Search

International audience; In this paper, we demonstrate that algebraic topology can be used to perform 2D+t object detection. After the construction of a topological complex for a 2D+t image sequence, we build a nested sequence of cell complexes on which relative persistent homology is computed. The relative homology adds to “absolute” homology the computation of classes related to the first and last frames of the sequence. By identifying 2D chains with large life spans, the most persistent classes are extracted. This allows for the identification of the interesting parts in a sequence and for the detection of the movement of objects despite continuous deformations in the image domain. The results obtained on a synthetic image and on two real biomedical images with moving vesicles recorded by a quantitative phase time‐lapse technique show the potential of this method. Comparing the method with a newly developed tracking tool proves that the strength of this method is its independence from prior parameters.

In recent years, driven by new standards and image brand, product quality has become highly critical in the food industry. Particularly concerning the beverage industry, the potential presence of very small foreign objects, especially glass fragments, has to be checked. The visible optical domain offers an advantageous alternative solution to the expensive and intrusive x-ray systems. However, such a solution requires the development of robust processing algorithms satisfying real-time industrial constraints, which is a very challenging task. A detection method based on kurtosis and local information that emphasizes connected components is proposed, as well as two robust criteria that give a greater robustness to the detector. A specific kurtosis is calculated using an inclined frame definition that is based on both time and spatial dimensions. Such kurtosis increases the detection capacity of the moving objects and allows the estimation of their direction without adding more costs. The kurtosis being very sensitive to outliers, noisy objects that are very small are filtered, allowing efficient detection of foreign objects, such as glass fragments, as long as they are bigger than the noisy objects. In case of big foreign objects, the low sensitivity of the kurtosis is compensated for by the large detected surface. The proposed detection method can be directly applied on video sequences acquired by a standard RGB camera in industrial environments. The experimental results show the effectiveness of the method in detecting real random foreign objects, regardless of their size or their transparency, in various semiopaque bottles.

International audience

International audience

Monitoring the health of ancient artworks requires adequate prudence because of the sensitive nature of these materials. Classical techniques for identifying the development of faults rely on acoustic testing. These techniques, being invasive, may result in causing permanent damage to the material, especially if the material is inspected periodically. Non destructive testing has been carried out for different materials since long. In this regard, non-invasive systems were developed based on infrared thermometry principle to identify the faults in artworks. The test artwork is heated and the thermal response of the different layers is captured with the help of a thermal infrared camera. However, prolonged heating risks overheating and thus causing damage to artworks and an alternate approach is to use pseudo-random binary sequence excitations. The faults in the artwork, though, cannot be detected on the captured images, especially if their strength is weak. The weaker faults are either masked by the stronger ones, by the pictorial layer of the artwork or by the non-uniform heating. This work addresses the detection and localization of the faults through a wavelet based subspace decomposition scheme. The proposed scheme, on one hand, allows to remove the background while, on the other hand, removes the undesired high frequency noise. It is shown that the detection parameter is proportional to the diameter and the depth of the fault. A criterion is proposed to select the optimal wavelet basis along with suitable level selection for wavelet decomposition and reconstruction. The proposed approach is tested on a laboratory developed test sample with known fault locations and dimensions as well as real artworks. A comparison with a previously reported method demonstrates the efficacy of the proposed approach for fault detection in artworks.

The estimation of sinkhole-induced ground displacement is an important issue for monitoring soil structures. Distributed optical fiber sensors (DOFSs) composed of an interrogator based on scattering effects in an optical fiber cable sensing element can be used to assess ground displacement. These sensors provide longitudinal strain measurements of the soil structure. This paper proposes a methodology that enables the estimation of displacement fields in the soil structure when a sinkhole appears. It also exposes an experiment that was carried out to create an artificial sinkhole instrumented by optical fiber sensors. This is the first time that those sensors are used to provide submillimeter vertical displacements. The first step of the methodology is to model the ground displacement under 2-D conditions. The longitudinal strain measured by a DOFS can thus be linked to the displacement of the structure. This model is described by the following parameters: 1) the spatial extent of the displacement signature; 2) a coefficient that depends on the interface between the optical fiber cable and the soil; 3) the depth of the sinkhole; and 4) the maximal vertical displacement. The second step consists of the estimation of each parameter independently. The spatial extension is given by fitting the measured strain signature with the empirical model. The depth of the sinkhole can be determined by the measurement of the spatial extension of the ground-displacement profile at several observation depths in the structure. Finally, the vertical maximal displacement is furnished with high precision.

In this paper, we develop a methodology originating from algebraic topology, and we demonstrate its capability of performing multidimensional object segmentation without the need of prior parameters. Persistent homology is a method used in algebraic topology to study qualitative features of data that persist across varying scales. The construction of a topological complex on the image is followed by a filtration scheme that consists of composing a nested sequence of cell complexes on which the persistent homology is computed. The most persistent homology classes are extracted by identifying 1D and 2D chains with large lifespans, which allows salient objects in 2D and 3D images to be segmented and detected. A comparison between this method and other segmentation techniques on a synthetic image shows the advantages of the proposed method. The strength of this technique is reflected in its insensitivity to continuous deformations and perturbations of the input function and in its independence of prior parameters. The results obtained on real and biomedical 2D and 3D images also demonstrate the potential of this method.

In this paper, we show that topological persistence can be employed in biomedical image processing to perform object segmentation. First we model the pixels of the image by combinatorial transformation into a cubical complex that we will call the pixels' complex. Then a nested sequence of complexes is built on which the persistent homology is computed. By identifying the 1D chains with large life spans, the most persistent classes are extracted. This allows to segment the salient objects in the biomedical image and to spot their components. An example was applied first on a toy image of coins that demonstrate the applicability of the method. Results on two real biomedical images, the first recorded by a quantitative phase technique and the second represents a classical image of cells show the potential of this technique. The insensitivity to continuous deformations and the independence to prior parameters reveals the strength of this method.

Infrared spectroscopy provides useful information on the molecular compositions of biological systems related to molecular vibrations, overtones, and combinations of fundamental vibrations. Mid-infrared (MIR) spectroscopy is sensitive to organic and mineral components and has attracted growing interest in the development of biomarkers related to intrinsic characteristics of lignocellulose biomass. However, not all spectral information is valuable for biomarker construction or for applying analysis methods such as classification. Better processing and interpretation can be achieved by identifying discriminating wavenumbers. The selection of wavenumbers has been addressed through several variable- or feature-selection methods. Some of them have not been adapted for use in large data sets or are difficult to tune, and others require additional information, such as concentrations. This paper proposes a new approach by combining a naïve Bayesian classifier with a genetic algorithm to identify discriminating spectral wavenumbers. The genetic algorithm uses a linear combination of an a posteriori probability and the Bayes error rate as the fitness function for optimization. Such a function allows the improvement of both the compactness and the separation of classes. This approach was tested to classify a small set of maize roots in soil according to their biodegradation process based on their MIR spectra. The results show that this optimization method allows better discrimination of the biodegradation process, compared with using the information of the entire MIR spectrum, the use of the spectral information at wavenumbers selected by a genetic algorithm based on a classical validity index or the use of the spectral information selected by combining a genetic algorithm with other methods, such as Linear Discriminant Analysis. The proposed method selects wavenumbers that correspond to principal vibrations of chemical functional groups of compounds that undergo degradation/conversion during the biodegradation of lignocellulosic biomass.

International audience; Early detection of leakages in hydraulic infrastructures is important to ensure their safety and security. Significant flow of water through the dike can be an indicator of internal erosion and results in a thermal anomaly. Temperature measurements are therefore capable of revealing information linked to leakage. Optical fiber-based distributed temperature sensors present an economically viable and reliable solution for recording spatio-temporal temperature data over long distances, with spatial and temperature resolutions of 1m and 0.05 C, respectively. The acquired data are influenced by several factors, among them water leakages, heat transfer through the above soil depth, seasonal thermal variations, and the geomechanical environment. Soil properties such as permeability alter the acquired signal locally. This article presents leakage detection methods based on signal processing of the raw temperature data from optical fiber sensors. The first approach based on source separation identifies leakages by separating them from the non-relevant information. The second approach presents a potential alarm system based on the analysis of daily temperature variations. Successful detection results for simulated as well as real experimental setups of Electricité de France are presented.

International audience; Distributed optical fiber sensors have gained an increasingly prominent role in structural-health monitoring. These are composed of an optical fiber cable in which a light impulse is launched by an opto-electronic device. The scattered light is of interest in the spectral domain: the spontaneous Brillouin spectrum is centered on the Brillouin frequency, which is related to the local strain and temperature changes in the optical fiber. When coupled with an industrial Brillouin optical time-domain analyzer (B-OTDA), an optical fiber cable can provide distributed measurements of strain and/or temperature, with a spatial resolution over kilometers of 40 cm. This paper focuses on the functioning of a B-OTDA device, where we address the problem of the improvement of spatial resolution. We model a Brillouin spectrum measured within an integration base of 1 m as the superposition of the elementary spectra contained in the base. Then, the spectral distortion phenomenon can be mathematically explained: if the strain is not constant within the integration base, the Brillouin spectrum is composed of several elementary spectra that are centered on different local Brillouin frequencies. We propose a source separation methodology approach to decompose a measured Brillouin spectrum into its spectral components. The local Brillouin frequencies and amplitudes are related to a portion of the integration base where the strain is constant. A layout algorithm allows the estimation of a strain profile with new spatial resolution chosen by the user. Numerical tests enable the finding of the optimal parameters, which provides a reduction to 1 cm of the 40-cm spatial resolution of the B-OTDA device. These parameters are highlighted during a comparison with a reference strain profile acquired by a 5-cm-resolution Rayleigh scatter analyzer under controlled conditions. In comparison with the B-OTDA strain profile, our estimated strain profile has better accuracy, with centimeter spatial resolut ion.

In this paper, a pulsed Infrared thermography technique using a homogeneous heat provided by a laser source is used for the non-destructive evaluation of paint coating thickness variations. Firstly, numerical simulations of the thermal response of a paint coated sample are performed. By analyzing the thermal responses as a function of thermal properties and thickness of both coating and substrate layers, optimal excitation parameters of the heating source are determined. Two characteristic parameters were studied with respect to the paint coating layer thickness variations. Results obtained using an experimental test bench based on the pulsed Infrared thermography laser technique are compared with those given by a classical Eddy current technique for paint coating variations from 5 to 130 μm. These results demonstrate the efficiency of this approach and suggest that the pulsed Infrared thermography technique presents good perspectives to characterize the heterogeneity of paint coating on large scale samples with other heating sources.

In this paper, a new treatment approach is proposed t o overcome the harmful effects of non-uniform heating in active infrared thermography test. It is based on a linear decomposition of the thermographic surface response evolution using a PLSR algorithm based on the SVD decomposition. The approach is then applied to coating layer thickness evaluation from thermographic response. Results show that the proposed technique allows to discard the influence of experimental conditions (positions, intensity) of heating sources and to enhance the robustness of the pulsed thermography technique for coating thickness heterogeneity evaluation.

This paper presents an automatic speaker recognition system based on the fusion of a speaker identification and an identity verification system. Two passwords are required for each user. The proposed system firstly performs a segmentation of the words based on a wavelet transform algorithm. The identification systems are GMM-based and use the first password in order to identify both speaker (text independent) and words (speaker independent). Each system provides an ordered list of likelihoods for both speakers and words which is combinated in order to identify a pair Speaker/Password. An identity verification system is applied on the second password. This paper evaluates each system, the global system and the noise influence. MOTS-CLES: reconnaissance du locuteur, reconnaissance des mots, fusion de decision, GMM/UBM, transformee en ondelettes.

Malignant melanoma (MM) is the most severe tumor affecting the skin and accounts for three quarters of all skin cancer deaths. Raman spectroscopy is a promising nondestructive tool that has been increasingly used for characterization of the molecular features of cancerous tissues. Different multivariate statistical analysis techniques are used in order to extract relevant information that can be considered as functional spectroscopic descriptors of a particular pathology. Paraffin embedding (waxing) is a highly efficient process used to conserve biopsies in tumor banks for several years. However, the use of non-dewaxed formalin-fixed paraffin-embedded tissues for Raman spectroscopic investigations remains very restricted, limiting the development of the technique as a routine analytical tool for biomedical purposes. This is due to the highly intense signal of paraffin, which masks important vibrations of the biological tissues. In addition to being time consuming and chemical intensive, chemical dewaxing methods are not efficient and they leave traces of the paraffin in tissues, which affects the Raman signal. In the present study, we use independent component analysis (ICA) on Raman spectral images collected on melanoma and nevus samples. The sources obtained from these images are then used to eliminate, using non-negativity constrained least squares (NCLS), the paraffin contribution from each individual spectrum of the spectral images of nevi and melanomas. Corrected spectra of both types of lesion are then compared and classified into dendrograms using hierarchical cluster analysis (HCA).

Resume La spectroscopie Raman est une technique efficace d’analyse de la composition moleculaire d’echantillons biologiques. Des methodes de separation de sources peuvent etre utilisees pour separer efficacement les informations denses melangees dans les spectres Raman acquis. Des deformations telles que le fond de fluorescence, le desalignement des pics ou l’heterogeneite de la largeur des pics remettent en cause le modele generatif lineaire requis par les methodes de separation de sources. Des pretraitements sont necessaires pour compenser ces effets perturbateurs et ainsi rendre les spectres Raman descriptibles par un modele generatif lineaire. Nous montrons dans cet article que l’efficacite des methodes de separation de sources est profondement dependante des pretraitements appliques aux donnees. L’etude du deparaffinage numerique du signal Raman d’une biopsie de peau humaine prouve les ameliorations ainsi apportees. Les methodes de separation de sources appliquees dans cet article sont : un algorithme classique d’analyse en composantes independantes (ACI), nomme joint approximate diagonalization of eigenmatrices (JADE) et deux methodes de separation de sources positives, appelees la factorisation en matrices non-negatives (FMN) et la separation de sources positives par maximum de vraisemblance (SSPMV).

Raman spectra provide wealthy but complex information about the chemical constituents of biological samples. Digital processing techniques are usually needed to extract the spectra of chemical constituents and their associated concentration profiles. However, spectral signatures may admit transformations from those recorded on pure constituents and these techniques require a priori knowledge of spectra to be estimated. We propose in this study to analyse paraffin-embedded skin biopsies of malignant and benign tumors dedicated to oncology researches by Raman spectroscopy and advanced signal processing methods. We show that the commonly used principal component analysis (PCA) does not give physically interpretable estimators of spectra and associated concentration profiles. Based on a linear model and taking into account the statistical properties of spectra, independent component analysis (ICA) is used to better estimate the spectra of chemical constituents. The estimators of associated concentration profiles are no longer orthogonal and have only positive values, contrary to PCA. ICA allows to model the paraffin by three Raman spectra and provides good estimators of underlying spectra of the human skin, which is of great interest in oncology since the retrieval of spectral features of different types of skin tumors is sufficient for their discrimination.

Infrared spectroscopy can provide useful information of the biomass composition and has been extensively used in several domains such as biology, food science, pharmaceutical, petrochemical, agricultural applications, etc. However, not all spectral information are valuable for biomarkers construction or for applying regression or classification models and by identifying interesting wavenumbers a better processing and interpretation can be achieved. The selection of optimal subsets has been addressed through several variable or feature selection methods including genetic algorithms. Some of them are not adapted on large data, others require additional information such as concentrations or are difficult to tune. This paper proposes an alternative approach by considering a weighted Euclidean distance. We show on real Mid-infrared spectra that this constrained nonlinear optimizer allows identifying the wavenumbers that best highlights the discrimination within the periods of the biodegradation process of the ligno-cellulosic biomass. These results are compared with previous ones obtained by a genetic algorithm.

In this paper, we propose a factor weighted fuzzy c-means clustering algorithm. Based on the inverse of a covariance factor, which assesses the collinearity between the centers and samples, this factor takes also into account the compactness of the samples within clusters. The proposed clustering algorithm allows to classify spherical and non-spherical structural clusters, contrary to classical fuzzy c-means algorithm that is only adapted for spherical structural clusters. Compared with other algorithms designed for non-spherical structural clusters, such as Gustafson-Kessel, Gath-Geva or adaptive Mahalanobis distance-based fuzzy c-means clustering algorithms, the proposed algorithm gives better numerical results on artificial and real well known data sets. Moreover, this algorithm can be used for high dimensional data, contrary to other algorithms that require the computation of determinants of large matrices. Application on Mid-Infrared spectra acquired on maize root and aerial parts of Miscanthus for the classification of vegetal biomass shows that this algorithm can successfully be applied on high dimensional data.

Mid-infrared (MIR) and near-infrared (NIR) spectroscopy provide useful information on the molecular composition of biological systems. Because they are sensitive to organic and mineral components, there is a growing interest in these techniques for the development of biomarkers that reflect intrinsic characteristics of plants and their mode of degradation. Due to their complexity and complementary nature, an important challenge is the combining of MIR and NIR information to identify discriminating wavenumbers in each wavenumber region, with the ultimate goal of assessing the biodegradation process of a lignocellulosic biomass at different time scales. This work investigates the potential of using the outer product to combine MIR and NIR spectra to highlight the connections between fundamental molecular vibrations and their combinations and bonds. Because this operation yields high-dimensional spectra, we propose to use a genetic algorithm to select the most discriminant wavenumbers within the degradation process. The results from two lignocellulosic biomasses with different biodegradation kinetics, miscanthus aerial parts and maize roots, confirm that the outer product combination of MIR and NIR spectral information allows a better discrimination of the biodegradation kinetic compared with the simple concatenation of MIR and NIR spectra or with the use of MIR or MIR spectral information separately. We show that the genetic algorithm selects wavenumbers that correspond to principal vibrations of chemical functional groups of compounds that undergo degradation/conversion during the biodegradation of the lignocellulosic biomass.

Multicomponent sensor arrays now are commonly used in seismic acquisition to record polarized waves. In this article, we use a three-mode model (polarization mode, distance mode, and temporal mode) to take into account the specific structure of signals that are recorded with these arrays, providing a data-structure-preserving processing. With the suggested model, we propose a multilinear decomposition named higher-order singular value decomposition and unimodal independent component analysis (HOSVD/unimodal ICA) to split the recorded three-mode data into two orthogonal subspaces: the signal and noise subspaces. This decomposition allows the separation and identification of polarized waves with infinite apparent horizontal propagation velocity. The HOSVD leads to a definition of a subspace method that is the counterpart of the well-known subspace method for matrices that is driven by singular value decomposition (SVD), a classic tool in monocomponent array processing. The proposed three-mode subspace decomposition provides a multicomponent wave-separation algorithm. To enhance the separation result, when the signal-to-noise ratio is low or when orthogonality constraints are not well adapted to the recorded waves, a unimodal-ICA step is included on the temporal mode. Doing this replaces the classic orthogonality constraints between estimated waves with independence constraints that might allow better recovery of recorded seismic waves. A simulation on realistic two-component (2C) geophysical data shows qualitative and quantitative improvements for the wavefield-separation results. The relative-mean-square errors between the original and estimated signal subspaces are, respectively, 52% for SVD applied on each component separately, 27.4% for HOSVD-based technique applied to the whole three-mode dataset, and 7.3% for HOSVD/unimodal-ICA technique. The efficiency of the three-mode subspace decompositions also is shown on real three-component (3C) geophysical data. These results emphasize the potential of the HOSVD/unimodal-ICA subspace method for multicomponent seismic-wave separation.

Raman spectroscopy is employed to record spectra which highlight vibrational information of biological structures. These spectra provide useful information about molecular composition of a tissue. However, the tissues are usually embedded into paraffin for a preservation purpose. The Independent Component Analysis (ICA) technique can be used to numerically dewax Raman spectra and to extract the specific information of the tissue. We show in this paper that the paraffin must be modeled by a three-source model in order to succeed. This linear model, confirmed also by analyzing paraffin blocks exclusively, should be preferred instead of commonly used one-source model.

International audience; In multisensor signal processing (underwater acoustics, geophysics, etc.), the initial dataset is usually separated into complementary subspaces called signal and noise subspaces in order to enhance the signal-to-noise ratio. The singular value decomposition (SVD) is a useful tool to achieve this separation. It provides two orthogonal matrices that convey information on normalized wavelets and propagation vectors. As signal and noise subspaces are on the whole well evaluated, usually the SVD procedure cannot correctly extract only the source waves with a high degree of sensor to sensor correlation. This is due to the constraint given by the orthogonality of the propagation vectors. To relax this condition, exploiting the concept of independent component analysis (ICA), we propose another orthogonal matrix made up of statistically independent normalized wavelets. By using this combined SVD-ICA procedure, we obtain a better separation of these source waves in the signal subspace. Efficiency of this new separation procedure is shown on synthetic and real datasets.

The analysis of lignocellulosic materials is crucial to optimizing the conversion efficiencies in biorefineries and to studying crop residue input to soil nutrient cycles. Mid-infrared (MIR) and near-infrared (NIR) spectroscopies are rapid, simple, and nondestructive methods for the determination of biomass compositions. However, the analysis of a small set of plant biomass is not generally possible with conventional methods of data processing, such as partial least squares. Additionally, IR spectra do not distribute spherically in the data space. To overcome these problems, we propose a weighted-covariance factor fuzzy C-means clustering method combined with bootstrapping. The algorithm can classify spherical and nonspherical clusters, in contrast to classic fuzzy C-means, which is only adapted to spherical clusters. Bootstrapping enables resampling of the available spectra to generate several datasets on which the classification is performed. This unsupervised clustering methodology was tested to classify a small set of maize roots in soil according to genotype or period of their biodegradation process based on their NIR and MIR spectra. This methodology is applied to determine the optimal pretreatment of IR spectra, to study the contribution of the combination of MIR and NIR spectra and to compare the results on spectral and chemical data. The results show that the best methods of pretreatment are the first-order Savitzky-Golay derivative followed by standard normal variate. The MIR spectra produce a better result than NIR spectra for the initial characterization and for dynamic samples, while MIR spectra acquired on raw samples, without soluble extraction, provided better classification than wet chemistry.

International audience; To ensure stability and durability of engineering structure in natural soil, optical fiber sensors have gained interest over last decade. In addition to conventional geophysical sensors, Brillouin spectra based sensor enables to perform distributed strain measurement. Its algorithm performs a strain measurement with a 40cm spatial sampling over several kilometers. The monitoring of engineering installations needs a centimeter spatial sampling and a better strain accuracy. Previous works highlighted that the industrialized algorithm has great limitation for the exploitation of the local information contained into Brillouin spectra. Indeed, based on its asymmetry and broadening, it is possible to estimate local Brillouin frequencies with a better strain accuracy. We propose here to apply a parametric inverse method using L-curve criterion to estimate the strain with a 5cm spatial sampling. To validate this method, a one-to-one scale experiment has been implemented by optical fiber cable at several depths. Comparing the distributed strain provided by the Brillouin based sensor and our algorithm with a reference strain sensor, the proposed algorithm successfully fulfills the combination of a 5cm spatial sampling over kilometers and a high strain accuracy.

As a complement to classical sensors, Distributed Optical Fiber Sensors now play a prominent role in several engineering fields and act as an antenna array. Depending of the devices used (Raman, Rayleigh, or Brillouin scattering), measurement record depend on temperature, strain, or pressure profile. As the wanted signal is often hidden by noise and other undesired sources, we can express the problem as a source separation problem. In this paper, we show that with the help of recent techniques based on data decomposition and source separation (PCA, ICA NMF techniques) from the virtual antenna, we can accurately identify water leakages from a noisy Raman spectra or a strain profile from Brillouin spectra with a spatial resolution of 1cm instead of 1 meter for classical devices.

Infrared spectral imaging coupled with clustering methods has been shown to be efficient to highlight the structural composition of human tissues and to provide clinically relevant diagnostic information for oncology. Especially, a new fuzzy C-means based algorithm was developed in order to automatically process an infrared image and to assess the transitions between structures (e.g. tumor/normal) in a tumor tissue. However, this algorithm is time consuming and the visualization of the transitions is not easy.

International audience; Distributed Optical Fiber Sensing systems (DOFSS) are composed by optical fibers wrapped in strain sensing cables, coupled with Brillouin interrogators. DOFSS are increasingly used for Structural Health Monitoring (SHM) as they can provide continuous strain profiles along the optical fiber localized in the structure. Raw Brillouin measurements consist in gain - frequency curves with a Lorentzian shape. Strain is generally assessed thanks to the abscissa of the maximum of the gain curve. Two new factors are introduced. They are sensitive to asymmetry and broadening of the Brillouin gain curve which can highlight strain gradient within the spatial resolution of the interrogator. These parameters could be used to detect more efficiently local events and improve instrument algorithm.

This study aims to develop a new FT–IR spectral imaging of tumoral tissue permitting a better characterization of tumor heterogeneity and tumor/surrounding tissue interface. Infrared (IR) data were acquired on 13 biopsies of paraffin-embedded human skin carcinomas. Our approach relies on an innovative fuzzy C-means (FCM)-based clustering algorithm, allowing the automatic and simultaneous estimation of the optimal FCM parameters (number of clusters K and fuzziness index m). FCM seems more suitable than classical ‘hard’ clusterings, as it permits the assignment of each IR spectrum to every cluster with a specific membership value. This characteristic allows differentiating the nuances in the assignment of pixels, particularly those corresponding to tumoral tissue and those located at the tumor/peritumoral tissue interface. FCM images permit to highlight a marked heterogeneity within the tumor and characterize the interconnection between tissular structures. For the infiltrative tumors, a progressive gradient in the membership values of the pixels of the invasive front was also revealed.

International audience; The development of automated monitoring systems for detection of singularities such as leakages in dikes is indispensable to avoid mass disaster. An efficient solution for dikes survey is the use of Distributed Temperature Sensors (DTS) based on optical fiber, offering multitude of advantages such as low-cost, extreme robustness, long range measurement, etc. However, the temperature data acquired with DTS, being not directly interpretable, require intervention of signal processing techniques. The current work addresses this signal processing aspect, exploiting the key idea that the temperature variations over the course of a day for singular zones are quite different from those for nonsingular zones. A daily reference temperature variation, representative of the nonsingular zones, is estimated using the Singular Value Decomposition (SVD). The residue subspace of SVD contains information linked to the deviations from this reference; thus allowing the degree of singularity to be quantified by a dissimilarity measure such as the L2 norm. In order to detect only the singularities in dikes, such as leakages or drains, a Constant False Alarm Rate (CFAR) detector is proposed by modeling each daily dissimilarity measure with a mixture of Gamma and Uniform distributions. The proposed automatic singularity detection system was validated under different scenarios on real data over periods from 2005 to 2007. The first scenario depicted the detection of percolation type artificial leakages with their detection strength depending on their flow rates. Another scenario allowed to detect the presence of a real water leakage at the site, previously unobserved during manual inspections. The repeatability of the system was also verified by periodic analysis.

International audience; We propose a method for the identification of leakages in dikes using the temperature data obtained through fiber optic distributed temperature sensors. We showed how it is possible to treat leakage identification as a source separation problem. The sources were considered as defining the response of the ground, the known structures in the path of the fiber sensors (drains), the seasonal variations, the precipitations and, of course, the leakages, the last ones being our desired signals. We showed that with the help of techniques based on data decomposition and source separation (by PCA and ICA techniques), we can identify the leakages.