On-line detection of soluble solids content of apples from different origins by visible and near-infrared spectroscopy

In order to realize fast, on-line, non-destructive testing of the Soluble Solids Content (SSC) of apples from different origins, and to reduce the effect of origin variability on NIR models, a universal model for predicting the SSC of apples from different origins is established. Firstly, the diffuse transmission spectra of Fuji apples from Qixia, Luochuan and Huining are collected with fruit dynamic online detection equipment. Then, 58 characteristic variables are selected and a UVE-PLS universal model for predicting the SSC of apples is established using the Partial Least Squares (PLS) algorithm combined with Uninformative Variable Elimination (UVE). The root mean square errors of single-origin prediction sets and the total-origin prediction set are 0.50~0.74° Brix and 0.63° Brix, respectively, which increase by 23.2%~44.4% and 35.7% respectively compared to the original individual model. Finally, a new external sample set is used to assess the performance of the model, showing a residual prediction deviation of 2.33 and ratios of the predicted values within the error range of ±1.0° Brix and ±1.5° Brix of 85% and 100%, respectively. Experimental results indicate that the establishment of a universal model for on-line detection of the SSC of apples, including those from multiple origins can improve the robustness of predicting the SSC of the samples from other origins. The results also show that an appropriate wavelength screening method can simplify the model. The development of a common model for the internal quality of fruit from different origins has strong potential for applications in wavelength-limited spectroscopy equipment.