Comparison between UAV and terrestrial LiDAR scans for high throughput phenotyping of architectural traits of a core collection of apple trees

Rojas bustos Juan Pablo, Branthomme Anne, Costes Evelyne, Boudon Frédéric. 2023. Comparison between UAV and terrestrial LiDAR scans for high throughput phenotyping of architectural traits of a core collection of apple trees. In : Proceedings of the III International Symposium on Mechanization, Precision Horticulture, and Robotics: Precision and Digital Horticulture in Field Environments. Sankaran S. (ed.), Rousseau D. (ed.). ISHS. Louvain : ISHS, 15-22. (Acta Horticulturae, 1360) ISBN 978-94-62613-59-1 International Horticultural Congress (IHC 2022): International Symposium on Mechanization, Precision Horticulture, and Robotics: Precision and Digital Horticulture in Field Environments. 31, Angers, France, 14 Août 2022/20 Août 2022.

https://doi.org/10.17660/ActaHortic.2023.1360.2

Communication avec actes

Version publiée - Anglais
Accès réservé aux personnels Cirad
Utilisation soumise à autorisation de l'auteur ou du Cirad.
Rojas-Bustos et al. - 2023 - Comparison between UAV and terrestrial LiDAR scans.pdf
Télécharger (1MB) | Demander une copie

Note générale : A l'occasion de ce congrès, s'est également déroulé le III International Symposium on Mechanization, Precision Horticulture, and Robotics: Precision and Digital Horticulture in Field Environments, du 17 au 19 août 2022, Angers, France

Résumé : In a context of climate change, the selection of fruit tree cultivars that perform well under sub-optimal growing conditions becomes essential. Architectural traits must be considered to assess the intrinsic production potential of cultivars, their interactions with the environment and the easiness of management. To phenotype such traits at high throughput on a core-collection of apple trees, we tested an approach based on UAV-LiDARs that allow rapid 3D scanning of an orchard and compared it to our previous approach, based on TLS. With the UAV-LiDAR different acquisition protocols were tested, with varying height or speed for the drone, that resulted in different densities and qualities of points. To process the point clouds, we built a pipeline composed of steps including the identification and removal of undesired elements (soil, pole, etc.), the segmentation of individual trees, and the characterization of architectural traits. For the first step, two methods were tested: CANUPO and RandLA-NET. For the tree segmentation, we used a semi-supervised method of label spreading. The initial seeds for the labels were determined from the GPS location of the trees. Architectural traits such as height, projected leaf area, convex and alpha volume, eccentricity were then determined and their broad sense heritabilities were estimated to assess genotypic variability and measure repeatability. The use of UAV-LiDAR scans was compared and validated with terrestrial LiDAR scans. The influence of the acquisition protocol on the resulting architectural traits was characterized. Correlations greater than or equal to 0.5 were found between the estimated indices from the different protocols, except for eccentricity. Indices from UAV scans (F2, F3) presented values similar to those obtained with the TLS. As a result, indices obtained with TLS can be approximated using UAV-LiDAR.

Agences de financement hors UE : Agence Nationale de la Recherche

Projets sur financement : (FRA) Institut Convergences en Agriculture Numérique

Auteurs et affiliations