论文部分内容阅读
应用支持向量机算法对实时土壤光谱数据进行处理,获得了土壤全氮和有机质的回归模型并研究了模型随参数变化的规律。从中国农业大学试验田采集了150个土样,用光谱仪获取了原始土壤样本的近红外光谱,用实验室分析法获取了各样本的全氮和有机质含量。以近红外光谱数据为自变量对2个土壤参数进行了回归建模并评价了算法各参数对模型的影响。研究表明土壤参数适合于全谱支持向量回归。对于土壤全氮,基于小波降噪NIR光谱的SVM回归模型的标定R2为0.9224,验证R2为0.3667;对于土壤有机质,基于原始NIR光谱的SVM回归模型的标定R2为0.9179,验证R2为0.4152;对经k-means聚类分析后的50个样本进行回归建模结果表明,标定R2和验证R2均有提高。
The support vector machine (SVM) algorithm was used to process real-time soil spectral data, and the regression model of soil total nitrogen and organic matter was obtained. The variation of the model with parameters was also studied. 150 soil samples were collected from the experimental field of China Agricultural University. The NIR spectra of the original soil samples were obtained by spectrometer. The contents of total nitrogen and organic matter in each sample were obtained by laboratory analysis. The two soil parameters were modeled by near-infrared spectral data as independent variables, and the influence of each parameter of the algorithm on the model was evaluated. Studies have shown that soil parameters are suitable for full-spectrum support vector regression. For soil total nitrogen, the calibration R2 of SVM regression model based on wavelet denoising NIR spectroscopy was 0.9224, and the validation R2 was 0.3667. For soil organic matter, the calibration R2 of SVM regression model based on the original NIR spectroscopy was 0.9179 and R2 was 0.4152. Regression modeling of 50 samples after k-means clustering analysis showed that both R2 and R2 were improved.