Optical properties of melon tissues and Monte Carlo simulation of light propagation through melon fruit

Visible/near-infrared (Vis/NIR) spectroscopy has been authenticated for quality evaluation of melon fruit. However, since melon is a multi-layered fruit with large size, the optical properties (OPs) of different tissue layers can influence the performance of the quality evaluation. Therefore, the OPs of different tissue layers (exocarp, green mesocarp and pulp) among different fruit positions (stem, equator and calyx) of 36 'Huanghemi' melons were measured using single integrating sphere measurements and the light propagation through melon fruit was investigated by the Monte Carlo (MC) simulation based on the OPs. Moreover, the relationship of the OPs with microstructure and internal quality properties (soluble solids content (SSC) and moisture content) was explored. In the absorption coefficient (mu a) spectra, five obvious absorption peaks for carotenoids (481 nm), chlorophyll (675 nm) and water (984, 1195 and 1423 nm) were observed. The exocarp had the highest mu a values at 481 nm owing to its orange-yellow peel and the lowest mu a values at 1423 nm, while the mu a values of green mesocarp and pulp had no significant differences at all absorption peaks. The differences in reduced scattering coefficient (mu s') spectra were obvious with a decreasing trend from the outermost (exocarp) to the innermost (pulp) layer and the stem to the calyx position in pulp layer, which mainly related to the different tissue structure. The MC simulation results indicated that the energy along the wavelength was mainly absorbed by the exocarp and green mesocarp, with a 'diagnostic window' appearing in 700-850 nm. As the source-detector distance increased, the mean optical pathlength and attenuation of detected photons in each tissue layer increased, while the contribution fraction to the detected diffuse reflectance signals by the exocarp, green mesocarp and pulp respectively decreased, increased and increased. To acquire more information about the pulp and ensure a good signal-tonoise ratio of the spectral signal, a source-detector distance of 7 mm with a suitable light source power would be a good choice. Furthermore, the best prediction models with good performance for SSC and moisture content were built based on the mu a spectra, with the correlation coefficient of prediction (Rp) of 0.78-0.80, 0.76-0.88 and the root mean square error of prediction (RMSEP) of 0.83-0.92%, 0.70-0.93%. This study was the first attempt to measure the OPs of melon tissues and investigate the light propagation through melon fruit, which would provide a reference to develop nondestructive testing techniques or instruments for melon fruit based on Vis/NIR spectroscopy.