Research on apple origin classification based on variable iterative space shrinkage approach with stepwise regression-support vector machine algorithm and visible-near infrared hyperspectral imaging

Effective classification of apples from different origins is important for horticultural processing and management in mechanized products. To obtain effective characteristic wavelengths and reduce operation time of the classification of apples produced in different areas, a method combining a variable iterative space shrinkage approach with stepwise regression (VISSA-SR) was proposed. The whole apple sphere was chosen as the region of interest for visible-near infrared (Vis-NIR) spectral acquisition. Savitzky-Golay smoothing, standard normalized variable, multiplicative scatter correction, and detrending were used to preprocess the spectral data, with their respective effects compared. VISSA, SR, and VISSA-SR were applied to extract the feature wavelengths. Furthermore, support vector machine (SVM) was utilized in the construction of the identification models. The modeling results demonstrated that a VISSA-SR-SVM model based on detrending spectral preprocessing performed best, exhibiting calibration and prediction accuracy rates of 100% and 97.14%, and a modeling time of 131.96 s. The proposed VISSA-SR algorithm could obtain characteristic wavelength effectively and reduce modeling process operation time. The optimal algorithm combined with Vis-NIR hyperspectral imaging could effectively non-destructively identify apples from different regions. Practical Applications It is important to classify the apples from different regions through nondestructive testing. In order to effectively realize the rapid nondestructive testing of apples produced in different areas, the optimal algorithm combined with Vis-NIR hyperspectral imaging was adopted in this study. VISSA-SR algorithm is proposed to obtain effective characteristic wavelength and reduce operation time. It is proved that VIS-NIR hyperspectral imaging technology is a feasible and effective method to identify apples from different regions.