Freeware alternatives to eCognition?

Question

Does anyone know any good freeware alternatives to eCognition?

I need to do image segmentation and classification. I tryed the ecognition trial and it was really good. I'm doing the job with SPRING, but it has a lot of limitations...

Is there any qgis plugin?

Answer 1

You might want to try Orfeo Toolbox.

OTB is based on the medical image processing library ITK and offers particular functionalities for remote sensing image processing in general and for high spatial resolution images in particular. Targeted algorithms for high resolution optical images (SPOT, Quickbird, Worldview, Landsat, Ikonos), hyperspectral sensors (Hyperion) or SAR (TerraSarX, ERS, Palsar) are available.

Among it's documented capabilities are:

• optimized read/write access for most of remote sensing image formats, meta-data access, visualization;
• standard remote sensing preprocessing: radiometric corrections, orthorectification; filtering: blurring, denoising, enhancement;
• feature extraction: interest points, alignments, lines;
• image segmentation: region growing, watershed, level sets
• classification: K-means, SVM, Markov random fields;
• change detection;
• information extraction for integration in GIS and mapping systems.

Answer 2

I was going to suggest SPRING. SPRING although it's a clumsy software it is very good at what it proposes. It has very interesting algorithms.

Maybe GRASS can handle the job, but AFAIK, GRASS is mostly a command line package.

Answer 3

In GRASS-GIS version 7.0 (a FOSS package -- not only freeware), still the development version, though fully working and more or less for production-workflows ready, there is i.segment. More upon the module and its implemenation in a dedicated GRASS-Wiki page (along with some sample screenshots).

At its current state, the module does (as per the module's manual):

This segmentation algorithm sequentially examines all current segments in the raster map. The similarity between the current segment and each of its neighbors is calculated according to the given distance formula. Segments will be merged if they meet a number of criteria, including:

 1. The pair is mutually most similar to each other (the similarity distance will be smaller than to any other neighbor), 
 2. The similarity must be lower than the input threshold. The process is repeated until no merges are made during a complete pass.

Answer 4

If I were you, I'd try to look for something that represents a GUI for OpenCV, which does have segmentation algorithms, but I'm not sure if they are accessible for non-developer. We used some of its functions for our classifier for QGIS (DTClassifier plugin).

Answer 5

SAGA GIS also provides "simple region growing", "fast region growing" and "watershed segmentation". The documentation is not available but I will prepare a video tutorial for my students in the next days. I will post here if ready.

download for SAGA GIS is here: http://sourceforge.net/projects/saga-gis/files/SAGA%20-%202.0/

2.08 is a stable older version 2.1 is still a bit unstable yet with more feature.

for "light" introductions to SAGA I suggest to check out youtube videos

Answer 6

The TerraLib software has an implementation of the Baatz & Schape (2000) segmentation algorithm used in eCognition. Once you have a segmentation built, you can bring the results into a statistical software to perform the classification. I imagine that QGIS has some plugins that could classify or cluster image object polygons.