A Comparative Assessment of the Performance of Individual Tree Crowns Delineation Algorithms from ALS Data in Tropical Forests

Abstract

Tropical forest canopies are comprised of tree crowns of multiple species varying in shape and height, and ground inventories do not usually reliably describe their structure. Airborne laser scanning data can be used to characterize these individual crowns, but analytical tools developed for boreal or temperate forests may require to be adjusted before they can be applied to tropical environments. Therefore, we compared results from six different segmentation methods applied to six plots (39 ha) from a study site in French Guiana. We measured the overlap of automatically segmented crowns projection with selected crowns manually delineated on high-resolution photography. We also evaluated the goodness of fit following automatic matching with field inventory data using a model linking tree diameter to tree crown width. The different methods tested in this benchmark segmented highly different numbers of crowns having different characteristics. Segmentation methods based on the point cloud (AMS3D and Graph-Cut) globally outperformed methods based on the Canopy Height Models, especially for small crowns; the AMS3D method outperformed the other methods tested for the overlap analysis, and AMS3D and Graph-Cut performed the best for the automatic matching validation. Nevertheless, other methods based on the Canopy Height Model performed better for very large emergent crowns. The dense foliage of tropical moist forests prevents sufficient point densities in the understory to segment subcanopy trees accurately, regardless of the segmentation method.

Geographic Generalization in Airborne RGB Deep Learning Tree Detection

“…Given our results, what are the strengths and limitations ecologists should consider when adding airborne-derived data? Our results, and prior works (Aubry-Kientz et al, 2019), suggest that small and subcanopy trees will likely be overlooked. We therefore expect that studies in which the results are driven by the upper canopy, sun-exposed trees will benefit the most from remote sensing at broad scales.…”

“…Identifying individual crowns in airborne imagery allows ecologists, foresters, and land managers to increase the extent of sampling compared to terrestrial surveys. While many LIDAR-based tree segmentation algorithms have been proposed (Aubry-Kientz et al, 2019), the field has been slow to adopt automated methods due to concerns over accuracy, transferability and transparency (Vaglio Laurin et al, 2019). As a result, existing methods are rarely evaluated on multiple forests simultaneously, making it unclear how they will perform in the novel contexts required for large scale application.…”

“…As a result of these limitations, most tree detection algorithms have been applied and tested on similar forest types with little exploration of how the algorithms generalize to other natural settings. Therefore, despite the intense work in airborne tree detection over the last decade (Coomes et al, 2017; Heinzel and Huber, 2018; Jakubowski et al, 2013; Li et al, 2012; Williams et al, 2019), there remains no clear consensus on best practices (Aubry-Kientz et al, 2019).…”

Tree Species Traits Determine the Success of LiDAR-Based Crown Mapping in a Mixed Temperate Forest

“…This may be because our validation method (manual delineation of UAV imagery) allowed for accurate surveying of canopy trees but did not attempt to address understory trees. Point cloud-based methods outperform CHM-based methods in detection and delineation of understory and suppressed crowns [58], but there is less of an advantage for delineation of canopy trees.…”

Quantifier les dimensions des houppiers à l’aide d’images aériennes à haute résolution pour estimer l’accroissement diamétrique des arbres dans les forêts d’Afrique centrale

“…Des techniques visuelles et automatiques de segmentation des houppiers ont été testées à partir des données LiDAR (Light and Detection Ranging) ou d'images satellitaires (IKO-NOS, 1-4 m/pixel) et ont montré des résultats prometteurs (Asner et al, 2002 ;Palace et al, 2008). L'efficacité de la segmentation automatique reste toutefois limitée à cause de la complexité de la structure des canopées en forêt dense tropicale (Aubry-Kientz et al, 2019 ;Broadbent et al, 2008). La précision de la détection visuelle des houppiers est quant à elle dépendante du niveau de résolution des images utilisées.…”