DEM generation from ASTER L1B imagery using FOSS

Question

Is there any (photogrammetric or other) FOSS package to support the generation of a DEM from ASTER L1B products (i.e. combining the spectral bands 3N and 3B)? Are there any practical examples or/and tutorials? It would certainly be preferable if the tools in question integrate easily in (workflows implemented in) well known GFOSS GIS/RS tools.

Relevant Information

Information about ASTER and its products specifications available at:

• ASTER User's Guide - NASA
• ASTER Level-1B Registered Radiance at the Sensor - LP DAAC
• ALGORITHM THEORETICAL BASIS DOCUMENT FOR ASTER DIGITAL ELEVATION MODELS (STANDARD PRODUCT AST14) -- references the PCI software.

Related ASTER data products

• On Demand Digital Elevation Model, AST14DEM
• On Demand Registered Radiance at the Sensor - Orthorectified AST14OTH
• On Demand Digital Elevation Model & Registered Radiance at the Sensor - Orthorectified, AST14DMO

DEM-generation related works

• DEM and orthoimage generation from ASTER L1B images for remote areas
• Comparison of automated digital elevation model extraction results using along-track ASTER and across-track SPOT stereo images
• ASTER DEM performance
• Digital elevation model (DEM) generation and accuracy assessment from ASTER stereo data

Various

• http://www.ipi.uni-hannover.de/fileadmin/institut/pdf/dowman.pdf

Answer 1

Some time has passed, the topic is still relevant since new software became available:

Did anyone try out the NASA Ames Stereo Pipeline (ASP, http://ti.arc.nasa.gov/tech/asr/intelligent-robotics/ngt/stereo/)? The source code is on github.

Furthermore, in the open source software Orfeo Toolbox (OTB, http://orfeo-toolbox.org) they offer "One-click DEM generation" in their Stereo Framework (announcement). In the manual, there is a chapter on "Stereoscopic reconstruction from VHR optical images pair". Did anyone try Orfeo?

Answer 2

Look into the recently developed open source code base for cleaning up ASTER jitter and successively producing DEMs from the ASTER Scenes (MMASTER, Luc Girod). A link to the paper is http://www.mdpi.com/2072-4292/9/7/704/html . Seems to quite effectively eliminate jitter problems.

Answer 3

In NASA's Ames Stereo Pipeline (ASP) latest documentation there is a section dedicated to ASTER describing how to process ASTER Level 1A VNIR imagery. It is also noted that

ASP cannot process ASTER Level 1B imagery, as those images lack camera information.

Essentially, ASTER L1B imagery lacks of the rational polynomial coefficients (RPCs) metadata that eventually enables orthocorrection and other 3D operations. Therefore, it is unlikely that any (other) FOSS tool would be able to generate a DEM for ASTER L1B.

ps- Thanks to http://www.mssl.ucl.ac.uk/imaging/members/jpm.html for the updated hint.

Answer 4

MicMac (https://github.com/micmacIGN) can extract a DSM from ASTER 3B/3N stereo bands. Furthermore, there is a pipeline available that uses full capability of the software and automates the procedure, MMASTER MMASTER .sh bash files are available here: https://github.com/luc-girod/MMASTER-workflows A more detailed description for the MMASTER procedure can be found here: https://www.mdpi.com/2072-4292/9/7/704

I've tried both Ames Stereo and MicMac/MMASTER, and results are better in MicMac. Final products are a 10m max digital surface raster file, and optional orthophotomosaic tif files and point cloud with the orthophotomosaic RGB values.

Answer 5

GRASS GIS, look at r.in.aster:

r.in.aster rectifies, georeferences, and imports Terra-ASTER imagery to current location using gdalwarp, hdf 4, and r.in.gdal, using projection parameters from g.proj. It can import Level 1A, Level 1B, and relative DEM products. The program may be run interactively or non-interactively from the command line. In either case, the user must specify an input *.hdf file name, the type of processing used, the image band to import, and an output GRASS raster map name. The type paremeter can take values of L1A, L1B, or DEM. The band parameter can take values of 1, 2, 3n, 3b, 4-14

see also Open Source Alternatives to Erdas ATCOR

Answer 6

Using QGIS, you can load any file format supported by GDAL (for raster data) and OGR (for vector). You can look at the following tutorial to see how to convert ASTER L1B data to other format using Python.

Anyway, in QGIS simply open your ASTER file as a raster data source and it will allow you to pick the bands you want as layers. Once done, you can explore the Analysis Tools there, but you simply need to export them to tif (in the menu: Raster -> Convert -> Translate). Then merge them:

• in the menu: Raster -> Miscellaneous -> Merge
• select the two new tif in Input
• click "Layer Stack"
• Type your desired output file name
• Click "Load into canvas when finished"

This will give you a single file with 2 bands that can be used to generate DEM. To generate a DEM follow the instruction:

• in the menu: Raster -> Analysis -> DEM
• select the new tif file and the band you want
• select the option you want
• Type your desired output file name
• Click "Load into canvas when finished"

Via command-line if you prefer:

$ gdalinfo path/to/your/file.hdf
$ gdal_translate 'HDF4_EOS:EOS_SWATH:"path/to/your/file.hdf":VNIR_Swath:ImageData3N' 3N.tif
$ gdal_translate 'HDF4_EOS:EOS_SWATH:"path/to/your/file.hdf":VNIR_Swath:ImageData3B' 3B.tif
$ gdal_merge.py -o combined.tif -of GTiff -seperate 3B.tif 3N.tif
$ gdaldem hillshade 3B.tif hillshade.tif -z 1.0 -s 1.0 -az 315.0 -alt 45.0 -b 1 -of GTiff

Note that the 2 gdal_translate command need the 'HDF4....' string to be taken from the output of gdalinfo.