Crop chlorophyll detection based on multiexcitation fluorescence imaging analysis

The chlorophyll content of wheat was assessed using multispectral fluorescence imaging (MSFI). Ultraviolet (UV) light (365 nm)-induced fluorescence images at 440, 520, 690, and 740 nm, and visible light (460, and 610 nm)induced fluorescence images at 690 and 740 nm were acquired while leaf chlorophyll content was measured using SPAD 520. The fluorescence images were processed after segmentation and channel extraction to calculate the parameters of each leaf based on fluorescence images (Fu440, Fu520, Fu690, and Fu740) obtained by UV excitation, and fluorescence images (Fu440, Fu520, Fu690, Fu740, Fb690, Fb740, and Fr740) obtained by three excitations of 365 nm, 460 nm, and 610 nm light. 12 fluorescence ratio parameters under UV excitation and 26 fluorescence ratio parameters under three excitations were calculated. The correlation analysis revealed that the fluorescence parameters (Fr740, Fu440, Fu520, Fu690, Fu740, Fb690, Fb740, Fu440/Fu520, Fu520/Fu690, and Fu740/Fr740) showed a strong correlation with the chlorophyll content. These parameters have the potential to measure the chlorophyll content. Subsequently, stepwise regression analysis (SRA) was employed to screen 16 fluorescence parameters under UV excitation and 33 fluorescence parameters under three excitations, with the objective of identifying and eliminating redundant variables. Finally, four variables (Fu520, Fu690, Fu740, and Fu690/Fu520) under UV excitation and five variables (Fr740, Fu520, Fb740, Fu740/Fu690, and Fb740/Fb690) under three excitations were selected. The partial least squares regression (PLSR) model, constructed using three excitations, demonstrated enhanced performance with an R2 c of 0.901, R2v of 0.904, root mean square error (RMSE) of calibration of 4.398, and RMSE of validation of 4.267. Multiexcitation fluorescence based on three excitations techniques has better performance for evaluating chlorophyll content.