论文部分内容阅读
可见光/近红外光谱模型是土壤属性预测的有效工具。波长优选在光谱建模过程中起着重要作用。文中首先利用从安徽省涡阳县采集的130个砂姜黑土土壤样本获得可见光/近红外光谱,然后利用平滑与多重散射校正联合的光谱预处理方式消除光谱中的无关变量和冗余信息以提高模型预测结果的相关性,再利用SPXY方法挑选建模集样本,分别利用连续投影算法和遗传算法进行波长优选,最后利用留一法进行交互验证建立有机质含量的主成分回归模型。研究结果显示:连续投影算法和遗传算法都可以有效地减少参与建模的波长数并提高模型的准确度,尤其是遗传算法能够更好地提高土壤有机质含量预测精度,其相关系数、预测均方根误差和相对分析误差分别达到0.931 6,0.214 2和2.319 5。通过合适的特征波长选取,不仅计算量可以大大减少,预测精度也会有效提高。
The visible / near infrared spectroscopy model is an effective tool for predicting soil properties. Wavelength optimization plays an important role in spectral modeling. In this paper, the visible / near infrared spectra of 130 samples of sago clinker soil collected from Guoyang County, Anhui Province were firstly used. Then the uncorrelated variables and redundant information in the spectra were removed by the combination of smoothing and multiple scattering correction pretreatment Then the SPXY method was used to select the modeling set samples. The continuous projection algorithm and the genetic algorithm were used to optimize the wavelength respectively. Finally, the principal component regression model of organic matter content was established by using the left-behind method. The results show that continuous projection algorithm and genetic algorithm can effectively reduce the number of wavelengths involved in modeling and improve the accuracy of the model, especially genetic algorithm can better improve soil organic matter content prediction accuracy, the correlation coefficient, the prediction of mean square The root and relative analysis errors were 0.931 6, 0.214 2 and 2.319 5, respectively. By selecting the appropriate characteristic wavelength, not only the amount of computation can be greatly reduced, but also the prediction accuracy can be effectively improved.