191 research outputs found
Supervised nonlinear spectral unmixing using a post-nonlinear mixing model for hyperspectral imagery
Get PDFThis paper presents a nonlinear mixing model for hyperspectral image unmixing. The proposed model assumes that the pixel reflectances are nonlinear functions of pure spectral components contaminated by an additive white Gaussian noise. These nonlinear functions are approximated using polynomial functions leading to a polynomial postnonlinear mixing model. A Bayesian algorithm and optimization methods are proposed to estimate the parameters involved in the model. The performance of the unmixing strategies is evaluated by simulations conducted on synthetic and real data
Parameter estimation for peaky altimetric waveforms
Get PDFMuch attention has been recently devoted to the analysis of coastal altimetric waveforms. When approaching the coast, altimetric waveforms are sometimes corrupted by peaks caused by high reflective areas inside the illuminated land surfaces or by the modification of the sea state close to the shoreline. This paper introduces a new parametric model for these peaky altimetric waveforms. This model assumes that the received altimetric waveform is the sum of a Brown echo and an asymmetric Gaussian peak. The asymmetric Gaussian peak is parameterized by a location, an amplitude, a width, and an asymmetry coefficient. A maximum-likelihood estimator is studied to estimate the Brown plus peak model parameters. The Cramér–Rao lower bounds of the model parameters are then derived providing minimum variances for any unbiased estimator, i.e., a reference in terms of estimation error. The performance of the proposed model and the resulting estimation strategy are evaluated via many simulations conducted on synthetic and real data. Results obtained in this paper show that the proposed model can be used to retrack efficiently standard oceanic Brown echoes as well as coastal echoes corrupted by symmetric or asymmetric Gaussian peaks. Thus, the Brown with Gaussian peak model is useful for analyzing altimetric easurements closer to the coast
Nonlinear unmixing of hyperspectral images using a generalized bilinear model
Get PDFNonlinear models have recently shown interesting properties for spectral unmixing. This paper studies a generalized bilinear model and a hierarchical Bayesian algorithm for unmixing hyperspectral images. The proposed model is a generalization not only of the accepted linear mixing model but also of a bilinear model that has been recently introduced in the literature. Appropriate priors are chosen for its parameters to satisfy the positivity and sum-to-one constraints for the abundances. The joint posterior distribution of the unknown parameter vector is then derived. Unfortunately, this posterior is too complex to obtain analytical expressions of the standard Bayesian estimators. As a consequence, a Metropolis-within-Gibbs algorithm is proposed, which allows samples distributed according to this posterior to be generated and to estimate the unknown model parameters. The performance of the resulting unmixing strategy is evaluated via simulations conducted on synthetic and real data
Deep learning applications for 3D imaging
No full textThis thesis investigates the application of Artificial Intelligence (AI) in combination with
Light Detection and Ranging (LiDAR) technology to enhance resolution and estimate
poses of humans across various sensors.
Initially, we create a specialized deep learning model to improve the resolution of a
research-grade Single-Photon Avalanche Diode (SPAD) array sensor. Employing a fusion
approach with intensity images and carefully chosen features from the SPAD raw data,
our method achieves fourfold super-resolution and maintains robustness in noisy environments. Extending this project to an extremely low spatial resolution commercial SPAD
sensor, we leverage its high depth resolution to generate detailed depth maps and estimate
poses for multiple individuals. Applying neural networks to this extremely low-resolution
sensor shows significant improvements in recovering high-resolution depth images compared to the sensor’s original spatial resolution, showing the ability of AI to automatically
identify patterns in data and reconstruct scenes from very low spatial resolution sensors.
We also extend our approach to radar data, which shares characteristics such as high-precision range information but low spatial resolution. Despite the radar’s limited resolution, our deep learning model successfully estimates poses in a restricted set of scenarios.
To enhance the interpretability of our models, we integrate an Explainable AI (XAI)
framework into the super-resolution work on SPAD sensors. This framework provides
insight into the relevant data features used by the model to work. We also adapt our
method to magnetoencephalography (MEG) data, to understand how a deep learning
model distinguishes various brain activities from MEG measurements in a simple case
study.
In summary, this research develops multiple deep-learning approaches for LiDAR imaging, complemented by XAI frameworks to provide intuition the decision-making processes
within neural networks. Beyond LiDAR, our research includes other data types like radar
and magnetoencephalography (MEG) data from the brain
Bayesian image reconstruction and adaptive scene sampling in single-photon LiDAR imaging
Get PDFThree-Dimensional multispectral Light Detection And Ranging (LiDAR) used
with time-correlated Single-Photon (SP) detection has emerged as a key imaging
modality for high-resolution depth imaging due to its high sensitivity and excellent surface-to-surface resolution. This allowed depth imaging through adversarial
conditions with a prime role in numerous applications. However, several practical
challenges currently limit the use of LiDAR in real-world conditions. Large data
volume constitutes a major challenge for multispectral SP-LiDAR imaging due to
the acquisition of millions of events per second that are usually gathered in large
histogram cubes. This challenge is more evident when the useful signal photons are
attenuated and the background noise is amplified as a result of imaging through a
scattering environment such as underwater or fog. Another limitation includes the
detection of multiple-surfaces-per pixel which usually occurs when imaging through
semi-transparent materials (e.g., windows, camouflage), or in long-range profiling.
This thesis proposes robust and fast computational solutions to improve the acquisition and processing of LiDAR data while measuring uncertainty on high-dimensional data. A smart task-based sampling framework
is proposed to improve the acquisition process and reduce data volume. In addition,
the processing was improved using a Bayesian approach to different types of inverse
problems (e.g. spectral classification, and scene reconstruction). The contributions
of this thesis enables fast and robust 3D reconstruction of complex scenes, paving
the way for the extensive use of single-photon imaging in real-world applications
Révision et évaluation de la qualité en traduction, Etude appliquée: Analyse d'un échantillon d'erreurs de traduction révisées sous l'angle du style
No full textLe présent travail comporte trois parties. La première partie est consacrée à la traduction et ses critères et à l'erreur en traduction et ses sources. La deuxième partie tente de définir les concepts de révision et d'évaluation en traduction. Elle est également consacrée au concept du style ainsi qu'à ses paramètres. Quant à la troisième partie, elle est dévolue à l'analyse d'un échantillon d'erreurs de traduction révisées sous l'angle du style
De l'altimétrie conventionnelle à l'altimétrie SAR/Doppler
No full textDepuis plus de vingt ans, les altimètres classiques comme Topex, Poseidon-2 ou Poséidon-3, ont fourni des formes d'onde qui sont utilisées pour estimer de nombreux paramètres tels que la distance entre le satellite et la scène observée, la hauteur des vagues et la vitesse du vent. L'amélioration de la qualité des paramètres altimétriques a nécessité le développement de plusieurs modèles d'échos et d'algorithmes d'estimation paramétrique. Par ailleurs, un grand effort est récemment dédié au traitement des échos côtiers afin d'analyser les mesures altimétriques le plus près possible des côtes. Cette thèse s'intéresse à la résolution de ces deux problèmes, à savoir, le traitement des formes d'onde côtières et l'amélioration de la qualité des paramètres océaniques estimés. La première partie de l'étude traite le problème des formes d'onde côtières en proposant un nouveau modèle altimétrique tenant compte de la présence éventuelle d'un pic sur l'écho altimétrique. Dans la seconde partie de notre travail, nous nous sommes intéressés à l'étude de l'altimétrie SAR/Doppler. Cette nouvelle technologie vise à réduire le bruit de mesure et à augmenter la résolution le long de la trace par rapport à l'altimétrie conventionnelle. Deux modèles altimétriques ont été développés afin d'estimer les paramètres associés aux échos SAR/Doppler. Ces modèles montrent une nette amélioration de la qualité des paramètres estimés par rapport à l'altimétrie conventionnelle.For more than twenty years, conventional altimeters like Topex, Poseidon-2 or Poseidon-3, have been delivering waveforms that are used to estimate many parameters such as the range between the satellite and the observed scene, the wave height and the wind speed. Several waveform models and estimation processing have been developed for the oceanic data in order to improve the quality of the estimated altimetric parameters. Moreover, a great effort has been devoted to process coastal echoes in order to move the altimetric measurements closer to the coast. In this thesis, we are interested in resolving these two problems, i.e., processing coastal waveforms and improving the quality of the estimated oceanic parameters. The first part of the study considers the problem of coastal waveforms and proposes a new altimetric model taking into account the possible presence of peaks affecting altimetric echoes. In a second part of our work, we have been interested in the delay/Doppler altimetry. This new technology aims at reducing the measurement noise and increasing the alongtrack resolution when compared to conventional altimetry. Two altimetric models have been developed in order to estimate the resulting delay/Doppler echoes. These models allow a clear improvement in parameter estimation when compared to conventional altimetry
From conventional to delay/Doppler altimetry
Get PDFFor more than twenty years, conventional altimeters like Topex, Poseidon-2 or Poseidon-3, have been delivering waveforms that are used to estimate many parameters such as the range between the satellite and the observed scene, the wave height and the wind speed. Several waveform models and estimation processing have been developed for the oceanic data in order to improve the quality of the estimated altimetric parameters. Moreover, a great effort has been devoted to process coastal echoes in order to move the altimetric measurements closer to the coast. In this thesis, we are interested in resolving these two problems, i.e., processing coastal waveforms and improving the quality of the estimated oceanic parameters. The first part of the study considers the problem of coastal waveforms and proposes a new altimetric model taking into account the possible presence of peaks affecting altimetric echoes. In a second part of our work, we have been interested in the delay/Doppler altimetry. This new technology aims at reducing the measurement noise and increasing the alongtrack resolution when compared to conventional altimetry. Two altimetric models have been developed in order to estimate the resulting delay/Doppler echoes. These models allow a clear improvement in parameter estimation when compared to conventional altimetry
Les jeunes maghrébins, entre rhétorique islamiste et contraintes de la mondialisation
No full textDer Autor zeigt, dass im Gegensatz zu der fundamentalistischen Rethorik auf Islamischen Boden die Zeichen der Expansion der Globalisiserung allgegenwärtig zu sein scheinen und zwar über verschiedene Kanäle wie zum Beispiel Medien, Internet, Musik, usw. Daher ergibt sich eine Reihe von Anpassungsversuchen, die durch einen Anziehungs-aber auch Abneigungseffekt gekennzeichnet sind.The author shows that in the land of Islamist fondamentalist rhetoric, signs of expanding globalization are omnipresent through various channels such as the media, Internet and music… Series of attempts are taking place to adapt to the new context under both an effect of attraction and of repulsion.Au rebours de la rhétorique fondamentaliste en terre d’Islam, les signes et les symboles de la mondialisation semblent omniprésents à travers des canaux divers, à l’instar des médias, de l’internet, de la musique, etc. D’où découle une série de tentatives d’adaptation marquées par un effet d’attraction mais aussi de répulsion.El autor muestra que, a contra corriente con la retórica fundamentalista en tierra del Islam, los signos de expansión de la globalización parecen omnipresentes a través de diversos canales, siguiendo el ejemplo de los medios de comunicación, del Internet, de la música, etc. De ahí resultan una serie de intentos de adaptación marcados por un efecto de atracción pero también de repulsión.Lamchichi Abderrahim. Les jeunes maghrébins, entre rhétorique islamiste et contraintes de la mondialisation. In: Agora débats/jeunesses, 19, 2000. Les jeunes et la mondialisation. pp. 57-70
fMRI BOLD signal decomposition using a multivariate low-rank model
Get PDFInternational audienceStandard methodologies for functional Magnetic Resonance Imaging (fMRI) data analysis decompose the observed Blood Oxygenation Level Dependent (BOLD) signals using voxel-wise linear model and perform maximum likelihood estimation to get the parameters associated with the regressors. In task fMRI, the latter are usually defined from the experimental paradigm and some confounds whereas in resting-state acquisitions, a seed-voxel time-course may be used as predictor. Nowadays, most fMRI datasets offer resting-state acquisitions, requiring multi-variate approaches (e.g., PCA, ICA, etc) to extract meaningful information in a data-driven manner. Here, we propose a novel low-rank model of fMRI BOLD data but instead of considering a dimension reduction in space as in ICA, our model relies on convolutional sparse coding between the hemodynamic system and a few temporal atoms which code for the neural activity inducing signals. A rank-1 constraint is also associated with each temporal atom to spatially map its influence in the brain. Within a variational framework, the joint estimation of the neural signals and the associated spatial maps is formulated as a non-convex optimization problem. A local minimizer is computed using an efficient alternate minimization algorithm. The proposed approach is first validated on simulations and then applied to task fMRI data for illustration purpose. Its comparison to a state-of-the-art approach suggests that our method is competitive regarding the uncovered neural fingerprints while offering a richer decomposition in time and space
- …
