About the Project
Radar Interferometry (InSAR) can provide measurements of surface displacement from Space, with millimetric accuracy. These measurements are used in the natural hazards community, e.g. for earthquake analysis and landslide monitoring, and for monitoring anthroprogenic activities, such as oil and gas extraction, drawdown of underground water storage, and movements of critical infrastructure. A key step in the InSAR processing chain is that of phase unwrapping, which is the estimation of the integer ambiguities that are inherent with any measurement of phase. Although there are several existing algorithms that exist to do this automatically, they all fail in some circumstances, and visual inspection accompanied by manual correction is currently the only failsafe way to achieve this. AI offers a novel way to solve this problem through the application of deep learning algorithms.
Objectives
Recent work on computing depth and surface orientation maps directly from single images is suggestive of an automated solution to unwrapping. Using deep networks with linked pipelines working at different spatial scales, the output maps are gradually refined, with information passing down from coarser to finer scales [Eigen and Fergus, 2015].
You will work with leading scientists at Leeds and SatSense Ltd to:
1) Assemble a database of interferograms that have been correctly unwrapped with human assistance.
2) Generate simulated interferograms based on simple deformation models as a way of increasing the amount of training data available.
3) Develop a deep convolutional network, structured to produce an iterative spatial refinement, and trained on pairs of wrapped and unwrapped interferograms.
Student profile
The student should have a strong background in a quantitative science (e.g. computing, maths, physics, engineering, earth sciences) an interest in earth sciences. Enthusiasm to develop new approaches to solving old problems is an advantage.
for more information visit http://www.nercdtp.leeds.ac.uk/projects/index.php?id=787
References
Chen, C.W. and Zebker, H., 2001. Two-dimensional phase unwrapping with use of statistical models for cost functions in nonlinear optimization, J. Opt. Soc. Am. A 18, 338–351.
Eigen, D. & Fergus, R., 2015. Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-scale Convolutional Architecture. 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, pp. 2650-2658. doi: 10.1109/ICCV.2015.304.
Hooper, A., Zebker, H., 2007. Phase unwrapping in three dimensions, with application to InSAR data, J. Opt. Soc. Am. A, 24, 2737-2747.
Hooper, A., 2010. A Statistical-Cost Approach to Unwrapping the Phase of InSAR Time Series, Proceedings 2009 FRINGE Workshop, Frascati.
Hussain, E., Hooper, A., Wright, T., Walters, R., Bekaert, D., 2016. Interseismic strain accumulation across the central North Anatolian Fault from iteratively unwrapped InSAR measurements, J. Geophys. Res.: Solid Earth, 121, pp.9000-9019.
Continue with Facebook
