Analysis of Near Infrared Spectra of Apple Soluble Solids Content Based on BP Neural Network

Abstract

Soluble solids content (SSC) is one of the most important factors determining the quality and price of fresh fruits. In the non-destructive testing of near-infrared spectra in apple SSC, BP neural network model was established on the basis of equal-spectral interval using spectral preprocessing method. First derivative, second derivative, standard normal variation (SNV) and multi-scatter calibration (MSC) spectral pretreatment methods were used. The structure of the BP neural network has sixteen inputs, the hidden layers were four layers, and contains five neurons of each layer, established by multivariate scatter correction spectroscopy has the best prediction results when the interval number is sixteen. The correlation coefficient (R) and the root mean square error (RMSE) were 0.9421 and 0.4653, respectively. Results illustrate that in near-infrared quantitative analysis, MSC is used for spectral pretreatment. After evenly dividing the spectral interval, the BP neural network is established, and the prediction accuracy is further improved and the modeling input variables are reduced.