基于约束非均匀形变特征的小尺度火焰识别方法研究

王向军; 杜志伟; 高超

基于约束非均匀形变特征的小尺度火焰识别方法研究

王向军^{1, 2,},
杜志伟^{1, 2},
高超³

1.
天津大学精密测试技术及仪器国家重点实验室，天津 300072
2.
天津大学微光机电系统技术教育部重点实验室，天津 300072
3.
天津大学智能与计算学部软件学院，天津 300072

详细信息

作者简介:
王向军（1955-），男，博士，教授，主要研究方向为光电探测与测量和图像测量与计算机视觉。E-mail：tjuxjw@126.com

中图分类号: TP391.4
计量
- 文章访问数: 208
- HTML全文浏览量: 120
- PDF下载量: 32
- 被引次数: 0
出版历程
- 收稿日期: 2019-10-11
- 修回日期: 2020-12-30
- 刊出日期: 2021-02-20

Small Scale Fire Identification Based on Constrained Inhomogeneous Deformation Feature

WANG Xiangjun^{1, 2
,},
DU Zhiwei^{1, 2},
GAO Chao³

1.
State Key Laboratory of Precision Measuring Technology and Instruments, Tianjin University, Tianjin 300072, China
2.
MOEMS Education Ministry Key Laboratory, Tianjin University, Tianjin 300072, China
3.
College of Intelligence and Computing, Tianjin University, Tianjin 300072, China

摘要

摘要: 基于视觉的火焰检测是一种灵活、低成本的火焰检测方式，但现阶段常用的火焰特征不能对火焰和干扰物进行有效的区分，有较大的误警率。本文基于目标轮廓的时序行为特征，将火焰的闪烁描述为一种有约束的非均匀形变过程，结合隐马尔可夫模型和传统几何特征对火焰和干扰物进行更准确地区分。实验表明，通过引入补充的火焰特征显著提高了火焰检测的准确率，有效减少了复杂环境下干扰物引起的虚警。
- 火焰检测 /
- 特征提取 /
- 主曲线 /
- 隐马尔可夫模型
Abstract: Video based fire detection(VFD) is a convenient, low-cost method widely used in fire detection. However, it's not credible enough to distinguish true fire from possible disruptors by traditional fire features. This paper extract two new features to analyze the time series behavior of fire based on the motion of edge pixels. The inter frame behavior of edge pixels is regarded as a nonuniformty constrained deformation procedure. Combined with HMM and additional geometric features to distinguish true fire from possible disruptors, the accuracy of fire detection is greatly improved and the false alarm rateis efficiently reduced.
- fire detection /
- feature extraction /
- principal curves /
- HMM

HTML全文

图 1 红外视频及火焰闪烁轮廓序列

Figure 1. The example of infrared video and flame flicker sequences

下载: 全尺寸图片幻灯片

图 2 火焰轮廓像素时域分布特性

Figure 2. Temporal distribution of flame contour pixels

下载: 全尺寸图片幻灯片

图 3 K主曲线拟合平均轮廓线

Figure 3. The fitting of average contour based on K-principal curve

下载: 全尺寸图片幻灯片

图 4 离散邻域像素提取模板

Figure 4. The extracting models of neighbor pixels

下载: 全尺寸图片幻灯片

图 5 拟合轮廓邻域点统计特征

Figure 5. The statistical characteristics of fitting contours

下载: 全尺寸图片幻灯片

图 6 部分样本轮廓时域分布及对应梯度特征

Figure 6. Contour pixel temporal distribution and gredient feature

下载: 全尺寸图片幻灯片

表 1 不同样本全局分布与去峰分布统计特征

Table 1. Statical features of all pixels and non-peak pixels of different sample

Sample No.	Peak Value	Flattened Peak	Variance	Flattened Var.	Kurtosis	Flattented Kurt.
1	38.5	10	1.194	0.1711	0.7987	-0.1611
2	115	105	90.67	117.8	0.1975	1.7610
3	20	16	0.2324	0.1712	1.452	1.105
4	8.7	7.5	0.0015	0.0022	1.325	1.133

下载: 导出CSV

表 2 不同特征检测结果对比

Table 2. Experimental results comparison of different features

Features	TPR	FPR	TNR	FNR
Inhomogeneous deformation	81.67%	10.00%	90.00%	18.33%
Combination of geometric features	70.00%	31.67%	68.33%	30.00%
Pixel flicker	71.67%	55.00%	45.00%	28.33%

下载: 导出CSV

参考文献(14)

[1]	Çetin A E, Dimitropoulos K, Gouverneur B, et al. Video fire detection -review[J]. Digital Signal Processing, 2013, 23(6): 1827-1843. doi: 10.1016/j.dsp.2013.07.003
[2]	Owrutsky J C, Steinhurst D A, Minor C P, et al. Long wavelength video detection of fire in ship compartments[J]. Fire Safety Journal, 2006, 41(4): 315-320. doi: 10.1016/j.firesaf.2005.11.011
[3]	Toreyin B U, Cinbis R G, Dedeoglu Y, et al. Fire detection in infrared video using wavelet analysis[J]. Optical Engineering, 2007, 46(6): 067204. doi: 10.1117/1.2748752
[4]	李涛, 向涛, 黄仁杰, 等. 基于新的运动特征的火焰检测方法[J]. 计算机仿真, 2014, 31(9): 392-396. doi: 10.3969/j.issn.1006-9348.2014.09.087 LI Tao, XIANG Tao, HUANG Renjie, et al. Fire detection method based on new moving feature[J]. Computer Simulation, 2014, 31(9): 392-396. doi: 10.3969/j.issn.1006-9348.2014.09.087
[5]	朱思思, 丁德红, 陈朝迎, 等. 基于图像处理的森林火灾识别方法研究[J]. 红外技术, 2016, 38(5): 440-446. http://hwjs.nvir.cn/article/id/hwjs201605014 ZHU Sisi, DING Dehong, CHEN Zhaoying, et al. The research of forest fire recognition method based on image processing[J]. Infrared Technology, 2016, 38(5): 440-446. http://hwjs.nvir.cn/article/id/hwjs201605014
[6]	陈娟. 基于多特征融合的视频火焰探测方法研究[D]. 北京: 中国科学技术大学, 2009. CHEN Juan. Study on Method of Multi-Feature Fusion Based Video Flame Detection[D]. Beijing: University of Science and Technology of China, 2009.
[7]	CHEN J H, Im H G. Correlation of flame speed with stretch in turbulent premixed methane/air flames[J]. Symposium (International) on Combustion, 1997, 27(1): 819-826.
[8]	Hamins A, Yang J C, Kashiwagi T, An experimental investigation of the pulsation frequency of flames[J]. Symposium(International) on Combustion, 1992, 24(1): 1695-1702. http://www.sciencedirect.com/science/article/pii/S0082078406801980
[9]	Chaconmurguia M I, Perezvargas F J. Thermal video analysis for fire detection using shape regularity and intensity saturation features[C]//Mexican Conference on Pattern Recognition, 2011: 118-126.
[10]	Takahashi N, Suzuki M, Dobashi R, et al. Behavior of luminous zones appearing on plumes of large-scale pool fires of kerosene[J]. Fire Safety Journal, 1999, 33(1): 1-10. doi: 10.1016/S0379-7112(99)00009-0
[11]	程鑫, 王大川, 尹东良. 图像型火灾火焰探测原理[J]. 火灾科学, 2005(4): 239-245, 196. https://www.cnki.com.cn/Article/CJFDTOTAL-HZKX200504007.htm CHENG Xin, WANG Dachuan, YIN Dongliang. Image type fire flame detecting principle[J]. Fire Safety Science, 2005(4): 239-245, 196. https://www.cnki.com.cn/Article/CJFDTOTAL-HZKX200504007.htm
[12]	Paresh P A, Parameswaran L. Vision-based algorithm for fire detection in smart buildings[C]//International Conference on ISMAC in Computational Vision and Bio-Engineering, 2018: 1029-1038.
[13]	Kegl B, Krzyzak A, Linder T, et al. A polygonal line algorithm for constructing principal curves[C]//Neural Information Processing Systems, 1998: 501-507.
[14]	WU Z, Fuller N W, Theriault D H, et al. A thermal infrared video benchmark for visual analysis[C]//Computer Vision and Pattern Recognition, 2014: 201-208.