基于视频的早期森林火灾识别算法研究  

Research on early forest fire recognition algorithm based on video

作　　者：林宏[1] 赵同林[1] 张雁[1] 禹玥昀[1] 

机构地区：[1]西南林业大学计算机与信息学院,云南昆明650224

出　　处：《消防科学与技术》2013年第4期418-421,共4页Fire Science and Technology

基　　金：西南林业大学科学基金项目(2010MS07)

摘　　要：针对传统火灾探测系统不适于森林火灾的室外、远距离、大面积特性的现实,提出一种基于林火颜色特征和形态特征的早期火灾识别算法。颜色特征采用基于RGB颜色空间和HSV颜色空间的火焰颜色模型和烟雾颜色模型识别。通过前景提取分离出图像中发生变化的区域。从不规则性判断和生长性判断等方面进行形态特征分析。实验结果表明,此算法可以有效地降低误报、漏报率,提高检测率。Traditional fire detection systems are not suitable for outdoor forest fires with long-distance and large-area character- istics. This paper presents an early fire recognition algorithm based on forest fire color model and shape characteristics. Color characteristics detection use flame color recognition and smoke color recognition based on the RGB color model and HSV color model. Area changing is separated by foreground extracting. Shape characteristics detection method contains the irregular and growing characteristics. Experiment result showed that, the algorithm can achieve early warning of a fire and can effectively reduce the false and negative rate of fire monitoring system.

关 键 词：森林火灾 火灾预警 火灾识别算法 

分 类 号：X924.4[环境科学与工程—安全科学] TP391.4[自动化与计算机技术—计算机应用技术]