Classification, change and anomaly detection

ALTB: Active Learning MATLAB(tm) Toolbox

Active Learning Toolbox Image

ALTB is a set of tools implementing state-of-the-art active learning algorithms for remote sensing applications.

References
  • Semisupervised classification of remote sensing images with active queries. Munoz-Mari, J., Tuia, D., and Camps-Valls, G. IEEE Transactions on Geoscience and Remote Sensing 50(10): 3751-3763, 2012.
  • Remote sensing image segmentation by active queries. Tuia, D., Muñoz-Marí, J., Camps-Valls, G. Pattern Recognition 45(6): 2180-2192, 2012.

BagSVM: Bag Support Vector Machine

Bag Support Vector Machine Image

A semi-supervised SVM method for the classification of remote sensing images, learning a kernel directly from the image and regularizing the representation with cluster kernels.

References
  • Semisupervised remote sensing image classification with cluster kernels. Tuia, D., Camps-Valls, G. IEEE Geoscience and Remote Sensing Letters 6(2): 224-228, 2009.
  • Spectral clustering with the probabilistic cluster kernel. Emma Izquierdo-Verdiguier, Robert Jenssen, Luis Gómez-Chova, Gustavo Camps-Valls. Neurocomputing 149(C): 1299-1304, 2015.

Graph Kernels for Spatio-Spectral Classification

Graph Kernels for Spatio-Spectral Classification Image

A graph kernel for spatio-spectral remote sensing image classification using support vector machines (SVM), incorporating higher-order relations in the neighborhood for improved classification accuracy.

References
  • Spatio-spectral remote sensing image classification with graph kernels. Camps-Valls, G., Shervashidze, N., and Borgwardt, K.M. IEEE Geoscience and Remote Sensing Letters 7(4): 741-745, 2010.

HyperLabelMe: A Web Platform for Benchmarking Remote-Sensing Image Classifiers

HyperLabelMe Platform Image

The Image and Signal Processing (ISP) group at the Universitat de València has harmonized a big database of labeled multi- and hyperspectral images for testing classification algorithms. We have harmonized 43 image datasets, both multi- and hyperspectral, for objective evaluation of algorithms and submitted papers. Researchers can train their algorithms off-line, and evaluate their accuracy on independent spectra test sets. The system returns accuracy and robustness measures, as well as a ranked list of the best methods.

References
  • J. Munoz-Mari et al., ‘HyperLabelMe : A Web Platform for Benchmarking Remote-Sensing Image Classifiers,’ in IEEE Geoscience and Remote Sensing Magazine, vol. 5, no. 4, pp. 79-85, Dec. 2017. doi: 10.1109/MGRS.2017.2762476.

Kernelized EC-ACD: Elliptically Contoured Anomaly Change Detection

Kernelized EC-ACD Image

A simple Toolbox for Anomaly Change Detection (ACD) with Gaussianity assumptions and Elliptically Contoured (EC) distributions, and their kernel-based versions.

References
  • A family of kernel anomaly change detectors. Longbotham, N. and Camps-Valls, G. IEEE Whispers, 2015.
  • Robustness analysis of elliptically contoured multi- and hyperspectral change detection algorithms. M. A. Belenguer, Longbotham, N. and Camps-Valls, G. Submitted, 2016.

Large Margin Filtering SVM

Large Margin Filtering SVM Image

A large margin SVM algorithm that learns convolutional filters, applicable to time series analysis and remote sensing image classification.

References
  • Large margin filtering. Flamary, R., Tuia, D., Labbé, B., Camps-Valls, G., Rakotomamonjy, A. IEEE Transactions on Signal Processing 60(2): 648-659, 2012.
  • Learning spatial filters for multispectral image segmentation. Tuia, D., Camps-Valls, G., Flamary, R., Rakotomamonjy, A. Proceedings of MLSP 2010.

Our Modified libSVM

Modified libSVM Image

Precomputed kernels, e-Huber cost function, accuracy assessment, and other useful features for support vector machine methods.

Semi-Supervised Graph-Based Classification

Semi-Supervised Graph-Based Classification Image

A graph-based method for semi-supervised learning, successfully applied to hyperspectral image classification. Incorporates contextual information via composite kernels and uses the Nyström method for scalability.

References
  • Semi-supervised graph-based hyperspectral image classification. Camps-Valls, G., Bandos Marsheva, T.V., Zhou, D. IEEE Transactions on Geoscience and Remote Sensing 45(10): 3044-3054, 2007.

simpleClass: Simple Classification Toolbox

simpleClass Toolbox Image

A set of train-test simple educational functions for data classification including LDA, QDA, SVM, decision trees, random forests, and Gaussian process classifiers.

References

UKC: Unsupervised Kernel Change Detection

UKC Image

Implements an automatic change detection algorithm using kmeans and gaussian kernel kmeans for clustering the difference image in feature spaces.

References
  • Unsupervised change detection by kernel clustering. Volpi, M., Tuia, D., Camps-Valls, G., Kanevski, M. Proceedings of SPIE 7830, 2010.
  • Unsupervised change detection in the feature space using kernels. Volpi, M., Tuia, D., Camps-Valls, G., Kanevski, M. IGARSS 2011.