processing earth observing images with ames stereo...

1
Newly added in our recent 2.3 release of ASP is the utility ‘pc_align’. This utility implements the iterative closest point (ICP) algorithm through the use of libpointmatcher library. This allows users to align their ASP created DEM against previously trusted dense sources like SRTM, USGS NED, or GMTED. This same utility can be used to align a DEM to sparse source like GLAS, LOLA, or MOLA via input from CSV. The example shown right aligns a DEM created from World View 2 imagery (2 m/px) of the Cathedral Range in California to a prior existing USGS NED map (30 m/px). The top left slice is the ASP created DEM while the bottom right is the trusted reference. Processing Earth Observing images with Ames Stereo Pipeline [email protected] http://irg.arc.nasa.gov/ngt/stereo Affine Script: > stereo <image1> <image2> \ <camera1> <camera2> \ <output prefix> \ --alignment affineepipolar Get Ames Stereo Pipeline Aligning Output to a Trusted Source Science Accomplished with ASP Methods of Ingestion Map Projecting Script: > mapproject -t rpc <dem prior> <image1> <camera1> <mimage1> > mapproject -t rpc <dem prior> <image2> <camera2> <mimage2> > stereo <mimage1> <mimage2> \ <camera1> <camera2> \ <output prefix> <dem prior> Script: > pc_align --max-displacement <meters> <dem prior> <asp dem> \ --save-transformed-source-points -o pc/pc > point2dem -tr 2.0 -t_srs "projection" pc/pc-trans_source.tif Script: Reference against geoid > dem_geoid <dem> Reference against ellipsoid > dem_geoid <dem> \ --reverse-adjustment Vertical Datum Support Script: Follow QR code in top right corner or go to: http://irg.arc.nasa.gov/ngt/stereo to find download link for your specific OS. > Download StereoPipeline.tar.bz2 > tar xf StereoPipeline.tar.bz2 > export PATH=$PATH: \ StereoPipeline/bin Read the Manual! > open StereoPipeline/*.pdf or > gvfs-open StereoPipeline/*.pdf Ames Stereo Pipeline (ASP) has for a long time been able to create dense elevation models for Planetary Science on non-terrestrial bodies. Now we’ve adapted our tools for Earth Science and can process imagery from Digital Globe, GeoEye, Astrium, and anything else with an RPC model readable by GDAL. You can download ASP to perform your own work from our website by following the instructions bellow. ASP is Apache 2 licensed (free as in beer) and operates on most forms of Linux and OSX. It is suitable for operation on your computer, the head-less server, or your university cluster. Processing satellite imagery into a point cloud is performed with ASP’s ‘stereo’ command. It requires two images with different perspectives in NTF or any other image format. It also accepts XML metadata that describes a rigorous camera model that is often shipped with Digital Globe data. The stereo command will by default try to correlate, match all pixels between images, without modifying the input imagery. However that can take a lot of time and will possibly fail if there is a rotation or scale difference between images. Instead it is recommended that the input images be epipolar rectified using the “--alignment affineepipolar” option. An example of its effect is shown right. Alternatively, for difficult stereo pairs, ASP can use prior existing low resolution DEM for map projection. The stereo command is smart enough when generate its point cloud to reverse the projection arithmetic and then triangulate with the more rigorous camera model. By default, all data created by ASP is referenced against an ellipsoid model. For Earth data this is usually WGS84. However this is not ideal for watershed analysis as an ellipsoidal model does not correctly model the changes in gravity across the surface of the Earth. Instead the solution is to use geoid, a lumpy model that describes mean sea level in the absence of weather and tides. ASP provides a utility, ‘dem_geoid’, which can be used to reference data against the commonly used EGM96, NAVD88, or if processing Mars data, MOLA MEGDRA. Our team desires to make elevation products for all imagery. However we don't have the time or experience to apply ASP to all applications. This is why NASA offers ASP completely free. We hope to enable the scientific community access to additional measurements that couldn't be had otherwise due to cost or data volume. For examples of ASP used by the community, please follow the sources on the right. Check out "Monitoring changes in polar ice sheets and sea ice using airborne and satellite remote sensing" by Shean et. al. Friday morning, C51A-0498. Samantha E. Peel, Caleb I. Fassett, "Valleys in pit craters on Mars: Characteristics, distribution, and formation mechanisms", Icarus, Volume 225, Issue 1, July 2013 N.P. Hammond, C.B. Phillips, F. Nimmo, S.A. Kattenhorn, "Flexure on Dione: Investigating subsurface structure and thermal history", Icarus, Volume 223, Issue 1, March 2013 Watters, W. A., L. Geiger, and M. Fendrock. "Shape Distribution of Fresh Martian Impact Craters from High-Resolution DEMs." LPI Contributions 1719 (2013): 3081. Lefort, Alexandra, et al. "Inverted fluvial features in the Aeolis-Zephyria Plana, western Medusae Fossae Formation, Mars: Evidence for post-formation modification." Journal of Geophysical Research: Planets (1991–2012) 117.E3 (2012).

Upload: others

Post on 26-Sep-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Processing Earth Observing images with Ames Stereo Pipelinelunokhod.org/other/papers/agu_2013_beyer_poster.pdf  \  \--alignment affineepipolar

Newly added in our recent 2.3 releaseof ASP is the utility ‘pc_align’. Thisutility implements the iterativeclosest point (ICP) algorithmthrough the use oflibpointmatcher library. Thisallows users to align theirASP created DEM againstpreviously trusted densesources like SRTM, USGSNED, or GMTED. This sameutility can be used to align aDEM to sparse source likeGLAS, LOLA, or MOLA viainput from CSV.

The example shown rightaligns a DEM created fromWorld View 2 imagery (2m/px) of the CathedralRange in California to aprior existing USGS NEDmap (30 m/px). The top leftslice is the ASP created DEMwhile the bottom right is thetrusted reference.

Processing Earth Observing images with Ames Stereo Pipeline

[email protected] http://irg.arc.nasa.gov/ngt/stereo

Affine Script:> stereo <image1> <image2> \

<camera1> <camera2> \<output prefix> \--alignment affineepipolar

Get Ames Stereo Pipeline

Aligning Output to a Trusted Source Science Accomplished with ASP

Methods of Ingestion

Map Projecting Script:> mapproject -t rpc <dem prior>

<image1> <camera1> <mimage1>> mapproject -t rpc <dem prior>

<image2> <camera2> <mimage2>> stereo <mimage1> <mimage2> \

<camera1> <camera2> \<output prefix> <dem prior>

Script:> pc_align --max-displacement <meters> <dem prior> <asp dem> \

--save-transformed-source-points -o pc/pc> point2dem -tr 2.0 -t_srs "projection" pc/pc-trans_source.tif

Script:Reference against geoid

> dem_geoid <dem>

Reference against ellipsoid> dem_geoid <dem> \

--reverse-adjustment

Vertical Datum Support

Script:Follow QR code in top rightcorner or go to:

http://irg.arc.nasa.gov/ngt/stereo

to find download link for yourspecific OS.

> Download StereoPipeline.tar.bz2> tar xf StereoPipeline.tar.bz2> export PATH=$PATH: \

StereoPipeline/bin

Read the Manual!

> open StereoPipeline/*.pdfor

> gvfs-open StereoPipeline/*.pdf

Ames Stereo Pipeline (ASP) has for along time been able to create denseelevation models for Planetary Scienceon non-terrestrial bodies. Now we’veadapted our tools for Earth Science andcan process imagery from Digital Globe,

GeoEye, Astrium, and anything elsewith an RPC model readable by GDAL.

You can download ASP to perform yourown work from our website by followingthe instructions bellow. ASP is Apache

2 licensed (free as in beer) and operateson most forms of Linux and OSX. It issuitable for operation on yourcomputer, the head-less server, or youruniversity cluster.

Processing satellite imagery into a pointcloud is performed with ASP’s ‘stereo’command. It requires two images withdifferent perspectives in NTF or any otherimage format. It also accepts XMLmetadata that describes a rigorouscamera model that is often shipped withDigital Globe data.

The stereo command will by default try tocorrelate, match all pixels betweenimages, without modifying the inputimagery. However that can take a lot oftime and will possibly fail if there is arotation or scale difference betweenimages. Instead it is recommended thatthe input images be epipolar rectifiedusing the “--alignment affineepipolar”option. An example of its effect is shownright.

Alternatively, for difficult stereo pairs, ASPcan use prior existing low resolution DEMfor map projection. The stereo commandis smart enough when generate its pointcloud to reverse the projection arithmeticand then triangulate with the morerigorous camera model.

By default, all data created by ASP isreferenced against an ellipsoid model. ForEarth data this is usually WGS84. Howeverthis is not ideal for watershed analysis as anellipsoidal model does not correctly modelthe changes in gravity across the surface ofthe Earth.

Instead the solution is to use geoid, a lumpymodel that describes mean sea level in theabsence of weather and tides. ASP providesa utility, ‘dem_geoid’, which can be used toreference data against the commonly usedEGM96, NAVD88, or if processing Marsdata, MOLA MEGDRA.

Our team desires to make elevation productsfor all imagery. However we don't have thetime or experience to apply ASP to allapplications. This is why NASA offers ASPcompletely free. We hope to enable thescientific community access to additionalmeasurements that couldn't be hadotherwise due to cost or data volume. Forexamples of ASP used by the community,please follow the sources on the right.

Check out "Monitoring changes in polar ice sheetsand sea ice using airborne and satellite remote

sensing" by Shean et. al. Friday morning, C51A-0498.

Samantha E. Peel, Caleb I. Fassett, "Valleys in pitcraters on Mars: Characteristics, distribution, andformation mechanisms", Icarus, Volume 225, Issue 1,July 2013

N.P. Hammond, C.B. Phillips, F. Nimmo, S.A.Kattenhorn, "Flexure on Dione: Investigatingsubsurface structure and thermal history", Icarus,Volume 223, Issue 1, March 2013

Watters, W. A., L. Geiger, and M. Fendrock. "ShapeDistribution of Fresh Martian Impact Craters fromHigh-Resolution DEMs." LPI Contributions 1719(2013): 3081.

Lefort, Alexandra, et al. "Inverted fluvial features in theAeolis-Zephyria Plana, western Medusae FossaeFormation, Mars: Evidence for post-formationmodification." Journal of Geophysical Research:Planets (1991–2012) 117.E3 (2012).