Multi-model multi-scale filter for landslide estimation

The monitoring and predicting of landslide is very important for geological disaster prevention, and landslide monitoring system based on multi-sensor information fusion can provide a much effective solution. To adapt to different type of observation noise from diverse sensors, and improve the precision of this approach, a multi-model multi-scale Kalman filter method is presented to estimate and predict the deformation of measuring points. By the method, according to the diversity of the landslide movement, multiple models are used to describe the landslide deformation, and the displacements of measuring point are acquired by multiple sensors. To each deformation model, the measurement is decomposed into multiple scales to get relevant estimation by Kalman filter, and multi-scale estimation corresponding to single model is gotten by weighted fusion. By synthesizing the results of multiple models with probabilistic combination, the final optimal estimation of deformation with higher accuracy is obtained. Experimental results show that comparing with the single model and multi-scale Kalman filter, estimation and prediction accuracy of the landslide deformation by method presented is improved.