项目来源：国家自然科学基金面上项目

项目名称：大数据支持下利用局部条件约束的滑坡自动识别问题可靠求解方法研究

项目负责人：钟成副研究员

项目时间：2018/01-2021/12

项目经费：55万元

项目简介：

遥感数据记录的滑坡表面光谱、形态和时间信息，不足以构成识别滑坡的充分条件，应当通过增加约束条件缩小问题的开放性。由于条件限制，相关研究未能系统考察和建立滑坡表面特征发生发展的约束函数，使得可能其指定的研究条件并不具备良好约束力，或者难以利用局部现有条件。此外，多数研究未考虑滑坡内部的异质性，识别误差较大。为此，项目提出一种在大数据支持下充分利用局部条件约束滑坡解构与识别过程的新方法。首先将滑坡解构为要素的组合，并抽取要素表面显著特征；然后通过正反两种推演方式，互相补充和借鉴，构建影响因素与要素表面特征间一般关系模型；最后，利用该模型在局部条件下的解和要素间空间关系，约束基于遥感数据识别要素和重建滑坡的求解过程。力图充分利用大数据和局部可观测条件，探测要素表面特征的分布模式，为滑坡自动识别提供科学、合理的约束条件，提高识别结果的精度和可靠性，为相关研究和应用提供有益的探索和借鉴。

The spectrum, morphology, and time information recorded by remote sensing data isnot sufficient for identifying landslide. Constrain conditions should be added tonarrow the openness of the problem. However, relative researches did notinvestigate and build the rules on the appearance and evolution of landslidesurface features systematically. Consequently, their assumptions could notconstrain the problems well, or their method can not make full use of existingconditions. Besides, most of them did not analyzing the heterogeneity betweenlandslides elements, introducing errors. This study presented a reliable methodfor automatically identifying landslide constrained by local conditions withlandslide big data. First, a landslide was decomposed as elements, and theirdistinct surface features were extracted with Random Forest; then, the generalrelationship between surface feature and their factors was built, in manners ofboth forward and backward derivation; at last, landslide elements could beidentified, constrained by the local solution of the function and the relationship

between elements. This study aimed at improving the accuracy and reliability ofidentifying landslides from remote sensing data, constrained by reasonable modelsof elements’surface features with big data and local conditions. It is believedthe study could be helpful to relative research and application.