摘要
提出基于过程特征信息的火灾早期探测数据融合方法.利用加窗函数的外推算法和最小二乘法,提取CO浓度上升速度与加速度、CO2的浓度上升速度作为火灾辨识的特征层信息,建立一个3输入,3输出的概率神经网络的系统决策层.选取58组具有代表性的火灾状况特征信息作为数据样本,对网络进行训练.仿真结果表明,经过50次训练,最终误差为10-3,方法能够快速、准确地对火灾的阴燃初期进行识别,正确判别明火,且对于干扰信号能够有效滤除.
In this paper, a method of data fusion based on procedure characteristics technology for early fire detection is issued. The rising velocity of CO and CO2 and the rising acceleration of CO extracted by extrapolation method and least square method are regarded as information of feature level fusion. 3-input and 3-output decision fusion of probabilistic neural network is built. 58 main typical groups of fire data-samples are used tO train the network. The simulation result shows that the convergence error is small than 10^-3 after 50 times training. This method can predict ignition and fire with high credibility and filter disturb signal effectively.
出处
《华侨大学学报（自然科学版）》
CAS
北大核心
2008年第1期11-13,共3页
Journal of Huaqiao University(Natural Science)
基金
福建省自然科学基金资助项目(D0610015)
华侨大学科研基金资助项目(06HZR09)
关键词
红外光谱探测
数据融合技术
过程特征
概率神经网络
infrared spectrum detection
data fusion technique
procedure characteristics
probabilistic neural network
