Automated Individual Tree Crown Detection and Segmentation using Simple Non-Iterative Clustering (SNIC) Algorithms and High-Resolution LiDAR

This study introduced an automated method for Individual Tree Crown Detection (ITCD) and Segmentation (ITCS) using high-resolution LiDAR data and the Simple Non-Iterative Clustering (SNIC) algorithm. The objective was to develop and evaluate an approach for detecting and segmenting individual tree crowns, emphasizing their reliability and robustness. The method was tested across different tree canopy cover (TCC) densities—classified as high (>70%), medium (40–70%), and low (<40%)—and various species categories. The performance of ITCD and ITCS was assessed using accuracy metrics and statistical analyses. Validation involved comparing the segmented tree crowns (STC) derived from the algorithm with visually tallied crowns (VTC) and additional ground-based verification. The algorithm achieved high precision, recall, and F1 scores, each exceeding 75%, resulting in an overall detection accuracy of approximately 64%. A one-way ANOVA and independent sample tests indicated no significant differences in detection accuracy across the TCC density and species categories. The method achieved an average Intersection-over-Union (IoU) of 65% for segmentation. A one-to-one matching analysis demonstrated a high correspondence between STC and VTC when evaluating crown area alignment, with a coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) exceeding 0.80. Additionally, the Wilcoxon signed-rank test indicated no statistical differences in the distribution of crown areas across most tree crown size categories, further validating the method's robustness. These results highlight the potential of the proposed approach as a reliable and scalable tool for forest inventory applications across varying forest conditions.