主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2022, Vol. 41 ›› Issue (11): 1604-1608.

• • 上一篇    下一篇

基于图像识别技术的汽油、无水乙醇燃烧火焰识别研究

李雅芝1,3,车 强2,3   

  1. （1.中国人民警察大学 研究生院,河北 廊坊 065000; 2.中国人民警察大学 侦查学院,河北 廊坊 065000;3. 中国人民警察大学 新型犯罪防控研究中心,河北 廊坊 065000）
  • 出版日期:2022-11-15 发布日期:2022-11-15
  • 作者简介:李雅芝（1994- ），女，河南焦作人，中国人民警察大学研究生院研究生，主要从事火灾调查研究，河北省廊坊市安次区西昌路220号，065000。
  • 基金资助:
    中国人民警察大学中青年科研创新计划课题（ZQN2021006）

Study on flame recognition of gasoline and anhydrous ethanol combustion based on image recognition technology

Li Yazhi1,3, Che Qiang2,3   

  1. （1. Graduate School， China People's Police University, Hebei Langfang 065000, China; 2. School of Criminal Investigation， China People's Police University, Hebei Langfang 065000, China; 3. Research Center for New Crime Prevention and Control, China People's Police University, Hebei Langfang 065000, China）
  • Online:2022-11-15 Published:2022-11-15

摘要： 为了实现基于视频图像对火灾现场存在助燃剂的分类识别，对燃烧火焰的特征进行分析，根据汽油和无水乙醇引燃后各自特有的燃烧现象，结合火焰的视频图像识别算法实现对汽油和无水乙醇燃烧火焰的识别。首先，基于图像的灰度阈值得到其疑似火焰区域，再提取其H、S、I颜色分量和面积变化特征；并提取燃烧图像的小波高频能量特征和LBP直方图特征；最后将特征向量输入SVM分类器进行分类识别。试验表明，SVM对汽油和无水乙醇燃烧火焰的识别分类准确率可达98.5%，可较好地实现对汽油、无水乙醇燃烧火焰的区分。

关键词: 汽油, 无水乙醇, 小波高频能量, LBP直方图

Abstract: In order to realize the classification of oxidant in the fire scene based on video images, the characteristics of the combustion flame were analyzed. In this paper, according to the unique combustion phenomena of gasoline and anhydrous ethanol, combined with the flame video image recognition algorithm, the flame recognition of gasoline and anhydrous ethanol is well realized. First, the suspected flame area can splited based on the grayscale threshold of the image, then its H, S, I color components and area change features are calculated; and the wavelet high-frequency energy features and LBP histogram features of the combustion image are extracted; finally, the feature vector is entered into the SVM classifier for classification and recognition. Experiments show that the classification accuracy of SVM for gasoline and anhydrous ethanol combustion flames is as high as 98.5%, this method can better distinguish the combustion flames of gasoline and anhydrous ethanol.

Key words: gasoline, anhydrous ethanol, wavelet high frequency energy, LBP Histogram