Comparison of Diffuse Reflectance and Diffuse Transmittance Vis/NIR Spectroscopy for Assessing Soluble Solids Content in Kiwifruit Coupled with Chemometrics

Abstract

Kiwifruit, as a climacteric fruit, undergoes rapid ripening and senescence after harvest, making it highly susceptible to softening, rotting, and spoilage. Therefore, monitoring the key quality parameters of kiwifruit, particularly the accurate detection of soluble solids content (SSC), is considered crucial. The performance of two spectral acquisition methods—diffuse reflectance and diffuse transmission—in detecting SSC in kiwifruit was compared. Various preprocessing methods and feature wavelength selection techniques were employed, and regression models were constructed using partial least squares (PLS) analysis. The stability and accuracy of the models were validated through an independent validation set. The results indicated that the spectral data acquired by the diffuse reflectance method, preprocessed using Savitzky–Golay smoothing and combined with competitive adaptive reweighted sampling (CARS), yielded a coefficient of determination (R2) of 0.98 for the prediction set, with a root-mean-square error (RMSE) of 0.66. In contrast, the spectral data obtained by the diffuse transmission method, preprocessed using multiplicative scatter correction and combined with CARS, achieved an R2 of 0.95 and an RMSE of 0.93 for the prediction set. This study demonstrated that both methods were effective for detecting SSC in kiwifruit, with the diffuse reflectance method showing the greater advantage.