Virginia Tech
Browse
Download file
Download file
Download file
1/1
3 files

Deep learning for efficient selection of measurement pixels in multi-temporal InSAR processing

Download all (2.37 GB)
dataset
posted on 2023-06-12, 19:31 authored by Ashutosh TiwariAshutosh Tiwari, Manoochehr ShirzaeiManoochehr Shirzaei

This dataset is an example dataset generated for training the convolutional long short memory network proposed by Tiwari et al. (2021) for measurement pixel selection in MT-InSAR processing. The dataset is generated by processing Sentinel-1 C-band SAR images (frames 79 and 84 of path 48) of the United States east coast from multi-temporal InSAR processing using the WabInSAR software developed by Shirzaei, M (2013). 


The file ph_im.mat contains time series of interforgrams generated from WabInSAR software v5.3.

The dimension is w*h*n, where w=width, h=height and n=number of SAR interferograms


The file elpx.mat contains pixel locations in image and list forms for the selected measurement points after the pixel selection step.

Funding

United States Geological Survey

History

Publisher

University Libraries, Virginia Tech

Location

Blacksburg, Virginia

Corresponding Author Name

Ashutosh Tiwari

Corresponding Author E-mail Address

ashutosh@vt.edu

Files/Folders in Dataset and Description

[Datasets]- Folder containing training datasets for deep learning model. [ph_im]- Interferometric stack of synthetic aperture radar images generated from WabInSAR software [elpx]- File containing pixel labels (elite pixel or not) and locations of these pixels in the SAR image. [WabInSAR_convlstm.jpeg]- Comparison of elite pixel selection results from Wavelet based InSAR (A) and Convolutional long short term memory network for interferometric semantic segmentation (CLSTM-ISS) model (B). The dataset is prepared from Sentinel-1 images of the United States east coast. X and Y axis denote the image coordinates (rows and columns respectively).

Usage metrics

    Geosciences

    Categories

    Keywords

    InSAR analysisDeep learning

    Licence

    CC0 1.0 Universal (CC0 1.0) Public Domain Dedication

    Exports