GeoImageRMP is a Rapid Miner extension for spatial data : remote sensing images, vector images, etc. It enables the use of data modelling operator of RapidMiner on spatial data.


Developpers :
  • Thomas Guyet, contact ( thomas_dot_guyet_at_irisa.fr)
  • Boussad Ghedamsi, 2012

Year : 2012-2013
Laboratory : AGROCAMPUS-OUEST/IRISA (Team DREAM)


Presentation [top]

RapidMiner

RapidMiner is one of the most used frontend for data mining, modelling and analysis. RapidMiner enable the user to design data processing tool chain interactively. A tool chain is flow chart of processing tools represented by boxes in the interface. This software is easily extendable by designing Plugins. The GeoImageRMP plugins is a plugin dedicated to the design of tool chain to process georeference images (raster and vector images). It is a practical and useful reponds to the analytic tasks of georeferenced data. This is the first plugins that is interested in including georeferenced data in RapidMiner and although the only userfriendly tool to create and compare geoereferenced data tool chains. It benefits from the large amount of data processing tools that are already implemented in RapidMiner (classification, clustering, frequent pattern mining, etc.). One of the main aims of this plugin is to quickly prototype machine learning tools chain for remote sensing classification task. The GeoImageRMP plugin provide several new processing boxes:

Sampling Figure
Figure : Convertion from georeferenced data to RapidMiner dataset

The geo-processing are supported by the open source (LGPL) GeoTools API. This API is OGC compliant. Our processes include management of heterogenous Spatial Reference Systems (SRS).

Process example

Learning Processes
Figure : Supervised machine learning tool chain. A SVM model is constructed from hyperspectral images. The learning set is based on a Shapefile layer that locates sampling points and their ground test class. The model is then applied to the full image.


Publications [top]


Download [top]


Contacts [top]

If you used this software, if you find it interesting but not appropriate for your data set or if you encounter any problem: do not hesitate to contact us.

Guyet Thomas, AGROCAMPUS-OUEST