Machine Learning Algorithms for Land Cover Classification

A good overview of machine learning techniques in R is the machine learning taskview. It offers a host of different algorithms, recommended by the experts.

I'd strongly recommend scikits-learn for Python. It supports supervised and unsupervised classification and the documentation is excellent (particularly check out the Machine Learning for Astronomical Data Analysis tutorial and the accompanying YouTube video (note: this is 3 hours long)).

The project is under active development, with the last version being 0.12 which was released in September.

As for what the package is capable of, see Nearest Neighbours, Random Forest (under Ensembe Methods), and Decision Trees to use the examples you gave.

Unfortunately no GUI unless you want to devote time to building one, but I'd recommend the iPython IDE as an excellent interactive scripting environment, including inline plots with matplotlib in the QT console.

Your question assumes that machine learning algorithms for land classification are somehow distinct from software used for other machine learning applications. There are some applications that require special treatment because of unusual characteristics, but there is no reason I know of to think that land use needs special treatment. If land use data can be put into a standard comma delimited form, existing tools such as R should do just fine. Now there may or may not be Land Use software that uses models discovered from machine learning techniques, but that's a different question.

Edited after the first response. -> Most of the major packages for machine learning do have some tools for spatial visualization, although of course they may not meet your particular needs. For example, are you familiar with the sp library for R which is intended for spatial data visualization? Let's see if I can find an appropriate link that gives the flavor of what you can do with it.

http://rwiki.sciviews.org/doku.php?id=tips:spatial-data:spatial_data_visualization For a more extensive listing of tools useful for spacial analysis in R you may want to look at http://cran.r-project.org/web/views/Spatial.html as this includes tools for Geostatistics, ecological analysis, and the like.

There is a group out of Duke University that have developed some interesting script tools for ArcGIS, including random forest models.

Marine Geospatial Ecology Tools

You can also do land classification with DTclassifier (Decision Tree classifier) plugin for QGIS. It provides simple interface for classification of raster data using decision trees, to perform within QGIS.

Did you have a look at eCognition? With their new Version (8.9) they provide Random Forests algorithm within a GUI environment. You can create nice process trees and include object features.

I highly recommend you check out Spacenet. As quoted on their website: "SpaceNet delivers access to high-quality geospatial data for developers, researchers, and startups. Before SpaceNet, computer vision researchers had minimal options to obtain free, precision-labeled, and high-resolution satellite imagery. SpaceNet focuses on four open source key pillars: data, challenges, algorithms, and tools."

They have open source labelled datasets that are covered under the creative commons license. They also hold regular Machine Learning Competitions on those datasets, and then open source and analyze the best models.

I have personally used their open source cresi road segmentation library, and was able to get some great results out of it. They also have an open source Geospatial Machine Learning Pipelining Library called Solaris which is also great.

While they do not cover Landcover Detection (Yet), I found that reading their blogposts really helped me into getting familiar with how to apply machine learning specifically on GIS data.

dzetsaka https://plugins.qgis.org/plugins/dzetsaka/ A Classification tool for QGIS may be considered for those who are not conversant with programming. It facilitates semi-automatic classification with Gaussian Mixture Model, Random Forest, and SVM classifiers.