Abstract
The condition of the dermis determines skin elasticity and is an important indicator of the condition and health of the skin. To overcome the limitations caused by the existing invasive or contact measurement methods, we propose a hyperspectral imaging technique for non-contact elasticity analysis. Since the skin layer has different thickness and composition depending on the body part, the surface information of each skin layer and the spectral information of the skin vary depending on the wavelength band. After acquiring hyperspectral images for 10 body parts, we divide the acquired hyperspectral images into three wavelength bands of light and extract seven texture features. We perform correlation analysis and ANOVA analysis to find wavelength bands and texture features that are highly correlated with the elasticity value depending on the body part. It was found that the texture feature types and wavelength bands of hyperspectral images that have a significant correlation with the skin elasticity value differ depending on the body part. In general, the average texture feature was identified as a parameter suitable for non-contact elasticity analysis that can significantly distinguish body parts in all wavelength bands.