基于BP神经网络的仓储火灾火焰识别研究  

The Research of Warehouse Fire Flame Recognition Based on BP Neural Network

作　　者：郭键[1] 张涛[1] 

机构地区：[1]北京物资学院信息学院,北京101149

出　　处：《数学的实践与认识》2015年第7期190-198,共9页Mathematics in Practice and Theory

基　　金：北京市教育委员会科技计划面上项目(KM201310037001);北京市属高等学校青年拔尖人才培育计划;北京物资学院青年运河学者;北京物资学院校级重大科研项目

摘　　要：仓储环境的特殊性限制了传统火灾探测设备的探测效果.为利用现有监控设备,可将图像火灾探测方法引入仓储领域.首先将采集的图像进行预处理,得出感兴趣的目标前景.然后对前景进行特征提取.最后采用基于BP神经网络的识别方法,以提取的图像特征作为输入,对网络进行训练与仿真.实验结果表明:BP神经网络对于火灾火焰图像具有很好的识别能力;作为其输入的图像特征对于火灾火焰图像有着较好的判别效果;为减少硬件投入,采用图像火灾探测方法弥补传统火灾探测设备在仓储应用中的不足是可行的.The particularity of warehouse environment limits the detection effect of the traditional fire detection equipment. Image fire detection method can be introduced into warehouse areas in order to make use of the existing monitoring equipment. The collected images would be preprocessed to get the interest target prospects firstly. Then extract features of the target prospects. At last, the recognition method based on BP neural network would be used, and BP neural network would be trained and simulated with the extracted image features as its inputs. The following is the experimental results： the BP neural network has the very good recognition ability for the fire flame image; image features as its inputs has a better discriminant result for fire flame image; using the image fire detection method to reduce the hardware investment and to make up the deficiency of traditional fire detection equipment in the warehouse application is feasible.

关 键 词：BP神经网络 仓储 火焰识别 图像特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]