Retrieval of high-level data structures from stereo vision dataFunded under: JRC-INDHAZ 2C
An intermediate level representation and processing of partial 3D wire frame data, obtained from edge-based stereo depth reconstruction systems, is described. A double representation schema, based both on CAD winged-edge-like data structures and on frames, is used. A goal-orientated grouping mechanism is also used, combining a data drive bottom-up indexing, mainly in the form of a potential connect graph, and a knowledge-driven top-down grouping, whose methods are distributed along the structural hierarchy of the knowledge-base. The best candidate instances of higher-level geometrical structures are sought using these grouping mechanisms. They are then matched against their template class and, when possible, completed using WFRL, a knowledge-based system capable of verifying and imposing constraints on incomplete wire frame structures.
Bibliographic Reference: Paper presented: 2nd Congress of AI*IA (Associazione Italiana per l'Intelligenza Artificiale, Palermo (IT), Oct. 28-31, 1991
Availability: Available from (1) as Paper EN 36341 ORA
Record Number: 199111101 / Last updated on: 1994-12-02
Original language: en
Available languages: en