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。

Visible-near infrared spectroscopy (Vis/NIRS) is proved to be an effective tool in the prediction of soil properties. Wavelength optimization plays an important role in the construction of Vis-NIRS prediction model. In this article, a total of 130 topsoil samples collected from Guoyang County, Anhui Province, China, were used to establish a Vis-NIRS model for the prediction of organic matter content (OMC) in line concretion black soils. Through comparison, the combined spectral pretreatments of smooth and multiplicative scatter correlation (MSC) were applied to minimize the irrelevant and useless information of the spectra and increase the correlation between spectra and the measured values, and subsequently, SPXY methods were used to select the representative training set. Successive projection algorithm (SPA) and genetic algorithm (GA) were then conducted for wavelength optimization. Finally, the principal component regression (PCR) model was constructed, in which the optimal number of principal components was determined using leave-one-out cross validation technique. Results show that: both SPA and GA can significantly reduce the wavelength and favorably increase the accuracy, especially, GA can greatly improve the prediction accuracy of soil OMC, with Rcc, RMSEP and RPD up to 0.931 6, 0.214 2, 2.319 5, respectively. Conclusively, using appropriate wavelength optimization methods, not only the computational load can be significantly reduced but also the prediction precision can be improved.