四川轻化工大学学报（自然科学版）2025，Vol.38Issue(3)：75-84,10.DOI:10.11863/j.suse.2025.03.09
基于改进YOLOv8n的红外影像森林火灾检测设计
Forest Fire Detection Based on Improved YOLOv8n with Infrared Spectrum
摘要
Abstract
Challenges in forest fire image detection and recognition,such as spectral differences,smoke occlusion,and insufficient detection accuracy,are addressed by proposing an improved YOLOv8n algorithm designed for infrared spectral forest fire images.The backbone of the YOLOv8n network is optimized through the incorporation of the Large Scale Knowledgeable Network(LSKnet)attention mechanism and the adoption of the GIOU loss function to enhance detection accuracy.The model's ability to capture target information and perceive thermal signals in infrared images are improved,as demonstrated by experimental results.The YOLOv8n model integrated with the LSKnet structure(hereinafter referred to as YOLOv8n+LSKnet)achieves accuracy rates of 93.1%and 64.7%on the mAP@0.50 and mAP@0.50-0.95 metrics,respectively,representing improvements of 1.0 and 1.9 percentage points over the original YOLOv8n model.These results indicate that the improved YOLOv8n-based algorithm is more effective in detecting and recognizing infrared spectral forest fire images.关键词
森林火灾检测/YOLOv8n/LSKnet/红外图像Key words
forest fire detection/YOLOv8n/LSKnet/infrared imagery分类
信息技术与安全科学引用本文复制引用
王志强,谭飞,杜正淼,鲁丹,张宗良..基于改进YOLOv8n的红外影像森林火灾检测设计[J].四川轻化工大学学报（自然科学版）,2025,38(3):75-84,10.基金项目
国家自然科学基金项目(61902268) （61902268）
四川省科技厅项目(21ZDYF4052 （）
2020YFH0124 （）
2021YFSY0060) （）
四川轻化工大学创新计划项目(cx2023193 （）
cx2023195 （）
cx2023198) （）
