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 Automatic Multi-Object Detection in Forest Areas Based on UAV Imagery

Download as PDF

DOI: 10.23977/autml.2026.070116 | Downloads: 4 | Views: 61

Author(s)

Yi Ren 1, Yongxin Yan 1, Yanhua Zhang 1, Suli Li 1, Dong Jing 1, Tianliang Zhang 2,3

Affiliation(s)

1 Weichang Manchu and Mongol Autonomous County State-Owned Luanhe Forest Farm, Weichang County, Chengde, China
2 Institute of Applied Mathematics, Hebei Academy of Sciences, No. 46 South Youyi Street, Shijiazhuang, China
3 Hebei Information Security Certification Technology Innovation Center, No. 46 South Youyi Street, Shijiazhuang, China

Corresponding Author

Tianliang Zhang

ABSTRACT

This study focuses on the Saihanba Mechanical Forest Farm in Hebei Province, China, and develops an automated multi-object detection system for forest areas—including graves, cow, and vehicles—based on visible-light imagery from consumer-grade drones and the YOLOv5 deep learning model. The system addresses core operational needs such as fire control during grave-side rituals, ecological monitoring of free-range cow herds, and safety patrols of vehicles in forest areas. A DJI Mavic 3 drone was used to acquire visible-light imagery of the forest area at various flight altitudes. A specialized multi-object dataset for forest areas was constructed, comprising 470 valid images and 1,467 annotated samples, covering the three target categories—graves, cow, and vehicles—across different scenarios. Data augmentation was performed through geometric transformations and mixed augmentation to enhance the model's generalization ability; To address the issues of low detection accuracy for small targets and feature loss in complex forest backgrounds, the YOLOv5 model was enhanced in multiple dimensions. This included adopting the GhostConv lightweight backbone network, embedding a Coordinate Attention (CA) mechanism, replacing the BiFPN with a weighted bidirectional feature fusion structure, and using the EIoU loss function to optimize bounding box regression. Model training was completed via transfer learning using COCO pre-training weights.Experimental results show that the improved YOLOv5 model achieves an average accuracy of 97.25% on the test set, capable of accurately identifying three types of targets in complex forest backgrounds, with excellent resistance to occlusion and background interference. This study validates the feasibility and superiority of combining consumer-grade drone visible-light imagery with deep learning models for automated multi-target monitoring in forest areas. It can provide efficient, low-cost technical support for forest fire early warning, forestry ecological management, and forest safety supervision, and holds significant application potential.

KEYWORDS

UAV remote sensing; YOLOv5; multi-object detection in forest areas; smart forestry

CITE THIS PAPER

Yi Ren, Yongxin Yan, Yanhua Zhang, Suli Li, Dong Jing, Tianliang Zhang. A Study on Automatic Multi-Object Detection in Forest Areas Based on UAV Imagery. Automation and Machine Learning (2026). Vol. 7, No. 1, 126-134. DOI: http://dx.doi.org/10.23977/autml.2026.070116.

REFERENCES

[1] Lan Y B, Deng X L, Zeng G L(2019). Advances in diagnosis of crop diseases, pests and weeds by UAV remote sensing. Smart Agriculture,1(02),1-19.
[2] Fan J X, Zhang W H, Zhagn L L, et al.(2023).Vehicle detection method of UAV imagery based on improved YOLOv5.Remote Sensing Information,38(03),114-121.
[3] WU Ｒ, LI J, Wang Z,et al.(2025). Vehicle Target Detection System from the Perspective of UAV Based on Improved YOLOv5. Urban Geotechnical Investigation & Surveying,(03),6-11.
[4] Guo X J, Shao Q Q.(2023).Population of kiangs and spatiotemporal variation of its habitat in Sanjiangyuan National Park baded on unmanned aerial vehicle remote sensing.ACTA ECOLOGICA SINICA,43(19),7886-7895.
[5] Wang Y C, Ma J R, Wang J J, et al(2025). Research on lightweight models for detection of large herbivores in the Yellow River Source Region based on UAV Images. PRATACULTURAL SCIENCE,42(06),1538-1551.
[6] Zhang L W, Zhou H, Zhu Q B.(2023). Multi-target tracking of group-housed pigs based on PigsTrack tracker. Transactions of the Chinese Society of Agricultural Engineering,39(16),181-190.
[7] Zhao Y X, Zhang G Q, Li D H, et al.(2025).A high-precision tracking and localization method for monitoring cows. Information and Control,54(01),137-149+160.
[8] Zang K.(2024). Lightweight UAV Remote Sensing Image Vehicle Detection Method Based on Improved YOLOv5s. GEOMATICS & SPATIAL INFORMATION TECHNOLOGY,47(09),86-89.
[9] Xian Y T, Li B, Wan T.(2025). UAV photography detection algorithm based on improved YOLOv5. Computer Measurement & Control,33(04),48-56+66. 

Subscription
E-Mail Alert
Most popular papers
Call for Special Issue
Publication Ethics & OA Statement
Plagiarism Policy
Contact Us
Downloads: 5198
Visits: 261009

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.