Vis/NIR spectroscopy coupled with wavelength-optimized Monte Carlo simulation for non-destructive SSC detection of apples of different sizes

Soluble solids content (SSC) is a critical determinant of apple quality and visible/near-infrared spectroscopy is commonly employed as a non-destructive method for SSC detection. However, the detection accuracy and its application robustness are interfered by skin and fruit size. In this study, the optical properties (OPs) of apple tissues were characterized by using a double integrating sphere system with inverse adding-doubling (IAD) and SSC prediction models were established based on different combinations of OPs. The light propagation in apples of different sizes was investigated using voxel-based Monte Carlo (MC) simulations at the most SSC-relevant wavelengths. Results indicated that the reduced scattering coefficient ( μ s ′ ) of skin generally exceeded that of flesh within 520–1600 nm (excluding 1400–1520 nm), showing stronger scattering. Among models based on different OPs, the flesh absorption coefficient ( μ a ) model performed best ( R P = 0.930). With competitive adaptive reweighted sampling (CARS) optimization, the integrated skin-flesh attenuation coefficient ( μ t ′ ) model improved most significantly, with R CV increasing from 0.871 to 0.955 using only 24 wavelengths, where 1030 nm and 1390 nm were the most SSC-relevant wavelengths. MC simulations revealed superior photon fluence, photon count, and photon weight at 1030 nm than 1390 nm. Optimal source-detector angles were 75°–90°, 90°–120° and 105°–150° for large, medium and small apple models at 1030 nm, achieving 90%–95% flesh optical contribution, while ≤ 15° was recommended for the highly absorptive 1390 nm band. This study provides theoretical and parametric guidance for developing non-destructive SSC detection system.