View in EDS

Efficient lightweight fire detection in UAV imagery: an improved YOLOv8 approach.

Saved in:
Bibliographic Details
Title: Efficient lightweight fire detection in UAV imagery: an improved YOLOv8 approach.
Authors: Hu, Jie1,2,3 (AUTHOR) jiehu@swjtu.edu.cn, Zhang, Jiaming1 (AUTHOR) jiamingz@my.swjtu.edu.cn, Pang, Ting1 (AUTHOR) tingpang@my.swjtu.edu.cn, Peng, Bo1,2,3 (AUTHOR) bpeng@swjtu.edu.cn, Li, Tianrui1,2,3 (AUTHOR) trli@swjtu.edu.cn
Source: Visual Computer. Mar2026, Vol. 42 Issue 4, p1-16. 16p.
Abstract: In forest fire monitoring, unmanned aerial vehicles (UAVs) equipped with visual sensors have emerged as a pivotal tool for early detection. However, mainstream fire detection models face challenges in balancing accuracy and efficiency on resource-constrained UAV platforms, as lightweight designs often lead to reduced sensitivity to small fire targets. To address these issues, this paper introduces an efficient and lightweight forest fire detection method based on an improved YOLOv8 architecture, designed for real-time processing on UAV platforms. Our approach integrates a lightweight coordinate attention cross-stage partial network module for accurate feature extraction, a dual-convolution-based cross-scale feature fusion network for enhanced multi-scale detection, and the WIoUv3 loss function for improved localization accuracy. Experimental results on the self-constructed DIFD dataset demonstrate that the proposed method achieves an m A P 50 of 63.9%, representing a significant improvement of 1.1% over state-of-the-art techniques. Furthermore, with a model size of only 7 MB, it achieves an optimal balance between accuracy and parameter count, making it well-suited for future deployment on edge computing devices. The code address of this paper is: . [ABSTRACT FROM AUTHOR]
Copyright of Visual Computer is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:In forest fire monitoring, unmanned aerial vehicles (UAVs) equipped with visual sensors have emerged as a pivotal tool for early detection. However, mainstream fire detection models face challenges in balancing accuracy and efficiency on resource-constrained UAV platforms, as lightweight designs often lead to reduced sensitivity to small fire targets. To address these issues, this paper introduces an efficient and lightweight forest fire detection method based on an improved YOLOv8 architecture, designed for real-time processing on UAV platforms. Our approach integrates a lightweight coordinate attention cross-stage partial network module for accurate feature extraction, a dual-convolution-based cross-scale feature fusion network for enhanced multi-scale detection, and the WIoUv3 loss function for improved localization accuracy. Experimental results on the self-constructed DIFD dataset demonstrate that the proposed method achieves an m A P 50 of 63.9%, representing a significant improvement of 1.1% over state-of-the-art techniques. Furthermore, with a model size of only 7 MB, it achieves an optimal balance between accuracy and parameter count, making it well-suited for future deployment on edge computing devices. The code address of this paper is: . [ABSTRACT FROM AUTHOR]
ISSN:01782789
DOI:10.1007/s00371-026-04373-6