Structured descriptions of complex curved objects for recognition and visual memory.
Ramakant Nevatia · 1975
Abstract : Description and recognition of three-dimensional objects from range data obtained by a laser triangulation technique are described. A complex object is described by decomposition into sub-parts and relations of these sub-parts. The individual parts are described by generalized cones, which are defined by a space curve known as the axis, and arbitrary shaped normal cross-sections along this axis. Techniques for segmenting an object into sub-parts and generating structured, symbolic, graph like descriptions are described. These symbolic descriptions are matched with stored descriptions and the best match is picked for recognition. A limited amount of indexing capability exists to efficiently retrieve a sub-class of similar objects from the models stored in the memory. Indexing is a necessity if a large number of visual models is to be used.