Extraction of line drawing features for object recognition

摘要

Geometrical criteria which define viewpoint-invariant features to be extracted from 2D line drawings of 3D objects are described. The extraction of these features forms the initial stage of a generic object recognition system, the Primal Access Recognition of Visual Objects (PARVO) system. In this system, part-based qualitative descriptions are built and matched to coarse 3D object models for recognition. The segmentation and labelling of the constituent parts of an object rely on the 3D properties inferred from the presence of its 2D features. The original motivation for PARVO is Recognition by Components (RBC), a theory of human image understanding from the field of psychology. An overview of the PARVO system is presented first. Next, definitions of the geometrical criteria defining the viewpoint-invariant features are introduced. Finally, examples of results obtained by applying these criteria to a typical line drawing are shown.