基于目标区域的卷积神经网络火灾烟雾识别  

Convolutional Neural Network Fire Smoke Detection Based on Target Region

作　　者：冯路佳 王慧琴[1] 王可[1] 卢英[1] 王钾 Feng Lujia;Wang Huiqin;Wang Ke;Lu Ying;Wang Jia(School of Information and Control Engineering,Xi'an University of Architecture and Technology,Xi'an,Shaanxi 710055,China)

机构地区：[1]西安建筑科技大学信息与控制工程学院,陕西西安710055

出　　处：《激光与光电子学进展》2020年第16期75-83,共9页Laser & Optoelectronics Progress

基　　金：陕西省文物局项目(Z20180301);陕西省教育厅重点科学研究计划(智库项目)(18JT006);西安市科技局项目(GXYD10.1)。

摘　　要：在场景复杂、干扰较多的情况下,传统的火灾烟雾识别方法的识别性能不高。针对该问题,提出了一种基于目标区域的卷积神经网络火灾烟雾识别方法,构建两层的火灾烟雾识别模型,利用目标区域定位层的运动检测算法对火灾烟雾图像进行烟雾目标区域的提取,快速去除复杂场景的大量无关干扰信息,并将提取的烟雾目标区域输入火灾烟雾识别层,通过卷积神经网络精细提取烟雾的深层特征后进行分类,完成火灾烟雾的识别。实验结果表明,所提方法在复杂环境下的数据集中,抗干扰能力较强,能够有效降低误检率,提高烟雾识别的准确率。The traditional fire smoke detection method has a degraded detection performance in the case of complex scenes and high interference.Aiming at this problem,this paper proposes a convolutional neural network fire smoke detection method based on target region.A two-layer fire smoke detection model is constructed.Using the motion detection algorithm of the target region positioning layer,the smoke target region is extracted from the fire smoke image,which can quickly remove a large amount of irrelevant interference information in complex scenes,and input the extracted smoke target region into the fire smoke recognition layer,and then extract the deep features of the smoke through the convolutional neural network to classify it to complete the fire smoke detection.Experimental results show that the proposed method has strong anti-interference performance in the data set under complex scenes,which effectively reduces the false detection rate and improves the accuracy of smoke detection.

关 键 词：图像处理 火灾烟雾识别 目标区域 卷积神经网络 运动检测 抗干扰能力 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]