Construction of spectral detection models to evaluate soluble solids content and acidity in Dangshan pear using two different sensors

Abstract

It is most desirable to have an reliable and effective nondestructive method to detect soluble solids content (SSC) and acidity (pH) in pear fruit for fruit quality evaluation. In this study, we used the non-imaging spectrometer PSR + 3500 and the imaging spectrometer SOC710E to collect the spectral data of SSC and pH in 'Dangshan' pear (Pyrus spp.) fruit. Their characteristic wavelengths were screened using Competitive Adaptive Reweighted Sampling (CARS) and Successive Projections Algorithm (SPA). Partial Least Squares (PLS) and Random Forest Regression (RFR) were used to establish different SSC and pH detection models. The effects of spectral range and use of the two sensors on the internal quality detection of pear fruit were compared and analyzed. The results show that the data processing process of the non-imaging spectrometer performed better. Taking the characteristic bands screened by CARS for SSC and pH as an input, the SSC and pH detection models, constructed by the RFR modeling method, had the highest accuracy, and the determination coefficients of calibration set and validation set were both above 0.9. The results of this study provide the important theoretical foundation for the development of real-time online fruit SSC and PH detection equipment.