Object Recognition with Stereo Vision and Geometric Hashing

The subject of this thesis is the automatic recognition of objects from digital images. The discussion is restricted to recognition of man made objects that can be described by deterministic, structural models. Applications of this kind of recognition tasks can be found in industry. Object recognition systems are for example used to inspect products and to equip robots with a detection system that allows them to find and handle objects. The thesis discusses a recognition system that is based on a combination of stereo vision and geometric hashing. This combination enables recognition of 3-D objects in a straightforward and relatively simple manner. The recognition relies on local features detected from the images. The stereo images allow these 2-D features to be combined to 3-D features. The 3-D features are matched to 3-D object models using a 3-D variant of the geometric hashing algorithm.