<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body text="#CC0000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix"><font face="Tempus Sans ITC">We've done
quite a bit of time series stacking of Landsat data for change
detection using the method you are testing. <br>
<br>
We determine the output frame of either the minimum bounding box
of all the input scenes (gdalinfo each scene, save corner
coordinates, compute </font>bounding box) or we have also
pre-defined tiling systems and used those corners. We then use
gdalwarp on each scene, specify the output frame with -te, specify
the resolution with -tr, and the output coordinate system with
-t_srs. This will produce a stack of images with the same number
of lines/samples and the same output extent that should overlay
correctly.<br>
<div class="moz-signature">
<p><font color="#cc0000" face="Tempus Sans ITC">__________________________________________________<br>
Kurtis Nelson<br>
Physical Scientist - Wildland Fire Science Team<br>
United States Geological Survey (USGS)<br>
Earth Resources Observation & Science (EROS) Center<br>
47914 252nd Street<br>
Sioux Falls, SD 57198<br>
(605) 594-2805 (phone)<br>
(605) 594-6529 (fax)<br>
<a class="moz-txt-link-abbreviated" href="mailto:knelson@usgs.gov">knelson@usgs.gov</a><br>
<br>
"For every complex problem there is an answer that is clear,
simple, and wrong"<br>
Henry Louis Mencken<br>
</font></p>
</div>
On 3/27/2014 5:37 PM, Jonathan Greenberg wrote:<br>
</div>
<blockquote
cite="mid:CABG0rftOzV9KmiZUOBSjXkZaBwZwwLOE9kXCPWzrRm9AJa+ing@mail.gmail.com"
type="cite">
<pre wrap="">Hi Tim:
Not quite -- all I'm trying to do is basically crop (or expand) one
raster to another's extent, but given two images may have their upper
left coordinates somewhat "out of sync" (so the pixels don't line up
perfectly with one another), I'm a bit unclear on how to do this --
the goal is to have two rasters with identical geographic information
(same projection, pixel size, row and column numbers) -- this then
allows me to "stack" them together to create an e.g. time series.
I'm thinking using gdalwarp with the te, t_srs, and ts/tr set to the
"reference" file may accomplish the task -- running a test now.
This is a common problem with a lot of raster analyses -- data from a
wide variety of sources/resolutions/projections need to be coerced to
a common set of spatial references so models can be applied to them
(e.g. species distribution models). This is a common enough task that
it would be nice to see this as a feature in e.g. gdalwarp, if it
isn't already.
--j
On Thu, Mar 27, 2014 at 5:27 PM, Tim Keitt <a class="moz-txt-link-rfc2396E" href="mailto:tkeitt@utexas.edu"><tkeitt@utexas.edu></a> wrote:
</pre>
<blockquote type="cite">
<pre wrap="">
On Thu, Mar 27, 2014 at 4:26 PM, Jonathan Greenberg <a class="moz-txt-link-rfc2396E" href="mailto:jgrn@illinois.edu"><jgrn@illinois.edu></a>
wrote:
</pre>
<blockquote type="cite">
<pre wrap="">
GDALers:
What is the most efficient way, given a "reference raster", and an
arbitrary raster (we'll call it "unsynced") synced together to allow
them to be stacked: the output of this should be the unsynced raster
with the same number of rows, columns, pixel size, upper left
coordinates, and projection as the reference raster.
Here are the assumptions:
- both rasters already have the same pixel size and projection, but
the offset of the upper left coordinate may not be "perfect"
- for each pixel location defined by the reference raster, the output
should have the closest (nearest neighbor) pixel from the unsynced
raster. It should have some NA value in regions where the reference
and unsynced do not overlap.
Thoughts? The application of this would be e.g. taking a set of
Landsat images from different time periods (each image has slightly
different numbers of rows and columns), and stacking them together to
perform time series analysis.
</pre>
</blockquote>
<pre wrap="">
If I read right you are trying to estimate the x,y offset error in one
image. You can assign GCPs by picking identical pixels in each image and
then warp one of them to the other. OSSIM has some capability for image
matching I believe. If you convolve the two images, you should get an
impulse at the correct offset, but this is tricky in practice. Panorama
software may be able to compute the exact matching of the images.
THK
</pre>
<blockquote type="cite">
<pre wrap="">
Cheers!
--j
--
Jonathan A. Greenberg, PhD
Assistant Professor
Global Environmental Analysis and Remote Sensing (GEARS) Laboratory
Department of Geography and Geographic Information Science
University of Illinois at Urbana-Champaign
259 Computing Applications Building, MC-150
605 East Springfield Avenue
Champaign, IL 61820-6371
Phone: 217-300-1924
<a class="moz-txt-link-freetext" href="http://www.geog.illinois.edu/~jgrn/">http://www.geog.illinois.edu/~jgrn/</a>
AIM: jgrn307, MSN: <a class="moz-txt-link-abbreviated" href="mailto:jgrn307@hotmail.com">jgrn307@hotmail.com</a>, Gchat: jgrn307, Skype: jgrn3007
_______________________________________________
gdal-dev mailing list
<a class="moz-txt-link-abbreviated" href="mailto:gdal-dev@lists.osgeo.org">gdal-dev@lists.osgeo.org</a>
<a class="moz-txt-link-freetext" href="http://lists.osgeo.org/mailman/listinfo/gdal-dev">http://lists.osgeo.org/mailman/listinfo/gdal-dev</a>
</pre>
</blockquote>
<pre wrap="">
--
<a class="moz-txt-link-freetext" href="http://www.keittlab.org/">http://www.keittlab.org/</a>
</pre>
</blockquote>
<pre wrap="">
</pre>
</blockquote>
<br>
</body>
</html>