Computer stereo vision for three-dimensional object location

Both the geometric and the structural information are used to remove the ambiguous features from two views. The structural information of each view is stored in a relational table in which the features of regions, edge segments and vertices are the primitive elements. A relaxation labelling process is employed to deduce two isomorphic relational tables in which only the consistent features of the two different views are stored. A matching line equation derived from the geometric relationships of two views is used as a constraint function in the relaxation labelling process. The corresponding points in the two views are found by searching along the consistent edge segments in two isomorphic tables using the matching line equation. The three-dimensional location of the object is computed from the extracted corresponding points.

The significance of this study is that a general rationale for three-dimensional object location is presented. Several examples for feature extraction and three-dimensional object location using computer stereo vision are also given.