中国安全科学学报 ›› 2020, Vol. 30 ›› Issue (4): 8-13.doi: 10.16265/j.cnki.issn1003-3033.2020.04.002

• 安全工程技术 • 上一篇    下一篇

可燃性粉尘云的图像检测方法

赵欣然1,2 高级工程师, 张琪3, 王卫东1 教授, 徐志强1 教授   

  1. 1.中国矿业大学(北京) 化学与环境工程学院,北京 100083;
    2.中国安全生产科学研究院 重大危险源监控中心,北京 100012;
    3.东北电力大学 计算机学院,吉林 长春 132000
  • 收稿日期:2020-01-04 修回日期:2020-03-21 出版日期:2020-04-28
  • 作者简介:赵欣然(1971—),男,山东蓬莱人,博士,高级工程师,主要从事安全监控与预警、人工智能、深度学习等方面的工作。E-mail:zhaoxr02@126.com。
  • 基金资助:
    国家重点研发计划项目(2017YFC0805900);十二五科技支撑计划项目(2015 BAK16B-JY003);中国安全生产科学研究院基本科研业务费专项资金资助(2016JBKY13);中国安全生产科学研究院科技创新研发基金资助(CXYF201902)。

Image detection method of combustible dust cloud

ZHAO Xinran1.2, ZHANG Qi3, WANG Weidong1, XU Zhiqing1   

  1. 1. School of Chemical & Environmental Engineering, China University of Mining and Technology(Beijing), Beijing 100083, China;
    2. Major Hazard Source Monitoring Center, China Academy of Safety Science and Technology, Beijing 100012, China;
    3. School of Computer Science, Northeast Electric Power University, Changchun Jilin 132000, China
  • Received:2020-01-04 Revised:2020-03-21 Published:2020-04-28

摘要: 近年来粉尘爆炸引起的安全生产事故频繁发生,在线检测粉尘易集聚场所的粉尘云浓度并进行预警,成为控制粉尘爆炸的关键手段,而目前粉尘浓度传感器在大空间粉尘云聚集场所存在安装与识别局限性。为此,提出基于深度学习的可燃性粉尘云图像检测方法;采用基于卷积神经网络(CNN)的Faster R-CNN模型,对可燃性粉尘云进行端到端的检测与识别;并通过建立的粉尘云标准浓度图像数据库验证模型的有效性。结果表明:Faster R-CNN模型具有较高的识别精度。

关键词: 可燃性粉尘云, 图像检测, 卷积神经网络(CNN), 深度学习, Faster R-CNN模型

Abstract: In recent years, production accidents caused by dust explosion occur frequently, and on-line detection and early warning of dust cloud concentration in dust gathering places has become a key means to control dust explosion. However, installation and identification of dust concentration sensors were limited in large space where dust cloud gathers. In order to address this, combustible dust cloud recognition method based on deep learning was proposed. End-to-end detection and identification of explosive dust cloud were conducted by using CNN-based Faster R-CNN model. Then, a standard concentration image database was established to verify experimental results. The results show that Faster R-CNN model can effectively detect and identify explosive dust clouds, and it has high recognition accuracy.

Key words: combustuble dust cloud, image detection, convolutional neural networks (CNN), deep learning, Faster R-CNN model

中图分类号: