Title: Machine Learning for Error Characterization and Correction in InSAR Satellite Data
Presenting Author: Victor Pankratius
Organization: Massachusetts Institute of Technology

Abstract:
Analyzing InSAR satellite data can be very challenging in the presence of atmospheric errors. This talk presents work in progress on a machine learning approach that aims to analyze individual patches of interferograms and characterize their degree and type of error. In addition to a training set based on real-world labeled data, our approach aims to integrate domain-specific model knowledge to enhance such characterizations and derive potential corrections by automatically synthesizing suitable processing pipelines. The presentation includes an early overview of PyINSAR, an open-source Python package for InSAR data processing that will integrate this type of functionality.