By Topic

Interpretation of 3D Scenes, 1989. Proceedings., Workshop on

Date 27-29 Nov. 1989

Filter Results

Displaying Results 1 - 25 of 27
  • Proceedings. Workshop on Interpretation of 3D Scenes (Cat. No.89CH2813-4)

    Save to Project icon | Request Permissions | PDF file iconPDF (75 KB)  
    Freely Available from IEEE
  • Scene description using range data

    Page(s): 138 - 144
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (564 KB)  

    A technique for describing scenes consisting of piles of simple but unknown objects using dense range images is discussed. The technique uses concepts such as symmetry, stability, viewpoint independence, and object solidity to hypothesize the unknown shapes and sizes of the objects. These hypotheses are analyzed using the known geometry of the range sensor to rule out the inconsistent configurations. The final result of the analysis is one or more descriptions of the 3D scene, each of which is consistent with the sensed data and with the constraints imposed by the physics of objects in contact View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Computing exact aspect graphs of curved objects: solids of revolution

    Page(s): 116 - 122
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (600 KB)  

    An approach to computing the exact orthographic aspect graph of curved objects is introduced. Curved corresponding to various visual events partition the Gaussian sphere into regions where the image structure is stable. A catalog of these events for piecewise-smooth objects is available from singularity theory. For a solid of revolution whose generator is an algebraic curve, each visual event is characterized by a system of polynomials whose roots can be computed by continuation methods. Within each region, the stable image structure is characterized by a variation of cylindrical algebraic decomposition and ray tracing. This approach has been implemented, and several examples are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A relational pyramid approach to view class determination

    Page(s): 177 - 183
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (644 KB)  

    Given a CAD model of an object, the authors would like to automatically generate a vision model and matching procedure that can be used in robot guidance and inspection tasks. They are building a system that can predict features that will appear in a 2D view of a 3D object, represent each such view with a hierarchical, relational structure, group together similar views into view classes, and match an unknown view to the appropriate view class to find its pose. They describe the relational pyramid structure for describing the features in a particular view or view class of an object, the summary structure that is used to summarize the relational information in the relational pyramid, and an accumulator-based method for rapidly determining the view class(es) that best match an unknown view of an object View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An analysis of stereo disparity for the recovery of three-dimensional scene geometry

    Page(s): 2 - 8
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (580 KB)  

    The recovery of three-dimensional scene geometry from binocular stereo disparity is discussed. In order to accomplish this goal an analysis of disparity is presented. The analysis makes explicit the geometric relations between a stereo disparity field and a differentially projected scene. These results show how it is possible to recover scene properties, such as relative surface depth and orientation, in a direct fashion from stereo disparity. As a particular application of the analysis, a method for recovering the discontinuities of surfaces from stereo disparity is presented. The results of applying this method to both natural and synthetic binocular stereo disparity information are also presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Kinematic and 3D motion prediction for gaze control

    Page(s): 145 - 151
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (544 KB)  

    Estimation and prediction of the state of dynamical systems operating in three dimensions, and of mechanisms acting in space, provide a way for multiple-loop, multirate, time-delayed control systems to cooperate. Animate vision systems, biological or robotic, employ gaze control systems to acquire, fixate, and stabilize images. The aim is to build robust gaze control behavior from cooperating lower level visual reflexes. Solutions are explored through simulation incorporating ten primitive gaze control capabilities, more or less comparable to subsystems in primate gaze and head control. Versions of several of the subsystems have been implemented on a binocular robot. Smith prediction is the basic paradigm, using kinematic simulation of the agent and optimal filtering to predict world state View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 3D scene reconstruction from multiple intensity images

    Page(s): 124 - 130
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (644 KB)  

    Methods for constructing face-edge-vertex models of objects directly from image data are presented. Each model bounds the actual positions of surfaces on the object. The techniques rely on topology significant image features to generate bounding volumes for the object surfaces. Potentially occupied volumes obtained from the multiple image viewpoints are intersected to form a cumulative description of potential surface positions. In conjunction with certain coherence and visibility assumptions, a model can be produced with little or no ambiguity. Typically, several well-chosen views will produce a good approximation to the object under consideration. Illustrative examples of building models from multiple intensity images are provided View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Accurate surface description from binocular stereo

    Page(s): 16 - 23
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (1104 KB)  

    The authors present a stereo vision system in which they attempt to achieve robustness with respect to scene characteristics, from textured outdoor scenes to environments composed of highly regular man-made objects. The system offers the advantages of both area-based (dense map) and feature-based (accurate disparity) processing by combining them whenever possible. The authors are able to geneate a disparity map that is sufficiently accurate to allow them to detect depth and surface orientation discontinuities, provided that the resolution is fine enough. They use an area-based cross-correlation, along with an ordering constraint and a weak surface smoothness assumption to produce an initial disparity map. Unlike other approaches, however, a match is accepted only if both views agree on a correlation peak and this peak is strong enough. This disparity map is a blurred version of the true one, however, because of the smoothing inherent in the correlation. The problem is most acute at C0 (depth) and C1 (crease) discontinuities but can be mitigated by introducing the edge information: the disparity map is adaptively smoothed subject to the constraint that the disparity at edges is fixed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Robust estimation of camera location and orientation from noisy data having outliers

    Page(s): 52 - 60
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (968 KB)  

    A solution and mathematical analysis are described for the problem of estimating camera location and orientation from a set of recognized landmarks (3D lines) appearing in the image under perspective projection. Given correspondences between these 3D lines and 2D lines found in the image, the goal is to find rotation (R) and transition (T) which map a world coordinate system to the camera coordinate system. Three algorithms are developed: R-then-T, R-and-T, and Med-R-and-T. Algorithms R-then-T and R-and-T are least square algorithms. Algorithm R-then-T solves for rotation first and uses the result to solve for translation. The second algorithm R-and-T solves for both rotation and translation simultaneously. The results from the second algorithm are much better. Some of the correspondences may be wrong. Algorithm Med-R-and-T can handle up to 49.9% outliers or gross errors in the date. It minimizes the median of the square of the same error measure used by the R-and-T algorithm. A closed-form expression function is developed for the uncertainty in the output parameters as a function of the variance of the noise in the input parameters View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Generating aspect graphs for curved objects

    Page(s): 109 - 115
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (568 KB)  

    The authors present an algorithm for computing the aspect graph of a curved object. The approach used to partition the viewpoint space is to compute boundary viewpoints from the shape descriptions of an object in a CAD database. These computations are formulated from the understanding of visual events and the locations of corresponding viewpoints. Also presented is a study of new visual events for piecewise smooth objects View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Navigation using image sequence analysis and 3-D terrain matching

    Page(s): 200 - 207
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (756 KB)  

    The application of digital elevation maps is extended to the field of computer vision. A terrain matching algorithm that makes possible a passive aircraft navigation system is presented. A sequence of aerial images is matched to a reference to a digital map of the 3-D terrain. Stereo analysis of successive images results in a recovered elevation map. A cliff map is then used as a novel, compact representation of the 3-D surfaces. The position and heading of the aircraft are determined via a terrain matching algorithm that locates the unknown cliff map within the reference cliff map. The robustness of the matching algorithm is demonstrated by experimental results using real terrain data View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Interpretation of 3D structure and motion using structured lighting

    Page(s): 84 - 90
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (796 KB)  

    The problem of analyzing time-varying image sequences using structured lighting is investigated. The goal is to estimate the structure and motion parameters of 3D objects in arbitrary motion from a sequence of structured light-coded images. The surface of the moving objects may be either planar or curved, and the mode of motion can be arbitrary. It is proposed that an active sensing principle in the analysis, in which a light pattern is projected onto the object surface to encode the surface for analysis, be adopted. A structured light-coded image is partitioned into small windows, and each window is assumed to be a view of a piece of a plane, of a sphere, or of a cylinder. And these windows can be processed in parallel. The distortion of the projected pattern observed in an image is related to the surface orientation and structure of the imaged object. Furthermore, as the object moves, the light pattern on the surface changes accordingly. Changes of the observed light pattern over time are utilized to compute the motion of the imaged object View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Darboux frames, snakes, and super-quadrics: geometry from the bottom-up

    Page(s): 170 - 176
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (660 KB)  

    A representation and a computational model for deriving three-dimensional articulated volumetric descriptions of objects from laser range-finder data are described. What differentiates this work from other approaches is that it is purely bottom up, relying on general assumptions cast in terms of differential geometry. Darboux frames, snakes, and superquadrics form the basis of this representation, and curvature consistency provides the computational framework. The organization is hierarchical. Darboux frames are used to describe the local surface, while snakes are used to interpolate between features, particularly those that serve to partition a surface into its constituent parts. Superquadrics are subsequently used to characterize the 3-D shape of each surface partition. The result is a set of connected volumetric primitives which serve to describe the overall shape of an object. A set of examples showing how the approach performs on data acquired with a laser range finder is included View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Invariant reconstruction of visual surfaces

    Page(s): 131 - 137
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (628 KB)  

    This study addresses a problem that arises in the early stages of visual processing. Low-level vision processing tasks generate information describing distances and orientations at sparse locations of visible surfaces in a scene. The information from these various low-level tasks need to be combined to form an estimate of the visual surface for all points in the scene. This step is commonly referred to as the surface reconstruction problem. The problem is cast as an ill-posed inverse problem, which must be stabilized using a priori information relative to image and constraint formation. A desirable property of any reconstruction algorithm is invariance with respect to rigid transformations of the surface in three-dimensional space. Two surface reconstruction algorithms that are based on invariant surface characteristics are proposed. These algorithms are studied in detail, and an example is presented to demonstrate their effectiveness. The discrete realization of the algorithms is also briefly discussed View full abstract»

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

    Page(s): 91 - 99
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (600 KB)  

    The use of a moving camera in stereo vision is validated, and the advantages and limitations of such an approach are demonstrated. The basic principle of 3D perception by a moving camera is presented and applied to the reconstruction of a 3D polyhedral scene. The problem is split into the following subproblems: (a) how to approximate an edge image by a set of 2D line segments; (b) how to match the set of 2D line segments within a sequence of images in order to obtain the dynamic parameters of 2D line segments; (c) how to rebuild the 3D line segments from the knowledge about the camera motion and the dynamic parameters of 2D line segments; and (d) how to interpret the set of 3D line segments in order to obtain a polyhedral representation of a scene View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A model-based vision system for manipulator position sensing

    Page(s): 186 - 193
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (844 KB)  

    The task and design requirements for a vision system for manipulator position sensing in a telerobotic system are described. Model-based analysis-by-synthesis techniques offer generally applicable methods with the potential to meet the system's requirement for accurate, fast, and reliable results. Edge-based chamfer matching allows efficient computation of a measure E of the local difference between the real image and a synthetic image generated from arm and camera models. Gradient descent techniques are used to minimize E by adjusting joint angles. The dependence of each link position on the position of the link preceding it allows the search to be broken down into lower dimensional problems. Intensive exploitation of geometric constraints on the possible position and orientation of manipulator components results in a correct and efficient solution to the problem. Experimental results demonstrate the use of the implemented prototype system to locate the boom, stick, and bucket of an excavator, given a single video image View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 3D visual information from vanishing points

    Page(s): 41 - 49
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (776 KB)  

    It is shown that when a priori information on the mutual direction of straight lines in the 3D scene is available, vanishing points can be extracted and located reliably on the image plane. Then, by using simple properties of vanishing points, it is possible to perform efficiently several visual tasks, such as estimating the rotational component of motion between the viewer and the scene, and identifying and recovering the orientation and relative position in space of the viewed planar patches. Extensive experimentation on real images shows that vanishing points are usually identified and located correctly, even in cluttered images. It is concluded that the proposed approach can be very helpful for the development of effective vision systems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Computing the orthographic projection aspect graph of solids of revolution

    Page(s): 102 - 108
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (588 KB)  

    The first algorithm to handle a defined class of curved-surface objects is presented. The algorithm automatically computes the partition of the Gaussian sphere, and thereby the aspect graph, for solids of revolution defined as right, circular, straight, homogeneous generalized cylinders. The algorithm demonstrates that it is possible to compute the exact aspect graph for at least some classes of curved-surface objects. The obvious extension of this work is to attempt to apply the same or a similar approach to larger and more varied classes of objects View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Face authentification or recognition by profile extraction from range images

    Page(s): 194 - 199
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (444 KB)  

    Work related to the authentification or recognition of human faces from their range images is described. The approach consists of comparing a part of the profile or the surface of two faces. For this purpose, the profile plane of each face is considered as a quasi-symmetry plane and is extracted by an iterative process which involves the similarity of the Gaussian curvature values of the face surface. Then an optimal matching of the two profiles is done, which allows the calculation of similarity indices of profiles or surfaces. Tests for robustness and the authentification or recognition results show the efficiency of the whole procedure View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Object recognition and localization using optical proximity sensor system: polyhedral case

    Page(s): 75 - 81
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (556 KB)  

    The authors present an algorithm for the recognition and localization of 3D polyhedral objects based on an optical proximity sensor system. In particular, the representation of a polyhedral object and the determination of the optimal sensor trajectory for the next probing are considered. The object representation is based on two levels of hierarchy: the description of a 3D structure by an intersurface relation description table (SDT) and the surface normal vector (SNV) distribution graph, and the description of individual surfaces by interedge relation description tables (EDTs). The partially filled SDT and EDTs of the test object are matched against the SDT and EDTs of a model object to extract all the possible interpretations. In order to achieve the maximum discrimination among all possible interpretations, the optimal sensor trajectory for the next probing is determined as follows: (1) select the optimal beam orientation on the basis of the SNV distribution graph of the multiple interpretation image (MII), where the MII is formed with reference to the hand frame by localizing the test object on the basis of individual interpretations, and (2) determine the optimal probing plane by projecting the MII onto the projection plane perpendicular to the beam orientation and deriving the optimal path on the probing plane. Simulation results are shown View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Quantitative and qualitative measures for the evaluation of the superquadric models

    Page(s): 162 - 169
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (652 KB)  

    Evaluation criteria for superquadric models recovered from the range data discussed. Arguments are presented to support the authors' belief that both quantitative and qualitative measures are required in order to evaluate a superquadric fit. The concept of superquadric contraction and dilation is introduced and used to derive a novel interpretation of the modified superquadric inside-outside function in terms of contraction/expansion factor. The same concept also gives a close initial guess for the numerical procedure computing the minimum Euclidean distance of a point from a superquadric model. The minimum Euclidean distance map is introduced as a qualitative criterion for interpretation of fit. View-dependent qualitative measures like the contour-difference map and the Z-distance map are shown to be essential for the complete evaluation of the models. Analytical solution and techniques for the contour generator on superquadric models are presented. Finally, examples of real objects are given to generate the measures View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Integrating multiresolution image acquisition and coarse-to-fine surface reconstruction from stereo

    Page(s): 9 - 15
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (776 KB)  

    The authors are concerned with the problems of reconstructing surfaces from stereo images for large scenes having large depth ranges. They present an approach to acquiring coarse structural information about the scene in the vicinity of the next fixation point during the current fixation and utilizing this information for surface reconstruction in the vicinity of the next fixation point. The approach involves the processing of peripheral low-resolution parts of the current images away from the image center, in additive to accurate surface estimation from the central high-resolution parts containing the fixated object. The processing of the low-resolution parts yields coarse surface estimates to be refined after the cameras have refixated and the parts of the scene around the new fixation point (currently at low resolution) are imaged more sharply. The coarse estimates are obtained from both stereo and focus. The choice as to which estimate is actually used depends on which is determined to be more accurate in the given situation. Experimental results demonstrate that the approach may be well suited for active data acquisition and surface reconstruction by an autonomous stereo system View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Acquisition and interpretation of 3-D sensor data from touch

    Page(s): 33 - 40
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (692 KB)  

    A description is given of the use of touch sensing as part of a larger system the authors are building for 3-D shape recovery and object recognition using touch and vision methods. The authors focus on three exploratory procedures (EPs) they have built to acquire and interpret sparse 3D touch data: grasping by containment, planar surface exploration, and surface contour exploration. Experimental results for each of these procedures are presented. The EPs can be used in a coarse-to-fine sensing strategy that tries to build shape descriptions at a number of levels. An important feature of this system is the multiple representations used in recovering and reasoning about shape View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Generic object recognition: building coarse 3D descriptions from line drawings

    Page(s): 68 - 74
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (648 KB)  

    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 View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Range imaging sensors development at NRC Laboratories

    Page(s): 154 - 160
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (824 KB)  

    Laser range finders can be used in a variety of geometries to match requirements of applications such as robot vision, inspection, prosthesis fabrication, CAD/CAM, computer graphics, and automatic assembly. A review is presented of the development of five prototypes which have specified advantages for a class of applications. One of these is based on the use of a mask with a conventional TV camera lens, and the others are variations of synchronized laser scanners. Also described is a segmentation technique based on Fourier transformations and a database of 3D images that are made available for computer vision researchers View full abstract»

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