Multi-sensor data fusion for measurement accuracy improvement for a landslide monitoring system

This research aims to propose techniques to integrate data from many sources to increase the accuracy of sensor data that can be used for further analysis in landslide monitoring systems. The research will investigate Kalman filtering to increase the accuracy of multi-sensor measurement. In this research, we aim to propose methods for correcting drifted values of target of interest sensor by divided neighboring sensors status into 2 cases: no drifted and all drifted. First, we proposed automatic error covariance adjustment into a nested Kalman filter. This method assumes that all the nearby sensors are in perfect functioning order. This method based on nested Kalman filter, cosine similarity and Euclidean distance. Second, continuing from the previous method, using nested Kalman filters with automatic error covariance adjustment and discrete Kalman filters, we propose the data fusion technique that uses two-layered Kalman filter algorithm. We assume that all the neighboring sensors have drifted values in this approach. We use a two-layered Kalman filter method in each sensor. The discrete Kalman filter is the first layer. The second layer is nested Kalman filter with automatic error covariance adjustment. The experimental results show that the first method reduced drifting value error by 14.85% when compared to no algorithm at all, while the second method reduced drifting value error by 9.57% when compared to the first method.