HONEINE , Paul

In this paper, we study the problem of decomposing spectra in hyperspectral data into the sum of pure spectra, or endmembers. We propose to jointly extract the endmembers and estimate the corresponding fractions, or abundances. For this purpose, we show that these abundances can be easily computed using volume of simplices, from the same...

De nombreuses études ont récemment montré l'avantage de l'approche géométrique en démélange de données hyperspectrales. Elle permet d'identifier les signatures spectrales des composants purs. Jusqu'ici, l'estimation des fractions d'abondance a toujours été réalisée dans un second temps, par résolution d'un problème inverse généralement. Dans...

The pre-image problem is a challenging research subject pursued by many researchers in machine learning. Kernel-based machines seek some relevant feature in a reproducing kernel Hilbert space (RKHS), optimized in a given sense, such as kernel-PCA algorithms. Operating the latter for denoising requires solving the pre-image problem, i.e....

In hyperspectral imaging, spectral unmixing is one of the most challenging and fundamental problems. It consists of breaking down the spectrum of a mixed pixel into a set of pure spectra, called endmembers, and their contributions, called abundances. Many endmember extraction techniques have been proposed in literature, based on either a...

Support vector machines have been investigated with success for hyperspectral data classification. In this paper, we propose a new kernel to measure spectral similarity, called the angular kernel. We provide some of its properties, such as its invariance to illumination energy, as well as connection to previous work. Furthermore, we show that...

While the nonlinear mapping from the input space to the feature space is central in kernel methods, the reverse mapping from the feature space back to the input space is also of primary interest. This is the case in many applications, including kernel principal component analysis (PCA) for signal and image denoising. Unfortunately, it turns out...

This paper deals with the linear unmixing problem in hyperspectral data processing, and in particular the estimation of the fractional abundances under sum-to-one and non-negativity constraints. For this purpose, we propose to adapt the reflect-then-combine iterative technique, initially derived by Cimmino. Several strategies are studied in...

The invention relates to a system and a method for locating at least one target using au array of transceivers or sensors, in which at least a portion has a known geographic location, each comprising data processing means implementing at least one algorithm for locating a target, means for transmitting/receiving a signal that decreases with the...

In this paper, we investigate a novel online one-class classification method. We consider a least-squares optimization problem, where the model complexity is controlled by the coherence criterion as a sparsification rule. This criterion is coupled with a simple updating rule for online learning, which yields a low computational demanding...

The estimation of fractional abundances under physical constraints is a fundamental problem in hyperspectral data processing. In this paper, we propose to adapt Kaczmarz's cyclic projections to solve this problem. The main contribution of this work is two-fold: On the one hand, we show that the non-negativity and the sum-to-one constraints can...

The one-class classification problemis often addressed by solving a constrained quadratic optimization problem, in the same spirit as support vector machines. In this paper, we derive a novel one-class classification approach, by investigating an original sparsification criterion. This criterion, known as the coherence criterion, is based on a...

Nonlinear unmixing of hyperspectral images has generated considerable interest among researchers, as it may overcome some inherent limitations of the linear mixing model. In this paper, we formulate the problem of estimating abundances of a nonlinear mixture of hyperspectral data based on a new multi-kernel learning paradigm. Experiments are...

In this paper, we study the multiclass classification problem. We derive a framework to solve this problem by providing algorithms with the complexity of a single binary classifier. The resulting multiclass machines can be decomposed into two categories. The first category corresponds to vector-output machines, where we develop several...

The one-class classification has been successfully applied in many communication, signal processing, and machine learning tasks. This problem, as defined by the one-class SVM approach, consists in identifying a sphere enclosing all (or the most) of the data. The classical strategy to solve the problem considers a simultaneous estimation of both...

In this paper, we derive an adaptive one-class classification algorithm. We propose a least-squares formulation of the problem, where the model complexity is controlled by a parsimony criterion. We consider the linear approximation criterion, and we couple it with a simple adaptive updating algorithm for online learning. We conduct experiments...

Spectral unmixing is an important issue to analyze remotely sensed hyperspectral data. Although the linear mixture model has obvious practical advantages, there are many situations in which it may not be appropriate and could be advantageously replaced by a nonlinear one. In this paper, we formulate a new kernel-based paradigm that relies on...

International audience

This paper deals with multi-class classification problems. Many methods extend binary classifiers to operate a multi-class task, with strategies such as the one-vs-one and the one-vs-all schemes. However, the computational cost of such techniques is highly dependent on the number of available classes. We present a method for multi-class...

In hyperspectral images, pixels are mixtures of spectral components associated to pure materials. Although the linear mixture model is the most studied case, nonlinear models have been taken into consideration to overcome some limitations of the linear model. In this paper, nonlinear hyperspectral unmixing problem is studied through...

Integrating spatial information into hyperspectral unmixing procedures has been shown to have a positive effect on the estimation of fractional abundances due to the inherent spatial-spectral duality in hyperspectral scenes. However, current research works that take spatial information into account are mainly focused on the linear mixing model....

En imagerie hyperspectrale, dans un contexte supervisé, chaque vecteur-pixel résulte d'un mélange de spectres de composants purs dont on voudrait estimer les proportions. Récemment, afin de résoudre ce problème en palliant les limitations des modèles linéaires, des méthodes de démélange non-linéaires des données hyperspectrales ont été...

Within the area of hyperspectral data processing, nonlinear unmixing techniques have emerged as promising alternatives for overcoming the limitations of linear methods. In this paper, we consider the class of post-nonlinear mixing models of the partially linear form. More precisely, these composite models consist of a linear mixing part and a...

International audience

Many processes exhibit exponential behavior. When kernel-based machines are applied on this type of data, conventional kernels such as the Gaussian kernel are not appropriate. In this paper, we derive kernels adapted to time series of exponential decay or growth processes. We provide a theoretical study of these kernels, including the issue of...

Cet article traite du problème de classification multi-classe en reconnaissance des formes. La résolution de ce type de problèmes nécessite des algorithmes au coût calculatoire souvent beaucoup plus élevé que les méthodes d'apprentissage dédiées à la classification binaire. On propose dans cet article une nouvelle formulation pour la conception...