Hi everyone.
As Hamish already posted, I am working on Object Based Image Analysis for my PhD.
I have developed a system based on mseg (http://sourceforge.net/projects/mseg/) for image segmentation, CLIPS for rule based classification and Support Vector Machines for supervised classification. I also use OpenCV for some advanced computer vision algorithms. All these programs are combined through scripts and have primitive CLI interfaces.
I have published some papers around this technology and I will release all my code under GPL soon.
Especially for the segmentation algorithm, it is still unoptimized for speed, but gives good results (of course it has to be improved a lot to meet my expectations...). It is written in C++ and is missing gdal integration right now (works only with ER-Mapper ers - BIL format) with my own implementation of image I/O. Initially I had in mind to include it on GRASS someday, but I will need support to know the internals of GRASS.
Regards,
Angelos Tzotsos
On 10/09/2010 03:51 AM, Hamish wrote:
Angelos Tzotsos (a.k.a. kalxas) wrote:
[...]Also, I have a remote sensing background and I have been very
interested in WKTRaster, Rasdaman and ZooWPS.
I will try to contribute to those parts of the live dvd when I
find some free time or I will try to find a volunteer among my
students to test and contribute.Also, I am developing a free algorithm/software for object
based image analysis (something similar to eCognition) and
plan to release some alpha version very soon.
http://sourceforge.net/projects/mseg/This will have to go way too much distance to be final (this
time is only cli, like gdal), but I hope some day I will have
it on OSGeo-Live for everyone to use.for use with GRASS, that level of interaction is perfect.
cheers,
Hamish