Centre dedocumentation
scientifique

A hierarchical deformable deep neural network and an aerial image benchmark dataset for surface multiview stereo reconstruction / Jiayi Li in IEEE Transactions on geoscience and remote sensing, vol 61 n° 1 (January 2023)  

Public [article]  inIEEE Transactions on geoscience and remote sensing > vol 61 n° 1 (January 2023) . - n° 5600812 Titre : A hierarchical deformable deep neural network and an aerial image benchmark dataset for surface multiview stereo reconstruction Type de document : Article/Communication Auteurs : Jiayi Li, Auteur ; Xin Huang, Auteur ; Yujin Feng, Auteur ; et al., Auteur Année de publication : 2023 Article en page(s) : n° 5600812 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] approche hiérarchique [Termes IGN] carte de profondeur [Termes IGN] déformation d'objet [Termes IGN] effet de profondeur cinétique [Termes IGN] extraction de traits caractéristiques [Termes IGN] image aérienne [Termes IGN] jeu de données [Termes IGN] modèle numérique de surface [Termes IGN] modèle stéréoscopique [Termes IGN] reconstruction d'image [Termes IGN] réseau neuronal profond [Termes IGN] segmentation sémantique Résumé : (auteur) Multiview stereo (MVS) aerial image depth estimation is a research frontier in the remote sensing field. Recent deep learning-based advances in close-range object reconstruction have suggested the great potential of this approach. Meanwhile, the deformation problem and the scale variation issue are also worthy of attention. These characteristics of aerial images limit the applicability of the current methods for aerial image depth estimation. Moreover, there are few available benchmark datasets for aerial image depth estimation. In this regard, this article describes a new benchmark dataset called the LuoJia-MVS dataset ( https://irsip.whu.edu.cn/resources/resources_en_v2.php ), as well as a new deep neural network known as the hierarchical deformable cascade MVS network (HDC-MVSNet). The LuoJia-MVS dataset contains 7972 five-view images with a spatial resolution of 10 cm, pixel-wise depths, and precise camera parameters, and was generated from an accurate digital surface model (DSM) built from thousands of stereo aerial images. In the HDC-MVSNet network, a new full-scale feature pyramid extraction module, a hierarchical set of 3-D convolutional blocks, and “true 3-D” deformable 3-D convolutional layers are specifically designed by considering the aforementioned characteristics of aerial images. Overall and ablation experiments on the WHU and LuoJia-MVS datasets validated the superiority of HDC-MVSNet over the current state-of-the-art MVS depth estimation methods and confirmed that the newly built dataset can provide an effective benchmark. Numéro de notice : A2023-117 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1109/TGRS.2023.3234694 En ligne : https://doi.org/10.1109/TGRS.2023.3234694 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=102488 [article]

Deep learning based 3D reconstruction: supervision and representation / François Darmon (2022)  

Public Titre : Deep learning based 3D reconstruction: supervision and representation Type de document : Thèse/HDR Auteurs : François Darmon, Auteur ; Pascal Monasse, Directeur de thèse ; Mathieu Aubry, Directeur de thèse Editeur : Champs-sur-Marne : Ecole des Ponts ParisTech Année de publication : 2022 Importance : 115 p. Format : 21 x 30 cm Note générale : Bibliographie Thèse de doctorat de l'Ecole des Ponts ParisTech, spécialité informatique Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] appariement d'images [Termes IGN] carte de profondeur [Termes IGN] classification par réseau neuronal convolutif [Termes IGN] extraction [Termes IGN] géométrie épipolaire [Termes IGN] maillage [Termes IGN] modèle stéréoscopique [Termes IGN] point d'intérêt [Termes IGN] Ransac (algorithme) [Termes IGN] reconstruction 3D [Termes IGN] reconstruction d'objet [Termes IGN] semis de points [Termes IGN] SIFT (algorithme) [Termes IGN] structure-from-motion [Termes IGN] voxel Index. décimale : THESE Thèses et HDR Résumé : (auteur) 3D reconstruction is a long standing problem in computer vision. Yet, state-of-the-art methods still struggle when the images used have large illumination changes, many occlusions or limited textures. Deep Learning holds promises of improving 3D reconstruction in such setups, but classical methods still produce the best results. In this thesis we analyse the specificity of deep learning applied to multiview 3D reconstruction and introduce new deep learning based methods.The first contribution of this thesis is an analysis of the possible supervision for training Deep Learning models for sparse image matching. We introduce a two-step algorithm that first computes low resolution matches using deep learning and then matches classical local features inside the matches regions. We analyze several levels of supervision and show that our new epipolar supervision leads to the best results.The second contribution is also a study of supervision for Deep Learning but applied to another scenario: calibrated 3D reconstruction in the wild. We show that existing unsupervised methods do not work on such data and we introduce a new training technique that solves this issue. We then exhaustively compare unsupervised approach and supervised approaches with different network architectures and training data.Finally, our third contribution is about data representation. Neural implicit representation were recently used for image rendering. We adapt this representation to the multiview reconstruction problem and we introduce a new method that, similar to classical 3D reconstruction techniques, optimizes photo-consistency between projections of multiple images. Our approach outperforms state-of-the-art by a large margin. Note de contenu : 1- Introduction 2- Background 3- Deep learning for guiding keypoint matching 4- Deep Learning based Multi-View Stereo in the wild 5- Multi-view reconstruction with implicit surfaces and patch warping 6- Conclusion Numéro de notice : 24085 Affiliation des auteurs : non IGN Thématique : IMAGERIE/INFORMATIQUE Nature : Thèse française Note de thèse : Thèse de Doctorat : Informatique : Ponts ParisTech : 2022 Organisme de stage : Laboratoire d'Informatique Gaspard-Monge LIGM DOI : sans En ligne : https://www.theses.fr/2022ENPC0024 Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=102473

DEM- and GIS-based analysis of soil erosion depth using machine learning / Kieu Anh Nguyen in ISPRS International journal of geo-information, vol 10 n° 7 (July 2021)  

Public [article]  inISPRS International journal of geo-information > vol 10 n° 7 (July 2021) . - n° 452 Titre : DEM- and GIS-based analysis of soil erosion depth using machine learning Type de document : Article/Communication Auteurs : Kieu Anh Nguyen, Auteur ; Walter Chen, Auteur Année de publication : 2021 Article en page(s) : n° 452 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications SIG [Termes IGN] apprentissage automatique [Termes IGN] bassin hydrographique [Termes IGN] carte de profondeur [Termes IGN] classification par forêts d'arbres décisionnels [Termes IGN] érosion [Termes IGN] Extreme Gradient Machine [Termes IGN] modèle de simulation [Termes IGN] modèle numérique de surface [Termes IGN] morphométrie [Termes IGN] système d'information géographique Résumé : (auteur) Soil erosion is a form of land degradation. It is the process of moving surface soil with the action of external forces such as wind or water. Tillage also causes soil erosion. As outlined by the United Nations Sustainable Development Goal (UN SDG) #15, it is a global challenge to “combat desertification, and halt and reverse land degradation and halt biodiversity loss.” In order to advance this goal, we studied and modeled the soil erosion depth of a typical watershed in Taiwan using 26 morphometric factors derived from a digital elevation model (DEM) and 10 environmental factors. Feature selection was performed using the Boruta algorithm to determine 15 factors with confirmed importance and one tentative factor. Then, machine learning models, including the random forest (RF) and gradient boosting machine (GBM), were used to create prediction models validated by erosion pin measurements. The results show that GBM, coupled with 15 important factors (confirmed), achieved the best result in the context of root mean square error (RMSE) and Nash–Sutcliffe efficiency (NSE). Finally, we present the maps of soil erosion depth using the two machine learning models. The maps are useful for conservation planning and mitigating future soil erosion. Numéro de notice : A2021-551 Affiliation des auteurs : non IGN Thématique : GEOMATIQUE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.3390/ijgi10070452 Date de publication en ligne : 01/07/2021 En ligne : https://doi.org/10.3390/ijgi10070452 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=98074 [article]

Structure-from-motion-derived digital surface models from historical aerial photographs: A new 3D application for coastal dune monitoring / Edoardo Grottoli in Remote sensing, vol 13 n° 1 (January-1 2021)  

Public [article]  inRemote sensing > vol 13 n° 1 (January-1 2021) . - n° 95 Titre : Structure-from-motion-derived digital surface models from historical aerial photographs: A new 3D application for coastal dune monitoring Type de document : Article/Communication Auteurs : Edoardo Grottoli, Auteur ; Mélanie Biausque, Auteur ; David Rogers, Auteur ; et al., Auteur Année de publication : 2021 Article en page(s) : n° 95 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications photogrammétriques [Termes IGN] analyse diachronique [Termes IGN] carte de profondeur [Termes IGN] données lidar [Termes IGN] dune [Termes IGN] érosion côtière [Termes IGN] filtrage de points [Termes IGN] image captée par drone [Termes IGN] image numérisée [Termes IGN] modèle numérique de surface [Termes IGN] reconstruction 3D [Termes IGN] semis de points [Termes IGN] structure-from-motion [Termes IGN] surveillance du littoral Résumé : (auteur) Recent advances in structure-from-motion (SfM) techniques have proliferated the use of unmanned aerial vehicles (UAVs) in the monitoring of coastal landform changes, particularly when applied in the reconstruction of 3D surface models from historical aerial photographs. Here, we explore a number of depth map filtering and point cloud cleaning methods using the commercial software Agisoft Metashape Pro to determine the optimal methodology to build reliable digital surface models (DSMs). Twelve different aerial photography-derived DSMs are validated and compared against light detection and ranging (LiDAR)- and UAV-derived DSMs of a vegetated coastal dune system that has undergone several decades of coastline retreat. The different studied methods showed an average vertical error (root mean square error, RMSE) of approximately 1 m, with the best method resulting in an error value of 0.93 m. In our case, the best method resulted from the removal of confidence values in the range of 0–3 from the dense point cloud (DPC), with no filter applied to the depth maps. Differences among the methods examined were associated with the reconstruction of the dune slipface. The application of the modern SfM methodology to the analysis of historical aerial (vertical) photography is a novel (and reliable) new approach that can be used to better quantify coastal dune volume changes. DSMs derived from suitable historical aerial photographs, therefore, represent dependable sources of 3D data that can be used to better analyse long-term geomorphic changes in coastal dune areas that have undergone retreat. Numéro de notice : A2021-079 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.3390/rs13010095 Date de publication en ligne : 30/12/2020 En ligne : https://doi.org/10.3390/rs13010095 Format de la ressource électronique : url article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=96821 [article]

Dense stereo matching strategy for oblique images that considers the plane directions in urban areas / Jianchen Liu in IEEE Transactions on geoscience and remote sensing, vol 58 n° 7 (July 2020)  

Public [article]  inIEEE Transactions on geoscience and remote sensing > vol 58 n° 7 (July 2020) . - pp 5109 - 5116 Titre : Dense stereo matching strategy for oblique images that considers the plane directions in urban areas Type de document : Article/Communication Auteurs : Jianchen Liu, Auteur ; Linjing Zhang, Auteur ; Zhen Wang, Auteur ; et al., Auteur Année de publication : 2020 Article en page(s) : pp 5109 - 5116 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] appariement automatique [Termes IGN] appariement d'images [Termes IGN] appariement dense [Termes IGN] appariement semi-global [Termes IGN] bati [Termes IGN] carte de profondeur [Termes IGN] corrélation épipolaire dense [Termes IGN] distorsion d'image [Termes IGN] erreur moyenne quadratique [Termes IGN] image oblique [Termes IGN] perspective [Termes IGN] planéité [Termes IGN] zone urbaine Résumé : (auteur) The perspective distortion of oblique images has a substantial impact on dense matching, i.e., it reduces the matching precision. In this article, a strategy of dense matching in which the object plane direction is considered is proposed. According to many regular planes in urban areas, epipolar rectification with minimum distortions relative to the selected reference planes can be generated. The matching results of epipolar images relative to various reference planes are weighted and fused into a single depth map, which is a better matching result. The experimental results demonstrate that the perspective distortion has a substantial influence on the dense matching performance. The root-mean-square error (RMSE) of the flatness for horizontal objects is increased by approximately 30%, and the RMSE of the flatness for façades is increased by approximately 40%. Hence, the proposed matching strategy, in which the object plane is considered, can effectively improve the matching results. Numéro de notice : A2020-394 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1109/TGRS.2020.2972312 Date de publication en ligne : 20/02/2020 En ligne : https://doi.org/10.1109/TGRS.2020.2972312 Format de la ressource électronique : url article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=95390 [article]

Cartographie sémantique hybride de scènes urbaines à partir de données image et Lidar / Mohamed Boussaha (2020)  

Permalink

Enhanced 3D mapping with an RGB-D sensor via integration of depth measurements and image sequences / Bo Wu in Photogrammetric Engineering & Remote Sensing, PERS, vol 85 n° 9 (September 2019)  

Permalink

Improved camera distortion correction and depth estimation for lenslet light field camera / Changkun Yang in Photogrammetric Engineering & Remote Sensing, PERS, vol 85 n° 3 (March 2019)  

Permalink

Entropy in image analysis / Amelia Carolina Sparavigna (2019)  

Permalink

Estimation de profondeur à partir d'images monoculaires par apprentissage profond / Michel Moukari (2019)  

Permalink

Geometric camera pose refinement with learned depth maps / Nathan Piasco (2019)  

Permalink

Learning scene geometry for visual localization in challenging conditions / Nathan Piasco (2019)    

Permalink

Perspective-n-learned-point: pose estimation from relative depth / Nathan Piasco (2019)    

Permalink

Vision-based localization with discriminative features from heterogeneous visual data / Nathan Piasco (2019)  

Permalink

EMVS : Event-based Multi-View Stereo : 3D reconstruction with an event camera in real-time / Henri Rebecq in International journal of computer vision, vol 126 n° 12 (December 2018)  

Permalink