摘要
火焰检测是识别复杂环境下火灾的重要方法,为提高火灾识别率,提出了一种基于RGB颜色空间统计模型的火焰识别算法,结合区域生长以及帧差法分割出疑似火焰区域后,侧重提取分析视频火焰的动态特征及分层特征,尤其是火焰闪烁特征,然后利用BP神经网络融合5个特征参量进行火灾的判决.实验结果表明:该方法在复杂场景下具有较好的鲁棒性,可有效识别火灾火焰,降低误报率.
Flame detection is an important method to recognize fire under complex circumstances. In order to improve the accuracy of fire detection,an algorithm of flame recognition using statistical model based on the RGB space is presented in this paper. The algorithm segments a suspected fire area in images that may contain flames and extracts a few dynamic and hierarchical features associated with the area,especially the flicker frequency of flames. Finally,five features of an area are processed and fused by a BP neural network for a decision. Experiments show that this algorithm is robust and efficient,and is significant for reducing false alarms.
出处
《江苏科技大学学报（自然科学版）》
CAS
北大核心
2017年第2期178-184,共7页
Journal of Jiangsu University of Science and Technology:Natural Science Edition
基金
江苏科技大学博士科研启动基金资助项目(635301202)
江苏省前瞻性联合研究项目(by2012176)
关键词
火灾探测
统计颜色模型
动态特征
闪烁频率
特征融合
fire detection
statistical color model
dynamic features
flicker frequency
feature fusion
