Isabelle Quidu - Academia.edu (original) (raw)
Papers by Isabelle Quidu
HAL (Le Centre pour la Communication Scientifique Directe), 2014
This book consists of the proceedings of the International Conference on Detection and Classifica... more This book consists of the proceedings of the International Conference on Detection and Classification of Underwater Targets which took place in Brest, France, in October 2012. This collection of academic papers represents the current state of the art of research and development in the areas of sensor technology, processing, modeling and automation for the purpose of detecting and classifying objects in the underwater environment, written by leading researchers in government, industry and academia. These articles should be of interest not only to those working on underwater target detection, but also to researchers in the related fields of remote sensing, robotic perception and medical imaging.
HAL (Le Centre pour la Communication Scientifique Directe), Jun 20, 2022
![Research paper thumbnail of Corrections to “Automatic Sea-Surface Obstacle Detection and Tracking in Forward-Looking Sonar Image Sequences” [Aug 15 4661-4669]](https://attachments.academia-assets.com/108550331/thumbnails/1.jpg)
](IEEE Transactions on Geoscience and Remote Sensing, Nov 1, 2015
HAL (Le Centre pour la Communication Scientifique Directe), Sep 5, 2017
National audienc
Ocean engineering & oceanography, 2018
HAL (Le Centre pour la Communication Scientifique Directe), Jun 22, 2014
We propose an automatic sea surface object detection from forward looking sonar images. The consi... more We propose an automatic sea surface object detection from forward looking sonar images. The considered sea surface obstacles are man-made objects: buoys, boats, ships (motorboats or sailboats). Their acoustic signature varies according to their type and state (fixed or moving). The proposed detection scheme is hierarchical in order to manage the various target signatures. The first step consists in detecting stationary self noise from ships. In case of detection, the strong-intensity strip corresponding to the ship direction is removed to avoid ship noise disturbance during other target detection processes. The next step consists in detecting the other types of obstacles. It is based on an adaptive CFAR (Constant False Alarm Rate) thresholding. The final step consists in analyzing the area around every detected position in order to state that this latter is a reliable obstacle and not a wake signature. Promising results are obtained using real data collected at sea with various objects and scenarios.
HAL (Le Centre pour la Communication Scientifique Directe), Jun 6, 2022
HAL (Le Centre pour la Communication Scientifique Directe), Jul 1, 2008
ABSTRACT Military underwater robots are designed to perform complex underwater missions in both k... more ABSTRACT Military underwater robots are designed to perform complex underwater missions in both known and unknown environments. To achieve these tasks, an Autonomous Underwater Vehicle (AUV) must be supplied with appropriate sensors to deal with unpredictable events that can put it in danger, and with a high degree of decisional autonomy. In this paper, we have studied the architecture of forward looking sensors to allow the creation of a 3D model of the environment presenting the seabed and the obstacles that are on the path of the AUV. Our approach is based on experimental trials using different and complementary ways (sonars with several configurations) to gather an information as complete as possible. This information will be processed by the vehicle during a survey mission. Practically, we create a reference model of a static environment using a multibeam system which produces bathymetric images at different grazing angles. In the same environment we then use a Forward Looking Sonar intended for the recognition of detected echoes in comparison with the reference model. If an echo cannot be related to a known object on the reference map, it is considered as an obstacle, and the map is updated.
Nous proposons une nouvelle methode de classification des mines sous-marine basee modele. Cette m... more Nous proposons une nouvelle methode de classification des mines sous-marine basee modele. Cette methode utilise l’information Echo dans la signature acoustique d’une cible nommee A-scan. Elle integre un module de filtrage adapte pour permettre un resultat de classification plus souple permettant d’introduire un degre de confiance lie au maximum de correlation obtenu. En pratique, la methode consiste a comparer l'A-scan d’une cible inconnue avec un ensemble d’A-scans simulees dans les memes conditions operationnelles. Afin d’entrainer le classificateur, une base des A-scans a ete creee en utilisant des signatures acoustiques des objets manufactures de formes plus ou moins complexes (mine Rockan, recifs artificiels). Lorsque la difference entre les coefficients de correlation n’est pas suffisante le resultat est considere ambigu. Une seconde phase est proposee afin de distinguer les deux classes les plus probables en etendant la base initiale par des nouvelles A-scans dans des con...
L'obstacle majeur dans le traitement des images sous-marines résulte des phénomènes d'absorption ... more L'obstacle majeur dans le traitement des images sous-marines résulte des phénomènes d'absorption et de diffusion dus aux propriétés optiques particulières de la lumière dans l'eau. Ces deux phénomènes auxquels s'ajoute le problème de turbidité, impose de travailler sur des images très bruitées, avec souvent, une illumination non uniforme, des contrastes faibles, des couleurs atténuées.. .Cet article présente une nouvelle méthode automatique de pré-traitement des images sous marines. L'algorithme proposé qui ne nécessite ni paramétrage manuel ni information a priori, permet d'atténuer les défauts précédemment cités et d'améliorer de façon significative la qualité des images. L'éclairage, le bruit, les contrastes puis les couleurs sont corrigés séquentiellement.
International audienceIn the context of sonar imagery, the image registration process is often th... more International audienceIn the context of sonar imagery, the image registration process is often the very first step to achieve further processings ranging from navigation correction to change detection through mosaicking. In this paper, we aim at providing intensity-based registration techniques through the consideration of various similarity metrics to rigidly align pairs of sonar tracks. Such similarity measures are assessed on real high-resolution synthetic aperture sonar tracks
OCEANS 2019 - Marseille, 2019
In the underwater mine warfare context, change detection is a principle consisting in comparing a... more In the underwater mine warfare context, change detection is a principle consisting in comparing a newly sensed seabed area, usually by means of a side scan sonar, to another one that has potentially been sensed several months or years ago. In this paper, we propose an approach to simultaneously register (i.e geometrically align) the reference and the repeated data while detecting new and missing objects between both datasets acquisition. This method is first evaluated on data provided by a simulator based on a model of navigation uncertainty as well as on error sources due to the imaging sonar, in order to assess its robustness against different parameters. We also provide results on datasets acquired at sea and demonstrate its efficiency to solve the change detection problem.
This book consists of the proceedings of the International Conference on Detection and Classifica... more This book consists of the proceedings of the International Conference on Detection and Classification of Underwater Targets which took place in Brest, France, in October 2012. This collection of academic papers represents the current state of the art of research and development in the areas of sensor technology, processing, modeling and automation for the purpose of detecting and classifying objects in the underwater environment, written by leading researchers in government, industry and academia. These articles should be of interest not only to those working on underwater target detection, but also to researchers in the related fields of remote sensing, robotic perception and medical imaging.
A l’instar des mammiferes marins, l’homme a developpe des systemes sonar capables de percevoir l’... more A l’instar des mammiferes marins, l’homme a developpe des systemes sonar capables de percevoir l’environnement marin sur des distances plus ou moins longues et avec plus ou moins de details en fonction de ses besoins. Le principal besoin traite dans ce travail prepare pour l’habilitation a diriger les recherches est la reconnaissance automatique de cibles sous-marines (plus connue sous l’acronyme ATR pour Automatic Target Recognition), qu’elles soient posees sur le fond ou flottantes entre deux eaux. La chronologie de la selection de travaux presentes suit le perfectionnement au fil des annees des systemes d’imagerie et des algorithmes de traitement associes pour repondre a l’evolution de la menace, depuis les mines furtives des annees 90 aux engins explosifs improvises d’aujourd’hui. Pour contrer ces menaces a partir de donnees sonar lateral ou sonar a antenne synthetique (dites donnees SAS pour Synthetic Aperture Sonar), deux points sont en particulier exposes : une nouvelle appro...
The seabed characterization from sonar images is a very hard task because of the produced data an... more The seabed characterization from sonar images is a very hard task because of the produced data and the unknown environment, even for an human expert. In this work we propose an original approach in order to combine binary classifiers arising from different kinds of strategies such as one-versusone or one-versus-rest, usually used in the SVM-classification. The decision functions coming from these binary classifiers are interpreted in terms of belief functions in order to combine these functions with one of the numerous operators of the belief functions theory. Moreover, this interpretation of the decision function allows us to propose a process of decisions by taking into account the rejected observations too far removed from the learning data, and the imprecise decisions given in unions of classes. This new approach is illustrated and evaluated with a SVM in order to classify the different kinds of sediment on image sonar.
The Journal of the Acoustical Society of America, 2008
In the context of mine warfare, detected objects can be classified from their cast shadow. A stan... more In the context of mine warfare, detected objects can be classified from their cast shadow. A standard solution consists in segmenting the image at first (we obtain binary from grey-level image giving the label zero for pixels belonging to the shadow and the label one elsewhere), and then in performing classification from features extracted from the 2D-shape of the segmented shadow. Other pre- or post-processings are generally used to make each step more robust by avoiding a mistake to be propagated through the following steps. In this paper, to focus on the actual goal, we propose a novel approach where a dynamic segmentation scheme is fully classification-oriented. Actually, classification is performed directly from raw image data. The approach is based on the combination of deformable models, genetic algorithms, and statistical image models.
Abstract: This work follows previous works on tracking of targets on Forward Looking Sonar images... more Abstract: This work follows previous works on tracking of targets on Forward Looking Sonar images. A Kalman filter based on a process model of the vehicle was implemented considering two strong assumptions: firstly, the obstacle is fixed in relation to the world reference frame and secondly, it lies proud on the seabed. Consequently, Kalman filtering leads to a biased estimation of successive positions of an obstacle floating in the water column. Starting with this observation a new algorithm has been developed to allow a deferred estimation of the z-coordinate (along the absolute z-axis) of the obstacle related to the vehicle. This is performed offline by minimizing at each step of the sequence the root mean squared deviation (RMSD) between measured sonar positions and predicted positions, i.e. by minimizing the innovation values of the Kalman filtering. Results are given on real data recorded in March 2009 and April 2010 during sea trials organized by GESMA involving the Rapid Env...
OCEANS 2016 MTS/IEEE Monterey
In mine warfare context, environmental effects are known to degrade performances of most of autom... more In mine warfare context, environmental effects are known to degrade performances of most of automatic target recognition (ATR) processes. In this study, we consider the environment as an information that can be used to design a robust ATR process. Hence, we investigate a way to extract and exploit information about the seafloor using an isotropic analysis of sidescan sonar images based on the monogenic signal. This tool provides an orthogonal separation between energetic, geometrical and structural information of the 2D signal in a scale-space framework. It also allows to efficiently compute the continuous intrinsic dimensionality scale-space. We propose to use these last descriptors to characterize the sidescan sonar images in terms of homogeneous, anisotropic and complex areas. In each of these areas it can be expected that adapted ATR processes could be defined to outperform classical global approaches.
HAL (Le Centre pour la Communication Scientifique Directe), 2014
This book consists of the proceedings of the International Conference on Detection and Classifica... more This book consists of the proceedings of the International Conference on Detection and Classification of Underwater Targets which took place in Brest, France, in October 2012. This collection of academic papers represents the current state of the art of research and development in the areas of sensor technology, processing, modeling and automation for the purpose of detecting and classifying objects in the underwater environment, written by leading researchers in government, industry and academia. These articles should be of interest not only to those working on underwater target detection, but also to researchers in the related fields of remote sensing, robotic perception and medical imaging.
HAL (Le Centre pour la Communication Scientifique Directe), Jun 20, 2022
IEEE Transactions on Geoscience and Remote Sensing, Nov 1, 2015
HAL (Le Centre pour la Communication Scientifique Directe), Sep 5, 2017
National audienc
Ocean engineering & oceanography, 2018
HAL (Le Centre pour la Communication Scientifique Directe), Jun 22, 2014
We propose an automatic sea surface object detection from forward looking sonar images. The consi... more We propose an automatic sea surface object detection from forward looking sonar images. The considered sea surface obstacles are man-made objects: buoys, boats, ships (motorboats or sailboats). Their acoustic signature varies according to their type and state (fixed or moving). The proposed detection scheme is hierarchical in order to manage the various target signatures. The first step consists in detecting stationary self noise from ships. In case of detection, the strong-intensity strip corresponding to the ship direction is removed to avoid ship noise disturbance during other target detection processes. The next step consists in detecting the other types of obstacles. It is based on an adaptive CFAR (Constant False Alarm Rate) thresholding. The final step consists in analyzing the area around every detected position in order to state that this latter is a reliable obstacle and not a wake signature. Promising results are obtained using real data collected at sea with various objects and scenarios.
HAL (Le Centre pour la Communication Scientifique Directe), Jun 6, 2022
HAL (Le Centre pour la Communication Scientifique Directe), Jul 1, 2008
ABSTRACT Military underwater robots are designed to perform complex underwater missions in both k... more ABSTRACT Military underwater robots are designed to perform complex underwater missions in both known and unknown environments. To achieve these tasks, an Autonomous Underwater Vehicle (AUV) must be supplied with appropriate sensors to deal with unpredictable events that can put it in danger, and with a high degree of decisional autonomy. In this paper, we have studied the architecture of forward looking sensors to allow the creation of a 3D model of the environment presenting the seabed and the obstacles that are on the path of the AUV. Our approach is based on experimental trials using different and complementary ways (sonars with several configurations) to gather an information as complete as possible. This information will be processed by the vehicle during a survey mission. Practically, we create a reference model of a static environment using a multibeam system which produces bathymetric images at different grazing angles. In the same environment we then use a Forward Looking Sonar intended for the recognition of detected echoes in comparison with the reference model. If an echo cannot be related to a known object on the reference map, it is considered as an obstacle, and the map is updated.
Nous proposons une nouvelle methode de classification des mines sous-marine basee modele. Cette m... more Nous proposons une nouvelle methode de classification des mines sous-marine basee modele. Cette methode utilise l’information Echo dans la signature acoustique d’une cible nommee A-scan. Elle integre un module de filtrage adapte pour permettre un resultat de classification plus souple permettant d’introduire un degre de confiance lie au maximum de correlation obtenu. En pratique, la methode consiste a comparer l'A-scan d’une cible inconnue avec un ensemble d’A-scans simulees dans les memes conditions operationnelles. Afin d’entrainer le classificateur, une base des A-scans a ete creee en utilisant des signatures acoustiques des objets manufactures de formes plus ou moins complexes (mine Rockan, recifs artificiels). Lorsque la difference entre les coefficients de correlation n’est pas suffisante le resultat est considere ambigu. Une seconde phase est proposee afin de distinguer les deux classes les plus probables en etendant la base initiale par des nouvelles A-scans dans des con...
L'obstacle majeur dans le traitement des images sous-marines résulte des phénomènes d'absorption ... more L'obstacle majeur dans le traitement des images sous-marines résulte des phénomènes d'absorption et de diffusion dus aux propriétés optiques particulières de la lumière dans l'eau. Ces deux phénomènes auxquels s'ajoute le problème de turbidité, impose de travailler sur des images très bruitées, avec souvent, une illumination non uniforme, des contrastes faibles, des couleurs atténuées.. .Cet article présente une nouvelle méthode automatique de pré-traitement des images sous marines. L'algorithme proposé qui ne nécessite ni paramétrage manuel ni information a priori, permet d'atténuer les défauts précédemment cités et d'améliorer de façon significative la qualité des images. L'éclairage, le bruit, les contrastes puis les couleurs sont corrigés séquentiellement.
International audienceIn the context of sonar imagery, the image registration process is often th... more International audienceIn the context of sonar imagery, the image registration process is often the very first step to achieve further processings ranging from navigation correction to change detection through mosaicking. In this paper, we aim at providing intensity-based registration techniques through the consideration of various similarity metrics to rigidly align pairs of sonar tracks. Such similarity measures are assessed on real high-resolution synthetic aperture sonar tracks
OCEANS 2019 - Marseille, 2019
In the underwater mine warfare context, change detection is a principle consisting in comparing a... more In the underwater mine warfare context, change detection is a principle consisting in comparing a newly sensed seabed area, usually by means of a side scan sonar, to another one that has potentially been sensed several months or years ago. In this paper, we propose an approach to simultaneously register (i.e geometrically align) the reference and the repeated data while detecting new and missing objects between both datasets acquisition. This method is first evaluated on data provided by a simulator based on a model of navigation uncertainty as well as on error sources due to the imaging sonar, in order to assess its robustness against different parameters. We also provide results on datasets acquired at sea and demonstrate its efficiency to solve the change detection problem.
This book consists of the proceedings of the International Conference on Detection and Classifica... more This book consists of the proceedings of the International Conference on Detection and Classification of Underwater Targets which took place in Brest, France, in October 2012. This collection of academic papers represents the current state of the art of research and development in the areas of sensor technology, processing, modeling and automation for the purpose of detecting and classifying objects in the underwater environment, written by leading researchers in government, industry and academia. These articles should be of interest not only to those working on underwater target detection, but also to researchers in the related fields of remote sensing, robotic perception and medical imaging.
A l’instar des mammiferes marins, l’homme a developpe des systemes sonar capables de percevoir l’... more A l’instar des mammiferes marins, l’homme a developpe des systemes sonar capables de percevoir l’environnement marin sur des distances plus ou moins longues et avec plus ou moins de details en fonction de ses besoins. Le principal besoin traite dans ce travail prepare pour l’habilitation a diriger les recherches est la reconnaissance automatique de cibles sous-marines (plus connue sous l’acronyme ATR pour Automatic Target Recognition), qu’elles soient posees sur le fond ou flottantes entre deux eaux. La chronologie de la selection de travaux presentes suit le perfectionnement au fil des annees des systemes d’imagerie et des algorithmes de traitement associes pour repondre a l’evolution de la menace, depuis les mines furtives des annees 90 aux engins explosifs improvises d’aujourd’hui. Pour contrer ces menaces a partir de donnees sonar lateral ou sonar a antenne synthetique (dites donnees SAS pour Synthetic Aperture Sonar), deux points sont en particulier exposes : une nouvelle appro...
The seabed characterization from sonar images is a very hard task because of the produced data an... more The seabed characterization from sonar images is a very hard task because of the produced data and the unknown environment, even for an human expert. In this work we propose an original approach in order to combine binary classifiers arising from different kinds of strategies such as one-versusone or one-versus-rest, usually used in the SVM-classification. The decision functions coming from these binary classifiers are interpreted in terms of belief functions in order to combine these functions with one of the numerous operators of the belief functions theory. Moreover, this interpretation of the decision function allows us to propose a process of decisions by taking into account the rejected observations too far removed from the learning data, and the imprecise decisions given in unions of classes. This new approach is illustrated and evaluated with a SVM in order to classify the different kinds of sediment on image sonar.
The Journal of the Acoustical Society of America, 2008
In the context of mine warfare, detected objects can be classified from their cast shadow. A stan... more In the context of mine warfare, detected objects can be classified from their cast shadow. A standard solution consists in segmenting the image at first (we obtain binary from grey-level image giving the label zero for pixels belonging to the shadow and the label one elsewhere), and then in performing classification from features extracted from the 2D-shape of the segmented shadow. Other pre- or post-processings are generally used to make each step more robust by avoiding a mistake to be propagated through the following steps. In this paper, to focus on the actual goal, we propose a novel approach where a dynamic segmentation scheme is fully classification-oriented. Actually, classification is performed directly from raw image data. The approach is based on the combination of deformable models, genetic algorithms, and statistical image models.
Abstract: This work follows previous works on tracking of targets on Forward Looking Sonar images... more Abstract: This work follows previous works on tracking of targets on Forward Looking Sonar images. A Kalman filter based on a process model of the vehicle was implemented considering two strong assumptions: firstly, the obstacle is fixed in relation to the world reference frame and secondly, it lies proud on the seabed. Consequently, Kalman filtering leads to a biased estimation of successive positions of an obstacle floating in the water column. Starting with this observation a new algorithm has been developed to allow a deferred estimation of the z-coordinate (along the absolute z-axis) of the obstacle related to the vehicle. This is performed offline by minimizing at each step of the sequence the root mean squared deviation (RMSD) between measured sonar positions and predicted positions, i.e. by minimizing the innovation values of the Kalman filtering. Results are given on real data recorded in March 2009 and April 2010 during sea trials organized by GESMA involving the Rapid Env...
OCEANS 2016 MTS/IEEE Monterey
In mine warfare context, environmental effects are known to degrade performances of most of autom... more In mine warfare context, environmental effects are known to degrade performances of most of automatic target recognition (ATR) processes. In this study, we consider the environment as an information that can be used to design a robust ATR process. Hence, we investigate a way to extract and exploit information about the seafloor using an isotropic analysis of sidescan sonar images based on the monogenic signal. This tool provides an orthogonal separation between energetic, geometrical and structural information of the 2D signal in a scale-space framework. It also allows to efficiently compute the continuous intrinsic dimensionality scale-space. We propose to use these last descriptors to characterize the sidescan sonar images in terms of homogeneous, anisotropic and complex areas. In each of these areas it can be expected that adapted ATR processes could be defined to outperform classical global approaches.