Vigor Detection for Naturally Aged Soybean Seeds Based on Polarized Hyperspectral Imaging Combined with Ensemble Learning Algorithm

To satisfy the increasing demand for soybeans, identifying and sorting high-vigor seeds before sowing is an effective way to improve yield. Polarized hyperspectral imaging (PHI) technology here proposed as a rapid, non-destructive method detecting vigor of naturally aged soybean seeds. First, spectrum 396.1–1044.1 nm was collected automatically extract region interest (ROI). Then, first derivative (FD), Savitzky–Golay (SG), multiplicative scatter correction (MSC), standard normal variate (SNV) preprocessed polarized data (0°, 45°, 90°, 135°) obtained. Finally, seed prediction model based on components such I, Q, U constructed, partial least squares regression (PLSR), back-propagation neural network (BPNN), generalized (GRNN), support vector (SVR), random forest (RF), blending ensemble learning were applied modeling analysis. The results showed that accuracy when using PHI improved 93.36%, higher than technique, with up 97.17%, 98.25%, 97.55% polarization component U, respectively.