Estimating Chlorophyll Fluorescence Parameters of Rice (Oryza sativa L.) Based on Spectrum Transformation and a Joint Feature Extraction Algorithm

The chlorophyll fluorescence parameter Fv/Fm plays a significant role in indicating the photosynthetic function of plants. existing technical methods used to measure are often inefficient and cumbersome. To realize fast non-destructive monitoring Fv/Fm, this study took rice under different fertilizer treatments measured hyperspectral reflectance information data leaves during whole growth period. Five spectral transformation were pre-process data. Then, characteristic wavelengths extracted by correlation coefficient method (CC) combined with competitive adaptative reweighted sampling (CARS) algorithm. Finally, based on combination from transformations, back propagation neural network (BPNN) models constructed evaluated. results showed that: (1) first derivative transform (FD), multiplicative scatter correction (MSC) standardized normal variation (SNV) could effectively highlight between Fv/Fm. most sensitive bands high coefficients concentrated range 650–850 nm, absolute values highest 0.84, 0.73, 0.72, respectively. (2) CC-CARS algorithm screen number FD, MSC, SNV pre-treatment 14, 13, 16 which only accounted for 2.33%, 2.16%, 2.66% total wavelength (the full is 601), (3) BPNN established above wavelengths, it was found that MSC-CC-CARS-BPNN had prediction accuracy, its testing set R2, RMSE RPD 0.74, 1.88% 2.46, can provide references pre-processing rapid parameters.