Well, I actually implemented this idea in Mirone long time ago. First
with SIFT and later with the SURF module in OpenCV, but I really never
tested much from the the point that "some cases work well, others not so
much".
Anyway, you might be interested to look also into ASIFT. It looks promising
http://www.cmap.polytechnique.fr/~yu/research/ASIFT/demo.html
Joaquim
Hi all,
I talked with Kshitij about this idea at length. I am sure that he can
do a good job. He has the programming skills and has academic
background in image processing and computer vision.
I am willing to co-mentor him.
Kshitij,
You did not include the extensions we discussed. Estimating the shift
between two georeferenced images, matching rasters with vector datasets.
On Mon, Jan 27, 2014 at 1:16 PM, Kshitij Kansal <kansa...@gmail.com
<mailto:kansa...@gmail.com>> wrote:
Hello Everyone
I am an undergrad student in my Junior year of study. I have been
involved with some research in field of Spatial Temporal Analysis
of Images. I have been using the GDAL for various purposes, for
quite some time now. I thought that now its time to contribute
something to Open Source Community. So I started to think about
the GSoC, 2014. As I started browsing the last year's ideas, one
idea that I could relate was "*Raster / Vector Geo-referencer on
the Web*".
*Motivation:*
The main motivation behind this is that, among the very first
things that are required in many remote sensing/geo-spatial
related projects are geo-referenced images. Most of the images
that we have are generally not referenced and hence we have to go
to some tool and MANUALLY reference them with the help of some
reference image. This generally takes a lot of time and most of
the times errors do creep in.
*Original Idea:*
What has been proposed in this(as a part of GSoC'13 idea) was that
we should provide two image, one reference image and other
un-referenced image. Now with the help of this we should be able
to geo-reference this unreferenced image by MANUALLY selecting the
Control Points. Some thing like this :
http://www.youtube.com/watch?feature=player_embedded&v=88gt1gj2dbs
<http://www.youtube.com/watch?feature=player_embedded&v=88gt1gj2dbs>
*Modified Idea:*
Now what I felt was that, this idea can be made more innovative
and useful if we can AUTOMATE this whole process of
geo-referencing along with providing the option for manual
selection. This is where the knowledge of Image Processing can
help. There are various algorithms(like SIFT/SURF) which can be
used for doing this. We can apply these algorithms and extract the
GCP's and then geo-reference the images. Also, this thing has not
been implemented in any of the other open source softwares(like
GRASS,QGIS). So if we can implement this, other communities can
build on these ideas or even adapt it. It would be "entirely based
on GDAL and completely reusable"
Furthermore, this idea can be expanded a lot in the sense that we
should not not put restrictions to the number of bands in the
images. Now what I mean by this is that reference image's bands
may not be the same order of unreferenced. It should work with
images of any number of bands.
Any Comments or Suggestions or Feedback ?
Anyone willing to help me in this?
Regards,
Kshitij Kansal
Lab For Spatial Informatics,
IIIT Hyderabad
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org <mailto:gdal-dev@lists.osgeo.org>
http://lists.osgeo.org/mailman/listinfo/gdal-dev
--
Best regards,
Chaitanya kumar CH.
+91-9494447584
17.2416N 80.1426E
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/gdal-dev
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/gdal-dev
