Polyhedral Object Recognition Research Articles

Reconstruction of images and recognition of polyhedral objects

An indirect way of reconstructing the coordinates of points on the surface of a 3D object by its planar parallel projections is proposed. The approach is based on the substitution of the object by another (virtual) object, for which this operation can be carried out simply, whereas the correctness of the obtained results is controlled. The specificities of obtaining a mathematical model of reconstructed objects with a polyhedral shape, the issues of normalization of the angular discrepancies between the recognized and the etalon objects, and the solution of the problem of their recognition based on the introduced model are considered.

다면체 인식을 위한 탐색 공간 감소 기법

본 논문에서는 다면체의 인식을 위하여 사용되는 여러-방향-보기 방법 (multiple-view approach)에서, ART-1 신경망을 이용하여 모델베이스의 탐색공간 크기를 줄이기 위한 방법을 제안한다. 이 방법에서 모델베이스는 물체를 둘러싸고 있는 보기 구체의 미리 정해진 시점에서 관측된 2차원 투영체에서 추출된 특징들로 구성된다. We suggest a method which reduces the search space of a model-base on multiple-view approach for polyhedral object recognition using the ART-1 neural network. In this approach, the model-base is consisted of extracted features from two-dimensional projections observed at the predetermined viewpoints of a viewing sphere enclosing the object.

Recognition of polyhedral objects using triplets of projected spatial edges based on a single perspective image

A novel and intuitive mathematical formulation is proposed for the interpretation of the geometric relationships of a triple of spatial edges and their perspective projection forming image lines. No restriction is imposed on the configuration of the triple of spatial edges. An eighth-degree polynomial equation explicitly defined by the space angles between the corresponding three spatial edges in an object frame is derived. The crux of this representation is that the angular attributes of pairs of spatial edges are expressed in an object-independent coordinate system. An effective hypothesis generation scheme is proposed which can take advantage of the representation. In the proposed recognition framework, triples of connected spatial edges and trihedral vertices are employed as key features for model invocation, hypothesis generation and pose estimation. Extensive experiments are performed to verify the plausibility and robustness of the proposed recognition methods. In general, the success rate for identifying and localising polyhedral objects using the proposed recognition method is relatively high.

Topological recognition of polyhedral objects from multiple views

The aspect graph (AG) has been proposed as a viewer-centered tool for object recognition. Although the aspects could be used for recognition from one image, the full information stored in the AG (aspects and visual events) suggests a multiple views, possibly active, approach. The topological nature of the AG also suggests a topological match of images of the unknown object and stored aspects. In this paper we present a theoretical investigation on the use of the AG for topological recognition from multiple views of polyhedral objects. First, we discuss the topological matching process, and give a suitable topological definition of aspect. Moreover, since topological identification is approximate, we tackle the problem of understanding the ability of the AG to topologically discriminate different objects. More precisely, we address the question: how “similar” are two polyhedra with the same AG? The isomorphism of polyhedra is chosen as the reference similarity condition. The cases of general and convex polyhedra are discussed under perspective and parallel projections. In practice, it could be difficult to identify the visual events from the images of the unknown object. Thus, we also compare isomorphism and the similarity induced by: (i) a topologically reduced aspect graph; (ii) the set of topologically different aspects, neglecting visual events.

Recognition of occluded polyhedra from range images

Occlusion remains a major hindrance for automatic recognition of 3-D objects. In this paper, we address the occlusion problem in the context of polyhedral object recognition from range data. A novel approach is presented for object recognition based on sound occlusion-guided reasoning for feature distortion analysis and perceptual organization. This type of reasoning enables us to maximize the amount of information extracted from the scene data, thus leading to robust and efficient recognition. The proposed approach is based on a multi-stage matching process, which attempts to recognize scene objects according to their order in the occlusion hierarchy (i.e., an object is recognized before those that are occluded by it). Such a strategy helps in resolving some occlusion-induced ambiguities in feature distortion analysis. Furthermore, it leads to verification of object/pose hypotheses with greater confidence. Matching is based on a hypothesize-cluster-and-verify approach. Hypotheses are generated using an occlusion-tolerant composite feature, a fork, which is a pair of non-parallel edges that belong to the same surface. Generated hypotheses are then clustered and verified using a robust pixel-based technique. Indexing is performed using distortion-adaptive bounds on a rich set of viewpoint-invariant fork attributes, for high selectivity even in the presence of heavy occlusion. Performance of the system is demonstrated using complex multi-object scenes.

Generalized Polyhedral Object Recognition and Localization Using Crossbeam Sensing

Object localization is a fundamental task in industrial automa tion. In this article, we present a recognition and localization technique based upon binary beam sensors. Binary beam sen sors, which consist of modulated LED light sources and de tectors, are appropriate for manufacturing applications due to their reliability, high accuracy, robustness, inexpensiveness, and ease of calibration. By extracting sufficient information, recognition and localization can be performed as fast (~ 0.1 s) and as accurately (~ 25 microns) as high-speed manufactur ing requires. Generalized polyhedral objects are recognized and localized by being passed through a crossbeam sensor which is a set of coplanar oriented binary light-beam sensors; robot positions are recorded when the beam sensors' outputs change. Recognition and localization share two subproblems: the correspondence problem (the problem of interpreting the sensed features in terms of the model features), and the pose- estimation problem (the problem of estimating the object's pose from the sensed data and an interpretation of the sensed fea tures in terms of model features). In Wallack and Manocha (1994), we described a pose-estimation technique for crossbeam sensor data. In this article, we present two methods for solving the correspondence problem for crossbeam sensor data: a lin ear time completely on-line method, and a constant-time on-line method that utilizes preprocessing.

Polyhedral object recognition by indexing

In computer vision, the indexing problem is the problem of recognizing a few objects in a large database of objects while avoiding the help of the classical image-feature-to-object-feature matching paradigm. In this paper we address the problem of recognizing three-dimensional (3-D) polyhedral objects from 2-D images by indexing. Both the objects to be recognized and the images are represented by weighted graphs. The indexing problem is therefore the problem of determining whether a graph extracted from the image is present or absent in a database of model graphs. We introduce a novel method for performing this graph indexing process which is based both on polynomial characterization of binary and weighted graphs and on hashing. We describe in detail this polynomial characterization and then we show how it can be used in the context of polyhedral object recognition. Next we describe a practical recognition-by-indexing system that includes the organization of the database, the representation of polyhedral objects in terms of 2-D characteristic views, the representation of this views in terms of weighted graphs and the associated image processing. Finally, some experimental results allow the evaluation of the system performance.

Matching and Clustering: Two Steps Toward Automatic Object Modeling in Computer Vision

In this article, we present a general frame for a system of au tomatic modeling and recognition of 3D polyhedral objects. Such a system has many applications for robotics: e.g., recog nition, localization, and grasping. Here we focus on one main aspect of the system: when many images of one 3D object are taken from different unknown viewpoints, how to recognize those that represent the same aspect of the object? Briefly, is it possible to determine automatically if two images are similar or not? The two stages detailed in the article are the matching of two images and the clustering of a set of images. Matching consists of finding the common features of two images while no information is known about the image contents, the motion, or the calibration of the camera. Clustering consists of regrouping into sets the images representing a same aspect of the modeled objects. For both stages, experimental results on real images are shown.

An algebraic method for detection and recognition of polyhedral objects from a single image

This paper proposes an algorithm to recognize rotated, scaled or overlapping polyhedral objects using only one orthographic projection. The procedure is not based on the numerical solution of nonlinear equations and does not require any a priori knowledge. It is an algebraic method for testing whether a cluster of edges and vertices, in a line drawing, correspond to an orthographic view of a given 3D model. It has the advantage of being based on simple algebraic equations, which are easy to solve and do not require time-consuming numerical procedures.

Global description of edge patterns using moments

Shape recognition has traditionally been accomplished using well bounded segmented regions or their closed contours. In this way, the internal edge details of the objects are ignored. There are cases such as in character recognition and in object recognition when patterns are not closed contours. In this paper, such patterns are defined as edge patterns and Edge Standard Moments (ESM) are developed which are invariant to location, scale and rotation. Results of patterns matching with character recognition and polyhedral object recognition are presented.

Pictorial drawing generation for polyhedral object recognition using intensity images

This paper presents a new approach to automatically generate the line drawings of polyhedral objects in intensity images. The intensity images are, through image processing techniques, converted into binary boundary images. However, the rounded edges of the objects causes smal intensity differences and cause some features to be missing from the image. Using the labeling technique, heuristic inference, and surface gradient analysis, the system may generate pictorial drawings. The pictorial drawing is viewpoint-dependent.

Recognition of polyhedral objects using triangle pair features

This paper is concerned with the problem of model based recognition of polyhedral objects from a single perspective view. A hypothesise-verify paradigm based on the use of high level knowledge constraints derived from local shape properties is presented. In the recognition system, two high level features, namely triangle pair and quadrilateral are employed as key features for model invocation and hypothesis generation. A verification process for performing a detailed check on the model to scene correspondences is developed. To reduce the number of implausible hypotheses generated from scene to model intermediate feature assignments, two geometrical contraints, namely distance and angle constraints are employed. A list of closed polygons and C-triple pairs extracted from a 2D intensity image by means of edge and intermediate feature detection process is used as an input to the matching system. The intermediate feature grouping process starts by identifying junctions created by pairs of line segments and then forms triples by combining pairs of junctions which share a common line. These triples are then scanned by procedure which connects them into meaningful geometric structures. As a byproduct of the recognition method, the relative pose of the 3D polyhedral objects with respect to the camera is recovered. Extensive experimental results are reported to confirm the feasibility of the proposed method.

INTEGRA — AN INTEGRATED SYSTEM FOR RANGE IMAGE UNDERSTANDING

Segmentation, feature extraction, recognition and localization are the four stages in range image understanding. Conventional approaches to range image understanding have treated these stages in isolation with a largely bottom-up flow of control and data through these various stages. Strictly bottom-up approaches have proved to be fragile in the face of errors in segmentation due to noise and limitations on sensor resolution and accuracy. Synergetic interaction of these various stages is essential for an image understanding system to exhibit robust behavior. This paper describes the design and implementation of INTEGRA, a range image understanding system that attempts to exploit the synergy between the various stages in the image understanding process. The salient features of INTEGRA are: (i) A synergetic combination of edge- and surface-based segmentation processes that results in more accurate segmentation than would have been possible with either of them alone and (ii) the ability to correct errors made during segmentation in the matching and localization stages. INTEGRA at this time, is limited to recognition and localization of polyhedral objects and is in the process of being enhanced to handle objects with curved surfaces of quadratic order such as spherical, ellipsoidal, cylindrical, and conical surfaces. Experimental results on real range images containing single and multiple polyhedral objects are presented. Future enhancements to INTEGRA are discussed.

Qualitative features and the generalized hough transform

In this paper we show how the use of qualitative features can enhance the performance of recognition and localization techniques, in particular, the Generalized Hough Transform. Qualitative features (i.e. scene features with qualitative attributes assigned to them) are shown to be effective in pruning the search space of possible scene interpretations and also reducing the number of spurious interpretations explored by the recognition and localization technique. The redundancy of the computed transform and the probability of spurious peaks of significant magnitude due to random accumulation of evidence are two criteria by which the performance of the Generalized Hough Transform is judged. The straightforward Generalized Hough Transform shows a high probability of spurious peaks of significant magnitude even for small values of redundancy and small magnitude of the search space of scene interpretations. The use of qualitative features enables us to come up with a weighted Generalized Hough Transform where each match of a scene feature with a model feature is assigned a weight based on the qualitative attributes assigned to the scene feature. These weights could be looked upon as membership function values for the fuzzy sets defined by these qualitative attributes. Analytic expressions for the probability of accumulation of random events within a bucket are derived for the weighted Generalized Hough Transform and compared with the corresponding expression for the straightforward Generalized Hough Transform. The weighted Generalized Hough Transform is shown to perform better than the straightforward Generalized Hough Transform. An experiment for the recognition of polyhedral objects from range images is described using dihedral junctions as features for matching and pose computation. The experimental results bring out the advantages of the weighted Generalized Hough Transform over the straightforward Generalized Hough Transform.

Geometric uncertainties in polyhedral object recognition

It is shown, by direct geometric construction, that previously published uncertainty bounds on the location of polygonal or polyhedral objects can be tightened considerably. The improvement of the bounds is a result of considering the cross-coupling between rotational and translation uncertainties in the interpretation of the sensor data. Both two-dimensional and three-dimensional uncertainty bounding are tested in simulation. The two-dimensional studies are implemented as part of a robotic system for tactile recognition. These implementations prove to be useful vehicles for exploring the computational issues involved in a detailed analysis of the geometric uncertainties present in model-based object recognition systems. Among other results it is found that rotational uncertainty is independent of the scale of the models, and translational uncertainty is highly dependent on the relative angles of the model components that are sensed. >

Sensitivity analysis for matching and pose computation using dihedral junctions

Recognition-via-localization is a popular approach in 3-D object recognition. This approach relies on the propagation of constraints that arise from the matches of local geometric features and could therefore be treated as a constraint satisfaction problem. Hough clustering, which verifies the consistency of local geometric constraints by determining the pose of the object in parameter space, is a popular technique owing to its conceptual simplicity and potential ease of parallelization. Our previous work has shown the usefulness of dihedral junctions for the recognition and localization of polyhedral objects and dihedral feature junctions for the recognition and localization of curved objects made up of piecewise combinations of conical, cylindrical and spherical surfaces. Experimental results from our previous work showed that the computed pose parameters are sensitive to the difference in the included angle between the scene and model dihedral junctions or the scene and model dihedral feature junctions. A formal analysis of the sensitivity of the computed pose to the difference in the included angle between the scene and model dihedral junctions or the scene and model dihedral feature junctions is presented in this paper. The results of the formal sensitivity analysis were found to be in conformity with the experimental results from our previous work and so the work presented in this paper could be treated as a sequel to our previous work. Based on the results of the sensitivity analysis, the rotation parameters were found to be more sensitive than the translation parameters which, in comparison, were far more robust. It is also shown how the introduction of redundancy in parameter space results in greater robustness in the computed pose. Although the analysis in this paper is based on the matching of dihedral junctions or dihedral feature junctions, the approach taken in the sensitivity analysis is general and can be applied to the matching based on other feature types.

Polyhedral object recognition with sparse data — validation of interpretations

The method of Grimson, Lozano Pérez et al., for the generation of feasible interpretations of scenes with sparse data, has been developed and implemented by the authors on a distributed array processor, the AMT DAP, which operates in SIMD mode. Measurements involving the location vectors and the surface normals at m data points, considered in pairs, are compared with the maximum and minimum values associated with the n × n pairs faces of a polyhedral object model, in a process that exploits n × n parallelism. The subsequent validation of the interpretations, in which data points have been assigned provisionally to object model faces, are discussed.

Polyhedral object recognition with sparse data in SIMD processing mode

The method of Grimson, Lozano Pérez and others for the recognition of polyhedral objects with sparse data, has been developed and implemented on a distributed array processor, the AMT DAP 500, which operates in single instruction-multiple data (SIMD) mode. Measurements involving the location vectors and the surface normals at m data points, considered in pairs, are compared with the corresponding maximum and minimum values associated with n × n pairs of object model faces in a process that exploits n × n parallelism. The overall processing time is essentiallv proportional to m × (m−1) 2 to explore the interpretation tree to its full depth. This paper discusses the nature of the comparisons between object models and data, together with the need to make these comparisons in a particular sequence, and results of test runs with a variety of object models and different geometric constraints are presented herein. Comparison is made with the corresponding sequential process, and with the more costly method of Flynn and Harris, in which n m processing elements are required to acheive, at best a processing time of the same order of magnitude.

Polyhedral object recognition using Hough-space features

We propose a new technique for three-dimensional (3-D) polyhedral object recognition on the basis of a single two-dimensional (2-D) view of a 3-D scene. The binary gradient image of the captured scene is converted into the Hough-space domain. The cluster patterns originating from straight-line features of the image are explored by reasoning in Hough space. This yields an attributed graph representation of the object to be recognized which is compared to similar representations of CAD-designed wire frame model objects by means of a new attributed subgraph isomorphism algorithm. Simulation experiments illustrate this promising new approach.

Using the orientation of fragmentary 3D edge segments for polyhedral object recognition

Several constraints are developed which impose consistency between model edge and sensed edge fragment geometries in order to collapse the search space in matching 3D edge data to 3D edge models. The constraints have been demonstrated on error-prone synthetic edge data and on 3D edge data obtained by stereo processing. An important feature of the constraints is their ability to attach sign labels to the sensed edge fragments during matching, providing consistency of edge directions between the model and data. This feature progressively increases the power of the constraints and assists in finding the transformation between model and sensor spaces.