基于可见光视频的森林火灾识别算法  

Forest fire identification algorithm based on a visible light video

作　　者：徐爱俊[1] 方陆明[1] 楼雄伟[1] 

机构地区：[1]浙江林学院信息工程学院

出　　处：《北京林业大学学报》2010年第2期14-20,共7页Journal of Beijing Forestry University

基　　金："863"国家高技术研究发展计划项目(2006AA12Z109);浙江省科技计划项目(2006C12109;2008C23036)

摘　　要：以森林火灾远程视频预警监控工程为依托,对森林火灾发生、发展的可见光视频图象进行研究,提出森林火灾识别算法,并进行处理。分析了基于视频的森林火灾火焰特征,指出火焰特征主要包括颜色变化、面积变化、边缘变化、形体变化、闪动规律和整体移动等。在此基础上,提出了视频图像中的森林火灾区域检测方法与森林火灾识别方法;根据图像区域分割匹配算法,以火焰颜色特征和面积变化为火灾判别依据,统计疑似火灾区域面积,定时地对其进行两两匹配,实现对森林火灾图像的实时检测和识别。经实际验证,该算法的查全率与查确率分别达到72.22%和92.86%。Based on the remote video surveillance system for forest fire,we studied the visible light video images during the forest fire and put forward an algorithm for forest fire identification.Flame features of forest fire from video images were analyzed,which include the change of color,area,edge,form, flickering law,and overall movement.On this basis,the paper proposed the methods of fire area checking and forest fire identification from video images.According to the algorithm of image region segmentation and matching, it adopts the fire color feature and flame area as criteria for fire identification,determines the coverage of potential forest fire,and matches them mutually at a regular time.The algorithm can realize real-time identification of forest fire,and testing results showed high recall ratio and precision ratio of the algorithm,which was 72.22% and 92.86% ,respectively.

关 键 词：可见光视频 森林火灾 颜色特征 区域分割 火灾识别 

分 类 号：S762.2[农业科学—森林保护学]