基于优化高光谱特征的低温胁迫下冬小麦SPAD值估算模型OA

[Objective]This study aimed to investigate the mechanism of low-temperature stress on chlorophyll content in winter wheat and its spectral response patterns.[Method]Based on spectral data from winter wheat leaves before and after chilling treatment,we systematically analyzed the correlation characteristics of raw spectra,smoothed spectra,first-order derivation spectra,and Hilbert transformed spectra with chlorophyll content(SPAD values).Four methods,kernel principal component analysis(KPCA),competitive adaptive reweighted sampling(CARS),variable combination population analysis(VCPA),and successive projection algorithm(SPA)were em-ployed to extract characteristic wavelengths.The optimal chlorophyll content prediction model was investigated by constructing three types of estimation models:back-propagation neural network(BPNN),random forest(RF),and least squares support vector machine(LSSVM).[Result]The first-order derivative spectra exhibited a significant positive correlation with SPAD values at 577 nm(r=0.884),while a significant negative correlation at 486 nm(r=-0.878).Among all combinations of the four feature selection methods with the three modeling algorithms,the ran-dom forest(RF)-based model demonstrated optimal performance,achieving a training set coefficient of determina-tion(R2)of 0.925 and root mean square error(RMSE)of 1.662 when combined with first-order derivative spectral bands showing positive correlation,along with a validation set R2 of 0.736 and an RMSE of 3.111.[Conclusion]The random forest algorithm can effectively characterize the response relationship between chlorophyll content and spectral features of winter wheat under low-temperature stress,providing a reliable spectral diagnostic method and a theoretical basis for frost damage monitoring and agricultural disaster mitigation in winter wheat cultivation.