中文摘要：

基于盐渍土修复过程中盐分含量和同步实测光谱数据,通过对原始光谱数据、平滑光谱数据及平滑后的不同变换光谱数据等八种光谱数据集,分别以相关系数的极值和不同相关系数范围两种方法分析其最佳敏感波段范围,深入分析了不同变换下土壤的光谱响应特征。在此基础上,运用偏最小二乘回归方法,以全波段（400-1 650nm）和分析获得的最佳敏感波段建立了基于修复过程的土壤盐含量和光谱反射率的关系模型。结果表明：针对八种光谱数据集,采用两种方法提取的土壤最佳敏感波段,均集中在947.11-949.31,1 340.27,1 394.11,1 419,1 457.81-1 461.31,1 537.68-1 551.39和1 602.32nm;且最佳波段的土壤盐含量反演模型,以模型评价参数的决定系数（R2）和均方根误差（RMSE）,以及赤池信息量准则（akaike’s information criterion,AIC）作为选择最佳模型的标准,均以SGSD（Log R）模型的建模和预测结果比其他光谱变换的模型更为显著。基于全波段的PLSR建模效果总体上稍优于最佳波段的模型,其中以SGSD的预测精度最为突出,其模型的决定系数R2与标准差RMSEP分别为0.673和1.256;基于两种方法获得的最佳波段的PLSR模型与全波段对比在模型精度方面虽有一定差距,但从模型的复杂程度比较,具有模型简单、变量更少及运算量小的特点。该研究可在土壤盐含量及其光谱特征的研究中,为实现土壤盐渍化定量、快速、便捷的监测和检测提供参考。

英文摘要：

In this paper,the soil salt content（SSC）and the associated spectral reflectance were measured and analyzed during the microbial remediation process of saline soil.The two methods including extremums of correlation coefficients and the different ranges of correlation coefficients were used to find the optimal sensitive bands of SSC for eight spectral data sets covering the raw spectral reflectance,the smoothed spectral reflectance and six different pre-processing transformations of spectral data of saline soil.With this basis,partial least squares regression（PLSR）was used to build relational models between SSC and spectral reflectance based on full bands（400-1 650nm）and optimal bands,respectively.The results showed that the optimal spectral bands for eight spectral data sets,concentrated on 947.11-949.31,1 340.27,1 394.11,1 457.81-1 461.31,1 537.68-1 551.39 and 1 602.32 nm.Taking the coefficient of determination（R2）,root mean squared error（RMSE）and akaike’s information criterion（AIC）as criteria to select the best model.For the PLSR predicting models of SSC based on optimal bands from two different ways,the SGSD（LogR）obtained more robust calibration and prediction accuracies than other pre-processing inversion models.Compared with optimal bands,the full bands using PLSR method could obtain better prediction accuracies on the whole.Among all of the eight spectral data sets in full bands,the prediction accuracy of SGSD was the best,the corresponding R2 and RMSEP of the predicted model were 0.673 and 1.256.For the inversion models based on optimal bands,although there was a slight gap in the prediction accuracy with that based on full bands,they also had their own merits：these models were much simpler and thus the reducing model computation and modelling speed were more important than improving prediction accuracy.The results of this study showed that the method had a great potential for diagnosing and monitoring soil salinization quickly and conveniently in researching the relation bet