不同波长选择方法在土壤有机质含量检测中对比研究

由于近红外光谱数据的多重共线性，特征波长选择一直是近红外光谱分析技术的重要研究内容。以108个土壤样本光谱数据和土壤有机质(SOM)含量为研究对象，以连续投影算法（SPA）、间隔偏最小二乘法（IPLS）、竞争自适应重加权采样法（CARS）三种典型的特征波长选择算法进行近红外光谱波长选择和土壤有机质含量建模。研究结果表明，基于上述三种方法提取的特征波长所建立的模型预测能力均优于全谱模型。其中，基于SPA算法的MLR预测模型精度最优，预测集相关系数（Rp）和均方根误差（RMSEP）分别为0970 2和1.214 4，模型参数只有6个。因此，SPA-MLR可以有效地应用近红外光谱的建模，并且简化模型的复杂度，提高模型的计算效率。

Comparison of Different Wavelength Selection Methods in SOM Content Detection

Because of the multicollinearity of near-infrared spectroscopy data, the selection of characteristic wavelength has been an important research for near-infrared spectroscopy analysis technology. Based on spectral data and the content of soil organic matter (SOM) of 108 soil samples, this paper used three typical characteristics wavelength selection algorithm, the successive projections algorithm (SPA), interval partial least squares (IPLS), competitive adaptive reweighted sampling (CARS), for wavelength selection of near-infrared spectroscopy and modeling of soil organic matter content. The results showed that the model based on the characteristic wavelength extracted by the three methods above had better prediction ability than that of the full-spectrum model. Among them, the accuracy of the MLR prediction model based on SPA algorithm was the best, and the correlation coefficient (Rp) and root mean square error (RMSEP) of the prediction set were 0.970 2 and 1.214 4, respectively, with only 6 model parameters. Therefore, SPA-MLR could effectively apply near-infrared spectroscopy modeling, simplify the complexity of the model, and improve the computational efficiency of the model.