625 USING LINE INVARIANTS FOR OBJECT RECOGNITION BY GEOMETRIC HASHING F. Tsai, February 1993 Geometric Hashing is a model-based object recognition technique for detecting objects which can be partially overlapping or partly occluded. It precompiles, from local geometric features, redundant transformation-invariant information of the models in a hash table and uses the invariants computed from a scene for fast indexing into the hash table to hypothesize possible matches between object instances and object models during recognition. In its simplest form, the geometric hashing method assumes relatively noise-free data and is applied to objects with points as local features. However, extracting of the locations of point features is inherently error-prone and the analysis of geometric hashing on point sets shows considerable noise sensitivity. Line features can generally be extracted with greater accuracy. We investigate the use of line features for geometric hashing applied to 2-D (or flat 3-D) object recognition and derive, from a combination of line features, invariants for lines under various geometric transformations.