This work proposes a segmentation method that isolates individual tree crowns using airborne LiDAR data. The proposed approach captures the topological structure of the forest in hierarchical data structures, quantifies topological relationships of tree crown components in a weighted graph, and finally partitions the graph to separate individual tree crowns. This novel bottom-up segmentation strategy is based on several quantifiable cohesion criteria that act as a measure of belief on weather two crown components belong to the same tree. An added flexibility is provided by a set of weights that balance the contribution of each criterion, thus effectively allowing the algorithm to adjust to different forest structures.The LiDAR data used for testing was acquired in Louisiana, inside the Clear Creek Wildlife management area with a RIEGL LMS-Q680i airborne laser scanner. Three 1ha forest areas of different conditions and increasing complexity were segmented and assessed in terms of an accuracy index (AI) accounting for both omission and commission. The three areas were segmented under optimum parameterization with an AI of 98.98%, 92.25% and 74.75% respectively, revealing the excellent potential of the algorithm. When segmentation parameters are optimized locally using plot references the AI drops to 98.23%, 89.24%, and 68.04% on average with plot sizes of 1000m2 and 97.68%, 87.78% and 61.1% on average with plot sizes of 500m2.More than introducing a segmentation algorithm, this paper proposes a powerful framework featuring flexibility to support a series of segmentation methods including some of those recurring in the tree segmentation literature. The segmentation method may extend its applications to any data of topological nature or data that has a topological equivalent

Autonomous Unmanned Aerial Vehicles (UAVs) have gained popularity due to their many potential application fields. Alongside sophisticated sensors, UAVs can be equipped with communication adaptors aimed for inter-UAV communication. Inter-communication of UAVs to form a UAV swarm raises questions on how to manage its communication structure and mobility. In this paper, we consider therefore the problem of establishing an efficient swarm movement model and a network topology between a collection of UAVs, which are specifically deployed for the scenario of high-quality forest-mapping. The forest environment with its highly heterogeneous distribution of trees and obstacles represents an extreme challenge for a UAV swarm. It requires the swarm to constantly avoid possible collisions with trees, to change autonomously the trajectory, which can lead to disconnection to the swarm, and to reconnect to the swarm after passing the obstacle, while continue collecting environmental data that needs to be fused and assessed efficiently. In this paper, we propose a novel solution to the formation flight problem for UAV swarms. The proposed method provides an adaptive and reliable network structure, which maintains swarm connectivity and communicability. These characteristics are needed to achieve a detailed and accurate description of the environment from the data acquired by the UAV swarm. The main characteristics of our approach are high scalability regarding the number of UAVs in the swarm and the adaptive network topology within the swarm.

The estimation of tree biomass and the products that can be obtained from a tree stem have focused forest research for more than two centuries. Traditionally, measurements of the entire tree bole were expensive or inaccurate, even when sophisticated remote sensing techniques were used. We propose a fast and accurate procedure for measuring diameters along the merchantable portion of the stem at any given height. The procedure uses unreferenced photos captured with a consumer grade camera. A photogrammetric point cloud (PPC) is produced from the acquired images using structure from motion, which is a computer vision range imaging technique. A set of 18 loblolly pines (Pinus taeda Lindl.) from east Louisiana, USA, were photographed, subsequently cut, and the diameter measured every meter. The same diameters were measured on the point cloud with AutoCAD Civil3D. The ground point cloud reconstruction provided useful information for at most 13 m along the stem. The PPC measurements are biased, overestimating real diameters by 17.2 mm, but with a reduced standard deviation (8.2%). A linear equation with parameters of the error at a diameter at breast height (d_1.3) and the error of photogrammetric rendering reduced the bias to 1.4 mm. The usability of the PPC measurements in taper modeling was assessed with four models: Max and Burkhart [¹], Baldwin and Feduccia [²], Lenhart et al. [³], and Kozak [⁴]. The evaluation revealed that the data fit well with all the models (R² ≥ 0.97), with the Kozak and the Baldwin and Feduccia performing the best. The results support the replacement of taper with PPC, as faster, and more accurate and precise product estimations are expected.

A plethora of information contained in full-waveform (FW) Light Detection and Ranging (LiDAR) data offers prospects for characterizing vegetation structures. This study aims to investigate the capacity of FW LiDAR data alone for tree species identification through the integration of waveform metrics with machine learning methods and Bayesian inference. Specifically, we first conducted automatic tree segmentation based on the waveform-based canopy height model (CHM) using three approaches including TreeVaW, watershed algorithms and the combination of TreeVaW and watershed (TW) algorithms. Subsequently, the Random forests (RF) and Conditional inference forests (CF) models were employed to identify important tree-level waveform metrics derived from three distinct sources, such as raw waveforms, composite waveforms, the waveform-based point cloud and the combined variables from these three sources. Further, we discriminated tree (gray pine, blue oak, interior live oak) and shrub species through the RF, CF and Bayesian multinomial logistic regression (BMLR) using important waveform metrics identified in this study. Results of the tree segmentation demonstrated that the TW algorithms outperformed other algorithms for delineating individual tree crowns. The CF model overcomes waveform metrics selection bias caused by the RF model which favors correlated metrics and enhances the accuracy of subsequent classification. We also found that composite waveforms are more informative than raw waveforms and waveform-based point cloud for characterizing tree species in our study area. Both classical machine learning methods (the RF and CF) and the BMLR generated satisfactory average overall accuracy (74% for the RF, 77% for the CF and 81% for the BMLR) and the BMLR slightly outperformed the other two methods. However, these three methods suffered from low individual classification accuracy for the blue oak which is prone to being misclassified as the interior live oak due to the similar characteristics of blue oak and interior live oak. Uncertainty estimates from the BMLR method compensate for this downside by providing classification results in a probabilistic sense and rendering users with more confidence in interpreting and applying classification results to real-world tasks such as forest inventory. Overall, this study recommends the CF method for feature selection and suggests that BMLR could be a superior alternative to classical machining learning methods.

The Douglas fir crown structure serves important ecological functions in regulating the ecosystem of the Pacific Northwest (PNW). Mapping and modeling of the Douglas-fir crown has traditionally focused on young plantations or old-growth forests. The crown description in natural regime forests is limited by data availability. Terrestrial laser scanning (TLS) enables the acquisition of crown structural attributes, even in dense forests, at a fine scale. The certical and horizontal distributions of the fine-scale branch attributes, such as branch diameter, branch length, and branch insertion angle, will reflect the crown behaviors towards light resources availability, as a result of neighborhood competition. The main objective of the study is to compare crown structural models of a group of neighboring trees developed with two TLS-based procedures, namely: semi-automatic (Cyclone software) and automatic (TreeQSM) procedures. The estimated crown attributes are the branch diameter, branch length, branch insertion angle, height of branch insertion point, and branch azimuth. The results show that branch azimuth distribution does not differ between TreeQSM and Cyclone for most of the sample trees. However, the TreeQSM and Cyclone identified branches exhibit different distributions of insertion height. A paired t-test indicates no difference between the mean branch diameter of Cyclone and TreeQSM at an individual tree level. However, Cyclone estimated that the branch length and branch insertion angle are 0.49 m and 9.9° greater than the TreeQSM estimates, respectively. Repeat measurements of the analysis of variance (ANOVA) suggest that the height along the stem is an influential factor of the difference between the Cyclone and TreeQSM branch diameter estimates. To test whether TLS-based estimates are within the ranges of the previous observations, we computed the tree crown attributes of second- and old-growth trees using Monte Carlo simulations for diameter at breast height (DBH) class 50–55 cm, 60–65 cm, and 85–105 cm. We found that the crown attributes estimated from both of the TLS-based methods are between the simulated second- and old-growth trees, except for DBH 85–105 cm. The TLS-based crown structural models show increasingly diverse distributions of branch insertion angles and increasing branch exclusion as DBH increases. Cyclone-based crown structural models are consistent with previous studies. However, TreeQSM-based crown structural models omitted a significant number of branches and generated crown structures with reduced plausibility.