In order to effectively control the saline soil, taking Songliao Basin for example, the Environmental Mitigation Satellite (HJ-1 A) hyperspectral data was used in this study. The most suitable quantitatively retrieve model of saline soil was selected by comparing the forecast results of the salt-bearing rate content retrieved by curvilinear regression and least squares support vector machine (LS-SVM) regression. Ultimately, LS-SVM regression was chosen to retrieve various saline soil indexes in Daqing where the soil was salinized seriously. The retrieve results were classified into several grades by binary decision tree. The results showed that, it was convenient and effective to acquire the saline soil information by using HJ-1 A. The accuracy of retrieve model based on LS-SVM was high. The saline soil grade classification, which was calculated by binary decision tree using the RS technology, was accurate and reliable. Soil salinization of Daqing was serious, most of that was alkali soil. The area of light, medium and server alkali soil was separately 345.03, 1 389.03, 869.94 km2, respectively. The research has great significance for saline soil rapid extracting and prevention in Songliao Basin.%为了进行盐碱地的有效防治,以松辽盆地为例,基于环境减灾卫星(HJ-1A)高光谱数据,对比曲线回归、最小二乘支持向量机回归二种非线性回归模型在含盐率反演中的预测效果,探索该区土壤盐碱化指标定量反演的最佳模型,最终采用最小二乘支持向量机(LS-SVM)回归预测的方法,在盐碱化较严重的大庆地区进行了多种盐碱地指标反演,并采用决策二叉树方法对试验区盐碱地进行了等级划分.结果表明:基于环境减灾卫星可以方便有效地对盐碱地进行信息提取；基于最小二乘支持向量机的反演模型精度最高；在遥感支持下采用决策二叉树的方法可以有效地对盐碱地进行等级划分,结果准确可靠；研究表明大庆地区盐碱化现象严重,绝大部分为碱化土,其中轻度、中度、重度碱化地面积分别为345.03、1389.03、869.94km2.该研究对盐碱地信息的快速提取与盐碱地的防治具有重要意义.
展开▼
