📄 http:^^www.cs.wisc.edu^computer-vision^pubs.html
字号:
Date: Tue, 05 Nov 1996 00:36:46 GMTServer: NCSA/1.5Content-type: text/htmlLast-modified: Tue, 03 Sep 1996 16:17:44 GMTContent-length: 54887<HTML><HEAD><title>Wisconsin Computer Vision Group Publications</title></HEAD><BODY><h1>Wisconsin Computer Vision Group Publications</h1><HR><BR>Click on any of the following topics to jump to that set of papersin this list of recent publications. Click on the title to view a (postscript) paper. If your browser supportsuncompressing of gzip'ed postscript, you'll prefer to click on thecompressed version to speed up downloading. Size of compressedfile is given in parentheses. <P><DD><!WA0><!WA0><!WA0><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/redball.gif"> <!WA1><!WA1><!WA1><A HREF="#exploration">Visual Exploration</A><DD><!WA2><!WA2><!WA2><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/redball.gif"> <!WA3><!WA3><!WA3><A HREF="#motion">Motion Analysis</A><DD><!WA4><!WA4><!WA4><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/redball.gif"> <!WA5><!WA5><!WA5><A HREF="#shape">3D Shape Representation</A><DD><!WA6><!WA6><!WA6><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/redball.gif"> <!WA7><!WA7><!WA7><A HREF="#snakes">Deformable Contours</A><DD><!WA8><!WA8><!WA8><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/redball.gif"> <!WA9><!WA9><!WA9><A HREF="#visualization">Visualization</A><P><HR><!WA10><!WA10><!WA10><A HREF="http://www.cs.wisc.edu/computer-vision/"> <!WA11><!WA11><!WA11><IMG SRC="http://www.cs.wisc.edu/~dyer/images/return.gif"> </A> Return to Wisconsin Computer Vision Group Home Page<HR><P><H2><A NAME="exploration">Visual Exploration</A></H2><UL><LI><!WA12><!WA12><!WA12><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/new.gif"> <B><A NAME="fest96-yu">Shape Recovery from Stationary Surface Contours by Controlled Observer Motion</A></B><BR> L. Yu and C. R. Dyer, in <CITE>Advances in Image Understanding: A Festschrift for Azriel Rosenfeld</CITE>, IEEE Computer Society Press, Los Alamitos, Ca., 1996, 177-193. (<!WA13><!WA13><!WA13><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/fest96-yu.ps">960K postscript</A>)<p><blockquote>The projected deformation of stationary contours and markings onobject surfaces is analyzed in this paper. It is shown that given amarked point on a stationary contour, an active observer can movedeterministically to the osculating plane for that point by observingand controlling the deformation of the projected contour. Reaching theosculating plane enables the observer to recover the object surfaceshape along the contour as well as the Frenet frame of thecontour. Complete local surface recovery requires either twointersecting surface contours and the knowledge of one principledirection, or more than two intersecting contours. To reach theosculating plane, two strategies involving both pure translation and acombination of translation and rotation are analyzed. Once the Frenetframe for the marked point on the contour is recovered, the sameinformation for all points on the contour can be recovered by stayingon osculating planes while moving along the contour. It is also shownthat occluding contours and stationary contours deform in aqualitatively different way and the problem of discriminating betweenthese two types of contours can be resolved before the recovery oflocal surface shape.</blockquote><LI> <B><A NAME="ijcv94-kutulakos">Recovering Shape by Purposive Viewpoint Adjustment</A></B><BR> K. N. Kutulakos and C. R. Dyer, <CITE>Int. J. Computer Vision</CITE> <B>12</B>, 1994, 113-136. (<!WA14><!WA14><!WA14><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/ijcv94-kutulakos.ps">postscript</A> or <!WA15><!WA15><!WA15><A NAME="ijcv94-kutulakos" HREF="ftp://ftp.cs.wisc.edu/computer-vision/ijcv94-kutulakos.ps.gz">570K gzip'ed postscript</A>)<br> (Earlier versions appeared in <CITE>Proc. Computer Vision and Pattern Recognition Conf.</CITE>, 1992, 16-22 (<!WA16><!WA16><!WA16><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cvpr92-kutulakos.ps">postscript</A> or <!WA17><!WA17><!WA17><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cvpr92-kutulakos.ps.gz">90K gzip'ed postscript</A>),<BR> and as Computer Sciences Department <CITE>Technical Report 1035</CITE> (<!WA18><!WA18><!WA18><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/tr1035-kutulakos.ps">postscript</A> or <!WA19><!WA19><!WA19><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/tr1035-kutulakos.ps.gz">160K gzip'ed postscript</A>).) <p> <blockquote> We present an approach for recovering surface shape from the occluding contour using an active (i.e., moving) observer. It is based on a relation between the geometries of a surface in a scene and its occluding contour: If the viewing direction of the observer is along a principal direction for a surface point whose projection is on the contour, surface shape (i.e., curvature) at the surface point can be recovered from the contour. Unlike previous approaches for recovering shape from the occluding contour, we use an observer that <EM>purposefully</EM> changes viewpoint in order to achieve a well-defined geometric relationship with respect to a 3D shape prior to its recognition. We show that there is a simple and efficient viewing strategy that allows the observer to align the viewing direction with one of the two principal directions for a point on the surface. This strategy depends on only curvature measurements on the occluding contour and therefore demonstrates that recovering quantitative shape information from the contour does not require knowledge of the velocities or accelerations of the observer. Experimental results demonstrate that our method can be easily implemented and can provide reliable shape information from the occluding contour.</blockquote><LI> <B><A NAME="cvpr94-1-kutulakos"> Occluding Contour Detection using Affine Invariants and Purposive Viewpoint Control</A></B><BR> K. N. Kutulakos and C. R. Dyer, <CITE>Proc. Computer Vision and Pattern Recognition Conf.</CITE>, 1994, 323-330. (Received Siemens Best Paper Award <!WA20><!WA20><!WA20><img alg="o" src="http://www.cs.wisc.edu/~dyer/images/award.gif">) (<!WA21><!WA21><!WA21><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cvpr94-1-kutulakos.ps">postscript</A> or <!WA22><!WA22><!WA22><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cvpr94-1-kutulakos.ps.gz">190K gzip'ed postscript</A>)<P><blockquote> We present an approach for identifying the occluding contour and determining its sidedness using an active (i.e., moving) observer. It is based on the <EM>non-stationarity property</EM> of the visible rim: When the observer's viewpoint is changed, the visible rim is a collection of curves that ``slide,'' rigidly or non-rigidly, over the surface. We show that the observer can deterministically choose three views on the tangent plane of selected surface points to distinguish such curves from stationary surface curves (i.e., surface markings). Our approach demonstrates that the occluding contour can be identified <EM> directly</EM>, i.e., without first computing surface shape (distance and curvature).</blockquote><LI> <B><A NAME="cvpr94-2-kutulakos"> Global Surface Reconstruction by Purposive Control of Observer Motion</A></B><BR> K. N. Kutulakos and C. R. Dyer, <CITE>Artificial Intelligence</CITE> <B>78</B>, No. 1-2, 1995, 147-177. (<!WA23><!WA23><!WA23><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/aij95-kutulakos.tar.gz">2.0M gzip'ed tar file</A>) <BR> (Earlier version appeared in <CITE>Proc. Computer Vision and Pattern Recognition Conf.</CITE>, 1994, 331-338. (<!WA24><!WA24><!WA24><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cvpr94-2-kutulakos.ps">postscript</A> or <!WA25><!WA25><!WA25><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cvpr94-2-kutulakos.ps.gz">370K gzip'ed postscript</A>).)<BR> (Longer version appears as Computer Sciences Department <CITE>Technical Report 1141</CITE> (<!WA26><!WA26><!WA26><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/tr1141-kutulakos.ps">postscript</A> or <!WA27><!WA27><!WA27><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/tr1141-kutulakos.ps.gz">1.1M gzip'ed postscript</A>).) <P><blockquote>What viewpoint-control strategies are important for performing globalvisual exploration tasks such as searching for specific surfacemarkings, building a global model of an arbitrary object, orrecognizing an object? In this paper we consider the task ofpurposefully controlling the motion of an active, monocular observerin order to recover a global description of a smooth,arbitrarily-shaped object. We formulate global surface reconstructionas the task of controlling the motion of the observer so that thevisible rim slides over the maximal, connected, reconstructiblesurface regions intersecting the visible rim at the initialviewpoint. We show that these regions are bounded by a subset of thevisual event curves defined on the surface.<P>By studying the epipolar parameterization, we develop two basicstrategies that allow reconstruction of a surface region around anypoint in a reconstructible surface region. These strategies controlviewpoint to achieve and maintain a well-defined geometricrelationship with the object's surface, rely only on informationextracted directly from images (e.g., tangents to the occludingcontour), and are simple enough to be performed in real time. Wethen show how global surface reconstruction can be provably achievedby (1) appropriately integrating these strategies to iteratively``grow'' the reconstructed regions, and (2) obeying four simplerules.</blockquote><LI> <B><A NAME="cbvw94-kutulakos"> Building Global Object Models by Purposive Viewpoint Control</A></B><BR> K. N. Kutulakos, W. B. Seales, and C. R. Dyer, <CITE>Proc. 2nd CAD-Based Vision Workshop</CITE>, 1994, 169-182. (<!WA28><!WA28><!WA28><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cbvw94-kutulakos.ps">postscript</A> or <!WA29><!WA29><!WA29><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/cbvw94-kutulakos.ps.gz">760K gzip'ed postscript</A>)<BR> (An earlier version appeared in <CITE>Proc. SPIE: Sensor Fusion VI</CITE>, 1993, 368-383 (<!WA30><!WA30><!WA30><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/spie93-kutulakos.ps">postscript</A> or <!WA31><!WA31><!WA31><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/spie93-kutulakos.ps.gz">740K gzip'ed postscript</A>).)<P><blockquote>We present an approach for recovering a global surface model of anobject from the deformation of the occluding contour using an active(i.e., mobile) observer able to control its motion. In particular, weconsider two problems: (1) How can the observer's viewpoint becontrolled in order to generate a dense sequence of images that allowsincremental reconstruction of an unknown surface, and (2) how can weconstruct a global surface model from the generated image sequence?Solving these two problems is crucial for automatically constructingmodels of objects whose surface is non-convex and self-occludes. Weachieve the first goal by <EM>purposefully</EM> and <EM>qualitatively</EM>controlling the observer's instantaneous direction of motion in orderto control the motion of the visible rim over the surface. We achievethe second goal by using a calibrated trinocular camera rig and amechanism for controlling the relative position and orientation of theviewed surface with respect to the trinocular rig.</blockquote><LI> <B><A NAME="thesis-kutulakos"> Exploring Three-Dimensional Objects by Controlling the Point of Observation</A></B><BR> K. N. Kutulakos, Ph.D. Dissertation, Computer Sciences Department Technical Report 1251, University of Wisconsin - Madison, October 1994. (<!WA32><!WA32><!WA32><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/thesis-kutulakos.ps">postscript</A> or <!WA33><!WA33><!WA33><A HREF="ftp://ftp.cs.wisc.edu/computer-vision/thesis-kutulakos.zip">5.1M zip-compressed postscript</A>)<P><blockquote>In this thesis we study how controlled movements of a camera can beused to infer properties of a curved object's three-dimensional shape.The unknown geometry of an environment's objects, the effects ofself-occlusion, the depth ambiguities caused by the projectionprocess, and the presence of noise in image measurements are a few ofthe complications that make object-dependent movements of the cameraadvantageous in certain shape recovery tasks. Such movements cansimplify local shape computations such as curvature estimation, allowuse of weaker camera calibration assumptions, and enable theextraction of global shape information for objects with complexsurface geometry. The utility of object-dependent camera movements isstudied in the context of three tasks, each involving the extractionof progressively richer information about an object's unknown shape:(1) detecting the occluding contour, (2) estimating surface curvaturefor points projecting to the contour, and (3) building athree-dimensional model for an object's entire surface. Our mainresult is the development of three distinct active vision strategiesthat solve these three tasks by controlling the motion of a camera.<P>Occluding contour detection and surface curvature estimation areachieved by exploiting the concept of a special viewpoint: Forany image there exist special camera positions from which the object'sview trivializes these tasks. We show that these positions can bedeterministically reached, and that they enable shape recovery evenwhen few or no markings and discontinuities exist on the object'ssurface, and when differential camera motion measurements cannot beaccurately obtained.<P>A basic issue in building three-dimensional global object models ishow to control the camera's motion so that previously-unreconstructedregions of the object become reconstructed. A fundamental difficultyis that the set of reconstructed points can change unpredictably(e.g., due to self-occlusions) when ad hoc motion strategies areused. We show how global model-building can be achieved for genericobjects of arbitrary shape by controlling the camera's motion onautomatically-selected surface tangent and normal planes so that theboundary of the already-reconstructed regions is guaranteed to"slide" over the object's entire surface.<P>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -