<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p><br>
    </p>
    <div class="moz-cite-prefix">Le 30/11/2021 à 12:52, Daniel Scheffler
      a écrit :<br>
    </div>
    <blockquote type="cite"
      cite="mid:c2cf3a49-200b-91a6-dd81-1b3059859558@gfz-potsdam.de">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      Thanks a lot for taking the time, Even, I got the transformation
      from cartesian to projected coordinates to work in memory with the
      GTiff driver. With MEM, NUMPY or VRT it does not work because
      these formats are either not readable from /vsimem/ or don´t have
      a regular file path which is needed to set the <font size="2">X_DATASET
        and </font><font size="2">Y_DATASET keys in the GEOLOCATION
        metadata</font>.<br>
    </blockquote>
    Ah indeed for X_DATASET/Y_DATASET, you can't use a MEM or NUMPY
    dataset. But a /vsimem/xxx dataset in another format should work.<br>
    <blockquote type="cite"
      cite="mid:c2cf3a49-200b-91a6-dd81-1b3059859558@gfz-potsdam.de">
      <div class="moz-forward-container"> <br>
        <font size="2">Regarding the inversion of this transformation,
          i.e., </font><font size="2">from projected coordinates to
          pixel/line:</font><br>
        <blockquote type="cite">The logic of <span class="pl-en">GDALCreateGenImgProjTransformer2()
            around <a class="moz-txt-link-freetext"
href="https://github.com/OSGeo/gdal/blob/master/alg/gdaltransformer.cpp#L1825"
              moz-do-not-send="true">https://github.com/OSGeo/gdal/blob/master/alg/gdaltransformer.cpp#L1825</a>
            which is for the source dataset should be ported a few lines
            after for the target dataset.  Probably with a new
            transformer option to be able to point to an auxiliary
            dataset, such as a VRT one of your example, to extract the
            geolocation metadata items from it, that would be different
            from the target dataset itself, because, except perhaps for
            the netCDF case, most GDAL datasets that expose a
            GEOLOCATION metadata domain must be read-only.</span></blockquote>
        <br>
        I don´t completely get what you mean here. To me, this sounds
        like there might be a way to do the inverted transformation
        using the C-API of GDAL. </div>
    </blockquote>
    No, I meant there's some missing code to do that.<br>
    <blockquote type="cite"
      cite="mid:c2cf3a49-200b-91a6-dd81-1b3059859558@gfz-potsdam.de">
      <div class="moz-forward-container">However, I am a Python
        developer and my C skills are a bit poor. Is there any way to
        use the Python bindings here?<br>
        <br>
        Kind regards,<br>
        Daniel<br>
        <br>
        <br>
                  <br>
        <br>
        <div class="moz-cite-prefix">Am 26.11.2021 um 12:38 schrieb Even
          Rouault:<br>
        </div>
        <blockquote type="cite"
          cite="mid:2c9d72a2-3950-f846-b3e4-ee80500179a1@spatialys.com">
          <meta http-equiv="Content-Type" content="text/html;
            charset=UTF-8">
          Daniel,<br>
          <blockquote type="cite"
            cite="mid:8e78eb30-6a89-104b-d2c4-f9ea8231aced@gfz-potsdam.de">
            <br>
            I am trying to convert image data from cartesian/image
            coordinates to projected coordinates AND vice versa using
            geolocation arrays in GDAL. I have two questions:<br>
            <ol>
              <li>Since this transformation is part of a processing
                chain implemented in Python, I try to transform the data
                directly in-memory, i.e, without any disk access. This
                saves IO time and avoids permission errors when trying
                to write temporary data on Windows. How can this be
                done? I got correct results with the code below,
                however, only when I temporarily write the data to disk.
                I tried to write the data to /vsimem/ using the MEM,
                GTiff and NUMPY drivers. However, gdal.Warp can´t find
                the data there (FileNotFoundError). I think, also the
                gdal.Transformer class might be useful and I found an
                interesting thread on that <a
href="https://lists.osgeo.org/pipermail/gdal-dev/2012-January/031502.html"
                  moz-do-not-send="true">here</a> and a related test in
                the GDAL autotest suite (<a
href="https://github.com/OSGeo/gdal/blob/master/autotest/alg/transformgeoloc.py"
                  moz-do-not-send="true">here</a>). However, I can´t get
                it to work for my specific case.</li>
            </ol>
          </blockquote>
          <p>There's no reason it won't work with a in-memory dataset.</p>
          <p>If you use a MEM dataset, then you need to provide the
            dataset object itself as the input dataset of gdal.Warp()
            (the name of a MEM dataset is completely ignored. a MEM
            dataset can't be opened, just created).</p>
          <p>Similarly with a NUMPY dataset.   With GTiff and /vsimem/,
            they behave as a regular file. If you pass it by name as
            input of gdal.Warp(), you need to make sure to close (ds =
            None typically) the dataset before, so it is properly
            flushed and can be opened. But you can also pass it as an
            object without that constraint.</p>
          <p>You can also create purely in-memory VRT files by assigning
            them an empty name. Then of course you need to provide them
            as objects to gdal.Warp()<br>
          </p>
          <p><br>
          </p>
          <blockquote type="cite"
            cite="mid:8e78eb30-6a89-104b-d2c4-f9ea8231aced@gfz-potsdam.de">
            <ol>
              <li>My second question is how I can invert the
                transformation, i.e., how can I transform an image with
                projected coordinates back to cartesian/image
                coordinates, given that a geolocation array tells GDAL
                where to put which pixel in the output? Background is a
                processing pipeline for satellite data where some
                processing steps are running in sensor geometry (image
                data as acquired by the sensor, without any geocoding
                and projection) and I need to provide corresponding AUX
                data which originally come with projected coordinates.<br>
              </li>
            </ol>
          </blockquote>
          The logic of <span class="pl-en">GDALCreateGenImgProjTransformer2()
            around <a class="moz-txt-link-freetext"
href="https://github.com/OSGeo/gdal/blob/master/alg/gdaltransformer.cpp#L1825"
              moz-do-not-send="true">https://github.com/OSGeo/gdal/blob/master/alg/gdaltransformer.cpp#L1825</a>
            which is for the source dataset should be ported a few lines
            after for the target dataset.  Probably with a new
            transformer option to be able to point to an auxiliary
            dataset, such as a VRT one of your example, to extract the
            geolocation metadata items from it, that would be different
            from the target dataset itself, because, except perhaps for
            the netCDF case, most GDAL datasets that expose a
            GEOLOCATION metadata domain must be read-only.<br>
          </span>
          <blockquote type="cite"
            cite="mid:8e78eb30-6a89-104b-d2c4-f9ea8231aced@gfz-potsdam.de">
            <ol>
              <li> <br>
              </li>
            </ol>
            <p>Here is the code I already have to convert a sample image
              from cartesian to projected coordinates:</p>
            <blockquote>
              <p><font size="2">import os<br>
                  from tempfile import TemporaryDirectory<br>
                  from osgeo import gdal, osr<br>
                  import numpy as np<br>
                  from matplotlib import pyplot as plt<br>
                  <br>
                  <br>
                  # get some test data<br>
                  swath_data = np.random.randint(1, 100, (500, 400))<br>
                  lons, lats = np.meshgrid(np.linspace(3, 5, 500),<br>
                                           np.linspace(40, 42, 400))<br>
                  <br>
                  with TemporaryDirectory() as td:<br>
                      p_lons_tmp = os.path.join(td, 'lons.tif')<br>
                      p_lats_tmp = os.path.join(td, 'lats.tif')<br>
                      p_data_tmp = os.path.join(td, 'data.tif')<br>
                      p_data_vrt = os.path.join(td, 'data.vrt')<br>
                      p_data_mapgeo_vrt = os.path.join(td,
                  'data_mapgeo.vrt')<br>
                  <br>
                      # save numpy arrays to temporary tif files<br>
                      for arr, path in zip((swath_data, lons, lats),
                  (p_data_tmp, p_lons_tmp, p_lats_tmp)):<br>
                          rows, cols = arr.shape<br>
                          drv = gdal.GetDriverByName('GTiff')<br>
                          ds = drv.Create(path, cols, rows, 1,
                  gdal.GDT_Float64)<br>
                          ds.GetRasterBand(1).WriteArray(arr)<br>
                          del ds<br>
                  <br>
                      # add geolocation information to input data<br>
                      wgs84_wkt = osr.GetUserInputAsWKT('WGS84')<br>
                      utm_wkt = osr.GetUserInputAsWKT('EPSG:32632')<br>
                      ds = gdal.Translate(p_data_vrt, p_data_tmp,
                  format='VRT')<br>
                      ds.SetMetadata(<br>
                      <br>
                          dict(<br>
                              SRS=wgs84_wkt,<br>
                              X_DATASET=p_lons_tmp,<br>
                              Y_DATASET=p_lats_tmp,<br>
                              X_BAND='1',<br>
                              Y_BAND='1',<br>
                              PIXEL_OFFSET='0',<br>
                              LINE_OFFSET='0',<br>
                              PIXEL_STEP='1',<br>
                              LINE_STEP='1'<br>
                          ),<br>
                          'GEOLOCATION'<br>
                      )del ds<br>
                  <br>
                      # warp from geolocation arrays and read the result<br>
                      gdal.Warp(p_data_mapgeo_vrt, p_data_vrt,
                  format='VRT', geoloc=True,<br>
                                srcSRS=wgs84_wkt, dstSRS=utm_wkt)<br>
                      data_mapgeo =
                  gdal.Open(p_data_mapgeo_vrt).ReadAsArray()<br>
                      <br>
                  # visualize input and output data<br>
                  fig, axes = plt.subplots(1, 4)<br>
                  for i, (arr, title) in enumerate(zip((swath_data,
                  lons, lats, data_mapgeo),<br>
                                                    ('swath data',
                  'lons', 'lats', 'projected data'))):<br>
                      axes[i].imshow(arr, cmap='gray')<br>
                      axes[i].set_title(title)<br>
                  plt.tight_layout()<br>
                  plt.show()</font></p>
              <p><font size="2"><br>
                </font></p>
            </blockquote>
            <p>Any help would be highly appreciated!</p>
            <p>Best,</p>
            <p>Daniel Scheffler<br>
            </p>
            <br>
            <pre class="moz-signature" cols="72">-- 

M.Sc. Geogr. Daniel Scheffler

Helmholtz Centre Potsdam
GFZ German Research Centre For Geosciences
Department 1 - Geodesy and Remote Sensing
Section 1.4  - Remote Sensing
Telegrafenberg, 14473 Potsdam, Germany

Phone:  +49 (0)331/288-1198
e-mail: <a class="moz-txt-link-abbreviated moz-txt-link-freetext" href="mailto:daniel.scheffler@gfz-potsdam.de" moz-do-not-send="true">daniel.scheffler@gfz-potsdam.de</a></pre>
            <br>
            <fieldset class="mimeAttachmentHeader"></fieldset>
            <pre class="moz-quote-pre" wrap="">_______________________________________________
gdal-dev mailing list
<a class="moz-txt-link-abbreviated moz-txt-link-freetext" href="mailto:gdal-dev@lists.osgeo.org" moz-do-not-send="true">gdal-dev@lists.osgeo.org</a>
<a class="moz-txt-link-freetext" href="https://lists.osgeo.org/mailman/listinfo/gdal-dev" moz-do-not-send="true">https://lists.osgeo.org/mailman/listinfo/gdal-dev</a>
</pre>
          </blockquote>
          <pre class="moz-signature" cols="72">-- 
<a class="moz-txt-link-freetext" href="http://www.spatialys.com" moz-do-not-send="true">http://www.spatialys.com</a>
My software is free, but my time generally not.</pre>
        </blockquote>
        <br>
        <pre class="moz-signature" cols="72">-- 

M.Sc. Geogr. Daniel Scheffler

Helmholtz Centre Potsdam
GFZ German Research Centre For Geosciences
Department 1 - Geodesy and Remote Sensing
Section 1.4  - Remote Sensing
Telegrafenberg, 14473 Potsdam, Germany

Phone:  +49 (0)331/288-1198
e-mail: <a class="moz-txt-link-abbreviated moz-txt-link-freetext" href="mailto:daniel.scheffler@gfz-potsdam.de" moz-do-not-send="true">daniel.scheffler@gfz-potsdam.de</a></pre>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <pre class="moz-quote-pre" wrap="">_______________________________________________
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="https://lists.osgeo.org/mailman/listinfo/gdal-dev">https://lists.osgeo.org/mailman/listinfo/gdal-dev</a>
</pre>
    </blockquote>
    <pre class="moz-signature" cols="72">-- 
<a class="moz-txt-link-freetext" href="http://www.spatialys.com">http://www.spatialys.com</a>
My software is free, but my time generally not.</pre>
  </body>
</html>