Volumetric, surface and contour-based models of object recognition
Figure 7.1 Volumetric primitives used to build simple 3-component objects. There are two possible top (A,D), two possible bottom (C,F), and6 possible middle (I,B,G,J,E,H) components.
In this chapter we summarize recent research from three separate studies (Christou et al, 1990; Wallace and Brodie, 1990; Brodie et al., 1991) to examine issues arising from component based models of object recognition. In Experiments 1 and 2 we consider the role of unary shape and pairwise relational geometry in a componential model. In Experiment 3, we examine the process of inference from 2D contour to three-dimensional (3D) volumetric components, and in Experiment 4 we assess the complementary role of surface characteristics.