Displacement updates in dynamic areas

点击翻译按钮获取中文摘要

Geological displacements can be caused by natural or man-made events, e.g., earthquakes, volcanic erruptions and mining activities. Interferometric Synthetic Aperture Radar (InSAR) technology, based on the analysis of microwave signals reflected back from the earth surface, is able to compute ground displacements with millimeter precision. In the past, researchers have applied computer vision, signal processing and other traditional algorithms to analyze InSAR. Our industrial partner 3vGeomatics Inc's optimization algorithm PtSel outperforms related work. However, the rapidly increasing volume of InSAR data has overloaded the processing pipeline. The amount of undesirable noise has also increased and affected the accuracy of displacement computation. Since July 2017, 3vG has been working with the University of Alberta (UA) team. We have introduced novel deep learning techniques to significantly reduce the amount of data required, while preserving similar accuracy. The goal of this CRD proposal is knowledge translation and operationalization by generalizing our findings on displacement updates in dynamic areas. Our three-year research plan aims to preserve high accuracy using reduced number of images (reducing from 12 to 6 to 3 to 1 month of data). Extreme weather conditions have caused more frequent occurrences of geological changes. Early prediction of ground displacements can help in preparation before disasters happen and thus save lives. The economic and technological impacts not only benefit Canada, but also the global community, and will strengthen Canada's leadership in the monitoring of geological activities.