基于卷积神经网络室内火焰烟雾识别

Indoor Flame Smoke Identification Based on Convolutional Neural Network

摘要: 人们在生活中面临的最大问题是室内火灾问题。近年来，基于图像模式识别的火灾探测技术引起了众多研究者的关注，并得到了迅速的发展。技术以计算机技术为核心，结合图像处理技术和模式识别技术，可以对火灾火焰和烟雾特征实现有效的检测和识别，在人流聚集的市场、企业事业单位的办公区域和生产车间区域火灾报警可以发挥出重要的作用。本文分析了烟雾和火焰模式识别的难点，使用运动学检测方法、特征提取、卷积神经网络等方法解决了传统方法对室内烟雾和火焰识别不准确的缺点。结果表明，该识别方案的准确率可以达到93%。

Abstract: The biggest problem people face in life is the problem of indoor fire. In recent years, fire detection technology based on image pattern recognition has attracted the attention of many researchers and developed rapidly. With computer technology as the core, combined with image processing technology and pattern recognition technology, it can effectively detect and identify fire flame and smoke characteristics, gather people in the market, and play an important role in fire alarm in office areas and production workshops of enterprises and institutions. This paper analyzes the difficulties of smoke and flame pattern recognition, and uses kinematics detection method, feature ex-traction, convolutional neural network and other methods to solve the shortcomings of the traditional method of indoor smoke and flame recognition which is not accurate. The result shows that the accuracy of the scheme can reach 93%.