Multi-temporal damage assessment of linear infrastructural objects using Dynamic Bayesian Networks

In this paper, a Dynamic Bayesian Network (DBN) is presented which assesses infrastructural objects concerning their functionality after natural disasters. The presented model combines multi-temporal observations from remote sensed images with simulations based on Digital Elevation Models (DEM). The inference in the DBN is established using the sum-product algorithm. The improved performance of DBN is shown compared to simpler pixel-based and topology-based graphical models. The paper shows results of the model assessing roads concerning their trafficability after flooding. In addition, an evaluation of the results with a reference is conducted.