Giona Matasci

Learning Relevant Image Features With Multiple-Kernel Classification

The increase in spatial and spectral resolution of the satellite sensors, along with the shortening of the time-revisiting periods, has provided high-quality data for remote sensing image classification. However, the high-dimensional feature space induced by using many heterogeneous information sources precludes the use of simple classifiers: thus, a proper feature selection is required for discarding irrelevant features and adapting the model to the specific problem. This paper proposes to classify the images and simultaneously to learn the relevant features in such high-dimensional scenarios. The proposed method is based on the automatic optimization of a linear combination of kernels dedicated to different meaningful sets of features. Such sets can be groups of bands, contextual or textural features, or bands acquired by different sensors. The combination of kernels is optimized through gradient descent on the support vector machine objective function. Even though the combination is linear, the ranked relevance takes into account the intrinsic nonlinearity of the data through kernels. Since a naive selection of the free parameters of the multiple-kernel method is computationally demanding, we propose an efficient model selection procedure based on the kernel alignment. The result is a weight (learned from the data) for each kernel where both relevant and meaningless image features automatically emerge after training the model. Experiments carried out in multi- and hyperspectral, contextual, and multisource remote sensing data classification confirm the capability of the method in ranking the relevant features and show the computational efficience of the proposed strategy.

Semisupervised Transfer Component Analysis for Domain Adaptation in Remote Sensing Image Classification

In this paper, we study the problem of feature extraction for knowledge transfer between multiple remotely sensed images in the context of land-cover classification. Several factors such as illumination, atmospheric, and ground conditions cause radiometric differences between images of similar scenes acquired on different geographical areas or over the same scene but at different time instants. Accordingly, a change in the probability distributions of the classes is observed. The purpose of this work is to statistically align in the feature space an image of interest that still has to be classified (the target image) to another image whose ground truth is already available (the source image). Following a specifically designed feature extraction step applied to both images, we show that classifiers trained on the source image can successfully predict the classes of the target image despite the shift that has occurred. In this context, we analyze a recently proposed domain adaptation method aiming at reducing the distance between domains, Transfer Component Analysis, and assess the potential of its unsupervised and semisupervised implementations. In particular, with a dedicated study of its key additional objectives, namely the alignment of the projection with the labels and the preservation of the local data structures, we demonstrate the advantages of Semisupervised Transfer Component Analysis. We compare this approach with other both linear and kernel-based feature extraction techniques. Experiments on multi- and hyperspectral acquisitions show remarkable cross- image classification performances for the considered strategy, thus confirming its suitability when applied to remotely sensed images.

SVM-Based Boosting of Active Learning Strategies for Efficient Domain Adaptation

We propose a procedure that efficiently adapts a classifier trained on a source image to a target image with similar spectral properties. The adaptation is carried out by adding new relevant training samples with active queries in the target domain following a strategy specifically designed for the case where class distributions have shifted between the two acquisitions. In fact, the procedure consists of two nested algorithms. An active selection of the pixels to be labeled is performed on a set of candidates of the target image in order to select the most informative pixels. Along the inclusion of the pixels to the training set, the weights associated with these samples are iteratively updated using different criteria, depending on their origin (source or target image). We study this adaptation framework in combination with a SVM classifier accepting instance weights. Experiments on two VHR QuickBird images and on a hyperspectral AVIRIS image prove the validity of the proposed adaptive approach with respect to existing techniques not involving any adjustments to the target domain.

Semi-supervised multiview embedding for hyperspectral data classification

In this paper, a method for semi-supervised multiview feature extraction based on the multiset regularized kernel canonical correlation analysis (kCCA) is proposed for the classification of hyperspectral images. The covariance matrix of this type of data is naturally composed of distinct blocks of spectral channels, which in turn compose the hypercube. To reduce the dimensionality of the data and extract discriminant features taking advantage of this particular structure, a multiview feature extraction method is applied prior to the classification. The proposed scheme exploits both the labels (as a distinct view on the data) and unlabeled pixels into the computation of cross-correlations and regularizations terms. First, we propose a technique to automatically obtain the segmentation of the spectral profile, based on the correlation between channels. Then, the multiset kernel canonical correlation analysis is applied to find a latent space which represents mutually correlated projected views and labels. Experiments on three real hyperspectral images with two linear classifiers and comparisons to state-of-the-art feature extraction methods show the benefits of this approach, which provides classification accuracies equal or superior to those obtained by training classifiers on the original input space but with only a fraction of the original data dimensionality.

Transfer component analysis for domain adaptation in image classification

This contribution studies a feature extraction technique aiming at reducing differences between domains in image classification. The purpose is to find a common feature space between labeled samples issued from a source image and test samples belonging to a related target image. The presented approach, Transfer Component Analysis, finds a transformation matrix performing a joint mapping of the two domains by minimizing a probability distribution distance measure, the Maximum Mean Discrepancy criterion. When predicting on a target image, such a projection allows to apply a supervised classifier trained exclusively on labeled source pixels mapped in this common latent subspace. Promising results are observed on a urban scene captured by a hyperspectral image. The experiments reveal improvements with respect to a standard classification model built on the original source image and other feature extraction techniques.

Domain separation for efficient adaptive active learning

This paper proposes a procedure aimed at efficiently adapting a classifier trained on a source image to a similar target image. The adaptation is carried out through active queries in the target domain following a strategy particularly designed for the case where class distributions have shifted between the two images. We first suggest a pre-selection of candidate pixels issued from the target image by keeping only those samples appearing to be lying in a region of the input space not yet covered by the existing ground truth (source domain pixels). Then, exploiting a classifier integrating instance weights, active queries are performed on the target image. As the inclusion to the training set of the samples progresses, the weights associated with the training pixels are updated using different criteria according to their origin (source or target domain). Experiments on a pair of QuickBird images of urban scenes prove the validity of the proposed approach if compared to existing benchmark methods.

Learning the relevant image features with multiple kernels

This paper proposes to learn the relevant features of remote sensing images for automatic spatio-spectral classification with the automatic optimization of multiple kernels. The method consists of building dedicated kernels for different sets of bands, contextual or textural features. The optimal linear combination of kernels is optimized through gradient descent on the support vector machine (SVM) objective function. Since a na¨ıve implementation is computationally demanding, we propose an efficient model selection procedure based on kernel alignment. The result is a weight — learned from the data — for each kernel where both relevant and meaningless image features emerge after training. Excellent results are observed in both multi and hyperspec-tral image classification, improving standard SVM and other spatio-spectral formulations.

Statistical assessment of dataset shift and model portability in multi-angle in-track image acquisitions

In this study we propose an evaluation of the angular effects altering the spectral response of the land-cover over multi-angle remote sensing image acquisitions. The shift in the statistical distribution of the pixels observed in an in-track sequence of WorldView-2 images is analyzed by means of a kernel-based measure of distance between probability distributions. Afterwards, the portability of supervised classifiers across the sequence is investigated by looking at the evolution of the classification accuracy with respect to the changing observation angle. In this context, the efficiency of various physically and statistically based preprocessing methods in obtaining angle-invariant data spaces is compared and possible synergies are discussed.

Enhanced change detection using nonlinear feature extraction

This paper presents an application of the kernel principal component analysis aiming at spectrally aligning optical images before the application of change detection techniques. The approach relies on the extraction of nonlinear features from a selected subset of pixels representing unchanged areas in the bi-temporal images. Both images are then projected into the new space defined by the eigenvectors associated to largest variance (eigenvalues). In the transformed space, unchanged pixels are mapped next to each other, thus reducing within-class variance. The difference image that results from subtracting the projected datasets is likely to provide a more suitable representation for detecting changes. A subset of two Landsat TM scenes validates the proposed approach. The new representation is studied thanks to the change vector analysis and to the support vector domain description.

A WebGIS platform for the monitoring of Farm Animal Genetic Resources (GENMON)

Background In 2007, the Food and Agriculture Organization of the United Nations (FAO) initiated the Global plan of action for Farm Animal Genetic Resources (FAnGR). The main goal of this plan is to reduce further loss of genetic diversity in farm animals, so as to protect and promote the diversity of farm animal resources. An important step to reach this goal is to monitor and prioritize endangered breeds in the context of conservation programs. Methodology/Web portal implementation The GENMON WebGIS platform is able to monitor FAnGR and to evaluate the degree of endangerment of livestock breeds. The system takes into account pedigree and introgression information, the geographical concentration of animals, the cryo-conservation plan and the sustainability of breeding activities based on socio-economic data as well as present and future land use conditions. A multi-criteria decision tool supports the aggregation of the multi-thematic indices mentioned above using the MACBETH method, which is based on a weighted average using satisfaction thresholds. GENMON is a monitoring tool to reach subjective decisions made by a government agency. It relies on open source software and is available at http://lasigsrv2.epfl.ch/genmon-ch. Results/Significance GENMON allows users to upload pedigree-information (animal ID, parents, birthdate, sex, location and introgression) from a specific livestock breed and to define species and/or region-specific weighting parameters and thresholds. The program then completes a pedigree analysis and derives several indices that are used to calculate an integrated score of conservation prioritization for the breeds under investigation. The score can be visualized on a geographic map and allows a fast, intuitive and regional identification of breeds in danger. Appropriate conservation actions and breeding programs can thus be undertaken in order to promote the recovery of the genetic diversity in livestock breeds in need. The use of the platform is illustrated by means of an example based on three local livestock breeds from different species in Switzerland.