支持向量机在火灾早期探测中的应用  

Research of early fire detection based on support vector machine

作　　者：谢理[1] 张认成[1] 蔡钟山[1] 

机构地区：[1]华侨大学机电及自动化学院,福建厦门361021

出　　处：《广西大学学报（自然科学版）》2011年第2期246-250,共5页Journal of Guangxi University（Natural Science Edition）

基　　金：福建省自然科学基金资助项目(2009J01290);国务院侨办科研基金资助项目(09QZR04);厦门市科技计划项目(3502Z20103028)

摘　　要：针对传统火灾探测算法及时性与准确性相互制约的问题,提出以CO与CO2体积分数的比值大小、变化速度以及加速度作为火灾过程特征信息向量,采用支持向量机算法对火灾源进行模式识别。根据国家标准火灾实验要求,选取了4种真实火灾源材料和3种虚假火灾源材料进行火灾实验,实验表明利用支持向量机可以对真假火灾源进行有效识别,其识别率达到98%。与传统火灾算法相比,支持向量机在保证较高准确性的同时,火灾识别的及时性也得到较大改善。Based on the mutual restraint between timeliness and accuracy of the traditional fire detection algorithm,SVM（support vector machine） was used to distinguish real fire source from non-fire nuisance sources.The early fire process eigenvector consists of the concentration ratio of CO to CO2 as well as its changing velocity and acceleration in detected environment.In addition,6 materials of true fire and 3 materials of fake fire were tested according to the national standard requirements on fire experiment.The result of experiments shows that SVM is feasible for early fire detecting with false positives and false negatives,and its recognition rate is 98%.In comparison with traditional fire detection algorithm,SVM improves the accuracy as well as timeliness of fire detection.

关 键 词：火灾早期探测 支持向量机 过程特征 身份识别 

分 类 号：X924.2[环境科学与工程—安全科学] TN911.23[电子电信—通信与信息系统]