基于PNN算法的多传感器火灾探测技术研究  

Study on the multi-sensor fire detection technology based on probabilistic neural network algorithm

作　　者：张志华[1,2] 许开立[1] 李增[2] 

机构地区：[1]东北大学资源与土木工程学院,辽宁沈阳110819 [2]中国人民武装警察部队学院,河北廊坊065000

出　　处：《消防科学与技术》2017年第10期1404-1406,共3页Fire Science and Technology

基　　金：河北省科技厅项目(16210339)

摘　　要：从传统算法和人工神经网络算法两方面,总结火灾探测中的多特征信息融合算法。以火焰、烟雾、CO和CO_2四种火灾特征的组合为例,基于MATLAB,运用PNN对实验采集到的数据进行训练和仿真测试。测试结果表明,采用PNN将多传感器信息融合后对火灾探测的准确度远高于单一种类火灾探测器;当扩展系数取0.3时,PNN对测试数据进行模式识别的准确度可高达98.95%。训练后的PNN可以更好地用于火灾的探测。In points of traditional algorithm and artificial neural net- work algorithm, the multi feature information fusion algorithms for fire detection were summarized. Taking the combination of flame, smoke, carbon monoxide and carbon dioxide as an example, and based on MATLAB, the data collected in the experiment were trained and tested using PNN. The testing results showed that fire detection technology combining the information from several sen sorsby PNN has higher accuracy than single type fire detector; when the expansion coefficient is 0.3, the accurate rate of the data pattern recognition using PNN can be as high as 98.95%. Therefore, the trained PNN network can be better used for fire detection.

关 键 词：消防 火灾探测 PNN 多传感器 

分 类 号：X924.4[环境科学与工程—安全科学]