Education, Science, Technology, Innovation and Life
Open Access
Sign In
  1. Home
  2. Journals
  3. Automation and Machine Learning
  4. Vol 7, Issue 1, 2026

A Study on a Single-Tree Segmentation Method Based on UAV Imagery and a YOLOX Threshold Cascade

Download as PDF

DOI: 10.23977/autml.2026.070117 | Downloads: 2 | Views: 8

Author(s)

Tianliang Zhang 1,2, Zerun Liu 1,2, Jia Dong 1,2, Tang Song 1,2, Huan Mei 1,2, Zhiqiang Wang 1,2

Affiliation(s)

1 Institute of Applied Mathematics, Hebei Academy of Sciences, No. 46 South Youyi Street, Shijiazhuang, China
2 Hebei Information Security Certification Technology Innovation Center, No. 46 South Youyi Street, Shijiazhuang, China

Corresponding Author

Zhiqiang Wang

ABSTRACT

Accurate individual tree segmentation is a core technical prerequisite for forest resource surveys, stand structure analysis, carbon stock estimation, and smart forestry monitoring. Traditional individual tree segmentation methods rely on manually designed features and global image processing, which are prone to over-segmentation, under-segmentation, and missed detections in scenarios with overlapping tree canopies, uneven lighting, and complex backgrounds. In contrast, pure deep learning instance segmentation models suffer from limitations such as high annotation costs, slow inference speeds, and insufficient edge accuracy. To address these challenges, this study uses planted forests in the Saihanba region of Hebei Province, China, as the research area. By constructing a dataset of high-resolution UAV imagery, we propose a single-tree segmentation framework that cascades the YOLOX object detection algorithm with a multi-stage optimized labeled watershed algorithm: First, a lightweight YOLOX model is used to perform precise single-tree detection and bounding box localization. Subsequently, within the detected boxes, grayscale enhancement, median filtering, morphological optimization, and label-guided watershed segmentation are executed to achieve fine-grained extraction of individual tree crowns. This method balances detection speed, segmentation accuracy, and practical applicability, providing lightweight, high-precision technical support for efficient surveys of planted forests, single-tree parameter inversion, and ecological monitoring. Comparisons of different IoU and confidence thresholds show that the optimal overall segmentation results are achieved when the confidence threshold is 0.1 and IoU is 0.5.

KEYWORDS

UAV Remote Sensing; Single-Tree Segmentation; YOLOX; Watershed; Cascaded Framework

CITE THIS PAPER

Tianliang Zhang, Zerun Liu, Jia Dong, Tang Song, Huan Mei, Zhiqiang Wang. A Study on a Single-Tree Segmentation Method Based on UAV Imagery and a YOLOX Threshold Cascade. Automation and Machine Learning (2026). Vol. 7, No. 1, 135-142. DOI: http://dx.doi.org/10.23977/autml.2026.070117.

REFERENCES

[1] Li D, Zhang J J, &Zhao M X.(2019). Extraction of Stand Factors in UAV Image Based on FCM and Watershed Algorithm. SCIENTIA SILVAE SINICAE, 55(05),180-187.
[2] Gu Z X, &Pei F R.(2022). Single Tree Recognition Algorithm Based on Multi-Layer K-means in Forest Point Cloud. FOREST RESOURCES MANAGEMENT,(01),124-131.https://doi.org/10.13466/j.cnki.lyzygl.2022.01.015.
[3] Wang X S & Yang G.(2018). A multi-objective optimization segmentation method for tree image based on fusion clustering and classification algorithm. JOURNAL OF BEIJING FORESTRY UNIVERSITY,40(12),124-131.https://doi.org/10.13332/j.1000-1522.20180160.
[4] Huo L N & Zhang X L.(2021). Individual Tree Information Extraction and Accuracy Evaluation Based on Airborne LiDAR Point Cloud by Multilayer Clustering Method. SCIENTIA SILVAE SINICAE, 57(01),85-94.
[5] Sun G, Xin Y & Yan C.(2019). Estimating Single Tree Height Based on Improved K-means Algorithm. FOREST ENGINEERING,35(06),31-35.https://doi.org/10.16270/j.cnki.slgc.2019.06.022.
[6] Zeng J.(2025). Research on deep learning-based individual tree segmentation methods for Liriodendron plantations using UAV high-resolution imagery(Master's Thesis, Nanjing Forestry University). Master's degree. https://doi.org/10.27242/d.cnki.gnjlu.2025.000459.
[7] Wu K, Lai T, Ye J M, et al.(2026). Individual Tree Detection and Crown Width Estimation Using UAV Imagery based on SSD and Faster R-CNN Algorithms. REMOTE SENSING TECHNOLOGY AND APPLICATION,41(01),166-177.
[8] Wang K, Chen Y L & Xu H L.(2024). Detection of individual trees of bayberry using improved YOLOv7. Transactions of the Chinese Society of Agricultural Engineering, 40(11),159-167.
[9] Ge Z, Liu S, Wang F, et al. YOLOX: Exceeding YOLO Series in 2021[J]. arXiv preprint arXiv:2107.08430, 2021.
[10] Chen H, Quan Y, Ma X T, et al.(2026) Individual-tree identification of Korean pine based on UAV imagery and Mask R-CNN. Chinese Journal of Applied Ecology,1-13.https://doi.org/10.13287/j.1001-9332.202605.007

Subscription
E-Mail Alert
Most popular papers
Call for Special Issue
Publication Ethics & OA Statement
Plagiarism Policy
Contact Us
Downloads: 5209
Visits: 261691

Sponsors, Associates, and Links

All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.