Centre de documentation
scientifique

PCEDNet: a lightweight neural network for fast and interactive edge detection in 3D point clouds / Chems-Eddine Himeur in ACM Transactions on Graphics, TOG, Vol 41 n° 1 (February 2022)  

Public [article]  inACM Transactions on Graphics, TOG > Vol 41 n° 1 (February 2022) . - n° 10 Titre : PCEDNet: a lightweight neural network for fast and interactive edge detection in 3D point clouds Type de document : Article/Communication Auteurs : Chems-Eddine Himeur, Auteur ; Thibault Lejemble, Auteur ; Thomas Pellegrini, Auteur ; Mathias Paulin, Auteur ; Loïc Barthe, Auteur ; Nicolas Mellado, Auteur Année de publication : 2022 Article en page(s) : n° 10 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie [Termes IGN] classification par réseau neuronal convolutif [Termes IGN] détection de contours [Termes IGN] données lidar [Termes IGN] données localisées 3D [Termes IGN] matrice [Termes IGN] semis de points Résumé : (auteur) In recent years, Convolutional Neural Networks (CNN) have proven to be efficient analysis tools for processing point clouds, e.g., for reconstruction, segmentation, and classification. In this article, we focus on the classification of edges in point clouds, where both edges and their surrounding are described. We propose a new parameterization adding to each point a set of differential information on its surrounding shape reconstructed at different scales. These parameters, stored in a Scale-Space Matrix (SSM), provide a well-suited information from which an adequate neural network can learn the description of edges and use it to efficiently detect them in acquired point clouds. After successfully applying a multi-scale CNN on SSMs for the efficient classification of edges and their neighborhood, we propose a new lightweight neural network architecture outperforming the CNN in learning time, processing time, and classification capabilities. Our architecture is compact, requires small learning sets, is very fast to train, and classifies millions of points in seconds. Numéro de notice : A2022-304 Affiliation des auteurs : non IGN Autre URL associée : vers ArXiv Thématique : IMAGERIE/INFORMATIQUE Nature : Article DOI : 10.1145/3481804 Date de publication en ligne : 10/11/2021 En ligne : https://doi.org/10.1145/3481804 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=100374 [article]

Multi-scale point cloud analysis / Thibault Lejemble (2020)  

Public Titre : Multi-scale point cloud analysis Titre original : Analyse multi-échelle de nuage de points Type de document : Thèse/HDR Auteurs : Thibault Lejemble, Auteur ; Loïc Barthe, Directeur de thèse Editeur : Toulouse : Université de Toulouse 3 Paul Sabatier Année de publication : 2020 Importance : 142 p. Format : 21 x 30 cm Note générale : bibliographie Thèse en vue du Doctorat de l'Université de Toulouse en Informatique et Télécommunications Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie [Termes IGN] analyse multiéchelle [Termes IGN] analyse multirésolution [Termes IGN] anisotropie [Termes IGN] approche hiérarchique [Termes IGN] données lidar [Termes IGN] données localisées 3D [Termes IGN] extraction automatique [Termes IGN] géométrie différentielle [Termes IGN] graphe [Termes IGN] reconnaissance de formes [Termes IGN] segmentation en plan [Termes IGN] segmentation en régions [Termes IGN] semis de points [Termes IGN] visualisation 3D Index. décimale : THESE Thèses et HDR Résumé : (auteur) 3D acquisition techniques like photogrammetry and laser scanning are commonly used in numerous fields such as reverse engineering, archeology, robotics and urban planning. The main objective is to get virtual versions of real objects in order to visualize, analyze and process them easily. Acquisition techniques become more and more powerful and affordable which creates important needs to process efficiently the resulting various and massive3D data. Data are usually obtained in the form of unstructured 3D point cloud sampling the scanned surface. Traditional signal processing methods cannot be directly applied due to the lack of spatial parametrization. Points are only represented by their 3D coordinates without any particular order. This thesis focuses on the notion of scale of analysis defined by the size of the neighborhood used to locally characterize the point-sampled surface. The analysis at different scales enables to consider various shapes which increases the analysis pertinence and the robustness to acquired data imperfections. We first present some theoretical and practical results on curvature estimation adapted to a multi-scale and multi-resolution representation of point clouds. They are used to develop multi-scale algorithms for the recognition of planar and anisotropic shapes such as cylinder sand feature curves. Finally, we propose to compute a global 2D parametrization of the underlying surface directly from the 3D unstructured point cloud. Note de contenu : Introduction 1- Multi-scale differential analysis of point clouds 2- Plane detection using persistence analysis of graph 3- An isotropic features detection using curvature lines 4- Point cloud parametrization Conclusion Numéro de notice : 28583 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse française Note de thèse : Thèse de Doctorat : Informatique et Télécommunications : Toulouse 3 : 2020 Organisme de stage : Institut de recherche en informatique de Toulouse En ligne : https://tel.archives-ouvertes.fr/tel-03170824/document Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=97923