Centre de documentation
scientifique

Convolutional neural networks for change analysis in earth observation images with noisy labels and domain shifts / Rodrigo Caye Daudt (2020)  

Public Titre : Convolutional neural networks for change analysis in earth observation images with noisy labels and domain shifts Type de document : Thèse/HDR Auteurs : Rodrigo Caye Daudt, Auteur ; Yann Gousseau, Directeur de thèse Editeur : Paris : Institut Polytechnique de Paris Année de publication : 2020 Importance : 151 p. Format : 21 x 30 cm Note générale : Bibliographie Thèse de doctorat de l’Institut Polytechnique de Paris préparée à Telecom Paris, spécialité Informatique, données, intelligence artificielle Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image [Termes IGN] analyse diachronique [Termes IGN] apprentissage dirigé [Termes IGN] apprentissage profond [Termes IGN] cartographie automatique [Termes IGN] classification par Perceptron multicouche [Termes IGN] classification par réseau neuronal convolutif [Termes IGN] détection de changement [Termes IGN] image multi sources [Termes IGN] réseau neuronal siamois [Termes IGN] segmentation sémantique [Termes IGN] vision par ordinateur Index. décimale : THESE Thèses et HDR Résumé : (Auteur) The analysis of satellite and aerial Earth observation images allows us to obtain precise information over large areas. A multitemporal analysis of such images is necessary to understand the evolution of such areas. In this thesis, convolutional neural networks are used to detect and understand changes using remote sensing images from various sources in supervised and weakly supervised settings. Siamese architectures are used to compare coregistered image pairs and to identify changed pixels. The proposed method is then extended into a multitask network architecture that is used to detect changes and perform land cover mapping simultaneously, which permits a semantic understanding of the detected changes. Then, classification filtering and a novel guided anisotropic diffusion algorithm are used to reduce the effect of biased label noise, which is a concern for automatically generated large-scale datasets. Weakly supervised learning is also achieved to perform pixel-level change detection using only image-level supervision through the usage of class activation maps and a novel spatial attention layer. Finally, a domain adaptation method based on adversarial training is proposed, which succeeds in projecting images from different domains into a common latent space where a given task can be performed. This method is tested not only for domain adaptation for change detection, but also for image classification and semantic segmentation, which proves its versatility. Note de contenu : 1. Introduction 1.1 Context 1.2 Domain 1.3 Objectives 1.4 Publications 2. Related Work 2.1 Computer Vision and Image Analysis 2.2 Machine Learning 2.3 Change Detection Using Remote Sensing Images 2.4 Evaluation Metrics 3. Supervised Change Detection 3.1 Introduction 3.2 ONERA Satellite Change Detection Dataset 3.3 Patch Based Architectures 3.4 Fully Convolutional Architectures 3.5 Experiments 3.6 Conclusion 4. Semantic Change Detection 62 4.1 High Resolution Semantic Change Detection Dataset 4.2 Methodology 4.3 Results 4.4 Conclusion 5. Weakly Supervised Change Detection 5.1 Change Detection with Unreliable Data 5.2 Method 5.3 Experiments 5.4 Analysis 5.5 Conclusion 6. Domain Adaptation for Change Detection 6.1 Motivation 6.2 Unsupervised Domain Adaptation 6.3 Formulation 6.4 Implementation 6.5 Results 6.6 Limitations and Discussion 6.7 Unpaired Translation of Change Detection Images 6.8 Conclusion 7. Conclusion Numéro de notice : 26557 Affiliation des auteurs : non IGN Thématique : IMAGERIE/INFORMATIQUE Nature : Thèse française Note de thèse : Thèse de Doctorat : Informatique, données, intelligence artificielle : Paris : 2020 Organisme de stage : Laboratoire Traitement et Communication de l'Information LTCI nature-HAL : Thèse DOI : sans Date de publication en ligne : 12/04/2021 En ligne : https://tel.archives-ouvertes.fr/tel-03105668/document Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=98101

Low level feature detection in SAR images / Chenguang Liu (2020)  

Public Titre : Low level feature detection in SAR images Type de document : Thèse/HDR Auteurs : Chenguang Liu, Auteur ; Florence Tupin, Directeur de thèse ; Yann Gousseau, Directeur de thèse Editeur : Paris [France] : Télécom ParisTech Année de publication : 2020 Importance : 138 p. Format : 21 x 30 cm Note générale : bibliographie Thèse de doctorat de l’Institut Polytechnique de Paris préparée à Télécom Paris, Spécialité de doctorat : Signal, Images, Automatique et robotique Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image radar et applications [Termes IGN] apprentissage profond [Termes IGN] classification par réseau neuronal convolutif [Termes IGN] détection de contours [Termes IGN] gradient [Termes IGN] image radar moirée [Termes IGN] modèle de Markov [Termes IGN] segment de droite Index. décimale : THESE Thèses et HDR Résumé : (auteur) In this thesis we develop low level feature detectors for Synthetic Aperture Radar (SAR) images to facilitate the joint use of SAR and optical data. Line segments and edges are very important low level features in images which can be used for many applications like image analysis, image registration and object detection. Contrarily to the availability of many efficient low level feature detectors dedicated to optical images, there are very few efficient line segment detector and edge detector for SAR images mostly because of the strong multiplicative noise. In this thesis we develop a generic line segment detector and an efficient edge detector for SAR images.The proposed line segment detector which is named as LSDSAR, is based on a Markovian a contrario model and the Helmholtz principle, where line segments are validated according to their meaningfulness. More specifically, a line segment is validated if its expected number of occurences in a random image under the hypothesis of the Markovian a contrario model is small. Contrarily to the usual a contrario approaches, the Markovian a contrario model allows strong filtering in the gradient computation step, since dependencies between local orientations of neighbouring pixels are permitted thanks to the use of a first order Markov chain. The proposed Markovian a contrario model based line segment detector LSDSAR benefit from the accuracy and efficiency of the new definition of the background model, indeed, many true line segments in SAR images are detected with a control of the number of false detections. Moreover, very little parameter tuning is required in the practical applications of LSDSAR. The second work of this thesis is that we propose a deep learning based edge detector for SAR images. The contributions of the proposed edge detector are two fold: 1) under the hypothesis that both optical images and real SAR images can be divided into piecewise constant areas, we propose to simulate a SAR dataset using optical dataset; 2) we propose to train a classical CNN (convolutional neural network) edge detector, HED, directly on the graident fields of images. This, by using an adequate method to compute the gradient, enables SAR images at test time to have statistics similar to the training set as inputs to the network. More precisely, the gradient distribution for all homogeneous areas are the same and the gradient distribution for two homogeneous areas across boundaries depends only on the ratio of their mean intensity values. The proposed method, GRHED, significantly improves the state-of-the-art, especially in very noisy cases such as 1-look images. Note de contenu : 1- Context 2- SAR basics, statistics of SAR images and data used in this thesis I Line segment detection in SAR images 3- Introduction 4- LSD, a line segment detector with false detection control 5- LSDSAR, a generic line segment detector for SAR images 6- Experiments II Edge detection in SAR images using CNNs 7- Introduction 8- Presentation of the HED method and of the training dataset 9- GRHED, introducing a hand-crafted layer before the usual CNNs 10- Experiments 11- Summary of the thesis Numéro de notice : 25878 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse française Note de thèse : Thèse de Doctorat : Spécialité : Signal, Images, Automatique et robotique : Paris : 2020 nature-HAL : Thèse DOI : sans En ligne : https://tel.archives-ouvertes.fr/tel-02861903/document Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=95689

SAR-SIFT : a SIFT-like algorithm for SAR images / Flora Dellinger in IEEE Transactions on geoscience and remote sensing, vol 53 n° 1 (January 2015)  

Public [article]  inIEEE Transactions on geoscience and remote sensing > vol 53 n° 1 (January 2015) . - pp 453 - 466 Titre : SAR-SIFT : a SIFT-like algorithm for SAR images Type de document : Article/Communication Auteurs : Flora Dellinger, Auteur ; Julien Delon, Auteur ; Yann Gousseau, Auteur ; et al., Auteur Année de publication : 2015 Article en page(s) : pp 453 - 466 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image radar et applications [Termes IGN] chatoiement [Termes IGN] image radar moirée [Termes IGN] SIFT (algorithme) Résumé : (Auteur) The scale-invariant feature transform (SIFT) algorithm and its many variants are widely used in computer vision and in remote sensing to match features between images or to localize and recognize objects. However, mostly because of speckle noise, it does not perform well on synthetic aperture radar (SAR) images. In this paper, we introduce a SIFT-like algorithm specifically dedicated to SAR imaging, which is named SAR-SIFT. The algorithm includes both the detection of keypoints and the computation of local descriptors. A new gradient definition, yielding an orientation and a magnitude that are robust to speckle noise, is first introduced. It is then used to adapt several steps of the SIFT algorithm to SAR images. We study the improvement brought by this new algorithm, as compared with existing approaches. We present an application of SAR-SIFT to the registration of SAR images in different configurations, particularly with different incidence angles. Numéro de notice : A2015-038 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1109/TGRS.2014.2323552 En ligne : https://doi.org/10.1109/TGRS.2014.2323552 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=75120 [article]

Exemplaires(1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
065-2015011	RAB	Revue	Centre de documentation	En réserve L003	Disponible