Given m viewpoints and n features, S is a rectangular similarity matrix whose generic i-th line is

where v,r,c and s are the four attributes of the i-th viewpoint defined above. Agglomerative hierarchial clustering is represented in the form of a dendogram which shows the progressive agglomeration of views as the correlation coefficient increases. As an example, Figure 11, below, shows the dendogram obtained from 80 views of the optical stand shown in Figure 12.

Five Characteristic Views (CVs) corresponding to a threshold level of 0.840 are shown in Figure 12 ( views 18, 47, 53, 73, and 50). Broadly speaking, these 5 views are sufficiently different to avoid unnecessary duplication of the matching process, yet show sufficient simultaneously visible combinations of primitives to allow location of the object in any arbitrary view.