A Study On The Surface Subsidence Prediction Of Changchun City Based On Kalman Filter

The rapid development of urbanization in China has also brought about a large number of urban surface subsidence problems,especially in large cities.As the capital of Jilin Province and the geographic center of Northeast China,Changchun has been rapidly urbanized,and disasters such as building subsidence and ground subsidence have also occurred,seriously threatening people's daily life and production safety.Due to the slow process of surface subsidence,it is not easy to be found in a short period of time,and once the surface subsidence occurs,the affected area is huge,and it is difficult to restore the subsidence area to its original appearance by artificial means.Therefore,the monitoring and early warning of surface subsidence is particularly important for disaster monitoring and forecasting.Synthetic Aperture Radar Interferometry(InSAR)technology solves the problems of low spatial resolution,small monitoring area and large workload in the subsidence monitoring process of leveling and Global Navigation Satellite System(GNSS)measurements.Differential interferometry is easily affected by many factors,and has certain limitations in application.PS-InSAR technology selects SAR images covering the same area and the same area,and uses the time series of images and the threshold of amplitude dispersion index to identify permanent scatterers.This method can maintain the scattering characteristics,overcome the shortcomings of differential synthetic aperture radar,and can Monitoring long-term regional subsidence processes with millimeter-level accuracy is widely used in urban surface subsidence monitoring,mine surveying,and foundation pit monitoring.Kalman filter uses random estimation theory to describe the state of objects containing noise at different times,establishes relationships and uses process and observation noise for filtering calculations.Kalman filter can estimate multidimensional noisy system and realize application on computer.Linear Kalman filtering,extended Kalman filtering and unscented Kalman filtering can be combined with PSInSAR technology for large-area prediction of settlement data obtained from monitoring,which overcomes common settlements such as regression analysis method,time series analysis method,grey model method and artificial neural network.The prediction model cannot make predictions for large-scale subsidence areas,and the spatial resolution is low.In this paper,the PS-InSAR technology is used to calculate the urban surface subsidence data,and three Kalman filters,namely linear Kalman filtering,extended Kalman filtering and unscented Kalman filtering,are used to model and predict the subsidence data.The prediction results under different filters are compared,and the optimal prediction model and the applicability of Kalman filter in settlement prediction are compared.The main research contents and achievements are as follows:(1)Based on PS-InSAR technology,the surface subsidence within the urban area of Changchun City is calculated by using sentinel-1B data of 50 scenes from 2016 to2020.The results show that most of the ground surface in Changchun City is stable without settlement,the maximum settlement rate is 20.45mm/a,and the average settlement rate is 2.98mm/a.The leveling monitoring value is used to verify the accuracy of the settlement value calculated by PS-InSAR.The verification results show that the average error of the monitoring values obtained by the two monitoring methods is 1.13mm/a and the root mean square error is 3.09mm/a,indicating that the settlement results obtained from the regional SAR image calculated by PS-InSAR technology are highly reliable.(2)The settlement process of PS points in the settlement area conforms to the linear and nonlinear variation law,with high data quality and no strong error fluctuation.It is suitable to use three filtering models: linear Kalman filter,extended Kalman filter and traceless Kalman filter for settlement prediction.(3)Linear Kalman filter,extended Kalman filter and unscented Kalman filter are used to smooth and predict the settlement process of PS point respectively.Three kinds of Kalman filters are implemented based on Matlab platform.The calculation results show that the three filter models have high applicability in the settlement prediction process,the unscented Kalman filter prediction model has the highest accuracy and reliability,and the extended Kalman filter prediction model has a slightly lower accuracy and reliability than the unscented Kalman filter prediction model.The accuracy and reliability of the linear Kalman filter prediction model are lower than the other two filter models.In the process of short-term settlement prediction,the three filter settlement prediction models have good accuracy and reliability.The extended Kalman filter and the unscented Kalman filter settlement prediction model can carry out long-term settlement prediction.When the prediction time reaches more than half a year,the three A large number of accidental errors have begun to appear in various filtering models,and the predicted results have begun to have no reference value.