VISUAL INSPECTION OF 3-DIMENSIONAL OBJECTS BY HUMAN OBSERVERS

Citation
T. Niemann et al., VISUAL INSPECTION OF 3-DIMENSIONAL OBJECTS BY HUMAN OBSERVERS, Perception, 25(9), 1996, pp. 1027-1042
Citations number
44
Categorie Soggetti
Psychology, Experimental",Psychology
Journal title
ISSN journal
03010066
Volume
25
Issue
9
Year of publication
1996
Pages
1027 - 1042
Database
ISI
SICI code
0301-0066(1996)25:9<1027:VIO3OB>2.0.ZU;2-E
Abstract
Eye movements are an important aid in active visual exploration of the environment and in performing behavioural tasks. Eye movements might also play a role in human perception of three-dimensional (3-D) object s. Eye-movement strategies were investigated when humans inspected and memorised 3-D objects. Subjects were instructed to memorise the 3-D s tructure of parts of statues of human figures placed on a turntable fr ee to rotate through 360 degrees. Eye movements and turning behaviour were recorded, Different turning and eye-movement strategies could be observed. Subjects showed individual turning behaviours that were repr oducible between trials. Turning strategies ranged from focusing on on ly a limited number of perspective views to almost continuously rotati ng the object with only short stops. On average twelve-thirteen views were inspected during memorising. Eye movements also revealed individu al strategies. Fixation locations within each inspection view ranged f rom either closely spaced on isolated parts of the object to distribut ed over the whole view with large saccades in between. Eye movements w ere often directed to the same details from different perspectives. Th e differences in turning and viewing strategy also resulted in differe nces in the ability to recognise parts of the object later on. In gene ral, successful later recognition required that the subject actually f ixated the part to be recognised. A strategy of thoroughly inspecting the object with a series of closely spaced fixations from only a limit ed number of viewpoints led to best recognition rates. This was especi ally true for two subjects trained in fine arts with prior experiences in modelling. The results support models of viewpoint-dependent objec t recognition with viewer-centred, two-dimensional representations of 3-D objects.