Comparison of different spectral collection methods for nondestructive on-line detection of soluble solids in apple

Abstract. Prior works have been suggestion that VIS/NIR spectroscopy is a useful method for nondestructive on-line detection of SSC in apple. At present, the nondestructive on-line detection methods of apple mainly include transmission and reflection. In order to explore that which one is more suitable for nondestructive on-line detection of SSC in apple, in this paper, two methods of transmission and reflection are used to collect spectra from 105 apples and one apple collects the spectra of four points evenly distributed at the equator. The transmission average spectra and the reflection average spectra are respectively preprocessed by FD, MSC, SNV, Nor and combined with the actual SSC of apples to establish a PLS model. 79 apples as the calibration set and 26 apples as the verification set. The results show that the reflection method adopts SNV pretreatment with the best result, where Rp is 0.9254, and RMSEP is 0.4915%. The transmission method adopts MSC pretreatment with the best result, where Rp is 0.9243 and the RMSEP is 0.3748%. Further, we try to use SPA and CARS to extract characteristic wavelengths, the performance of the model is improved. At the same time, the effects of single-point spectral modeling and average spectral modeling are also compared and average spectral modeling is better Through the above comparison can see that the transmission is better for nondestructive on-line detection of SSC in apple, in spite of reflection is better for the SSC prediction of a certain position of the apple.