改进型YOLOv8算法在火灾探测中的应用  

Fire detection algorithm based on an improved YOLOv8 model

作　　者：邓力 周进 刘全义 DENG Li;ZHOU Jin;LIU Quanyi(College of Civil Aviation Safety Engineering,Civil Aviation Flight University of China,Guanghan 618307,Sichuan,China;Civil Aircraft Fire Science and Safety Engineering Key Laboratory of Sichuan Province,Guanghan 618307,Sichuan,China;Sichuan Key Technology Engineering Research Center for All-electric Navigable Aircraft,Guanghan 618307,Sichuan,China)

机构地区：[1]中国民用航空飞行学院民航安全工程学院,四川广汉618307 [2]民机火灾科学与安全工程四川省重点实验室,四川广汉618307 [3]四川全电动导航飞机关键技术工程研究中心,四川广汉618307

出　　处：《安全与环境学报》2025年第3期888-897,共10页Journal of Safety and Environment

基　　金：国家自然科学基金民航联合研究基金项目(U2033206);四川省重点实验室项目(MZ2022JB01);航空科学基金项目(ASFC-20200046117001)。

摘　　要：为了提高YOLOv8n算法在火灾探测方面的性能,给出了一种改进方法,通过集成上下文聚合架构Container和轻量级网络GhostNet来优化YOLOv8n网络结构。消融试验和对比试验的结果表明,所提方法能够有效改善YOLOv8n算法检测火灾的效果。该算法的平均精度达92.8%,探测速度达95.24帧/s,查准率达95%,具备更高的探测性能,可以为火灾探测器的研发提供参考。To significantly improve the performance of the YOLOv8n algorithm for fire detection,a carefully designed enhancement is proposed.The YOLOv8n network structure is optimized by integrating the context aggregation architecture,Container,along with the lightweight GhostNet network.The Container architecture excels at efficiently extracting image features of flames and smoke.It achieves this by effectively capturing and analyzing the context of fire elements,allowing for a more comprehensive understanding of fire scenes.To reduce the dispersion of contextual information and preserve more comprehensive feature details,YOLOv8n incorporates the Container architecture into the neck,forming a Container-PAFPN structure.This enhancement improves YOLOv8n's ability to focus on fire features,allowing it to capture rich fire information over long distances.Simultaneously,GhostNet plays a crucial role in enhancing the computational efficiency of YOLOv8n.Its lightweight design reduces computational complexity without sacrificing accuracy.A series of ablation and comparative experiments were conducted using both the fire dataset and the COCO128 dataset.The experimental results clearly demonstrate that the proposed method significantly enhances the effectiveness of the YOLOv8n algorithm for fire detection.Specifically,the mean Average Precision(mAP)and precision of the algorithm reached 92.8%and 95%,respectively,reflecting increases of 2.8 percentage point and 0.6 percentage point compared to the original YOLOv8n algorithm.Furthermore,the detection speed reaches an impressive 95.24 frames per second,indicating a significant improvement in performance.When compared to other advanced fire detection algorithms,the experimental results demonstrate that this algorithm achieves both higher mean Average Precision(mAP)and faster detection speeds.Overall,this algorithm provides a valuable reference for the development of fire detection systems and can significantly enhance fire detection capabilities,offering a reliable and efficient soluti

关 键 词：安全工程 改进YOLOv8算法 深度学习 火灾探测 

分 类 号：X92[环境科学与工程—安全科学]