黄棕壤性水稻土有机质含量高光谱反演研究OACSTPCD

The yellow-brown paddy soil was taken as the research object,and the soil hyperspectral reflectance curve was obtained by the ASD FieldSpec® 4 spectrophotometer.The characteristics of soil organic matter content(SOM)and hyperspectral characteristics in the study area were analyzed.Based on six spectral data,including original spectrum(R),first-order differentiation(FD),second-order differentiation(SD),logarithm of reciprocal(LR),first-orderderivative of reciprocal(FDR)and first-order differentiation of logarithm(FDL),partial least squares regression models(PLSR),support vector machine models(SVM)and BP neural network models(BPNN)were established for SOM content predicting in yellow-brown paddy soil.The accuracy of the model was compared.The results show that:(1)there is a weak correlation between SOM content and original spectral reflectance.After FD treatment,the spectral curve features are prominent,the spectral FD,SD,FDR and FDL transformations can effectively improve the correlation between spectral reflectance and SOM content.(2)SVM and BP models have poor predictive performance for low SOM content(1.98%).Mathematical statistics can help evaluate model accuracy.(3)The SVM model has better predictive performance than PLSR and BPNN models.PLSR.The SVM model for spectral FD transformation has the best predictive performance for SOM content,and R2,RMSE and RPD are 0.902,0.257 and 2.287.The results of this study can provide new model references and technical ideas for the rapid and accurate determination of SOM content in yellow-brown paddy soil.