Abstract:Accurate identification and segmentation of individual tree points are crucial for assessing forest spatial distribution, understanding tree growth and structure, and managing forest resources. Traditional methods based on Canopy Height Models (CHM) are simple yet prone to over- and/or under-segmentation. To deal with this problem, this paper introduces a novel approach that combines marker-controlled watershed segmentation with a spectral clustering algorithm. Initially, we determined the local maxima within a series of variable windows according to the lower bound of the prediction interval of the regression equation between tree crown radius and tree height to preliminarily segment individual trees. Subsequently, using this geometric shape analysis method, the under-segmented trees were identified. For these trees, vertical tree crown profile analysis was performed in multiple directions to detect potential treetops which were then considered as inputs for spectral clustering optimization. Our experiments across six plots showed that our method markedly surpasses traditional approaches, achieving an average Recall of 0.854, a Precision of 0.937, and an F1-score of 0.892.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to accurately distinguish individual trees using airborne LiDAR point cloud data in broad - leaved forest and coniferous forest scenes, while overcoming the problems of over - segmentation and under - segmentation existing in traditional methods**. ### Specific problem description: 1. **Limitations of traditional methods**: - **CHM (Canopy Height Model) method**: Although simple and efficient, in complex forest environments (such as broad - leaved forests), due to the diverse shapes of tree canopies and the mutual occlusion of adjacent canopies, it is prone to over - segmentation or under - segmentation. - **Point cloud method**: It can directly process 3D point cloud data, but has high computational complexity and is difficult to be applied to large - scale forest scenes. - **Deep learning method**: Although it has high precision, it needs to convert 3D point cloud into 2D images, resulting in information loss, and has a large demand for labeled data. 2. **Research objectives**: - Propose a hybrid method that combines the advantages of the CHM method and the point cloud method to achieve more accurate individual tree segmentation. - Improve the detection ability of non - dominant small trees by optimizing the under - segmented areas. ### Core solutions in the paper: The paper proposes a hybrid method that combines **Marker - Controlled Watershed Segmentation** and **Spectral Clustering Optimization**. The specific steps are as follows: 1. **Coarse segmentation stage**: - In the rasterized CHM image, use a variable window size to identify local maxima (tree - top positions). - Take these tree - tops as markers and input them into the Marker - Controlled Watershed Segmentation algorithm to generate the initial individual - tree segmentation results. 2. **Optimization stage**: - For the under - segmented areas, perform multi - directional vertical canopy profile analysis to infer potential tree - top positions. - Use these inferred tree - tops as seed points and apply the spectral clustering algorithm to optimize the segmentation results. ### Summary of mathematical formulas: - **Regression model of canopy radius and tree height**: $$ y = a+bx + cx^{2} $$ where $y$ represents the canopy radius, $x$ represents the tree height, and $a, b, c$ are regression coefficients. - **Lower curve of the prediction interval**: $$ y_{\text{lower}}=a_{\text{lower}}+b_{\text{lower}}x + c_{\text{lower}}x^{2} $$ For example, for Plot_1, the regression model is: $$ y=- 0.301+0.344x - 0.005x^{2} $$ The lower bound of the prediction interval is: $$ y_{\text{lower}}=-4.781+0.470x - 0.008x^{2} $$ Through the above methods, the paper aims to provide a robust and efficient individual - tree segmentation scheme for airborne LiDAR point cloud data in complex forest environments.

Segmentation of Individual Tree Points by Combining Marker-Controlled Watershed Segmentation and Spectral Clustering Optimization

Individual tree segmentation of airborne and UAV LiDAR point clouds based on the watershed and optimized connection center evolution clustering

Individual Tree Segmentation Based on Seed Points Detected by an Adaptive Crown Shaped Algorithm Using UAV-LiDAR Data

Individual-Tree Segmentation from UAV–LiDAR Data Using a Region-Growing Segmentation and Supervoxel-Weighted Fuzzy Clustering Approach

A Hierarchical Region-Merging Algorithm for 3-D Segmentation of Individual Trees Using UAV-LiDAR Point Clouds

An Individual Tree Segmentation Method That Combines LiDAR Data and Spectral Imagery

A Hierarchic Method of Individual Tree Canopy Segmentation Combing UAV Image and LiDAR

Study on Single-Tree Segmentation of Chinese Fir Plantations Using Coupled Local Maximum and Height-Weighted Improved K-Means Algorithm

Improved Tree Segmentation Algorithm Based on Backpack-LiDAR Point Cloud

Individual tree crown segmentation from airborne LiDAR data using a novel Gaussian filter and energy function minimization-based approach

Segmenting Individual Trees from Terrestrial LiDAR Data Using Tree Branch Directivity

Individual Tree-Crown Delineation and Treetop Detection in High-Spatial-Resolution Aerial Imagery

A New Method for Segmenting Individual Trees from the Lidar Point Cloud

A Novel Vegetation Point Cloud Density Tree-Segmentation Model for Overlapping Crowns Using UAV LiDAR

Delineating Individual Tree Crowns in an Uneven-Aged, Mixed Broadleaf Forest Using Multispectral Watershed Segmentation and Multiscale Fitting

Adaptive Mean Shift-Based Identification of Individual Trees Using Airborne LiDAR Data

Individual tree segmentation and tree species classification in subtropical broadleaf forests using UAV-based LiDAR, hyperspectral, and ultrahigh-resolution RGB data

Airborne LiDAR Remote Sensing for Individual Tree Forest Inventory Using Trunk Detection-Aided Mean Shift Clustering Techniques.

An Improved Method for Individual Tree Segmentation in Complex Urban Scenes Based on Using Multispectral LiDAR by Deep Learning

Tree Species Classification Using Airborne LiDAR Data Based on Individual Tree Segmentation and Shape Fitting

A Multi-Threshold Segmentation for Tree-Level Parameter Extraction in a Deciduous Forest Using Small-Footprint Airborne LiDAR Data