Road Damage Extraction from High-Resolution SAR Image Based on GIS Data and Bayes Network

Synthetic aperture radar (SAR) permits all-time, all-weather observation and strong penetration, and is widely used in disaster monitoring and evaluation, resource exploration, and military reconnaissance. However, speckle noise seriously affects the quality of SAR images, thus limiting the use of SAR image for quick access to information. In this paper, a new road damage extraction method for high-resolution SAR images based on GIS data and Bayes network is proposed. Guided by GIS data, suspected damaged roads are extracted using the fusion of level-set segmentation and an improved D1 line detection. A Bayes network is applied to further confirm real damaged roads based on multi-evidence and the observed values from suspected damage road, to eliminate the false-alarms from the SAR images. Experimental results indicate that our proposed method can extract road damage information quickly and accurately.