Aspect Graph Matching -- Ikeuchi

Chapter Contents (Back)
Object Recognition. Matching, Volumes. Aspect Graphs. Matching, Aspect Graphs.

Ikeuchi, K.[Katsushi], Kanade, T.[Takeo],
Modeling Sensors: Toward Automatic Generation of Object Recognition Program,
(sic), CVGIP(48), No. 1, October 1989, pp. 50-79.
Elsevier DOI BibRef 8910
Applying Sensor Models to Automatic Generation of Object Recognition Programs,
Modeling Sensors and Applying Sensor Model to Automatic Generation of Object Recognition Program,
DARPA88(697-710). BibRef
Modeling Sensor Detectability and Reliability for Model-Based Vision,
CVWS87(288-294). Recognize Three-Dimensional Objects. More on the basic bin-picking task, how to model the appearance to aid in generating constraints. See also Towards an Assembly Plan from Observation, Part I: Task Recognition with Polyhedral Objects. BibRef

Sato, K.[Kosuke], Ikeuchi, K.[Katsushi], Kanade, T.[Takeo],
Model Based Recognition of Specular Objects Using Sensor Models,
CVGIP(55), No. 2, March 1992, pp. 155-169.
Elsevier DOI BibRef 9203
Earlier: CADBV91(2-10). Generating and recognizing specular objects using aspect graph techniques. Use a different aspect based on changes in specular reflections. BibRef

Ikeuchi, K.[Katsushi], Kanade, T.[Takeo],
Automatic Generation of Object Recognition Programs,
PIEEE(76), No. 8, August 1988, pp. 1016-1035. BibRef 8808
Towards Automatic Generation of Object Recognition Programs,
CMU-CS-TR-88-138, CMU CS Dept., May 1988. Similar to the above papers. BibRef

Wheeler, M.D., and Ikeuchi, K.,
Sensor Modeling, Probabilistic Hypothesis Generation, and Robust Localization for Object Recognition,
PAMI(17), No. 3, March 1995, pp. 252-265.
IEEE DOI BibRef 9503
Sensor Modeling, Markov Random Fields, and Robust Localization for Reconstructing Partially Occluded Objects,
DARPA93(811-818). BibRef
Towards a Vision Algorithm Compiler for Recognition of Partially Occluded 3-D Objects,
CMU-CS-TR-92-185, CMU CS Dept. Hypothesize and Verify. Region matching of model and image. Develop from the Vision Algorithm Compiler for occluded objects. BibRef

Wheeler, M.D.[Mark Damon],
Automatic Modeling and Localization for Object Recognition,
CMU-CS-TR-96-188, October 1996. BibRef 9610 Ph.D.Thesis BibRef

Gremban, K.D., and Ikeuchi, K.,
Planning Multiple Observations for Object Recognition,
IJCV(12), No. 2-3, April 1994, pp. 137-172.
Springer DOI BibRef 9404
Earlier: CMU-CS-TR-92-146, CMU CS Dept., December 1992. BibRef
Appearance-Based Vision and the Automatic Generation of Object Recognition Plans,
CMU-CS-TR-92-159, CMU CS Dept., July 1992. How to generate the models needed for matching. Generate the model using ananlytical technique and using an appearance simulator to synthesize images. BibRef

Hong, K.S., Ikeuchi, K., and Gremban, K.D.,
Minimum Cost Aspect Classification: A Module of a Vision Algorithm Compiler,
ICPR90(I: 65-69).
IEEE DOI BibRef 9000
And: CMU-CS-TR-90-124, April 1990. BibRef

Ikeuchi, K., and Hong, K.S.,
Determining Linear Shape Change: Towards Automatic Generation of Object Recognition Programs,
CVGIP(53), No. 2, March 1991, pp. 154-170.
WWW Link. BibRef 9103
Earlier: CVPR89(450-457).
And: CMU-CS-TR-88-188, CMU CS Dept., December 1989. Recognize Three-Dimensional Objects. Automatically compile the object descriptions into the recognition programs! BibRef

Ikeuchi, K.,
Generating an Interpretation Tree from a CAD Model for 3D-Object Recognition in Bin-Picking Tasks,
IJCV(1), No. 2, 1987, pp. 145-166.
Springer DOI BibRef 8700
Precompiling a Geometric Model into an Interpretation Tree for Object Recognition in Bin-Picking Tasks,
DARPA87(321-339). CAD. Recognize Three-Dimensional Objects. The 3D model is transformed into a set of general 2D views to allow matching. These views are based on the visible surfaces. First determine which of the groups of views is valid, then determine the precise attitude of the object, then pick it up. BibRef

Ikeuchi, K., and Shirai, Y.,
A Model Based Vision System for Recognition of Machine Parts,
AAAI-82(18-21). Recognize Three-Dimensional Objects. Needle maps (extended Gaussian sphere) are generated for the model at a given viewing angle (determined from the image) and the image based on the shading information (called photometric stereo - just shape from shading). Three light sources are used to eliminate ambiguities in surface orientation - it becomes a table lookup. Attitude (of object) is reduced by analysis of possibilities - only a few stable states. Thus a few models to check. BibRef 8200

Chapter on Matching and Recognition Using Volumes, High Level Vision Techniques, Invariants continues in
3-D Object Recognition Using Invariants .

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