基于深度学习的早期火灾检测方法  

Research on Early Fire Monitoring Based on Deep Learning

作　　者：任国凤 张帆 李少琼 梁小瑞 REN Guofeng;ZHANG Fan;LI Shaoqiong;LIANG Xiaorui(Department of Electronics,Xinzhou Normal University,Xinzhou 034000,China;Xinzhou Fire Rescue Detachment,Xinzhou 034000,China)

机构地区：[1]忻州师范学院电子系,山西忻州034000 [2]忻州市消防救援支队,山西忻州034000

出　　处：《洛阳理工学院学报（自然科学版）》2024年第2期65-70,共6页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

基　　金：山西省虚拟仿真实验教学项目(X2020048);山西省教学改革研究项目(J2021572);山西省大学生创新实践项目(20220912)。

摘　　要：为了减少火灾的危害,对火灾隐患进行早期检测是十分必要的。利用深度学习算法构建数学模型,设计一种基于Python平台的火灾图片特征提取方法,构建CNN网络对早期火灾图片进行识别,通过Python平台提取火焰特征,选择具有良好辨识性能的特征指数用于ResNet50网络模型的训练,建立早期火灾检测系统并进行实验仿真分析。In order to reduce the fire harm,it is necessary to quickly monitor the fire risk.The mathematical model is studied based on deep-learning algorithms,designing a fire feature detection based on Python platform.CNN network is built to identify the early fire picture,extract flame features through Python,and select the feature index with good identification performance for the training of ResNet50 network model.An early fire monitor system is established,conducting experimental simulation analysis.

关 键 词：火灾检测 CNN ResNet50网络 PYTHON 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]