Research On Multi-source Heterogeneous Monitoring Data Fusion Algorithm Of Loess Landslide

The frequent occurrence of landslide disasters in China’s loess area has seriously damaged the safety of local people’s lives and property,and caused irreparable losses to the ecological environment and natural resources.Landslide deformation is affected by a variety of influencing factors.Relying only on the monitoring information of a single type of sensor to judge the state of the landslide,the accuracy of the judgment results is reduced to a certain extent.Multi-source data fusion technology can comprehensively analyze each monitoring data and effectively improve the accuracy of landslide state judgment.Therefore,based on the indepth study of loess landslide deformation monitoring theory and data processing method,taking the Dangchuan landslide in Heifangtai,Yongjing County,Gansu Province as an example,through the analysis of multi-source heterogeneous monitoring data for loess landslides,the preprocessing methods for multi-source heterogeneous monitoring data were studied,and two fusion algorithms suitable for multi-source heterogeneous monitoring data of loess landslides are proposed,and the results after fusion are analyzed.The main research contents and results of this article are as follows:(1)This paper introduces the basic theory of multi-source data fusion in the field of landslide deformation monitoring,including the definition,basic principle,time and space of multi-source data fusion,and analyzes the structure,level and some problems of multi-source data fusion.(2)In this paper,the principle and characteristics of five wireless sensors commonly used in landslide monitoring are studied.Taking BD-LX10 displacement meter as an example,the original data stream of this type displacement meter is decoded by using C++ programming language.Through the practical application in the deformation monitoring of Dangchuan landslide in Hefangtai,Yongjing county,Gansu province,the deformation information at the DCF10 monitoring site was obtained in real time,and the early warning information was issued 2 days in advance.(3)Aiming at the problems of inconsistency,incompleteness,noise and differences in multi-source data of landslide monitoring,the preprocessing methods of multi-source heterogeneous monitoring data were studied,including the removal of abnormal data,the completion of missing data,the smoothing of data,and the standardized processing of data.And analyzed using the measured data.The results show that the Pauta criterion and the Chauvenet criterion are superior to the Grubbs criterion in removing abnormal displacement data;the Lagrange interpolation method has the best completion effect on missing displacement data;Simple moving average method and weighted moving average method have better smoothing effect on displacement data than Savitzky-Golay smoothing method;The three methods of Z-Score standardization,Min-max standardization,and Decimal scaling decimal scaling standardization have the same effect on the processing of temperature data.(4)To solve the problem of multi-source heterogeneous monitoring data fusion in loess landslide,a multi-source heterogeneous monitoring data fusion model based on principal component stepwise regression analysis and a multi-source heterogeneous monitoring data fusion model based on BP neural network are proposed.The experimental results show that the curve of principal component stepwise regression model can reflect and predict the variation trend of landslide to a certain extent,and the BP neural network data fusion model can be used in the landslide displacement prediction with multi-source heterogeneous monitoring data.After filtering the environmental factor variables with calculating the correlation and significance of the environmental factor variables and the landslide displacement changes,the determination coefficient of the BP neural network data fusion model can achieve 0.985 and the RMSE can achieve 0.4787 mm.Thus the accuracy of deformation prediction can be effectively improved.