Dear friends,

I have been suggested to write this post to point your attention to a free software i'm being developing for automatic image (and 3d model!!!) georeferencing.

It is called AutoGR-Toolkit and it is based on image matching. It provides thousands of ground control points in seconds.  These points can then be used in ArcGIS or QGIS (for 2D matching) and also in Agisoft Photoscan, VSfm or sfm_georef (for 3D model georeferencing).

It also gives the possibility to download high-resolution orthophotos from freely accessibly WMS server.

It is free and it will stay free of charge, but I need your feedback to improve it and make it more useful for you too.

If you think you are interested, please feel free to visit and download the last version of AutoGR-Toolkit.

Looking forward to hearing from you!


Views: 385

Comment by Jonathan Dandois on March 13, 2015 at 1:59pm

Hi Gianluca, so after reviewing the AutoGR website a bit, it sounds like you are georeferencing to existing Google Earth imagery using SIFT, is that correct?  That is super cool and a very interesting approach for georeferencing.  Do you have any demo papers or examples out for this?  Have you compared the results of this with georeferencing from drone GPS?  The EcosynthAerial pipeline uses the APM/Pixhawk telemetry log to georeference the output of Ecosynther, Photoscan or other SFM which is a pretty fast approach, but is obviously limited by the quality of the GPS.  I think both this tecniqhue and yours are probably much easier than ground control points.

Comment by Gianluca Cantoro on March 17, 2015 at 5:13am

Dear Jonathan,

thank you for your interest and sorry for my late (and perhaps too long) reply.

My software does use SIFT (multiple algorithms) to match aerial photographs and 3D models with -respectively- orthophotos and DTMs (the last version still makes use only of the elevation values provided by Google Api, but I am working on the possibility to use other user-provided dtm).

If the orthophoto is accurate (and accurately georeferenced) and the matching with aerials is successful, the final georeferencing is much more accurate than any other method. Of course this is the ideal... in reality, the matching may not work (or it may need some intermediate steps to work out) or the orthophoto may not be available. In that case, I provide the user with the possibility to use freely accessible WMS images as base orthophoto (not only Google, but also Bing and a number of national orthophotos). If you use google, the images it provides are not intended for accuracy, but one may need to compare newly acquired images with already available satellite images. In that case, the software works perfectly.

The ratio behind the software is to try and avoid the manual input of ground control points, since this operation brings a number of problems, related with precision and accuracy (and time consumption!).

The advantage of my software (I believe) is the possibility to control the output. It is not like photogrammetry software, where your GPS exif is used no matter if it is accurate or not (and it is up to you to decide how much the output fits your needs). With AutoGR-Toolkit you get thousands of GCP in seconds and this is especially valuable in mountainous landscapes, as you can imagine, where the georeferencing of a single photo (whether perfectly vertical or slightly oblique) may take forever.

I hope I gave you the information you were looking for... and I don't mind discussing about the above if anybody is interested :)




You need to be a member of Ecosynth to add comments!

Join Ecosynth


© 2017   Created by Erle Ellis. Content is made available under CC BY 4.0.   Powered by

Badges  |  Report an Issue  |  Terms of Service