Assessment of multiregion local models for detection of SSC of whole peach (Amygdalus persica L.) by combining both hyperspectral imaging and wavelength optimization methods

Soluble solids content (SSC) is an important quality attribute representing the internal quality of fruits. In this study, the feasibility of different types of local region models used to measure SSC of whole peach by Vis-NIR hyperspectral imaging coupling with effective wavelengths optimization methods was investigated. Three types of local region models namely V-local model, H-local model, and V&H-local model were established based on different data sets (Set-I, Set-II, and Set-III), respectively. For optimizing the models, effective wavelengths were chosen using three types of wavelength selection methods including Monte-Carlo-uninformative variable elimination (MC-UVE), competitive adaptive reweighted sampling (CARS) and random frog (RF), respectively. Model analysis demonstrated that all multispectral local region models have the similar or better performance than the corresponding models with full wavelengths, and RF-PLS model did much better performance than MC-UVE-PLS and CARS-PLS models in terms of each type of dataset. Also, RF-PLS model was tried to measure whole SSC of the single peach using independent samples. Results showed that RF-PLS model could obtain the best SSC assessment ability with r(p) of 0.8469 and RMSEP of 0.4260 based on Set-III and 31 wavelengths. However, as an alternative, V-local model was also useful for SSC assessment of peach. Practical applicationsConventional SSC assessment is usually performed by destructive way and it is very time-wasting. This way is only helpful for sampling several fruits from the whole batch. In a global competitive marketplace, the noncontacting and fast detection of internal quality of fruit is very important for fruit processing factory. Based on the hyperspectral imaging, this study develops the effective multi-region local models for SSC prediction of whole peach. It is meaningful because it could provide the useful references for establishing a more robust multispectral global model used to detect the internal quality of other kind of fruit.