perm filename RES81.VW[INS,TOB] blob sn#603065 filedate 1982-02-03 generic text, type C, neo UTF8
COMMENT āŠ—   VALID 00012 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	\|\\viewgraphs of results of research - brag
C00003 00003	\F0
C00004 00004	\F0
C00005 00005	\F0
C00006 00006	\F0
C00007 00007	\F0
C00008 00008	\F0
C00009 00009	\F0
C00010 00010	\F0
C00011 00011	\F0
C00012 00012	\F0
C00013 ENDMK
CāŠ—;
\|\\;viewgraphs of results of research - brag
\|\\M0FIX25;\M1NONH;\M2FIX40;\M3FIX30;
\F0
\F0


\CSUMMARY

ACRONYM
INFERRING SURFACES FROM IMAGES
USING SHADOWS
EDGES AND CURVES
DESCRIPTION AND RECOGNITION OF COMPLEX OBJECTS
STEREO CONSTRAINTS
GENERALIZED CYLINDERS
MORAVEC
GENNERY
HANNAH
CLARKSON
MILLER
LIEBES
\F0


\CACRONYM

IDENTIFICATION OF AIRCRAFT
    results:
        aerial: L-1011, aircraft
    monocular, edge input
    aircraft models
    symbolic prediction: viewpoint-independent
    rules not special to aircraft or aerial views
    3d interpretation with constraints
    rule-based geometric reasoning
\F0


\CARCHITECTURE OF IMAGING OPERATIONS

VLSI PROCESSOR
    result:  fabricated and tested single processor
    plan:   10x10 array;     100x100 array
    fault-tolerant discretionary design
    general-purpose computer
    bit-serial

ANALYSIS OF SEGMENTATION OPERATORS
\F0


\CDESCRIPTION AND RECOGNITION 
\COF COMPLEX OBJECTS

\CNevatia

results:
    recognition of doll, toy horse, tools, ...
    circular cross-sections: ring, cones, dolls, snake
learning descriptions
curved, complex objects
using depth input from laser ranging 
extracting generalized cylinder descriptions
attachment hierarchy
recognition in large visual memory
    indexing into related subclass of shapes
\F0


\CGENERALIZED CYLINDERS


representation of 3d shape
relating 2d shape to 3d shape
    quasi-invariants for 2d shape
large class of objects
    locally-generative
compact
symbolic predictions
volume representation
    natural part/whole segmentation
\F0



\CGENNERY

RESULTS: 
    GROUND LEVEL:   MARS, PARKING LOT, 
    AERIAL: APARTMENT HOUSE
GROUND PLANE FINDER
BOX FINDER
ELLIPSE FINDER
CAMERA TRANSFORM SOLVER
    EQUIVALENT TO RIGID BODY MOTION SOLUTION
INTERPOLATING CORRELATOR
STEREO MAPPING SYSTEM
\F0



\CMORAVEC


RESULTS:
    CART NAVIGATION BY VISION

\F0


\CARNOLD

EDGE-BASED STEREO VISION
    results:
        aerial: cars, aircraft, apartment house
    description of car as box, dimensions to 5-10%
    theoretical constraints:
        angle, surface interval
    dynamic programming
    new results in alternate solutions
\F0


\CBAKER

HIGH-RESOLUTION STEREO MAPPING
    results: 
       aerial: building complex, terrain
    comparison with Grimson
    edge correspondence, zero-crossings, coarse-fine
    surface interval match
    dynamic programming within epipolar line
    consistency between epipolar lines
\F0