以北京市东部地区96个潮土土样的土壤参数及对应光谱数据为数据源, 采用连续小波多尺度分析处理与分析。 首先将土壤光谱进行初步处理, 生成小波系数, 其次将土样的有机质含量与小波分解系数开展相关性分析, 提取特征波段, 最后采用特征波段建立预测耕层有机质含量的模型。 结果表明: 经连续小波处理后, 光谱对耕层有机质含量的预测能力明显优于传统光谱变换技术; 经连续小波分解后, 对土壤有机质含量的预测能力随光谱分辨率降低呈先降后升再降的趋势; 连续小波分析算法可提升土壤光谱对有机质含量的估测能力, 与土壤高光谱反射率相比, 基于连续小波变换的土壤有机含量最佳的精度提高19%; 由于光谱分辨率为80 nm建立的模型精度较高, 其R2达到0.632, 这表明在连续小波算法下, 光谱分辨率较低的宽波段数据可用于土壤有机质含量的监测。

In this study, the data sourced from hyperspectral data of 96 tidal soil samples in Miyun, Tongzhou and Shunyi Districts of Beijing are processed and analyzed by means of continuous wavelet multiscale analysis technique. Firstly, the hyperspectral data are decomposed to generate wavelet coefficients and the correlation between the coefficients and soil organic matter content is analyzed, and the characteristic band is selected. Finally, the model to estimate soil organic matter content is constructed by using the characteristic band. The research results show that the estimation of soil organic matter by the reflectivity of soil spectrum is better than that of the traditional spectral transformation technology after continuous wavelet transformation. The ability of estimating soil organic matter by continuous wavelet decomposition decreases first and then increases with the reduction of spectral resolution. The results of continuous wavelet analysis can improve the ability to estimate the content of organic matter by the soil spectrum. Compared with the high spectral reflectivity of soil, the accuracy of soil organic content based on continuous wavelet is improved by 19%. Since the model accuracy is higher when built with the spectral resolution of 80 nm, its R2 reaches 0.632, which indicates that the wide band data can be used for the monitoring of soil organic matter content by using the continuous wavelet technique.