摘要
早期火灾烟雾不仅具有动态特征,同时具有静态特征,为提高早期火灾识别准确率,提出一种典型相关分析和BP神经网络相结合的早期火灾识别算法(CCA-BPNN)。首先提取烟雾主方向性状、烟雾面积增长、颜色和亮度等特征,然后采用典型相关分析算法对这些特征进行融合,最后将特征输入到BP神经网络进行训练,建立早期火灾识别模型。仿真结果表明,相对于对比算法,CCA-BPNN算法提高了早期火灾识别准确率,降低了虚警率和漏报率,能够满足早期火灾识别的准确性和实时性要求。
The early fire smoke not only has the dynamic features and static features,this paper puts forward a early fire recognition algorithm based on canonical correlation analysis and BP neural network combined to o improve the early fire recognition accuracy firstly,the main direction extraction smoke,smoke area growth traits,color and brightness features are extracted,and then the canonical correlation analysis algorithm is used to fuse these features,finally the features are input into BP neural network to train and establish the early fire recognition model.The simulation results show that,compared with to the comparison algorithm,CCA-BPNN algorithm has improved the early fire recognition accuracy,reduced the false alarm rate and the rate of missing report,it can satisfy the early fire recognition accuracy and real-time requirements.
出处
《科技通报》
北大核心
2013年第5期126-129,共4页
Bulletin of Science and Technology
基金
国家自然科学基金
课题号11047026
关键词
早期火灾识别
典型相关分析
神经网络
特征融合
early fire recognition
canonical correlation analysis
neural network
feature fusion
