mgkst1@unix.cis.pitt.edu (Michael G Koopman) (04/19/91)
I am interested in the latest references regarding attempts to define, classify and recognize three-dimensional solid features. I am not interested in image analysis and should not need to be concerned about permuted images, besides isomerism. The problem I wish to address is the classification of solid objects for which a boundary representation, or preferably a constructive solids geometry representation exists. I am aware of the work by K. Spies (1957 Ph.D. Thesis, Technical University Hannover), primarily through the work of Taylan Altan and other principals at Battelle. The concept of 'shape groups' is then one avenue in which I have interest. I am looking for algorithms for the determination of the types of descriptions of solids which people can determine. For example, long and slender, fat and short, long and wide, thin and twisted come to mind as types of classifications (albeit with no apparent formalism in design). Of importance, the application of scaling techniques to reduce the number features (primitives?) in the description without losing the overall shape and the major characteristics of a solid. One of the types of results would be identification of an axisymmetric, or nearly so, solid or solid segment. Also, the determination of local features of a solid is of interest. An article by Hiroshi Sakurai and David C. Gossard (MIT), Recognizing Shape Features in Solid Models, IEEE Computer Graphics and Applications, Sept. 1990 conveys the meaning of what I have identified as "local features" better than I am able. In this case, a combination of facts from a B-rep graph (e.g. solid angles, adjacency) is graphed and such graphs matched to known features. (I have done grave injustice to this work, my apologies). A type of result to which this leads is identification of such features as through holes or sharp edges. Cuningham and Dixon at the University of Amherst and Mark Henderson of the University of Arizona have made progress in these areas. I am interested in any references, comments, "unbending laws of nature", suggestions, etc. which you are willing to share. I will attempt to create a compendium of these ideas for posting to the net. With all of the good, free ideas I expect to collect this compendium of commentaries should need to be compressed and 'ftp'-ed from a server :-) With Thanks, Mike Koopman (<mgkst1@unix.cis.pitt.edu>) MTI 1450 Scalp Avenue Johnstown, PA 15904 USA