A New Strategy for Individual Tree Detection and Segmentation from Leaf-on and Leaf-off UAV-LiDAR Point Clouds Based on Automatic Detection of Seed Points

Abstract

Accurate and efficient estimation of forest volume or biomass is critical for carbon cycles, forest management, and the timber industry. Individual tree detection and segmentation (ITDS) is the first and key step to ensure the accurate extraction of detailed forest structure parameters from LiDAR (light detection and ranging). However, ITDS is still a challenge to achieve using UAV-LiDAR (LiDAR from Unmanned Aerial Vehicles) in broadleaved forests due to the irregular and overlapped canopies. We developed an efficient and accurate ITDS framework for broadleaved forests based on UAV-LiDAR point clouds. It involves ITD (individual tree detection) from point clouds taken during the leaf-off season, initial ITS (individual tree segmentation) based on the seed points from ITD, and improvement of initial ITS through a refining process. The results indicate that this new proposed strategy efficiently provides accurate results for ITDS. We show the following: (1) point-cloud-based ITD methods, especially the Mean Shift, perform better for seed point selection than CHM-based (Canopy Height Model) ITD methods on the point clouds from leaf-off seasons; (2) seed points significantly improved the accuracy and efficiency of ITS algorithms; (3) the refining process using DBSCAN (density-based spatial clustering of applications with noise) and kNN (k-Nearest Neighbor classifier) classification significantly reduced edge errors in ITS results. Our study developed a novel ITDS strategy for UAV-LiDAR point clouds that demonstrates proficiency in dense deciduous broadleaved forests, and this proposed ITDS framework could be applied to single-phase point clouds instead of the multi-temporal LiDAR data in the future if the point clouds have detailed tree trunk points.

Automatic Detection and Counting of Stacked Eucalypt Timber Using the YOLOv8 Model

“…Research has indicated that object detection can be applied to a range of variables, enabling precise classification of scenes from remote sensing data. This includes tasks such as detecting trees [17] and dead trees in forest management [54], thereby enhancing our understanding of human impacts on natural ecosystems and biodiversity.…”

“…While statistical comparisons found no significant differences between the evaluated measurement techniques, practical considerations emphasize the potential financial impact of volume discrepancies [15]. Furthermore, LiDAR-based methods are a trend in these times; it is possible that these methods will complement, or even replace, traditional measurement techniques in the future [3,16,17].…”

3D point cloud global denoising algorithm based on different noise characteristics

“…If the absolute value of the cosine of the normal of the point p and the normal of the point q i in the neighborhood is less than the threshold σ 3 , the effect of point q i is not considered when smoothing the noise. m i can be obtained from equation (16):…”

A Hybrid Method for Individual Tree Detection in Broadleaf Forests Based on UAV-LiDAR Data and Multistage 3D Structure Analysis

“…Reitberger et al [26] proposed a tree segmentation approach that first applied watershed segmentation on a CHM, then detected tree trunks in each segment using a hierarchical clustering scheme, and finally segmented individual trees using a normalized cut segmentation method. Focusing on poplar plantations, Pu et al [27] detected trunk seed points from leaf-off LiDAR data using three methods (local maximum, local minimum, and mean shift algorithm), and applied seed-based segmentation methods to segment individual trees from leaf-on LIDAR data. The segmentation result was refined in a horizontal plane based on three tree-crown features.…”

“…Furthermore, trunk detection is usually based on the LiDAR data collected in the leaf-off season (hereafter referred to as leaf-off data) because more laser pulses penetrate the canopy in the leaf-off season [35]. On the contrary, the crown segmentation is commonly performed on the LiDAR data acquired in the leaf-on seasons (hereafter referred to as leaf-on data) [27]. Since it takes more time to acquire multi-season data, it is necessary to analyze the influence of the data acquisition season on the tree detection result and evaluate the possibility of using the hybrid method to segment crowns and detect individual tree trunks based on single-season data.…”