Permalink
Browse files

docs: Expand DG tutorial, other refactoring

docs: Expand DG tutorial, other refactoring
  • Loading branch information...
1 parent 530c724 commit 43e4d96d9dc92f4102cc0187a49c9351efcfda4b @oleg-alexandrov oleg-alexandrov committed Dec 25, 2013
@@ -676,6 +676,13 @@ @unpublished{digital-globe:samples
url = {http://www.digitalglobe.com/product-samples}
}
+@unpublished{digital-globe:camera,
+ author = {Digital Globe},
+ title = {{Radiometric Use of WorldView 2 Imagery}},
+ note = "Description of the WV02 camera",
+ url = {http://www.digitalglobe.com/sites/default/files/Radiometric_Use_of_WorldView-2_Imagery\%20\%281\%29.pdf}
+}
+
@unpublished{geoeye:samples,
author = {GeoEye},
title = {{Sample Imagery Request Form}},
@@ -180,7 +180,7 @@ \subsection{Processing Mars Orbital Camera Imagery}
What follows is an example of bundle adjustment using two \ac{MOC}
images of Hrad Vallis. We use images E02/01461 and M01/00115, the same
-as used in Chapter~\ref{ch:tutorial}. These images are available from
+as used in Chapter~\ref{ch:moc_tutorial}. These images are available from
NASA's \ac{PDS} (the \ac{ISIS} \texttt{mocproc} program will operate
on either the IMQ or IMG format files, we use the \texttt{.imq} below
in the example). For reference, the following \ac{ISIS} commands are
@@ -357,7 +357,7 @@ \subsection{Processing Mars Orbital Camera Imagery}
1065 px and the final solution had an error of 1.1 px.
Producing a DEM using the newly created camera corrections is the same
-as covered in the Tutorial on page \pageref{ch:tutorial}. When using
+as covered in the Tutorial on page \pageref{ch:moc_tutorial}. When using
\texttt{jigsaw}, it modifies a copy of the spice data that is stored
internally to the cube file. Thus when we want to create a DEM using
the correct camera geometry, no extra information needs to be given to
View
@@ -45,7 +45,7 @@ \section{Pre-Processing}
the original cameras which took the pictures, (3) a 3D rotation that achieves
epipolar rectification {\it(only implemented for Pinhole sessions for
missions like MER or K10)} or (4) map-projection of both the left
-and right images using the \ac{ISIS} \texttt{cam2map} command or
+and right images using the \ac{ISIS} \texttt{cam2map} command, or
through \texttt{mapproject} for Digital Globe and GeoEye images (see
section \ref{mapproj} for the latter).
The first three options can be applied automatically by the Stereo
@@ -106,6 +106,7 @@ \section{Pre-Processing}
your results are sub-optimal.
\section{Disparity Map Initialization}
+\label{d-sub}
Correlation is the process at the heart of the Stereo Pipeline. It is
a collection of algorithms that compute correspondences between pixels
@@ -185,14 +186,16 @@ \section{Disparity Map Initialization}
\texttt{\textit{output\_prefix}-D\_sub.tif}, we are allowed to
process beyond that limitation.
-Any large failure in
-the low-resolution disparity image will be detrimental to the
-performance of the higher resolution disparity. In the event that the
-low-resolution disparity is completely unhelpful, it can be skipped by
-adding \texttt{corr-seed-mode 0} in the \texttt{stereo.default}
-file. This should only be considered in cases where the texture in an
-image is completely lost when subsampled. An example would be
-satellite imagery of fresh snow in the Arctic.
+Any large failure in the low-resolution disparity image will be
+detrimental to the performance of the higher resolution disparity. In
+the event that the low-resolution disparity is completely unhelpful, it
+can be skipped by adding \texttt{corr-seed-mode 0} in the
+\texttt{stereo.default} file. This should only be considered in cases
+where the texture in an image is completely lost when subsampled. An
+example would be satellite imagery of fresh snow in the
+Arctic. Alternatively, \texttt{\textit{output\_prefix}-D\_sub.tif} can
+be computed at a sparse set of pixels at full resolution, as described
+in section \ref{sparse-disp}.
An alternative to computing \texttt{\textit{output\_prefix}-D.tif}
from sub-sampled images (\texttt{corr-seed-mode 1}) or skipping it
View
@@ -324,7 +324,7 @@ \subsubsection*{stereo.default}
\clearpage
\section{Mars Global Surveyor MOC-NA}
-In the Stereo Pipeline Tutorial in Chapter~\ref{ch:tutorial}, we
+In the Stereo Pipeline Tutorial in Chapter~\ref{ch:moc_tutorial}, we
showed you how to process a narrow angle \ac{MOC} stereo pair that
covered a portion of Hrad Vallis. In this section we will show you
more examples, some of which exhibit a problem common to stereo
@@ -387,7 +387,6 @@ \subsubsection*{stereo.default}
\texttt{none} when using projected imagery. If using non-projected use
\texttt{homography} or \texttt{affineepipolar}.
-\newpage
\section{Mars Exploration Rovers MER}
The MER rovers have several cameras on board and they all seem to have
@@ -759,200 +758,12 @@ \subsubsection*{stereo.default}
\end{Verbatim}
\end{minipage}\end{center}
-\newpage
\section{Digital Globe Imagery}
-\label{sec:digital_globe_imagery}
-
-Digital Globe provides imagery from the Quick Bird and the three World
-View satellites. These are the hardest images to process with Ames
-Stereo Pipeline because they are exceedingly large, much larger than
-HiRISE imagery. There is also a wide range of terrain challenges and
-atmospheric effects that can confuse ASP. Trees are particularly
-difficult for us since their texture is nearly nadir and perpendicular
-to our line of sight. It is important to know that the driving force
-behind our support for Digital Globe imagery is to create models of
-ice and bare rock. That is the type of imagery that we have tested
-with and have focused on. If we can make models of wooded or urban
-areas, that is a bonus, but we can't provide any advice for how to
-perform or improve the results if you choose to use ASP in that way.
-
-ASP can only process Level 1B imagery and cannot process Digital Globe's
-aerial imagery. We can pull a camera model from the RPC coefficients or
-from their linear camera model described in the provided XML files. We
-won't be discussing the RPC method in this section, however you can
-learn more about it in the later example (section \ref{rpc}) where we
-discuss processing GeoEye imagery which comes only with RPC
-coefficients.
-
-Our implementation of the linear camera model only
-models the geometry of the imaging hardware itself and velocity
-aberration. We do not currently model refraction due to light bending
-in Earth's atmosphere. It is our understanding that this could
-represent misplacement of points up to a meter for some
-imagery. However this is still smaller error than the error from
-measurement of the spacecraft's position and orientation. We do not
-provide facilities for correcting spacecraft attitude either. So you
-will have to perform some manual shifting of the data to get it into
-the correct location. These errors are fortunately much less than
-found with extra-terrestrial missions largely due to the availability
-of GPS and high bandwidth comms with the satellite.
-
-In the next 2 sections we will show how to process unmodified and
-map-projected variants of World View 1 imagery. This steps will be the
-same for Digital Globe's other satellites. The imagery we are using are
-from the free stereo pair example of Lucknow, India available from
-Digital Globe's website \cite{digital-globe:samples}. These images
-represent a non-ideal problem for us since this is an urban location,
-but at least you should be able to download this imagery yourself and
-follow along.
-
-\subsection{Processing Raw}
-\label{rawdg}
-
-After you have downloaded the example stereo imagery of India, you
-will find a directory titled\newline
-\texttt{052783824050\_01\_P001\_PAN}. It has a
-lot of files and many of them contain redundant information just
-displayed in different formats. We are interested only in the TIF or NTF
-imagery and the similarly named XML file.
-
-Further investigation of the files downloaded will show that there are
-in fact 4 image files. This is because Digital Globe breaks down a
-single observation into multiple files for what we assume are size
-reasons. These files have a pattern string of ``\_R[N]C1-'', where N
-increments for every subframe of the full observation. The tool named
-\texttt{dg\_mosaic} can be used to mosaic (and optionally reduce the
-resolution of) such a set of sub-observations into a single image file
-and create an appropriate camera file
+\label{digital_globe_data}
-\begin{verbatim}
- > dg_mosaic 12FEB12053305*TIF --output-prefix 12FEB12053305 --reduce-percent 50
-\end{verbatim}
-
-and anologously for the second set. See section \ref{dgmosaic} for more
-details. The \texttt{stereo} program can use either the original images
-or the mosacked pair.
-
-Since we are ingesting these images raw, it is strongly recommended that
-you use an affine epipolar alignment to reduce the search range. The
-\texttt{stereo} command and a rendering in QGIS are shown below.
-
-\begin{verbatim}
- > stereo -t dg \
- 12FEB12053305-P1BS_R2C1-052783824050_01_P001.TIF \
- 12FEB12053341-P1BS_R2C1-052783824050_01_P001.TIF \
- 12FEB12053305-P1BS_R2C1-052783824050_01_P001.XML \
- 12FEB12053341-P1BS_R2C1-052783824050_01_P001.XML dg/dg
-\end{verbatim}
-
-\begin{figure}[h!]
-\centering
- \includegraphics[width=2.0in]{images/examples/dg/DigitalGlobeContext.png}
- \includegraphics[width=2.0in]{images/examples/dg/DigitalGlobeCloseUp.png}
- \includegraphics[width=2.0in]{images/examples/dg/DigitalGlobeCloseUpDRG.png}
-\caption{Example colorized height map and ortho image output.}
-\label{fig:dg-nomap-example}
-\end{figure}
-
-\subsubsection*{stereo.default}
-
-The stereo.default example file works generally well with all Digital
-Globe pairs. Just set \texttt{alignment-method} to
-\texttt{affineepipolar} or \texttt{homography} if using non-projected
-imagery.
-
-\subsection{Processing Map-Projected Imagery}
-\label{mapproj}
-
-Eventually you will run into Digital Globe imagery that has too much
-parallax to be processed in a reasonable time. (That wasn't the case
-for Lucknow, India because it is so flat.) We can speed up the result
-by processing map-projected versions of the imagery. The catch is, you
-are not allowed to use any map-projection software you like (such as
-GDAL). We need to have complete control of the process since ASP will
-have to work backwards through this math and interpolations in order
-to make the final elevation model. So we have provided a utility
-called \texttt{mapproject} whose commands closely resemble
-\texttt{point2dem}. The purpose of this tool is to use the simplified
-RPC model to map-project the imagery onto a low resolution and
-hole-less digital elevation model. Later, ASP will then work backwards
-through the RPC model and then forward through the linear camera model
-to calculate the final result.
-
-The hardest part of this whole process is getting the input
-low-resolution 3D model. In this example we will use a variant of NASA
-SRTM data with no holes. Other choices might be GMTED2010 or USGS's NED
-data.
-
-It is important to note that ASP expects DEM images to be in
-reference to a datum ellipsoid, such as WGS84 or NAD83. If the
-low-resolution DEM is in respect to either the EGM96 or NAVD88 geoids,
-the ASP tool \texttt{dem\_geoid} can be used to convert the DEM to WGS84
-or NAD83 (section \ref{demgeoid}). (The same tool can be used to convert
-back the final output ASP DEM to be in reference to a geoid, if
-desired.)
-
-Not applying this conversion might not properly negate the parallax
-seen between the two images, though it will not corrupt the
-triangulation results. In other words, sometimes one may get by
-ignoring the vertical datums on the input but we do not recommend
-doing that. Also, you should note that the geoheader attached to those
-types of files usually do not describe the vertical datum they
-used. That can only be understood by careful reading of your
-provider's documents.
-
-The NASA SRTM square for our example spot in India is N26E080. Below are
-the commands for map-projecting the input and then running through
-stereo. You can use any projection you like as long as it preserves
-detail in the imagery. Note also that we have added a seventh parameter
-to the stereo call where we provide the input low-resolution DEM.
-
-\begin{figure}[h!]
-\centering
- \includegraphics[width=2.0in]{images/examples/dg/MappedContext.png}
- \includegraphics[width=2.0in]{images/examples/dg/MappedCloseUp.png}
- \includegraphics[width=2.0in]{images/examples/dg/MappedCloseUpDRG.png}
-\caption{Example colorized height map and ortho image output.}
-\label{fig:dg-map-example}
-\end{figure}
-
-\subsubsection*{Commands}
-
-The first step is downloading a low-resolution DEM file without holes to
-map-project on to. In this example we use \texttt{srtm\_53\_07.tif}, a
-90 meter resolution tile from the CGIAR-CSI modification of the original
-NASA product \cite{cgiar:srtm90m}.
-
-\begin{verbatim}
- > mapproject -t rpc --t_srs "+proj=eqc +units=m +datum=WGS84" \
- --tr 0.5 srtm_53_07.tif \
- 12FEB12053305-P1BS_R2C1-052783824050_01_P001.TIF \
- 12FEB12053305-P1BS_R2C1-052783824050_01_P001.XML \
- left_mapped.tif
- > mapproject -t rpc --t_srs "+proj=eqc +units=m +datum=WGS84" \
- --tr 0.5 srtm_53_07.tif \
- 12FEB12053341-P1BS_R2C1-052783824050_01_P001.TIF \
- 12FEB12053341-P1BS_R2C1-052783824050_01_P001.XML \
- right_mapped.tif
- > stereo -t dg left_mapped.tif right_mapped.tif \
- 12FEB12053305-P1BS_R2C1-052783824050_01_P001.XML \
- 12FEB12053341-P1BS_R2C1-052783824050_01_P001.XML \
- dg/dg srtm_53_07.tif
-\end{verbatim}
-
-If the \texttt{--t\_srs} option is not specified, it will be read from
-the low-resolution input DEM.
-
-The complete list of options for \texttt{mapproject} is described in
-section \ref{mapproject}.
-
-\subsubsection*{stereo.default}
-
-The stereo.default example file works generally well with all Digital
-Globe pairs. Just set
-\texttt{alignment-method} to \texttt{none} or \texttt{homography}.
+Processing of Digital Globe images is described extensively in the
+tutorial in chapter \ref{ch:dg_tutorial}.
-\newpage
\section{GeoEye and Astrium Imagery / RPC Imagery}
\label{rpc}
@@ -1012,7 +823,6 @@ \subsubsection*{stereo.default}
pairs. Just set \texttt{alignment-method} to \texttt{affineepipolar}
or \texttt{homography}.
-\newpage
\section{Dawn (FC) Framing Camera}
This is a NASA mission to visit two of the largest objects in the
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@@ -82,7 +82,7 @@ \subsection{Quick Start for ISIS Users}
\hspace*{2em}\texttt{source \$ISISROOT/scripts/isis3Startup.csh}
\item [{Try~It~Out}] ~\\
-See the next chapter (Chapter~\ref{ch:tutorial}) for an example.
+See the next chapter (Chapter~\ref{ch:moc_tutorial}) for an example.
\end{description}
\subsection{Quick Start for Digital Globe Users}
@@ -96,11 +96,8 @@ \subsection{Quick Start for Digital Globe Users}
\texttt{tar xvfz StereoPipeline-\textit{VERSION-ARCH-OS}.tar.gz}
\item [{Try~It~Out}] ~\\
-Processing Earth imagery is similar in principle to processing any of
-the NASA imagery. We encourage you to still read the next chapter
-(Chapter~\ref{ch:tutorial}) that demos processing Mars
-imagery. Afterwards you should skip to the data processing example
-shown in section \ref{sec:digital_globe_imagery}.
+Processing Earth imagery is described in the data processing tutorial
+in chapter \ref{ch:dg_tutorial}.
\end{description}
View
@@ -642,14 +642,18 @@ \section{dem\_geoid}
\section{dg\_mosaic}
\label{dgmosaic}
-This tool can be used when processing Digital Globe Imagery (section
-\ref{sec:digital_globe_imagery}). A Digital Globe satellite may take a
-picture, and then split it into several images and corresponding camera XML
-files. \texttt{dg\_mosaic} will mosaic these images into a single file,
-and create the appropriate combined camera XML file.
-
-Digital Globe camera files contain, in addition to the original camera models, their RPC approximations
-(section \ref{rpc}). \texttt{dg\_mosaic} outputs both types of combined models. The combined RPC model can be used to map-project the mosaiced images with the goal of computing stereo from them (section \ref{mapproj}).
+This tool can be used when processing Digital Globe Imagery (chapter
+\ref{ch:dg_tutorial}). A Digital Globe satellite may take a
+picture, and then split it into several images and corresponding camera
+XML files. \texttt{dg\_mosaic} will mosaic these images into a single
+file, and create the appropriate combined camera XML file.
+
+Digital Globe camera files contain, in addition to the original camera
+models, their RPC approximations (section
+\ref{rpc}). \texttt{dg\_mosaic} outputs both types of combined
+models. The combined RPC model can be used to map-project the mosaiced
+images with the goal of computing stereo from them (section
+\ref{mapproj}).
The tool needs to be applied twice, for each of the left and right image sets.
Oops, something went wrong.

0 comments on commit 43e4d96

Please sign in to comment.