autoRIFT (autonomous Repeat Image Feature Tracking)

A Python module of a fast and intelligent algorithm for finding the pixel displacement between two images

autoRIFT can be installed as a standalone Python module (does not support radar coordinates) where both manual and conda installs (https://github.com/conda-forge/autorift-feedstock) are supported or with the InSAR Scientific Computing Environment (ISCE: https://github.com/isce-framework/isce2) software that supports handling Cartesian and radar coordinates

Use cases include all dense feature tracking applications, including the measurement of surface displacements occurring between two repeat satellite images as a result of glacier flow, large earthquake displacements, and land slides

autoRIFT can be used for dense feature tracking between two images over a grid defined in an arbitrary geographic Cartesian (northing/easting) coordinate projection when used in combination with the sister Geogrid Python module (https://github.com/leiyangleon/Geogrid). Example applications include searching radar-coordinate imagery on a polar stereographic grid and searching Universal Transverse Mercator (UTM) imagery at a specified geographic Cartesian (northing/easting) coordinate grid

Copyright (C) 2019 California Institute of Technology. Government Sponsorship Acknowledged.

Link: https://github.com/leiyangleon/autoRIFT

Acknowledgement:

This effort was funded by the NASA MEaSUREs program in contribution to the Inter-mission Time Series of Land Ice Velocity and Elevation (ITS_LIVE) project (https://its-live.jpl.nasa.gov/) and through Alex Gardner’s participation in the NASA NISAR Science Team

v1.0.7 Updates:

- eased Geogrid output handling (the program will automatically check the optional inputs provided by the user and only generate the meaningful output based on the input combinations)

- bug of grid offset in Geogrid outputs has been fixed

- bug of erroneous large values (due to radar geometry's loss of sensitivity to geocode pixel displacement to velocity at large surface slopes) in Geogrid outputs has been fixed

- autoRIFT was successfully compiled with OpenCV v3 and OpenCV v4

- netcdf output packaging (not publicly available at this moment) for autoRIFT final products is significantly improved, where error metrics are calculated at the stable surfaces wherever available, and radar range-projected velocities onto a priori flow (e.g. surface slope parallel) are also provided when ionosphere-induced azimuth error are much larger.