Abstract:This paper presents an automated pipeline for detecting tree whorls in proximally laser scanning data using a pose-estimation deep learning model. Accurate whorl detection provides valuable insights into tree growth patterns, wood quality, and offers potential for use as a biometric marker to track trees throughout the forestry value chain. The workflow processes point cloud data to create sectional images, which are subsequently used to identify keypoints representing tree whorls and branches along the stem. The method was tested on a dataset of destructively sampled individual trees, where the whorls were located along the stems of felled trees. The results demonstrated strong potential, with accurate identification of tree whorls and precise calculation of key structural metrics, unlocking new insights and deeper levels of information from individual tree point clouds.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to use deep - learning methods to automatically detect the positions of tree whorls and their geometric features, thereby providing more detailed and accurate information for tree growth patterns, wood quality assessment, and forest resource management. Specifically, the authors proposed a deep - learning model based on pose estimation (BranchPoseNet) to process tree point - cloud data in close - range laser - scanning data in order to identify the key points of tree whorls. This method aims to improve the traditional manual or semi - automated ways of analyzing LiDAR data. By introducing a pose - estimation model to capture the spatial relationships of branch arrangements on the trunk, it provides a more detailed and biologically - reasonable representation of the tree structure. ### Main problems and solutions: 1. **Detection of tree - structure features**: - **Problem**: Traditional methods have difficulty in efficiently and accurately extracting the specific positions and geometric properties of tree whorls from large - scale LiDAR data. - **Solution**: Use a pose - estimation model in deep learning (such as YOLOv8) to predict the key - point positions of tree whorls, thereby achieving automated, high - precision detection. 2. **Compatibility of multi - platform data**: - **Problem**: Data obtained from different types of laser - scanning platforms (such as ground - based, mobile, and unmanned - aerial - vehicle) are different, resulting in limited generalization ability of the model. - **Solution**: Increase the amount of training data and include data from multiple tree species and scanning platforms to improve the universality and adaptability of the model. 3. **Extraction of geometric features**: - **Problem**: It is necessary to extract more useful geometric parameters from the detected whorls, such as branch - insertion angles and branch lengths. - **Solution**: Utilize the knowledge of fixed geometric structures to calculate these additional structural parameters from the detection results, providing more information about tree morphology. ### Research background and significance: - **Tree - growth patterns**: Accurate whorl detection helps in understanding the growth history and future trends of trees, and supports forest management and carbon - capture research. - **Wood - quality assessment**: The presence of whorls is related to internal wood properties (such as knots and fiber disturbances), which affect the quality of wood. - **Competition analysis**: The size of the tree canopy and the number of visible whorls reflect the light - competition situation among trees. - **Traceability**: The whorl pattern can be used as a biomarker of tree identity for subsequent identification in the wood - processing chain. By solving the above problems, this research provides new tools and technical means for refined management and scientific research in the forestry field.

BranchPoseNet: Characterizing tree branching with a deep learning-based pose estimation approach

Advancing fine branch biomass estimation with lidar and structural models

Data-Driven Synthetic Modeling of Trees

Retrieval of Tree Branch Architecture Attributes from Terrestrial Laser Scan Data Using a Laplacian Algorithm

Automatic Reconstruction of Tree Skeletal Structures from Point Clouds

SegmentAnyTree: A sensor and platform agnostic deep learning model for tree segmentation using laser scanning data

CherryPicker: Semantic Skeletonization and Topological Reconstruction of Cherry Trees

L 1 -Tree: A novel algorithm for constructing 3D tree models and estimating branch architectural traits using terrestrial laser scanning data

L1-Tree: A Novel Algorithm for Constructing 3D Tree Models and Estimating Branch Architectural Traits Using Terrestrial Laser Scanning Data

Automatic marker-free registration of tree point-cloud data based on rotating projection

3D Branch Point Cloud Completion for Robotic Pruning in Apple Orchards

Fast Automatic Precision Tree Models from Terrestrial Laser Scanner Data

Development of a Precise Tree Structure from LiDAR Point Clouds

A Deep Learning-Based Method for Extracting Standing Wood Feature Parameters from Terrestrial Laser Scanning Point Clouds of Artificially Planted Forest

Vertical Optical Scanning with Panoramic Vision for Tree Trunk Reconstruction

Treegraph: tree architecture from terrestrial laser scanning point clouds

Online Tree Reconstruction and Forest Inventory on a Mobile Robotic System

Using Synthetic Tree Data in Deep Learning-Based Tree Segmentation Using LiDAR Point Clouds

Tree stem extraction from TLS point-cloud data of natural forests based on geometric features and DBSCAN

TreePartNet

Tree Branch Skeleton Extraction from Drone-Based Photogrammetric Point Cloud