The Nondestructive Model of Near-Infrared Spectroscopy with Different Pretreatment Transformation for Predicting “Dangshan” Pear Woolliness Disease

The “Dangshan” pear woolliness response is a physiological disease that mostly occurs in the growth process. appearance of not obvious, and it difficult to detect with naked eye. Therefore, finding way quickly nondestructively identify great significance. In this paper, near-infrared spectral (NIR) data samples were collected at 900–1700 nm reflectance spectra using handheld miniature NIR spectrometer, modelled analysed random forest (RF), support vector machine (SVM) boosting algorithms under processing 24 pretreatment methods. Considering variations between different methods, work determined relative optimality index methods by evaluating their effects on model accuracy Kappa selected best-performing first derivative standard normal variate Savitzky–Golay multiplicative scatter correction as best With method, all five models three categories showed good stability after parameter debugging, F1 greater than 0.8 floating approximately 0.7, reflecting classification ability proving spectroscopy (NIRS) rapid identification was feasible. By comparing performance differences before found ensemble-learning such RF more stringent identifying machines, ensemble learning significantly improved appropriate This experiment provided relatively stable detection method for nonideal conditions analysing impact prediction result.