Video-based smoke detection with histogram sequence
of LBP and LBPV pyramids
Feiniu Yuan
n
School of Information Technology, Jiangxi University of Finance and Economics, Nanchang 330032, Jiangxi, China
article info
Article history:
Received 4 January 2010
Received in revised form
31 October 2010
Accepted 5 January 2011
Available online 20 January 2011
Keywords:
Video-based smoke detection
Local binary pattern
Multi-scale analysis
Neural network
abstract
Video surveillance systems are widely applied in a variety of fields. Hence, video-based smoke
detection is regarded as an effective and inexpensive way for fire detection in an open or large spaces.
In order to improve the efficiency of the video-based smoke detection, a novel video-based smoke
detection method is proposed by using a histogram sequence of pyramids. The method involves four
steps. Firstly, through multi-scale analysis, a 3-level image pyramid is constructed. Secondly, local
binary patterns (LBP), which are insensitive to image rotation and illumination conditions, are
extracted at each level of the image pyramid with uniform pattern, rotation invariance pattern and
rotation invariance uniform pattern to generate an LBP pyramid. Thirdly, local binary patterns based on
variance (LBPV) with the same patterns are also adopted in the same way to generate an LBPV pyramid.
And fourthly, histog rams of the LBP and LBPV pyramids are computed, and then all the histograms are
concatenated into an enhanced feature vector. A neural network classifier is trained and used for
discrimination of smoke and non-smoke objects. Experimental results show that the features are
insensitive to rotation and illumination, and that the method is feasible and effective for video-based
smoke detection at interactive frame rates.
& 2011 Elsevier Ltd. All rights reserved.
1. Introduction
Traditional smoke detectors usually detect the presence of
combustion products through an ionization or photometry based
sensors. But it takes a long time for combustion products to reach
these sensors in outdoor or open spaces and in the case of a strong
wind, combustion products may even be blown away, thus failing
to give fire alarms. Therefore, traditional smoke detectors are not
suitable in such cases. In recent years, researchers have used
computer vision technology in the field of fire detection in order
to overcome the aforementioned deficiencies of traditional sen-
sors. Video-based fire detection is one of the computer vision-
based methods and can be classified into two categories: video-
based flame and video-based smoke detections.
1.1. Video fire and smoke detection
Many approaches to the video-based flame detection have
been discussed by different worldwide researchers. Yamagishi
and Yamaguchi [1] presented a flame detection algorithm based
on the spatio-temporal fluctuation data of flame contours. Noda
and Ueda [2] proposed a flame detection system based on gray
scale images for tunnels. Phillips III et al. [3] implemented a video
flame detection method, using a Gaussian-smoothed color histo-
gram. T
¨
oreyin et al. [4] presented the video flame detection
method based on motion, flicker, edge blurring and color features.
Yuan et al. [5] proposed a video flame detection method, using the
mixture Gaussian model to extract temporal features. Ko et al. [6]
presented a fire detection method using techniques of moving
detection and fire-colored pixels.
In many cases of fire, smoke is usually visible before the flame
can be sighted, so video-based smoke detection is able to give fire
alarms earlier than the video-based flame detection. Most meth-
ods of video-based smoke detection often extract motion, edge,
color and texture features from video for the discrimination of
smoke and non-smoke objects. Toreyin et al. [7] used the features
of motion, flicker, edge blurring and color to detect smoke. In the
method presented by Gubbi et al. [8], some statistical features,
such as arithmetic mean, geometric mean, standard deviation,
skewness, kurtosis and entropy, were computed on each sub-
band of 3-level wavelet transformed images. Then, the SVM light
implementation of support vector machines was used for detec-
tion of smoke. Guillemant and Vicente [9] proposed a method of
smoke detection applied in the forests. Ferrari et al. [10] proposed
a real-time image processing technique for the detection of steam
in videos. They used Hidden Markov Tree (HMT), which was
derived from the coefficients of the dual-tree complex wavelet
transform (DT-CWT) in small local regions, to characterize the
steam texture pattern, and an SVM classifier was used to detect
the steam. This approach has the referential significance to smoke
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/firesaf
Fire Safety Journ al
0379-7112/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.firesaf.2011.01.001
n
Tel.: +86 7913820232.
E-mail address: yfn@ustc.edu
Fire Safety Journal 46 (2011) 132–139
剩余7页未读，继续阅读
资源评论
