Mapping of Chlorophyll and Spad Distribution in Pepper Leaves During Leaf Senescence Using Visible and Near-Infrared Hyperspectral Imaging
K. -Q. Yu,Y. -R. Zhao,F. -L. Zhu,X. -L. Li,Y. He
DOI: https://doi.org/10.13031/trans.59.10536
2016-01-01
Transactions of the ASABE
Abstract:Hyperspectral imaging covering the visible and near-infrared region (420-1000 nm) was used to determine the spatial distribution of chlorophyll and SPAD (soil and plant analyzer development) in pepper leaves during leaf senescence. Hyperspectral images of pepper leaves from three positions (upper, middle, and lower) of the plant were acquired, and their corresponding chlorophyll a, b, (a+ b), and SPAD values were measured using chemical analysis and a SPAD meter, respectively. Spectral reflectance values of the samples were then extracted from the hyperspectral images based on a leaf mask using image segmentation. Chlorophyll a, b, (a+ b), and SPAD in leaves from the upper, middle, and lower positions were analyzed by one-way analysis of variance (ANOVA). Microstructure images of pepper leaves from the upper and lower positions are shown to illustrate the chlorophyll changes in mesophyll cells. Effective wavelengths (EWs) carrying masses of valuable information were then selected by the regression coefficients (RCs) of partial least square regression (PLSR) models. As a result, five wavelengths (503, 551, 690, 717, and 770 nm) were selected as EWs for chlorophyll a, b, and (a+ b) prediction, and ten wavelengths (517, 536, 545, 555, 584, 603, 610, 648, 699, and 728 nm) were selected for SPAD prediction. After that, PLSR predictive models were established using full spectra and the selected EWs. PLSR models established by EWs with coefficients of correlation for prediction (R-p) of 0.779, 0.743, 0.770, and 0.907 were considered optimal models for predicting chlorophyll a, b, (a+ b), and SPAD, respectively. Finally, these optimal models were used to estimate the chlorophyll a, b, (a+ b), and SPAD of each pixel within pepper leaf images for generating spatial distribution maps. The results indicated that hyperspectral imaging combined with chemometric analysis could determine chlorophyll and SPAD values in pepper leaves during leaf senescence, which provides a reference for monitoring plant growth.