13.6.3 University of Massachusetts VISIONS System

Chapter Contents (Back)
Knowledge-Based Vision. Recognition, Model Based. Model Based Recognition. Object Recognition. Matching, Models. VISIONS. Model Based Recognition. System: VISIONS. See also Complete Systems Derived from the Univ. Massachusetts Work.

Hanson, A.R., and Riseman, E.M.,
VISIONS: A computer System for Interpreting Scenes,
CVS78(303-333). Multiple Resolutions. System: VISIONS. The basic outline of their system. For the full set of papers and a more complete description: See also University of Massachusetts VISIONS System. BibRef 7800

Hanson, A.R., and Riseman, E.M.,
The VISIONS Image-Understanding System,
ACV88(I: 1-114). BibRef 8800
And: COINS-TR-86-62, December 1986. The work on the VISIONS System at the Univ. of Massachusetts has extended over a number of years. Throughout the effort the work has addressed the mapping of image regions to objects names in the context of outdoor scenes. This work combines bottom-up (data driven) processing with top-down (knowledge driven) approaches to identify many object regions. Initial object hypotheses are formed based on the appearance (intensity, color, etc.) using a trapezoidal weighting function (similar to that discussed in section 4) that gives a maximum rating (1) when the value is within the good range and a minimum value (0) when it is in the bad range. Between these ranges, it is a linear combination. These threshold values are derived from the histograms of the feature value for known regions in a training set (e.g. grass, sky, etc.). The initial hypotheses are extended by finding other regions in the current image that are similar to the image region that best matches the given values from the model. This allows for shifts in the value of properties of objects between images, with some controls provided by the model. The final choice of the object identification depends on domain knowledge such as constraints on relationships of the different objects (grass is below sky). BibRef

Draper, B.A., Hanson, A.R., Riseman, E.M.,
Knowledge-Directed Vision: Control, Learning, and Integration,
PIEEE(84), No. 11, November 1996, pp. 1625-1637. 9611

PS File. BibRef

Draper, B.A., and Hanson, A.R.,
An Example of Learning in Knowledge-directed Vision,
PS File. BibRef 9100

Draper, B.A., Collins, R.T., Brolio, J., Hanson, A.R., and Riseman, E.M.,
The Schema System,
IJCV(2), No. 3, January? 1989, pp. 209-250.
Springer DOI BibRef 8901
And: COINSTR 88-76, UMass. Domain knowledge in VISIONS is represented by schemas, which provide a means to represent structured descriptions using features, relations, uncertain links, etc. Each schema while operating in parallel, tries to find evidence to support (or reject) the interpretation of a region as an object, by communicating with schema for conflicting interpretations. The whole schema system gives a means of resolving conflicting interpretations much like other tree searching or graph matching systems.
HTML Version. BibRef

Draper, B.A., Hanson, A.R., Riseman, E.M.,
Learning Blackboard-Based Scheduling Algorithms for Computer Vision,
PRAI(7), 1993, pp. 309-328. BibRef 9300
And: COINSTR-92-51, August 1992.
PS File. BibRef

Hanson, A.R., and Riseman, E.M.,
From Image Measurements to Object Hypotheses,
COINSTR 87-129, UMass., December 1987. Color. Description of generating symbolic interpretations of outdoor scenes by combining many different techniques. The mapping of color parameters is a 0, slope, 1 function. BibRef 8712

Draper, B.A., Collins, R.T., Brolio, J., Griffith, J.S., Hanson, A.R., and Riseman, E.M.,
Tools and Experiments in the Knowledge-Directed Interpretation of Road Scenes,
DARPA87(178-193). BibRef 8700
And: COINSTR-87-05, January 1987. More of the UMass system work. Region hypothesis aids in merging, etc. This looks a lot like Yakimovsky's work(?). BibRef

Lehrer, N.B., Reynolds, G., Griffith, J.,
A Method for Initial Hypothesis Formation in Image Understanding,
ICCV87(578-585). BibRef 8700

Lehrer, N.B., Reynolds, G., Griffith, J.,
Initial Hypothesis Formation in Image Understanding Using an Automatically Generated Knowledge Base,
DARPA87(521-237). BibRef 8700

Draper, B.A.,
Statistical Properties of Learning Recognition Strategies,
DARPA93(557-565). BibRef 9300
Learning Object Recognition Strategies,
UMassCS TR-93-50, May 1993. BibRef Ph.D.Thesis (CS), University of Massachusetts, 1993. BibRef

Draper, B.A.[Bruce A.],
Learning from the Schema Learning System,
AAAI-MLCV93(75-79). University of Massachusetts.
PS File. BibRef 9300

Draper, B.A., and Riseman, E.M.,
Learning 3D Object Recognition Strategies,
IEEE DOI BibRef 9000

Draper, B.A., Hanson, A.R., and Riseman, E.M.,
Learning Knowledge-Directed Visual Strategies,
DARPA92(933-940). More on the use of Schemas. BibRef 9200

Brolio, J., Draper, B.A., Beveridge, J.R., and Hanson, A.R.,
ISR: A Database for Symbolic Processing in Computer Vision,
Computer(22), No. 12, December 1989, pp. 22-30. BibRef 8912
And: COINSTR-89-111, November 1989. Issues of the database for the VISIONS system. BibRef

Draper, B.A., Beveridge, J.R., Brolio, J., Hanson, A.R., Heller, R., and Williams, L.R.,
ISR2: User's Guide,
COINSTR 90-52, UMass., July 1990. New version of ISR database system for VISIONS system. BibRef 9007

Draper, B.A., Hanson, A.R., and Riseman, E.M.,
ISR3: A Token Database for Integration of Visual Modules,
DARPA93(1155-1161). BibRef 9300

Draper, B.A., Kutlu, G., Riseman, E.M., Hanson, A.R.,
ISR3: Communication and Data Storage for an Unmanned Ground Vehicle,
IEEE DOI BibRef 9400

Kutlu, G., Draper, B.A., Moss, J.E.B., Riseman, E.M., Hanson, A.R.,
Persistent Data Management for Visual Applications,
ARPA96(1519-1524). How to manage persistent objects
PS File. BibRef 9600

Reynolds, G., and Beveridge, J.R.,
Searching for Geometric Structure in Images of Natural Scenes,
DARPA87(257-271). BibRef 8700
And: COINSTR-87-03, January 1987. Generate basic groupings based on transitive relations of parallel, corners, collinear, etc. The clusters are then used for more analysis. Schema are used to describe the objects and their appearance in the image. BibRef

Belknap, R.[Robert], Riseman, E.M., Hanson, A.R.,
The Information Fusion Problem and Rule Based Hypothesis Applied to Complex Aggregations of Image Events,
CVPR86(227-234). BibRef 8600
And: DARPA85(279-292). BibRef
And: COINS-TR-85-39, December 1985. Data Fusion. This year's paper on the overall system. Edges and lines which support the same interpretation. BibRef

Reynolds, G.[George], Irwin, N.[Nancy], Hanson, A.R.[Allen R.], and Riseman, E.M.[Edward M.],
Hierarchical Knowledge-Directed Object Extraction using a Combined Region and Line Representation,
DARPA84(195-204). BibRef 8400
And: CVWS84(238-247). Application, Cartography. Cartography application using very large images. The program transforms the detailed pixel level representation into intermediate general representations that combine regions and lines. Only some areas are analyzed, based on the location of probable features by the coarse analysis. BibRef

Weymouth, T.E., Griffith, J.S., Hanson, A.R., and Riseman, E.M.,
Rule Based Strategies for Image Interpretation,
DARPA83(193-202). BibRef 8300
And: AAAI-83(429-432). Interpretation system, use knowledge about the appearance of objects (houses) and how they combine to interpret the outdoor scenes. BibRef

Weymouth, T.E.,
Experiments in Knowledge-Driven Interpretation of Natural Scenes,
IJCAI81(628-630). BibRef 8100

Williams, T.D., and Lowrance, J., Hanson, A.R., Riseman, E.M.,
Model Building in the Visions System,
IJCAI77(644-645). BibRef 7700

Williams, T.D., Glazer, F.,
Comparison of Feature Operators for Use in Matching Image Pairs,
ISPDSA83(395-423). BibRef 8300

Draper, B.A., Brolio, J., Collins, R.T., Hanson, A.R., Riseman, E.M.,
Image Interpretation by Distributed Cooperative Processes,
IEEE DOI BibRef 8800

Parma, C.C., Hanson, A.R., Riseman, E.M.,
Experiments in Schema-Driven Interpretation of a Natural Scene,
COINSTR 80-10, 1980. BibRef 8000

Wesley, L.P., and Hanson, A.R.,
The Use of an Evidential Based Model for Representing Knowledge and Reasoning about Images in the VISIONS System,
CVWS82(14-25). BibRef 8200
And: COINSTR 82-29, December 1982. System: VISIONS. Outlines some of the ideas behind Shafer and Dempster to combine evidence. Basically evidence is a pair [support, plausibility], minimum and maximum amount that the evidence confirms the proposition. See also Mathematical Theory of Evidence, A. BibRef

Konolige, K.G., York, B.W.[Bryant W.], Hanson, A.R., Riseman, E.M.,
Between Regions and Objects: Surfaces and Volumes,
IJCAI77(646-647). BibRef 7700

Riseman, E.M., Hanson, A.R.,
Design of a Semanitcally Directed Vision Processor,
COINSTR 74C-1, January 1974. BibRef 7401

Kohler, R.R., Hanson, A.R.,
The VISIONS Image Operating System,
ICPR82(71-74). BibRef 8200

Chapter on Matching and Recognition Using Volumes, High Level Vision Techniques, Invariants continues in
Constraint Based Matching .

Last update:Mar 13, 2017 at 16:25:24