Geomatics Department, Humboldt-Universität zu Berlin

Other Projects of the Geomatics Lab
Image SVM

imageSVM is an IDL based tool for the support vector machine classification and regression analysis of remote sensing image data. Its workflow allows a flexible and transparent use of the support vector machine (SVM) concept for both simple and advanced classification/regression approaches.

imageSVM advances the use of the SVM in the field of remote sensing image analysis by

- offering a platform and license independent implementation for SVM classification and regression,

- enabling the use of common image file formats for data in- and output,

- integrating a widely accepted, powerful algorithm for the training of the SVM that is open-source and updated by machine learning specialist on a regular basis,

- offering alternative workflows for automized parameterization (by default values) and user-defined parameterization (limited to parameters relevant for remote sensing),

- visualizing training parameters and intermediate results in a transparent workflow to increase the understanding and acceptance of the support vector approach in the remote sensing.

imageSVM is developed as a non-commercial product at the Geomatics Lab of Humboldt-Universität zu Berlin. By distributing the code, the authors hope to enlarge the number of applications with imageSVM and this way learn more about its performance, its strengths and weaknesses. Users of imageSVM are hence strongly encouraged (- expected) to report their experiences and problems.

imageSVM uses LIBSVM by Chih-Chung Chang and Chih-Jen Lin (Taiwan) for the training of the SVM. http://www.csie.ntu.edu.tw/~cjlin/libsvm/

imageSVM is programmed in IDL. Users can easily run imageSVM with its full capacity by - using IDL (with Development License) - using ENVI (Runtime Application or Development License) - using the freely available IDL Virtual Machine ™ (version 7 or higher)
http://www.ittvis.com/ProductServices/IDL/VirtualMachine.aspx

Download imageSVM

Use the link below to obtain the current version of imageSVM.
--> Download imageSVM


			Projects
	overview Research Projects of the Geomatics Lab: Data integration and data mining DHAKA-INNOVATE DeSurvey EnMAP-Box EnMAP Core Science Team Environmental justice Graduate School on Urban Ecology Land changes in Albania and Kosovo Linking urban land use characteristics and mental illness Metrik Modeling cropland dynamics in Romania Modeling with domain-specific languages Risk model of Dengue Disease in Malaysia Social and health characteristics in urban areas Urban Environmental Monitoring Urban Environmental Monitoring II Urban growth in Greater Tirana Research Collaborations: ESF Exploratory Workshop: EuCaRe EARSeL workshop Post-USSR land cover Rapid urbanization Other Projects of the Geomatics Lab: Geodateninfrastruktur imageSVM		Other Projects of the Geomatics Lab Image SVM imageSVM is an IDL based tool for the support vector machine classification and regression analysis of remote sensing image data. Its workflow allows a flexible and transparent use of the support vector machine (SVM) concept for both simple and advanced classification/regression approaches. imageSVM advances the use of the SVM in the field of remote sensing image analysis by - offering a platform and license independent implementation for SVM classification and regression, - enabling the use of common image file formats for data in- and output, - integrating a widely accepted, powerful algorithm for the training of the SVM that is open-source and updated by machine learning specialist on a regular basis, - offering alternative workflows for automized parameterization (by default values) and user-defined parameterization (limited to parameters relevant for remote sensing), - visualizing training parameters and intermediate results in a transparent workflow to increase the understanding and acceptance of the support vector approach in the remote sensing. imageSVM is developed as a non-commercial product at the Geomatics Lab of Humboldt-Universität zu Berlin. By distributing the code, the authors hope to enlarge the number of applications with imageSVM and this way learn more about its performance, its strengths and weaknesses. Users of imageSVM are hence strongly encouraged (- expected) to report their experiences and problems. imageSVM uses LIBSVM by Chih-Chung Chang and Chih-Jen Lin (Taiwan) for the training of the SVM. http://www.csie.ntu.edu.tw/~cjlin/libsvm/ imageSVM is programmed in IDL. Users can easily run imageSVM with its full capacity by - using IDL (with Development License) - using ENVI (Runtime Application or Development License) - using the freely available IDL Virtual Machine ™ (version 7 or higher) http://www.ittvis.com/ProductServices/IDL/VirtualMachine.aspx Download imageSVM Use the link below to obtain the current version of imageSVM. --> Download imageSVM