By Topic

CAD-Based Vision Workshop, 1994., Proceedings of the 1994 Second

Date 8-11 Feb. 1994

Filter Results

Displaying Results 1 - 25 of 34
  • Feature based aspects-trees. Generation and interpretation

    Publication Year: 1994 , Page(s): 192 - 201
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (976 KB)  

    Presents a new approach in the field of CAD-based vision starting up from the retrieval of the data from a CAD system. A viewer-centered approach is used to generate a set of normalized views. They serve as a basis for an optimal selection of features. The aspect idea is used for grouping the values of the features into aspect-trees. The sophisticated interpretation of the trees proves the ability of the image interpretation system to recognize complex objects in different environments robustly and efficiently.<> View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Reconstructing CAD models of articulated objects

    Publication Year: 1994 , Page(s): 98 - 105
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (882 KB)  

    We present an algorithm for recovering a CAD model from a time sequence of range images of an articulated object. Each recovered part is described using an augmented boundary representation which includes the actually observed surfaces of the object part, implied surfaces between it and other connected parts, and occlusion limits to the unseen portion of the part. The algorithm first develops feature correspondences across the frames of the input sequence. Then, after potential connections are inferred between parts based on relative motion, the boundary of each part is computed. This is followed by the calculation of the parameters of validated connections, including any potential causal relations.<> View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Industrial inspection and reverse engineering

    Publication Year: 1994 , Page(s): 228 - 235
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (580 KB)  

    Proposes a new design for inspection and reverse engineering environments. The authors have designed and experimented with such an environment for capturing sense data of mechanical parts in an intelligent way. They construct a sensing → CAD interface for the automatic reconstruction of parts from visual data. They briefly discuss the use of the dynamic recursive finite state machine (DRFSM) as a new discrete event dynamic system (DEDS) tool for controlling inspection and exploration. They also implement a graphical interface for designing DRFSM DEDS controllers View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Merging multiple views using a spherical representation

    Publication Year: 1994 , Page(s): 124 - 131
    Cited by:  Papers (4)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (780 KB)  

    This paper proposes a new method for building a 3-D model from a set of range images. The method can merge data of free-form surfaces obtained from arbitrary viewing directions, with no prior knowledge of the poses. Our approach is based on matching the spherical representations of an object between the views. To obtain the spherical representation, we deform a discrete mesh to fit the object surface. A variation of the Gaussian curvature metric, which we call simplex angle, is computed at each node on the deformed mesh and mapped to a coordinate on the unit sphere. The transformation of the objects is computed by comparing the simplex angle measure at each node on the unit sphere. The transformation which produces the minimum errors is selected as the best match. We have implemented this method, applied the method to range images of objects from arbitrary viewpoints, and demonstrated the applicability for modeling from observation View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A visual database system for data and experiment management in model-based computer vision

    Publication Year: 1994 , Page(s): 64 - 72
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (732 KB)  

    We present the design of a visual database system for data and experiment management. Our system was designed as a general scientific database system, but motivated by and intended for use in model-based computer vision. We provide a unified data model, a highly graphical user interface, an advanced query facility, and an interactive laboratory notebook. We hope that the system, when completed, will aid in scientific experimentation and will promote data sharing in the computer vision research community View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Generic object recognition using CAD-based multiple representations

    Publication Year: 1994 , Page(s): 202 - 209
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (596 KB)  

    Real-world applications of computer vision usually involve a variety of object models making a single model representation somewhat inadequate for object recognition. Multiple representations, on the other hand, allow different matching strategies to be applied for the same object, or even for different parts of the same object. The paper is concerned with the use of CAD-derived hierarchical models having multiple representations-concave/convex edges and straight homogeneous generalized cylinder-for generic object recognition in outdoor visible imagery. The authors also present a refocused matching algorithm that uses a hierarchically structured model database to facilitate generic object recognition. Experimental results demonstrating generic recognition of objects in perspective, aerial images are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Optimal geometric model matching under full 3D perspective

    Publication Year: 1994 , Page(s): 54 - 63
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (848 KB)  

    Matching algorithms use random-start local search and a 3D pose recovery algorithm to find optimal matches between 3D object models and 2D image features. An algorithm using only a a weak-perspective approximation to full 3D perspective solves a subset of the test problems presented. A second algorithm always uses an iterative 3D pose algorithm to account for 3D perspective and solves all test problems including those with varying 3D perspective. A third hybrid algorithm uses weak-perspective to direct search and 3D pose to periodically correct for perspective. It is faster than the second. A fourth algorithm is a hybrid which also uses a technique called `subset-convergence ' to escape from some local optima. It performs best on the most difficult matching problems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fitting a surface to 3-D points using an inflating balloon model

    Publication Year: 1994 , Page(s): 266 - 273
    Cited by:  Papers (6)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (716 KB)  

    We address the problem of fitting a surface to a cloud of 3-D points, as obtained from a set of registered range images. Our approach is based on a dynamic balloon model represented using a triangulated mesh. The vertices in the mesh are linked to their neighboring vertices through springs to simulate the surface tension and to keep the shell smooth. Unlike other dynamic models proposed by previous researchers, our balloon model is purely driven by an applied inflation force towards the object surface from inside of the object, until all the triangles have reached the surface. We present results on simple as well as nonstar-shaped objects from real range images View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Building three-dimensional CAD/CAM models from image sequences

    Publication Year: 1994 , Page(s): 116 - 123
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (652 KB)  

    We present the results from a working system designed to reconstruct a complete CAD/CAM surface model for objects that have both smooth and sharp surface boundaries. The input is a sequence of images of the object; the output is a 3D surface mesh that can be manipulated and rendered with standard CAD/CAM tools. It is now well-known that complete surface information (second order differential surface properties) can be recovered at edges generated by the extremal boundary of a 3D surface. We present new results in applying this theoretical framework to many views of a real object to build a complete 3D model. Our experiments place these multiple frames in a common coordinate system using known motion, if available, or by otherwise automatically computing object motion based on our classification of edges in the reconstruction process. Experimental results are shown for both real and synthetic data View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Unified 3D models for multisensor image synthesis

    Publication Year: 1994 , Page(s): 161 - 168
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (728 KB)  

    A unified approach to modelling three dimensional objects for multisensory imagery and feature prediction is presented. The object representation is a modified volume-surface octree. This representation is well adapted for the simulations of the physical processes that affect the generation of the different imagery. Specifically, the model supports visual, thermal and laser radar imagery. It is shown that the volume-surface octree representation is well suited for thermal modeling of complex objects with non-homogeneities and heat generation. An octree intersection technique that incorporates non-homogeneities in the material composition of the object is described. A computationally efficient, implicit finite difference method is described for the simulation of energy flow within objects and between the object surface and the environment. The prediction of laser radar (ladar) images uses a statistical model to predict speckle noise, which allows the synthesis of realistic imagery. Examples of the multisensory imagery produced by this scheme are presented. The imagery and features values predicted by this unified modeling scheme may be used to train multisensor-based object recognition systems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Rapid object recognition from a large model database

    Publication Year: 1994 , Page(s): 28 - 35
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (612 KB)  

    A design for a system to perform object recognition from a large model data base is presented, focusing on efficient indexing. The authors propose a decision-theoretic approach using a Bayesian framework to achieve efficient indexing of model objects. A decision-theoretic measure of the discriminatory power of a feature for a model object is defined in terms of posterior probability. Domain-specific knowledge compiled off-line from CAD model data is used in order to estimate posterior probabilities that define the discriminatory power of features for model objects. Detectability of a feature defined as a function of the feature itself, viewpoint, sensor characteristics, and the feature detection algorithm(s) is also considered in the computation of discriminatory power. In order to speed up the indexing or selection of the correct objects, the authors generate and verify the object hypotheses for the features detected in the scene in the order of the discriminatory power of these features for model objects View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Sensor modeling, probabilistic hypothesis generation, and robust localization for object recognition

    Publication Year: 1994 , Page(s): 46 - 53
    Cited by:  Papers (5)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (656 KB)  

    In an effort to make object recognition efficient and accurate enough for applications, the authors have developed three techniques; sensor modeling, probabilistic hypothesis generation, and robust localization-which form the basis of a probabilistic object recognition algorithm. To minimize recognition time, these techniques exploit prior knowledge to reduce the number of verifications (the most expensive and critical part of the algorithm) required during recognition. The approach utilizes statistical constraints generated by modeling the entire sensing process, resulting in more accurate constraints on matches. Hypotheses are pruned by a probabilistic algorithm which selects matches based on image evidence and prior statistical constraints. The reliability of the verification decision is increased by robust localization. The authors have implemented these techniques in a system for recognizing polyhedral objects in range images. The results demonstrate accurate recognition while greatly limiting the number of verifications performed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Matching curved 3D object models to 2D images

    Publication Year: 1994 , Page(s): 210 - 218
    Cited by:  Papers (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (640 KB)  

    Presents a method of locating known rigid 3D objects with arbitrary curved surfaces using a single image. A 3D object is modeled by a covering set of 2D silhouettes together with important internal edges. The model silhouette is derived by the curvature method of Basri and Ullman. Internal edges are computed using a stereo matching strategy. The pose of the observed object is determined by fitting the edgemap derived from the model images to the edgemap of the object. No salient matching primitives are used: correspondence is guided by the minimization of the over-all Euclidean distance between the model edgemap and the observed edgemap. Bench tests and simulations show that the matching technique converges for a broad range (entire aspect) of starting poses View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Object recognition and categorization using modal matching

    Publication Year: 1994 , Page(s): 258 - 265
    Cited by:  Papers (6)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1020 KB)  

    Recently the authors (1993) have proposed a new method for establishing correspondences and computing canonical descriptions, based on the idea of describing objects by their generalized symmetries, as defined by the object's vibration of deformation modes. In this paper they extend this method to the problems of object recognition and categorization. Examples of recognizing hand tools and airplanes are shown View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Surface approximation of a cloud of 3D points

    Publication Year: 1994 , Page(s): 274 - 281
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (976 KB)  

    Presents an implementation of deformable models to approximate a 3-D surface given by a cloud of 3-D points. It is an extension of the authors previous work on “B-snakes” (1990), (1992) which approximates curves and surfaces using B-splines. The user (or the system itself) provides an initial simple surface, such as a closed cylinder, which is subject to internal forces (describing implicit continuity properties such as smoothness) and external forces which attract it toward the data points. The problem is cast in terms of energy minimization. The authors solve this non-convex optimization problem by using the well known Powell algorithm which guarantees convergence and does not require gradient information. Tire variables are the positions of the control points. The number of control points processed by Powell at one time is controlled. This methodology leads to a reasonable complexity, robustness, and good numerical stability View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Planning focus of attention for multi-fingered hand with consideration of time-varying aspects

    Publication Year: 1994 , Page(s): 151 - 160
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (592 KB)  

    Visual monitoring of an object manipulated by a multi-fingered hand has a problem of occlusions caused by the fingers. We propose a new planning method, based on a 3-dimensional CAD model of the environment, to focus attention in such a monitoring task. We formulate the planning to focus attention as a scheduling problem for the whole task period rather than for a specific time. We introduce a genetic algorithm to cope with combinatorial explosion of the image features considered during time-sequential aspects in the task. The genetic algorithm is a population-based search method which efficiently finds solutions in a large search space by simulating biological evolutionary processes. The algorithm consists of selection, crossover, and mutation procedures applied to a population of chromosomes. To solve the problem of planning a focus of attention by using the genetic algorithm, we employed a coding scheme for a chromosome to produce consistent offspring after the crossover. Experimental results show the usefulness of the proposed method by finding the pseudo optimal plan within a short computation time View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multiple representation approach to geometric model construction from range data

    Publication Year: 1994 , Page(s): 132 - 139
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (688 KB)  

    This paper concentrates on techniques for constructing geometric models from data provided by 3-D imaging sensors. Such techniques allow for rapid modeling of sculptured free-form shapes and generation of geometric models for existing parts. Also, the problem of merging images in order to obtain a complete 3-D data set is addressed. The shape representations are closely related to modeling primitives commonly used in computer aided geometric design. In particular, Non-Uniform Rational B-Spline (NURBS), optimal triangulation and superellipsoid surface representations are used in the shape estimation. Computational solutions are provided for shape estimation. The surface approximations are refined by the algorithms to meet a user defined tolerance value. The resulting geometric model is in a procedural form which can convey structural information in addition to low level geometric primitives. Experimental results on view registration and shape estimation are shown View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive sampling and reconstruction for discontinuity preserving texture-mapped triangulation

    Publication Year: 1994 , Page(s): 298 - 303
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (804 KB)  

    The objective of this work is to use input data to construct a texture-mapped polyhedral representation of surfaces that can adapt to intrinsic surface properties, preserving their discontinuities, and might be used efficiently as an accurate 3D object model in graphical rendering. The authors present an adaptive sampling and reconstruction model for hierarchical triangulation of 3D objects. They have developed a parallel algorithm of adaptive mesh generation that recursively bisects mesh elements by increasing the number of mesh nodes according to local surface properties, such as surface orientation, curvature and color. The recursive subdivision based on such a viewpoint-invariant feature yields the hierarchical surface triangulation that is intrinsic to the surface, and satisfies the absolute accuracy criterion, because nodes are generated as many times as required until the entire surface has been approximated with a given threshold. The authors apply the algorithm to the adaptive sampling and reconstruction of range and color-texture images of human faces and fine antique dolls View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Building a B-rep from a segmented range image

    Publication Year: 1994 , Page(s): 74 - 81
    Cited by:  Papers (4)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (648 KB)  

    A segmented range image contains a set of regions representing the projections of the surfaces imaged. Many methods have been proposed to perform the segmentation and surface fitting processes; that is not what this paper is about. Instead, we consider a problem very little analysis has been given to: finding boundary relationships among the regions/surfaces. Since a segmented range image gives us a set of surface patches, it is natural to search for the topology of the 3D surface patches (that is, how the surface patches are connected together). A solution to this problem gives a true 3D boundary representation of the scene View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A robust method for registration and segmentation of multiple range images

    Publication Year: 1994 , Page(s): 106 - 113
    Cited by:  Papers (6)  |  Patents (3)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (648 KB)  

    Registration and segmentation of multiple range images are one of the most important problems in range image analysis. This problem has been investigated by a number of researchers, but most of existing methods are easily affected by outlying points (outliers) like noise and occlusion. We first propose a robust method of estimating rigid motion parameters from a pair of range images. This method is an integration of the iterative closest point (ICP) algorithm with the random sampling and the least median of squares (LMS) estimator. We then detect the outliers by thresholding the residuals in the LMS estimation, and finally we classify each pixel into one of five categories to obtain a segmentation. We experimented on real range images taken by two kinds of rangefinders, and observed that our method worked successfully even for noisy data. The proposed method has another advantage of reducing the computational cost View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • CAD-based object tracking with distributed monocular camera for security monitoring

    Publication Year: 1994 , Page(s): 291 - 297
    Cited by:  Papers (5)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (500 KB)  

    Describes a method for tracking moving objects using several monocular cameras which are mounted in various locations in a building. The authors propose a distributed vision agent model in which each agent has a monocular camera, an image processor and a communication link to other agents. Although the eye of each agent is monocular, it can estimate the location of unknown moving objects by using a CAD solid model of the building. The communication among agents enables the whole system harmoniously to track multiple objects. One application of the system is building security. For that purpose, the system has the ability to recognize and track a deformable object, that is, a walking person. The experimental results showed that the proposed distributed vision agent model can realize an accurate and wide area of object tracking for building security purposes. The merging of image data obtained by inexpensive distributed monocular cameras with a solid model of a building can provide useful 3D information, thus making an expensive active 3D sensing or stereo vision system unnecessary View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Creating random structural descriptions of CAD models and determining object classes

    Publication Year: 1994 , Page(s): 38 - 45
    Cited by:  Papers (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (488 KB)  

    Addresses two problems related to organizing CAD models represented as random parametric structural descriptions (RPSDs). First, the authors present a method to estimate the probability information for RPSDs from CAD models. The randomness arising from viewpoint variation is captured in the random variables corresponding to the attributes of the primitives and the relationship tuples. Next, and more significantly, they present a method to subdivide a large, heterogeneous set of models into smaller, structurally homogeneous subsets based on the distribution of the eigenvalues of property matrices derived from 2½ descriptions of the CAD models. Once this is done, each sublibrary of RPSDs can be hierarchically organized View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • IVIS: an integrated volumetric inspection system

    Publication Year: 1994 , Page(s): 220 - 227
    Cited by:  Patents (2)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (792 KB)  

    Presents a novel approach to automated visual inspection based on comparing volumetric models of a reference object to a model of an actual object iteratively created from sensor data. The use of volumetric models gives this approach a number of distinct advantages over more traditional surface-based methods. First, there is no need to identify “features” which is important when inspecting objects whose features are difficult to identify. Second, volumetric inspection lends itself naturally to multi-sensor applications. Finally, true 3-D comparisons of the reference and sensed objects can be easily carried out using volumetric models View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Integration of multiple feature groups and multiple views into a 3D object recognition system

    Publication Year: 1994 , Page(s): 184 - 191
    Cited by:  Papers (1)  |  Patents (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (588 KB)  

    Proposes two approaches for utilizing multiple-feature group (triples) and multiple-view information to reduce the number of hypotheses passed to the verification stage in an invariant feature indexing (IFI)-based object recognition system. The first approach is based on a majority voting scheme that keeps track of the number of consistent votes cast by prototype hypotheses for particular object models. The second approach examines the consistency of estimated object pose from multiple scene-triples of a single view or multiple views. Monte Carlo experiments employing 500 single-view synthetic range images and 195 pairs of synthetic range images with a large CAD-based 3D object database show that a significant number of hypotheses can be eliminated by using these approaches. The proposed approaches have also been tested on real range images of several objects. A salient feature of this system and experiment design compared to most existing 3D object recognition systems is the use of a large object data base and a large number of test images View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Geometric hashing with attributed features

    Publication Year: 1994 , Page(s): 9 - 16
    Cited by:  Papers (1)
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (540 KB)  

    Geometric hashing systems for object recognition have typically made use of point features in order to describe models and objects. When lines have been included as primitive features, they have been used to generate collections of points from pairwise intersections. In the experiments described in the paper, the authors use line features that include location and orientation information. These features, for which the orientation information is an attribute, are incorporated into a geometric hashing system using weighted voting in order to effect a Bayesian-based maximum likelihood object recognition system. The authors show results of this system which is the first example of the use of attributed features (features with more than coordinate position information) in a geometric hashing application View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.