition making complete surface hypotheses surface clusters description of three-dimensional structures object representation model invocation hypothesis construction hypothesis verification.

The trajectory primal sketch: a multi-scale scheme for representing motion characteristics

Abstract: Conventional approaches to dynamic scene analysis do not use motion itself explicitly for recognition. The authors propose a different approach for the use of motion in a computer vision system which uses the motion characteristics of moving objects without actually recovering the structure. In this approach, the extended trajectories followed by the objects are considered. It is argued that in many cases, where an object has a fixed and predefined motion, the trajectory of several points may serve to uniquely identify the object. In this approach, the trajectories are analyzed at multiple scales to identify important events corresponding to discontinuities in direction, speed, and acceleration using scale space. These important events are recorded in a presentation called trajectory primal sketch. Experimental results are presented graphically, demonstrating the potential value of this approach. >

An object model for image recognition

Abstract: The use of object-oriented database principles to help model an image for computer vision, specifically, for line-image analysis, is described. The resulting representation, called thin line code (TLC), is general across known applications and extensible to new applications. TLC's advantages, and also some difficulties it has in strictly adhering to traditional notions of object orientation, are addressed. A review of relevant aspects of object modeling is included. >

Modeling sensors: toward automatic generation of object recognition program

Abstract: One of the most important and systematic methods of building model-based vision systems is that of generating object recognition programs automatically from given geometric models. Automatic generation of object recognition programs requires several key components to be developed: object models to describe the geometric and photometric properties of the object to be recognized, sensor models to predict object appearances from the object model under a given sensor, strategy generation using the predicted appearances to produce a recognition strategy, and program generation converting the recognition strategy into an executable code. This paper concentrates on sensor modeling and its relationship to strategy generation, because we regard it as the bottleneck to automatic generation of object recognition programs. We consider two aspects of sensor characteristics: sensor detectability and sensor reliability. Sensor detectability specifies what kinds of featuers can be detected and under what conditions the features are detected; sensor reliability is a confidence level for the detected features. We define a configuration space to represent sensor characteristics. Then, we propose a representation method for sensor detectability and reliability in the configuration space. Finally, we investigate how to use the proposed sensor model in automatic generation of object recognition programs.

Using Local Orientational Information As Image Primitive For Robust Object Recognition

Abstract: Orientational information can replace the traditional edges as basic image features ("primitives") for object recognition. A comparison of five different orientation operators on 3 x 3 windows is carried out, and it is found that these operators have similar performance. A first attempt at object recognition searches for minimum root mean square deviation of orientation in a picture. This technique shows better object discrimination than the traditional normalized cross-correlation of grey-level images. Additionally the parameters of the Gaussian distribution of orientational correlation can be accurately predicted by a simple theoretical model. The orientational correlation technique shows difficulties in recognizing geometrically distorted and partially occluded objects. For this reason a very robust algorithm for the recognition of simple objects is developed, based only on orientationl information as image feature, and local polar coordinates for the model of the object. Practical examples taken under difficult, natural conditions illustrate the reliable performance of the proposed algorithm, and it becomes apparent that orientational information is indeed a powerful image primitive.

Generic object recognition: building coarse 3D descriptions from line drawings

Abstract: PARVO, a computer vision system that addresses the problems of fast and generic recognition of unexpected 3D objects from single 2D views, is introduced. Recently, RBC (recognition by components), a new human image understanding theory, has been proposed on the basis of the results of various psychological studies. However, no systematic computational evaluation of its many aspects has been reported yet. The object recognition system the authors have built is a first step toward this goal, since its design respects and makes explicit the main assumptions of the proposed theory. It analyzes single-view 2D line drawings of 3D objects typical of the ones used in human image understanding studies. The main issues related to generic object recognition are discussed, original algorithms and techniques specific to the author's systems are described, and results of the different processing stages of the system are presented. >

Object recognize apparatus

Abstract: An object recognize apparatus suitable for recognition of a complex object constituted with a plurality of elements. When specifying a recognition object, a plurality of elements constituting the finally required recognition object are respectively designated as recognition objects, and when specifying an upper-level recognition object, lower-level recognition objects thereof are designated. Furthermore, for each lower-level recognition object, geometric features necessary for the recognition are specified. In an analysis of image data, the lower-level recognition objects are first detected by use of the geometric features so as to next attain the upper-level recognition object based on results of the detection. As a result, a complex object existing at an arbitrary position can be judged and the position thereof is obtained.

Multisensor Object Recognition From 3D Models

Abstract: Multisensor fusion for object recognition, particularly when multiple sensor platforms and phenomenologies are considered, stresses the state of the art in model-based Image Understanding. The discrimination power of the algorithms depends on accumulation of evidence from diverse sources and on properly applying known model constraints to sensor data. In this paper we describe a model-based multiple-hypothesis Bayesian approach to recognition that has roots in detection and estimation theory. We also describe an approach to object modeling that utilizes an object-based representation that allows multiple geometric representations and multiple, alternative decompositions of the object model. Initial implementations of these ideas have been incorporated into a model-based vision testbed and are currently undergoing testing and evaluation.

Applying uncertainty reasoning to model based object recognition

Abstract: An architecture for reasoning with uncertainty about the identities of objects in a scene is described. The main components of this architecture create and assign credibility to object hypotheses based on feature-match, object, relational, and aspect consistencies. The Dempster-Shafer formalism is used for representing uncertainty, so these credibilities are expressed as belief functions which are combined using Dempster's combination rule to yield the system's aggregate belief in each object hypothesis. One of the principal objections to the use of Dempster's rule is that its worst-case time complexity is exponential in the size of the hypothesis set. The structure of the hypothesis sets developed by this system allow for a polynomial implementation of the combination rule. Experimental results affirm the effectiveness of the method in assessing the credibility of candidate object hypotheses. >

Distributed sensor data fusion with binary decision trees

Abstract: A distributed sensor object recognition scheme that uses object features collected by several sensors is presented. Recognition is performed by a binary decision tree generated from a training set. The scheme does not assume the availability of any probability density functions, thus it is practical for nonparametric object recognition. Simulations have been performed for Gaussian feature objects, and some of the results are presented. >

Determining linear shape change: toward automatic generation of object recognition programs

Abstract: A 3-D object-localization task may be divided into two parts. First, one visible region is classified into one of the aspects of the 3-D object where an aspect is defined as a topologically equivalent class of appearances. Then, the precise attitude and position of the object are determined within one aspect. The authors generate a program to determine the precise attitude and position of an object within one aspect, provided that the face correspondences are given as the result of aspect classification. They establish rules (to define each free coordinate system at each aspect) and correspondences between model edges and image edges, and iteratively solve the transformation equation to determine the object's attitude and position using these correspondences. To turn the strategy into a runnable program, an object library and a geometric compiler are prepared. >

Model-based Recognition of Human Posture Using Single Synthetic Images.

Abstract: Model-based vision has been predominantly concerned with the recognition of single component, rigid objects. This paper describes work attempting to recover the 3D structure of a multi-component, highly articulated object the human body. A major goal of this research has been to go beyond the level of basic object recognition, and attempt to reach a semantic level of description regarding the object's behaviour. Previous research on the recognition of human figures has assumed that the behaviour of the figure in the image is known a priori, for example, "walking", or has made use of motion information derived from image sequences. This research shows that accurate 3D structure can be recovered without such knowledge, and that descriptions of a human figure's behaviour can be obtained in terms of static posture descriptions such as; "sitting", "kneeling", "standing".

Objects, object types and object identification

Abstract: The usefulness of category-theoretic concepts for a theory of object-oriented programming is advocated. Objects (in the latter sense) are defined as objects (in the category-theoretic sense) of a category OB. Colimits in OB are used to express aggregation of objects into complex objects as well as interaction between objects. Object types consist of an identification system, the object universe, and an instantiation system, describing the instances of the type. The main result of this paper gives a semantic basis for database-like identification by keys: the object universe can be specified uniquely (up to isomorphism) employing general principles of preservation of data, distinguishability by keys, and representability by keys.

3D Euclidean versus 2D non-Euclidean: two approaches to 3D recovery from images

Abstract: Methods of 3D recovery in computer vision for computing the shape and motion of an object from projected images when an object model is available are classified into two types: the 3D Euclidean approach, which is based on geometrical constraints in 3D Euclidean space, and the 2D non-Euclidean space. Implications of these two approaches are discussed, and some illustrating examples are presented. >

Knowledge Based Vision For Terrestrial Robots

Abstract: The Knowledge Based Vision Project1,2 is concerned with developing terrain recognition and modeling capabilities for an au land vehicle. For functioning in realistic outdoor environments, we are assuming a vehicle with a laser range finder, controllable cameras, and limited inertial sensing. The range finder is used for mapping and navigating through the immediate environment. The cameras are used for object recognition and recognizing distant landmarks beyond the access of the range sensor. We are assuming the vehicle has realistically limited perceptual and object recognition capabilities. In particular, it will see things that it won't be familiar with and can't recognize, but which can be described as stable visual perceptions. The vehicle will not always be able to recognize the same object as being identical from very different points of view. It will have limited, inexact, and undetailed a prior terrain information generally in the form of labeled grid data. One of the basic functions of the vehicle is to elaborate this terrain map of the environment. Another is to successfully navigate through the environment using landmarks.

Recognizing and locating partially occluded 2-D objects: symbolic clustering method

Abstract: Recognition of objects using a model can be formulated as the finding of affine transforms such that the locations of all object features are consistent with the projected position of the model from a single view A method for the computing of the transform using the symbolic clustering method is described The advantage of this approach is that the desired transform can be computed efficiently and in parallel Experiments support this observation >

Low-level numerical characteristics and inductive learning methodology in texture recognition

Abstract: A method for applying inductive learning to the texture recognition problem is proposed. The method is based on a three-level generalization for the description of texture classes. The first step, scaling interface, is to transform local texture features into their higher symbolic representation as numerical intervals. The second step is the incorporation of the AQ inductive learning algorithm in order to find description rules. The third step is to apply the SG-TRUNC method for rule optimization. The medium recognition ratio for this method was over 90%, and all classes of texture were recognized. In comparison, the k-NN pattern recognition method failed to recognize all classes of textures and had a recognition ratio of 83%. >

Automatic generation of binary feature detectors

Abstract: The authors discuss the automatic generation of feature detectors, which is the major task in the design of classical pattern recognition systems. They present a software environment that, in place of human intuition, utilizes learning strategies and stochastic search procedures to guide the generation process. The environment allows the exploration of evolutionary learning processes and adaptive control mechanisms. A preliminary experiment with a two-class recognition system is described, and initial observations are discussed. The recognition task requires the classification of upper case English letters into two categories: target and nontarget. >

An inexpensive robot arm for computer vision applications

Abstract: The authors have developed a simple, economical robotic arm with both image acquisition and range sensing capabilities for machine vision applications It is connected via an interface board to an IBM PC/AT microcomputer, offering maximum flexibility in programming the arm for different tasks A sonar ranging device and a video camera are mounted on the arm, and give the system 3-D vision capabilities The sonar ranging device and associated clock/counter circuitry is connected to the computer via the same board as the robotic arm, whereas the video camera is connected to an image digitization and display board To illustrate its capabilities, the arm was programmed to scan the immediate environment and attempt to locate and identify the nearest object Standard image processing and computer vision techniques such as binary thresholding, histogram analysis, signatures and pattern recognition were applied, and excellent object recognition was achieved under a wide variety of lighting and distance conditions >

Pose determination using vertex-pairs on the Connection Machine

Abstract: A model-matching scheme for three-dimensional object recognition on a fine-grained data parallel machine is described. The feature that is used for model matching is the vertex pair, introduced by J.L. Mundy et al. (1987). The transformation between the features in the three-dimensional model and the two-dimensional scene is described by an affine transformation, which is an approximation to the perspective transformation. Coarse-to-fine histogramming on an n-dimensional grid is used to compute the best transformation between the model and the scene. Performance results for an implementation on the Connection Machine are presented. >

Model-based three-dimensional object recognition and localization using properties of surface curvatures

Abstract: The ability to recognize three-dimensional (3-D) objects accurately from range images is a fundamental goal of vision in robotics. This facility is important in automated manufacturing environments in industry. In contrast to the extensive work done in computer-aided design and manufacturing (CAD/CAM), the robotic process is primitive and ad hoc. This thesis defines and investigates a fundamental problem in robot vision systems: recognizing and localizing multiple free-form 3-D objects in range images. An effective and efficient approach is developed and implemented as a system Free-form Object Recognition and Localization (FORL). The technique used for surface characterization is surface curvatures derived from geometric models of objects. It uniquely defines surface shapes in conjunction with a knowledge representation scheme which is used in the search for corresponding surfaces of an objects. Model representation has a significant effect on model-based recognition. Without using surface properties, many important industrial vision tasks would remain beyond the competence of machine vision. Knowledge about model surface shapes is automatically abstracted from CAD models, and the CAD models are also used directly in the vision process. The knowledge representation scheme eases the processes of acquisition, retrieval, modification and reasoning so that the recognition and localization process is effective and efficient. Our approach is to recognize objects by hypothesizing and locating objects. The knowledge about the object surface shapes is used to infer the hypotheses and the CAD models are used to locate the objects. Therefore, localization becomes a by-product of the recognition process, which is significant since localization of an object is necessary in robotic applications. One of the most important problems in 3-D machine vision is the recognition of objects from their partial view due to occlusion. Our approach is surface-based, thus, sensitive to neither noise nor occlusion. For the same reason, surface-based recognition also makes the multiple object recognition easier. Our approach uses appropriate strategies for recognition and localization of 3-D solids by using the information from the CAD database, which makes the integration of robot vision systems with CAD/CAM systems a promising future.