Farid Melgani

Classification of hyperspectral remote sensing images with support vector machines

This paper addresses the problem of the classification of hyperspectral remote sensing images by support vector machines (SVMs). First, we propose a theoretical discussion and experimental analysis aimed at understanding and assessing the potentialities of SVM classifiers in hyperdimensional feature spaces. Then, we assess the effectiveness of SVMs with respect to conventional feature-reduction-based approaches and their performances in hypersubspaces of various dimensionalities. To sustain such an analysis, the performances of SVMs are compared with those of two other nonparametric classifiers (i.e., radial basis function neural networks and the K-nearest neighbor classifier). Finally, we study the potentially critical issue of applying binary SVMs to multiclass problems in hyperspectral data. In particular, four different multiclass strategies are analyzed and compared: the one-against-all, the one-against-one, and two hierarchical tree-based strategies. Different performance indicators have been used to support our experimental studies in a detailed and accurate way, i.e., the classification accuracy, the computational time, the stability to parameter setting, and the complexity of the multiclass architecture. The results obtained on a real Airborne Visible/Infrared Imaging Spectroradiometer hyperspectral dataset allow to conclude that, whatever the multiclass strategy adopted, SVMs are a valid and effective alternative to conventional pattern recognition approaches (feature-reduction procedures combined with a classification method) for the classification of hyperspectral remote sensing data.

An unsupervised approach based on the generalized Gaussian model to automatic change detection in multitemporal SAR images

We present a novel automatic and unsupervised change-detection approach specifically oriented to the analysis of multitemporal single-channel single-polarization synthetic aperture radar (SAR) images. This approach is based on a closed-loop process made up of three main steps: (1) a novel preprocessing based on a controlled adaptive iterative filtering; (2) a comparison between multitemporal images carried out according to a standard log-ratio operator; and (3) a novel approach to the automatic analysis of the log-ratio image for generating the change-detection map. The first step aims at reducing the speckle noise in a controlled way in order to maximize the discrimination capability between changed and unchanged classes. In the second step, the two filtered multitemporal images are compared to generate a log-ratio image that contains explicit information on changed areas. The third step produces the change-detection map according to a thresholding procedure based on a reformulation of the Kittler-Illingworth (KI) threshold selection criterion. In particular, the modified KI criterion is derived under the generalized Gaussian assumption for modeling the distributions of changed and unchanged classes. This parametric model was chosen because it is capable of better fitting the conditional densities of classes in the log-ratio image. In order to control the filtering step and, accordingly, the effects of the filtering process on change-detection accuracy, we propose to identify automatically the optimal number of despeckling filter iterations [Step 1] by analyzing the behavior of the modified KI criterion. This results in a completely automatic and self-consistent change-detection approach that avoids the use of empirical methods for the selection of the best number of filtering iterations. Experiments carried out on two sets of multitemporal images (characterized by different levels of speckle noise) acquired by the European Remote Sensing 2 satellite SAR sensor confirm the effectiveness of the proposed unsupervised approach, which results in change-detection accuracies very similar to those that can be achieved by a manual supervised thresholding.

Toward an Optimal SVM Classification System for Hyperspectral Remote Sensing Images

Recent remote sensing literature has shown that support vector machine (SVM) methods generally outperform traditional statistical and neural methods in classification problems involving hyperspectral images. However, there are still open issues that, if suitably addressed, could allow further improvement of their performances in terms of classification accuracy. Two especially critical issues are: 1) the determination of the most appropriate feature subspace where to carry out the classification task and 2) model selection. In this paper, these two issues are addressed through a classification system that optimizes the SVM classifier accuracy for this kind of imagery. This system is based on a genetic optimization framework formulated in such a way as to detect the best discriminative features without requiring the a priori setting of their number by the user and to estimate the best SVM parameters (i.e., regularization and kernel parameters) in a completely automatic way. For these purposes, it exploits fitness criteria intrinsically related to the generalization capabilities of SVM classifiers. In particular, two criteria are explored, namely: 1) the simple support vector count and 2) the radius margin bound. The effectiveness of the proposed classification system in general and of these two criteria in particular is assessed both by simulated and real experiments. In addition, a comparison with classification approaches based on three different feature selection methods is reported, i.e., the steepest ascent (SA) algorithm and two other methods explicitly developed for SVM classifiers, namely: 1) the recursive feature elimination technique and 2) the radius margin bound minimization method

Classification of Electrocardiogram Signals With Support Vector Machines and Particle Swarm Optimization

The aim of this paper is twofold. First, we present a thorough experimental study to show the superiority of the generalization capability of the support vector machine (SVM) approach in the automatic classification of electrocardiogram (ECG) beats. Second, we propose a novel classification system based on particle swarm optimization (PSO) to improve the generalization performance of the SVM classifier. For this purpose, we have optimized the SVM classifier design by searching for the best value of the parameters that tune its discriminant function, and upstream by looking for the best subset of features that feed the classifier. The experiments were conducted on the basis of ECG data from the Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) arrhythmia database to classify five kinds of abnormal waveforms and normal beats. In particular, they were organized so as to test the sensitivity of the SVM classifier and that of two reference classifiers used for comparison, i.e., the k-nearest neighbor (kNN) classifier and the radial basis function (RBF) neural network classifier, with respect to the curse of dimensionality and the number of available training beats. The obtained results clearly confirm the superiority of the SVM approach as compared to traditional classifiers, and suggest that further substantial improvements in terms of classification accuracy can be achieved by the proposed PSO-SVM classification system. On an average, over three experiments making use of a different total number of training beats (250, 500, and 750, respectively), the PSO-SVM yielded an overall accuracy of 89.72% on 40438 test beats selected from 20 patient records against 85.98%, 83.70%, and 82.34% for the SVM, the kNN, and the RBF classifiers, respectively.

Automatic Analysis of GPR Images: A Pattern-Recognition Approach

In this paper, we propose a novel pattern-recognition system to identify and classify buried objects from ground-penetrating radar (GPR) imagery. The entire process is subdivided into four steps. After a preprocessing step, the GPR image is thresholded to put under light the regions containing potential objects. The third step of the system consists of automatically detecting the objects in the obtained binary image by means of a search of linear/hyperbolic patterns formulated within a genetic optimization framework. In the genetic optimizer, each chromosome models the apex position and the curvature associated with the candidate pattern, while the fitness function expresses the Hamming distance between that pattern and the binary image content. Finally, in the fourth step, the problem of the recognition of the material type of the identified objects is approached as a classification issue, which is solved by means of an opportune feature-extraction strategy and a support vector machine classifier. To illustrate the performances of the proposed system, we conducted a thorough experimental study based on GPR images generated by a GPR simulator based on the finite-difference time-domain method so as to construct different acquisition scenarios by varying the number of buried objects, their position, their size, their shape, and their material type. In general, the obtained experimental results show that the proposed system exhibits promising performances both in terms of object detection and material recognition.

Robust support vector regression for biophysical variable estimation from remotely sensed images

This letter introduces the epsiv-Huber loss function in the support vector regression (SVR) formulation for the estimation of biophysical parameters extracted from remotely sensed data. This cost function can handle the different types of noise contained in the dataset. The method is successfully compared to other cost functions in the SVR framework, neural networks and classical bio-optical models for the particular case of the estimation of ocean chlorophyll concentration from satellite remote sensing data. The proposed model provides more accurate, less biased, and improved robust estimation results on the considered case study, especially significant when few in situ measurements are available

Nearest Neighbor Classification of Remote Sensing Images With the Maximal Margin Principle

In this paper, we present a new variant of the k-nearest neighbor (kNN) classifier based on the maximal margin principle. The proposed method relies on classifying a given unlabeled sample by first finding its k-nearest training samples. A local partition of the input feature space is then carried out by means of local support vector machine (SVM) decision boundaries determined after training a multiclass SVM classifier on the considered k training samples. The labeling of the unknown sample is done by looking at the local decision region to which it belongs. The method is characterized by resulting global decision boundaries of the piecewise linear type. However, the entire process can be kernelized through the determination of the k -nearest training samples in the transformed feature space by using a distance function simply reformulated on the basis of the adopted kernel. To illustrate the performance of the proposed method, an experimental analysis on three different remote sensing datasets is reported and discussed.

Unsupervised Change Detection in Multispectral Remotely Sensed Imagery With Level Set Methods

In this paper, the unsupervised change-detection problem in remote sensing images is formulated as a segmentation issue where the discrimination between changed and unchanged classes in the difference image is achieved by defining a proper energy functional. The minimization of this functional is carried out by means of a level set method which iteratively seeks to find a global optimal contour splitting the image into two mutually exclusive regions associated with changed and unchanged classes, respectively. In order to increase the robustness of the method to noise and to the choice of the initial contour, a multiresolution implementation, which performs an analysis of the difference image at different resolution levels, is proposed. The experimental results obtained on three different multitemporal remote sensing images acquired by low- as well as high-spatial-resolution optical remote sensing sensors suggest a clear superiority of the proposed approach compared with state-of-the-art change-detection methods.

Clustering of Hyperspectral Images Based on Multiobjective Particle Swarm Optimization

In this paper, we present a new methodology for clustering hyperspectral images. It aims at simultaneously solving the following three different issues: 1) estimation of the class statistical parameters; 2) detection of the best discriminative bands without requiring the a priori setting of their number by the user; and 3) estimation of the number of data classes characterizing the considered image. It is formulated within a multiobjective particle swarm optimization (MOPSO) framework and is guided by three different optimization criteria, which are the log-likelihood function, the Bhattacharyya statistical distance between classes, and the minimum description length (MDL). A detailed experimental analysis was conducted on both simulated and real hyperspectral images. In general, the obtained results show that interesting classification performances can be achieved by the proposed methodology despite its completely unsupervised nature.

Contextual reconstruction of cloud-contaminated multitemporal multispectral images

The frequent presence of clouds in passive remotely sensed imagery severely limits its regular exploitation in various application fields. Thus, the removal of cloud cover from this imagery represents an important preprocessing task consisting in the reconstruction of cloud-contaminated data. The intent of this study is to propose two novel general methods for the reconstruction of areas obscured by clouds in a sequence of multitemporal multispectral images. Given a cloud-contaminated image of the sequence, each area of missing measurements is reconstructed through an unsupervised contextual prediction process that reproduces the local spectro-temporal relationships between the considered image and an opportunely selected subset of the remaining temporal images. In the first method, the contextual prediction process is implemented by means of an ensemble of linear predictors, each trained over a local multitemporal region that is spectrally homogeneous in each temporal image of the selected subset. In order to obtain such regions, each temporal image is locally classified by an unsupervised classifier based on the expectation-maximization (EM) algorithm. In the second method, the local spectro-temporal relationships are reproduced by a single nonlinear predictor based on the support vector machines (SVM) approach. To illustrate the performance of the two proposed methods, an experimental analysis on a sequence of three temporal images acquired by the Landsat-7 Enhanced Thematic Mapper Plus sensor over a total period of four months is reported and discussed. It includes a detailed simulation study that aims at assessing with different reconstruction quality criteria the accuracy of the methods in different qualitative and quantitative cloud contamination conditions. Compared with two techniques based on compositing algorithms for cloud removal, the proposed methods show a clear superiority, which makes them a promising and useful tool in solving the considered problem, whose great complexity is commensurate with its practical importance.