Virginia Tech
Browse
1/1
17 files

Detecting deformation hotspots from InSAR derived deformation maps

software
posted on 2024-03-20, 13:09 authored by Ashutosh TiwariAshutosh Tiwari, Manoochehr ShirzaeiManoochehr Shirzaei

This module gives a semi-supervised machine learning based information mining approach to detect deformation hotspots from InSAR derived displacement maps.

The set of programs can be used for post-processing InSAR deformation maps obtained from multi-temporal InSAR processing.

Related paper citation:
Tiwari, Ashutosh; Shirzaei, Manoochehr (2024). A novel machine learning and deep learning semi-supervised approach for automatic detection of InSAR-based deformation hotspots, International Journal of Applied Earth Observation and Geoinformation Volume 126, DOI: [10.1016/j.jag.2023.103611]

Funding

NSF and USGS have provided funding for this work. Authors thank Department of Energy for supporting this study.

History

Publisher

University Libraries, Virginia Tech

Location

Blacksburg

Corresponding Author Name

Ashutosh Tiwari

Corresponding Author E-mail Address

ashutosh22@vt.edu

Files/Folders in Dataset and Description

insar_deformation_hotspots-main Programs p01_time_series_unsupervised_clustering_rts.py: This program perform time series clustering for InSAR time series displacement using time series k-means algorithm p02_Individual_cluster_dbscan_clustering.py: This program performs spatial clustering over individual temporal clusters detected using p01. p03a_LSTM_supervised_classification_dbscan_out.py: Program for supervised classification using LSTM networks with clusters generated from results of p02 acting as training labels. p03b_LSTMP_supervised_classification_dbscan_out.py: Program for supervised classification using LSTM and perceptron networks with clusters generated from results of p02 acting as training labels. p03_plot_dbscan_clusters.m: A program for plotting the resulting clusters from p02. p04_finding_spatial_correlation_lengths.m: Finding spatial correlation lengths for displacement in the study area, to find out a threshold for removing widespread unwanted clusters. p05_spatial_dispersion.m: Finding spatial dispersion of individual clusters generated from p02. Those with large dispersions would be removed. LICENSE: the license appended to this software README.md: guidance on how to use this software requirements.txt: requirements to run this software Data elpx_ll.mat: Coordinates for elite pixels MVel.mat: One dimensional line of sight velocities MVelq.mat: Standard deviation for line of sight velocities Pco_m.mat: Mean coherence R_TS_den.mat: Displacement time series