软件导刊2025，Vol.24Issue(8)：27-37,11.DOI:10.11907/rjdk.241581
基于模拟数据的古建筑火灾检测改进算法研究
Research on Improved Fire Detection Algorithms for Ancient Buildings Based on Simulated Data
摘要
Abstract
In the research on fire prevention algorithms for ancient buildings,the scarcity of datasets has become a major challenge.To ad-dress this issue,this study proposes a novel method for fire detection and prevention in ancient buildings.First,a virtual platform was devel-oped to generate simulated fire data,which can model the spread of flames and smoke in ancient structures.On the algorithmic level,this study introduces an improved algorithm model,YOLOv5s-STI,which incorporates the Involution convolution computation method,allowing the model to more flexibly adapt to the characteristics of the input data.The backbone of the model integrates the Swin Transformer,which ef-fectively captures the spatial distribution of flames and smoke,enhancing the model's ability to recognize fire dynamics.The C3 module is en-hanced with the SimAM self-attention mechanism,which adaptively adjusts the spatial correlation of features,improving the model's accura-cy in identifying fire regions within images.Evaluation experiments on the simulated dataset show that the improved model achieved a 1.2 per-centage increase in mean Average Precision(mAP@0.5)and demonstrated strong performance in detecting real-world flames and smoke.These results validate the effectiveness of this method in the fire prevention of ancient buildings.关键词
火灾检测/YOLOv5s/虚拟仿真/注意力机制/卷积算子/TransformerKey words
fire detection/YOLOv5s/virtual simulation/attention mechanism/convolution operator/Transformer分类
信息技术与安全科学引用本文复制引用
江金松,史东辉,王鹏林,周思恒,王飞龙..基于模拟数据的古建筑火灾检测改进算法研究[J].软件导刊,2025,24(8):27-37,11.基金项目
安徽省高校人文社会科学研究重点项目(2022AH050224) （2022AH050224）
智能建筑与建筑节能安徽省重点实验室主任基金项目(IBES2024ZR01) （IBES2024ZR01）
中国建设教育协会教育教学科研课题重点项目(2023003) （2023003）
安徽省研究生教育质量工程项目(2023cx-cysj129,2024zyxwjxalk126) （2023cx-cysj129,2024zyxwjxalk126）
