General model of multi-quality detection for apple from different origins by Vis/NIR transmittance spectroscopy

Visible Near-infrared (Vis/NIR) spectroscopy is widely used to evaluate fruit quality due to its fast and non-destructive advantage, but the diffuse reflectance mode only obtains information of the surface, and the traditional model is difficult to meet multi-origin detection in practical applications. A portable Vis/NIR transmittance prototype was designed and developed to acquire Vis/NIR spectra of apple samples from different origins. Soluble solids content (SSC), firmness, pH and vitamin C (VC) content were determined as the internal quality parameter. The partial least square (PLS) model was established by optimizing the best from different spectral preprocessing and feature selection algorithms. The results showed that the competitive adaptive weighted PLS (CARS-PLS) achieved the best prediction performance, with correlation coefficient of prediction (R p ), and root mean square error of prediction (RMSEP) values of 0.940, 0.542 °Brix for SSC, 0.789, 7.018 N for firmness, 0.698, 0.119 for pH, and 0.804, 10.363 mg kg −1 for VC content, respectively. The general model of CARS-PLS was verified by the independent prediction sets from 7 origins. The establishment of general models expanded the prediction range, improved the prediction stability of models between different cultivars, and reduced the complexity of the models through appropriate preprocessing and feature variable selection methods. The development of the general model of different origins and cultivars for predicting the internal quality of apple has a great potential application using Vis/NIR transmittance spectroscopy.