知识引导的滑坡监测数据粗差定位与剔除方法

为了避免灾情误判和误报，准确探测和剔除滑坡形变监测数据中的粗差已经成为提高监测数据质量亟待解决的问题。已有方法主要针对单一传感器数据独立处理，且过度依赖数据变化本身的突变-平滑关系，难以有效区分粗差和外界因素突变引起的奇异值。介绍了一种知识引导的滑坡监测数据粗差剔除方法，通过粗糙集属性约简筛选具有相关关系的多源滑坡观测数据，并结合多元统计理论挖掘粗差影响因素间的时空约束关系，利用不同类型滑坡监测数据变化间的相关性规律，将多因素影响下的滑坡形变抽象为多模式的组合，根据不同模式自适应选择多因子模型以此引导卡尔曼滤波模型更新，从而实现滑坡形变监测粗差的定位与剔除。实验证明，该方法不仅能够有效甄别因环境变化引起的突变，并且能显著提高滑坡形变监测数据粗差自适应剔除的准确性、可靠性与智能化水平。

Knowledge-guided Gross Errors Detection and Elimination Approach of Landslide Monitoring Data

In order to avoid the disaster misjudgment and incorrect reports of landslide disaster, the detection and elimination of the gross errors of landslide monitoring data, has become a critical issue for the observational data quality control. The traditional data filtering methods using curve characteristics of single data source, which are limited by the characteristic of mutations-smooth relations and it is also hard to effectively distinguish the gross error and singular value induced by external factors. To overcome these problems, an approach for gross errors detection and elimination guided by landslide knowledge is proposed in this paper. Experimental results prove that more accurate and reliable landslide deformation information can be available. And proposed method can improve the automation and intelligent level of the gross errors detection and elimination for landslide monitoring data.