多特征融合视频火灾识别研究  

Research on Video Fire Detection Based on Multi-feature Fusion

作　　者：金肖 叶锦华[1] 杨素珍 JIN Xiao;YE Jinhua;YANG Suzhen(College of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350116, China;Mechanical Engineering, Zhangzhou Institute of Technology, Zhangzhou 363000, China)

机构地区：[1]福州大学机械工程及自动化学院,福建福州350116 [2]漳州职业技术学院机械工程学院,福建漳州363000

出　　处：《机械制造与自动化》2019年第4期163-167,共5页Machine Building & Automation

基　　金：福建省自然科学基金(2015J05121);福建省中青年教师教育科研项目(JAT160877)

摘　　要：针对目前的视频火焰检测算法容易受到复杂场景、光照条件的影响,算法的实时性较差且容易误判等问题,提出一种红外视频图像自适应背景更新疑似火焰区域检测与改进层次分析法多特征融合火灾火焰识别方法。利用改进型自适应背景更新运动目标检测来获取红外视频图像中疑似火焰区域,消除背景干扰,提高火焰静动态特征的提取效率,将火焰的多种特征用三标度改进层次分析法进行融合,实验结果表明该方法对火焰的识别效果更加快速、精确、有效。The current video flame detection algorithm is easily affected by complex scenes, lighting conditions, so its real-time reliability is poor, Its misjudgement and other issues commonly occur in the operation. This article proposes a method of adaptive background update based on infrared video images which is used to detect suspected flame areas and introduces the flame detection method of multi-feature fusion fire detection based on the analysis method for improving hierarchical process. The improved adaptive background updating model is used to obtain the target of the suspected flame in the infrared video image, then, the background interference is eliminated and the extraction efficiency of flame static and dynamic characteristics is improved. Then, the analysis method for the improving hierarchical process based on three-scale is used to fuse the multiple characteristics of the flame. Experimental results show that this method is use to more rapidly accurately and effectively identify the flame.

关 键 词：火焰检测 红外图像 自适应背景更新 改进层次分析法 特征融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]