Rapid Detecting Soluble Solid Content of Pears Based on Near-Infrared Spectroscopy

Abstract

N243 detecting of soluble solid content (SSC) of fruits is important for improvement of internal quality control of fruits in global fresh produce markets. Currently, consumers pay more and more attention to fruits' quality and safety rather than appearance. In this paper, a hybrid approach combined with successive projections algorithm (SPA) and Extreme Learning Machine (ELM) was proposed for effective SSC determining of pears based on Near-Infrared (NIR) spectroscopy. SPA was used for variable selection while ELM for model establishment. In addition, ELM model with full spectra and PCA-ELM model (Principal Component Analysis (PCA) was used on spectra for dimensional reduction) were also developed for comparison to explore the robustness of the SPA-ELM model. The results showed that ELM models with variable selection algorithms perform better than ELM model on full spectra, and SPA-ELM model outperformed PCA-ELM model. It is feasible to determine the SSC of pear fruits using SPA-ELM model based on NIR spectroscopy.