Centre de documentation
scientifique

Matching plot-level tree maps with 3D remote sensing data for assessing and estimating forest parameters / Cédric Vega (2017)

Public Titre : Matching plot-level tree maps with 3D remote sensing data for assessing and estimating forest parameters Type de document : Article/Communication Auteurs : Cédric Vega , Auteur ; Maryem Fadili , Auteur Editeur : Saint-Mandé : Institut national de l'information géographique et forestière - IGN (2012-) Année de publication : 2017 Conférence : SilviLaser 2017, 15th conference on Lidar Applications for Assessing and Managing Forest Ecosystems 10/10/2017 12/10/2017 Blacksburg Virginie - Etats-Unis OA Abstracts only Langues : Anglais (eng) Descripteur : [Termes IGN] appariement de données localisées [Termes IGN] erreur de positionnement [Termes IGN] inventaire forestier (techniques et méthodes) [Termes IGN] modèle numérique de surface de la canopée [Termes IGN] programmation par contraintes [Termes IGN] structure d'un peuplement forestier [Vedettes matières IGN] Inventaire forestier Résumé : (auteur) 3D remote sensing data from either Lidar or Photogrammetric means are recognized as valuable sources of information for assessing and estimating forest structure and related parameters. Both data types have been used with field inventory data for both mapping forest parameters and supporting multisource inventories. However, such a combination requires the data to be accurately matched in the spatial domain. While 3D remote sensing data might provide metric accuracy, the spatial accuracy of field plots remain largely constrained by the limited precision of GPS measurements under forest canopies. Different approaches have been proposed to improve this data registration issue, mainly through matching algorithms aiming to spatially adjust height information from field inventory with remote sensing-based models of canopy heights (CHM). State of the art approaches rely on either point to surface or point to point matching algorithms. However, the former did not make any hypothesis on the tree position on the CHM and could lead to inappropriate match. And the later relies on strong assumptions on the spatial distribution of trees and are thus sensitive to the quality of the tree apices detected on the CHM. We propose an algorithm taking advantage of both approaches. The algorithm is based on a point to surface matching algorithm constraints by local maxima (LM) extracted from the CHM. A search algorithm moved the field tree map in a given neighborhood, ensuring that the highest field tree is located over a LM. The best position is defined using both the correlation and the height error. The algorithm was tested on 91 plots including different forest types and a range of forest structure. Initial positions were shifted in average by 2.18 m (±1.95 m SD) and led to an average error of 1.61 m (±1.07 m). The higher the tree number, the better the registration. Numéro de notice : C2017-061 Affiliation des auteurs : LIF (2012-2019) Thématique : FORET/MATHEMATIQUE Nature : Communication nature-HAL : ComSansActesPubliés-Unpublished DOI : sans Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=99224