Author(s)

FuLin Li, WenFa Xu*, Zhen Min, XuePeng Wu, ChangYu Xiang, YuDong Liu

Affiliation(s)

School of Information Science and Engineering, Wuchang Shouyi University, Wuhan 430064, Hubei, China.

Corresponding Author

WenFa Xu

ABSTRACT

Forest fires, as a global ecological disaster, pose a serious threat to the stability of environmental systems, biodiversity security, and socio-economic development. In response to the shortcomings of traditional fire point detection methods in small target recognition accuracy, model robustness, and real-time performance, this paper proposes an intelligent forest fire point detection model based on an improved YOLOv8n framework. This model integrates the Fast Super Resolution Convolutional Neural Network feature enhancement module, Enhanced Squeeze Excitation attention mechanism, and an improved Minimum Point Distance Intersection over Union bounding box regression algorithm, aiming to improve the detection ability of early fire points and overall system performance. Through systematic experiments on a self built multi scene forest fire image dataset, the results showed that compared with Faster R-CNN, YOLOv5s, and the standard YOLOv8n model, the proposed method performed well in comprehensive detection performance, with mAP reaching 84.7%, Precision reaching 84.2%, Recall reaching 80.4%, and also possessing high real-time processing capabilities. This study not only provides effective technical support for intelligent monitoring of forest fires, but also proposes a multi module collaborative optimization framework, which provides new theoretical references and practical paths for research and application in the field of small target detection.

KEYWORDS

Forest fire point recognition; Super-resolution; YOLOv8n; Object detection; Deep learning

CITE THIS PAPER

FuLin Li, WenFa Xu, Zhen Min, XuePeng Wu, ChangYu Xiang, YuDong Liu. Forest fire point recognition based on super-resolution techniques. Journal of Computer Science and Electrical Engineering. 2025, 7(4): 29-36. DOI: https://doi.org/10.61784/jcsee3065.

REFERENCES

[1] Pan YD, Birdsey R, Fang JY, et al. A large and persistent carbon sink in the world’s forests. Science, 2011, 333(6045): 988–993.

[2] Li J, Hu F, Du L. Forest Fire Detection Based on Image Processing Technology: A Review. Fire Technology, 2020, 56(5): 1945-1972.

[3] Yuan F, Wan W, Wei H, et al. Review of remote sensing methods for forest fire detection. Remote Sensing, 2018, 10(5): 763.

[4] Koltunov A, Ustin S L. Early Fire Detection Using Non-Linear Multispectral Analysis: International Journal of Wildland Fire, 2007, 16(1): 1-10.

[5] Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks: IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.

[6] Yuan Z, Zhang Y, Liu Y, et al. A Real-Time Forest Fire Detection Algorithm Based on UAV Video: Remote Sensing, 2021, 13(12): 2305.

[7] Dong C, Loy C C, He K, et al. Accelerating the Super-Resolution Convolutional Neural Network: IEEE Transactions on Image Processing, 2016, 25(12): 5534-5545.

[8] Gao Z, He W, Luo H, et al. ESNet: Efficient Squeeze-and-Excitation Network for Channel Attention: IEEE Transactions on Neural Networks and Learning Systems, 2022, 33(6): 2567-2579.

[9] Li Y, Huang T, Liu Y, et al. Bounding box regression method based on improved dynamic IoU. Journal of Software, 2022, 33(11): 4667–4680.

[10] Dong C, Loy C C, He K, et al. Accelerating the Super-Resolution Convolutional Neural Network: IEEE Transactions on Image Processing, 2016, 25(12): 5534-5545.