Hyperspectral Inversion Model of Relative Heavy Metal Content in Pennisetum sinese Roxb via EEMD-db3 Algorithm

Detection rapidity and model accuracy are the keys to hyperspectral nondestructive testing technology, especially for Pennisetum sinese Roxb (PsR) due its extremely high adsorptive heavy metal content. The study of resolution PsR is conducive analysis accumulated content in different parts. In this paper, contents Cd, Cu Zn old leaves, young upper stem, middle stem lower as well data corresponding parts, were measured simultaneously both fresh dry states. To begin, spectral preprocessed by using Ensemble Empirical Mode Decomposition-Daubechies3 (EEMD-db3), Savitzky–Golay (SG), Symlet3 (sym3), Symlet5 (sym5), multiplicative scatter correction (MSC). 40 samples divided into 32 training sets 8 validation sets. transformed first derivative (FD) reciprocal logarithm (log(1/R)) highlight singularities binary wavelet decomposition. After screening significant bands from correlation curve, competitive adaptive reweighted sampling (CARS) successive projection algorithm (SPA) applied extract characteristic variables, which used establish partial least-squares (PLS) regression multiple stepwise linear (MSLR) inversion models Cu, contents. Based on EEMD-db3 pretreatment, (fresh) state had R2 values 0.884 (0.880), NRMSE 0.179 (0.253) RPD 3.191 (3.221), indicating excellent stability predictive performance. findings can not only aid rapid detection adsorption various parts PsR, but also be guide development use animal feed.