计算机应用 ›› 2012, Vol. 32 ›› Issue (07): 1894-1898.DOI: 10.3724/SP.J.1087.2012.01894

• 图形图像技术 • 上一篇    下一篇

基于图像的火焰检测中无量纲动态特征研究

黄正宇,缪小平,芮挺   

  1. 解放军理工大学 工程兵工程学院,南京210007
  • 收稿日期:2011-12-20 修回日期:2012-02-04 发布日期:2012-07-05 出版日期:2012-07-01
  • 通讯作者: 黄正宇
  • 作者简介:黄正宇(1986-),男,湖南宁乡人,硕士研究生,主要研究方向:图像识别及智能化应用;缪小平(1957-),男,江苏靖江人,教授,博士生导师,主要研究方向:国防工程内部设备及智能化;芮挺(1972-),男,天津人,副教授,博士,CCF高级会员,主要研究方向:图像与视频分析、智能技术与控制系统。

Study on dimensionless dynamic characteristics of flame based on image recognition

HUANG Zheng-yu,MIAO Xiao-ping,RUI Ting   

  1. Engineering Institute of Engineer Crops, PLA University of Science and Technology, Nanjing Jiangsu 210007, China
  • Received:2011-12-20 Revised:2012-02-04 Online:2012-07-05 Published:2012-07-01
  • Contact: HUANG Zheng-yu

摘要: 对火焰动态特征难以统一描述的问题,提出一种无量纲的检测方法。用“搜寻”的方法分割出火焰的可疑区域,分析火焰初期的特性,提取出三个无量纲特征因子并在贝叶斯分类器中训练,最后实现对火灾火焰的检测。其中“火焰动态常数”因子具有“稳定”的特性,其统计取值区间为[-0.003,0.003],突破了传统研究的时空局限性,不受火焰发展阶段、空间探测尺度以及监控设备种类的影响。对于实验中的不同远近的200帧序列的火焰检测,无量纲特征识别同一般特征识别结果相比较,正确识别率均超过90%。实验结果表明,无量纲动态特征因子能更好地描述火焰的特征,提高火焰识别的效率,增强火检系统的鲁棒性和可靠性。

关键词: 火焰特征, 火焰检测, 图像分割, 无量纲

Abstract: Concerning the difficulty in consistent description of flame dynamic characteristics, a dimensionless detection method was proposed. A "search" method was made to segment the suspicious districts. Then three dynamic dimensionless characteristic factors were extracted to train a Bayesian classifier after analyzing the characteristics of flame and the final detection result was realized. Of all three characteristic factors, the "flame dynamic constant" factor has "stable" characteristics and its statistical values range from -0.003 to 0.003. It has broken the spatio-temporal restriction of traditional research and is free from the influence of the flame development stages, space-scales of detection and varieties of surveillance devices. Under the flame detection experiments on different distance sequences of 200 frames, the correct recognition rate based on dimensionless features, compared with the general features, was almost over 90%. The experimental results show that the proposed factors are better able to describe the flame characteristics and they can greatly improve the efficiency of flame detection and boost the robustness of fire detection system.

Key words: flame characteristic, flame detection, image segmentation, dimensionless

中图分类号: