摘要
为减少火灾探测中的误报警,基于信息融合技术对火灾传感器输出的信息进行处理。充分利用火灾探测系统的在线和离线数据,采用改进的主元分析法(PCA)、粗糙集(RS)理论、支持向量机(SVM)等3种方法的融合与互补,通过对系统的输入数据进行简化,消除原有信息的各分量之间的相关性,降低特征信息维数;实施最优最小约简,特征提取优化;构造自适应核函数,确定最优分类超平面,进行样本训练,获得火灾探测结果。从数据级、特征级、决策级3个层次上实现火灾信息融合。结果表明:该方法减少了融合过程中的信息损失,降低了计算的复杂性,有效地提高了火灾探测系统的可靠性和准确度。
Fire signal processing is the key technology of fire detection.To completely utilize the online and off-line data of fire detection system,a new method combining and fusing PCA,RS and SVM is proposed based on information fusion technology.This method can reduce the dimension of characteristic data by simplifying the input data and eliminating the correlations of original data.Meanwhile,this method can also fulfill the least and optimal reduction and the optimization of feature extraction,and obtain the results of fire detection by constructing self-adaptive kernel function,defining optimal classification hyper-plane and training samples.Thus,the fire information fusion is eventually achieved from data level,the characteristic level and the decision-making level.The results show that this method reduces the computational complexity and the information loss in the fusion process,and enhances the reliability and accuracy of the fire detection system effectively.
出处
《中国安全科学学报》
CAS
CSCD
北大核心
2011年第6期94-98,共5页
China Safety Science Journal
基金
山东省自然科学基金资助(ZR2010FL021)
