Purchase this article with an account.
Bei Xiao; Inferring cloth stiffness from dynamic shape cues. Journal of Vision 2017;17(15):22. doi: 10.1167/17.15.22.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Mechanical properties of non-rigid materials determine how they respond to external forces. Little is understood of how human perceive mechanical properties in comparison to optical properties. Here, we showed that shape could provide powerful cues for estimating bending stiffness of cloth. We rendered cloth with different intrinsic mechanical properties under an oscillating wind force. Observers viewed a triplet of animations and are asked to rate the similarities of the stiffness of the cloth. Using Maximum Likelihood Difference Scaling, we reconstructed the perceptual scale for perceived bending stiffness as a function of physical stiffness. The perceptual scales had a sigmoid shape, which indicated that observers could distinguish stiff cloth from flexible ones. We computed 20 2D shape statistics (Paulun et al. 2014) from the temporal window of interests within which the winds blew directly at the cloth. The final descriptor has 11(number of videos) *20 (number of shape descriptors) dimensions. We first standardized the shape descriptors to z-scores and then calculated the Principle Component Analysis (PCA) in a common 20-dimensional feature space. Within the PCA space, we calculated the Euclidian distance between the cloth with the smallest stiffness and cloths with other stiffness values. The distances were normalized to 0–1 and serve as the predictions. Finally, we calculated the correlation between the predictions and the normalized perceptual score and used the R-value as the fitness measure. The correlation is high (R>0.89) for all observers. Our results suggested that humans might exploit simple shape statistics to estimate bending stiffness of deformable materials.
This PDF is available to Subscribers Only