Centre dedocumentation
scientifique

Topology-based individual tree segmentation for automated processing of terrestrial laser scanning point clouds / Xin Xu in International journal of applied Earth observation and geoinformation, vol 116 (February 2023)  

Public [article]  inInternational journal of applied Earth observation and geoinformation > vol 116 (February 2023) . - n° 103145 Titre : Topology-based individual tree segmentation for automated processing of terrestrial laser scanning point clouds Type de document : Article/Communication Auteurs : Xin Xu, Auteur ; Federico Iuricich, Auteur ; Kim Calders, Auteur ; et al., Auteur Année de publication : 2023 Article en page(s) : n° 103145 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie [Termes IGN] analyse de groupement [Termes IGN] données lidar [Termes IGN] données localisées 3D [Termes IGN] extraction d'arbres [Termes IGN] houppier [Termes IGN] segmentation d'image [Termes IGN] semis de points [Termes IGN] topologie Résumé : (auteur) Terrestrial laser scanning (TLS) is a ground-based approach to rapidly acquire 3D point clouds via Light Detection and Ranging (LiDAR) technologies. Quantifying tree-scale structure from TLS point clouds requires segmentation, yet there is a lack of automated methods available to the forest ecology community. In this work, we consider the problem of segmenting a forest TLS point cloud into individual tree point clouds. Different approaches have been investigated to identify and segment individual trees in a forest point cloud. Typically these methods require intensive parameter tuning and time-consuming user interactions, which has inhibited the application of TLS to large area research. Our goal is to define a new automated segmentation method that lifts these limitations. Our Topology-based Tree Segmentation (TTS) algorithm uses a new topological technique rooted in discrete Morse theory to segment input point clouds into single trees. TTS algorithm identifies distinctive tree structures (i.e., tree bottoms and tops) without user interactions. Tree tops and bottoms are then used to reconstruct single trees using the notion of relevant topological features. This mathematically well-established notion helps distinguish between noise and relevant tree features. To demonstrate the generality of our approach, we present an evaluation using multiple datasets, including different forest types and point densities. We also compare our TTS approach with open-source tree segmentation methods. The experiments show that we achieve a higher segmentation accuracy when performing point-by-point validation. Without expensive user interactions, TTS algorithm is promising for greater usage of TLS point clouds in the forest ecology community, such as fire risk and behavior modeling, estimating tree-level biodiversity structural traits, and above-ground biomass monitoring. Numéro de notice : A2023-129 Affiliation des auteurs : non IGN Thématique : FORET/IMAGERIE Nature : Article DOI : 10.1016/j.jag.2022.103145 Date de publication en ligne : 12/12/2022 En ligne : https://doi.org/10.1016/j.jag.2022.103145 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=102517 [article]