基于小波变换与偏最小二乘耦合模型估测北方潮土有机质含量

基于北京市通州、顺义两区52个潮土样品高光谱数据，利用离散小波多尺度分析技术对其进行处理分析。首先将光谱按六种尺度进行分解，然后将各尺度分解数据与土壤有机质含量进行相关性分析，并筛选敏感波段，最后利用偏最小二乘法构建土壤有机质含量估测模型。结果表明：土壤光谱反射率经小波变换后，在参与建模的特征波段中，近红外波段居多，即近红外波段估测有机质含量的贡献高于可见光波段;低频信息对有机质含量的估测能力优于高频信息;高频信息对土壤有机质含量的估测精度随光谱分辨率降低而降低;与常用光谱变换算法相比，小波变换分析法在一定程度上提高了土壤光谱对有机质含量的估测能力，其低频信息与高频信息构建的最优模型预测精度均较高，低频信息的R2=0.722，RMSE=0.221，高频信息的R2=0.670，RMSE=0.255。

Estimation of Organic Matter Content of North Fluvo-Aquic Soil Based on the Coupling Model of Wavelet Transform and Partial Least Squares

For improving the estimation accuracy of soil organic matter content of the north fluvo-aquic soil, wavelet transform technology is introduced. The soil samples were collected from Tongzhou district and Shunyi district in Beijing city. And the data source is from soil hyperspectral data obtained under laboratory condition. First, discrete wavelet transform efficiently decomposes hyperspectral into approximate coefficients and detail coefficients. Then, the correlation between approximate coefficients, detail coefficients and organic matter content was analyzed, and the sensitive bands of the organic matter were screened. Finally, models were established to estimate the soil organic content by using the partial least squares regression (PLSR). Results show that the NIR bands made more contributions than the visible band in estimating organic matter content models; the ability of approximate coefficients to estimate organic matter content is better than that of detail coefficients; The estimation precision of the detail coefficients fir soil organic matter content decreases with the spectral resolution being lower; Compared with the commonly used three types of soil spectral reflectance transforms, the wavelet transform can improve the estimation ability of soil spectral fir organic content; The accuracy of the best model established by the approximate coefficients or detail coefficients is higher, and the coefficient of determination (R2) and the root mean square error (RMSE) of the best model for approximate coefficients are 0.722 and 0.221, respectively. The R2 and RMSE of the best model for detail coefficients are 0.670 and 0.255, respectively.