Abstract
An important issue in object recognition concerns the nature of shape primitives used to represent 3D objects. One hypothesis is that object shape is represented in terms of its volumetric components and their spatial configuration. Another hypothesis predicts that 3D shape is represented in terms of regions of edge contour corresponding to surfaces. Recently Leek et al (2005) showed that there is a performance advantage in whole-part matching for configurations of edge contour that corresponds to surfaces or volumetric components as opposed to nonvolumetric shape regions. In this study we examined the role of viewpoint and the effect of viewing time on whole-part matching performance. There were three types of shape primitives: volumetric components, surfaces and non-volumetric regions in the representation of 3D object shape. In two experiments, participants were required to match object parts to novel surface-rendered 3D objects. In Exp. 1 the parts were presented either in the same or a plane-rotated orientation relative to the whole object. Volumetric and surface parts led to faster matching than contour parts for both orientations. There was no difference in matching between volumetric and surface parts in either orientation, suggesting that volumetric effects do not arise even under viewpoint changes. In Exp.2 the whole object were presented for 1,200, 2,000 or 5,000 msecs. The results showed matching benefits for volumetric and nonvolumetric parts over edge contour parts in all three viewing times. The results are discussed in terms of the volumetric and nonvolumetric theories of 3D object shape representation.