NIR hyperspectral imaging coupled with chemometrics for nondestructive assessment of phosphorus and potassium contents in tea leaves

By Yujie Wang, Ge Jin, Luqing Li, Y. Liu, Y. K. Kalkhajeh, Jingming Ning, Zheng-Zhu Zhang. 2020

61 citations recorded by the source corpus.

Abstract To date, studies using hyperspectral imaging (HSI) to monitor plant nutrient status have focused mainly on the prediction of water and nitrogen (N) contents, and few have attempted to monitor phosphorus (P) and potassium (K). In this study, we aimed to predict P and K contents in tea plant leaves of five cultivars using HSI coupled with chemometrics. Hyperspectral data within the spectral range of 908.15–1735.68 nm were extracted for 87 leaf samples by mask method. To eliminate noise interference in raw spectral data, different spectral pre-processing methods were tried and compared. Standard normal variate (SNV) showed better model performance than others both for the prediction of P and K contents. Successive projections algorithm (SPA), and regression coefficients (RC) of partial least squares regression (PLSR) model were employed to select the optimal wavelengths. After variable selection, four models of SPA-PLSR, RC-PLSR, SPA-multiple linear regression (MLR), and RC-MLR were established. SPA-MLR models yielded the best performance with correlation coefficients of prediction (RP) of 0.9423 (P), and 0.9168 (K).Whereas, the root mean square errors in prediction (RMSEP) were 0.0927, 0.4941, residual predictive deviation (RPD) of 2.76, 2.50, for P and K, respectively. Our results suggested that HSI combined with chemometrics is a rapid and accurate approach to potentially predict the contents of P and K in tea plants.

Subfields: Analytical Chemistry, Pathology and Forensic Medicine

View Paper: https://www.semanticscholar.org/paper/ba89cbfac5785355b793b503820aa37a37f422cc