SSRG International Journal of Computer Science and Engineering - (JCSE) – Volume 4 Issue4 April 2017

Image Processing Based Fire Detection

System using Rasperry Pi System
1
R.Dhanujalakshmi, 2B.Divya, 3C.Divya@sandhiya, 4A.Robertsingh
1,2,3
Dept of CSE, 4Assistant Professor/CSE,
Kalasalingam University,
AnandNagar, Krishnankoil, India.

Abstract
system. And then it search for the heat signatures
Internet of things (IoT) is the network of
and fire patterns, if it is a fire then it will on the
entities that consists of electronics, programmable
emergency mode. On detecting fire ,the system will
software, sensors, and communication facility that
send the MMS to the remote user.
enables these entities to gather and transfer data. The
objective of the proposed system is to alert the remote
user while the fire accidents occur. This system can II. RELATED WORKS
be installed at any remote premise which has threat In this section discusses various
of fire accidents. Using this system we can detect the fire detection methods using image processing and
fire by camera. So sensors are not required to detect using raspberry pi Md Rifat Hasan [1] They designed
fire. The Raspberry Pi controller processes the by using sensor, fuzzy logic, Data fusion. The
camera input and detects fire using heat signatures. purpose of this system is to avoid panic inside the
By using image processing method, the report is building.the main disadvantage in this system is
automatically generated and sends to the person mulitiple sensor, false alarm and false message.
immediately after the fire is detected in any part of M.Malathi [2] designed by using Raspberry pi,
the frame using Wi-Fi/GSM. On detecting fire, the artificial neural network, RGB colors and the Purpose
system will go into emergency mode. The major of this system is to avoid the false alarm. The main
advantages in this method are: sending the disadvantage in this method is it will not send any
information to the person at any time, any place and message to the person and also it does not find the
remote monitoring for immediate actions. location of the fire. Cao Shunxia et.al [3] designed a
system with single chip microcomputer (SCM)
Keywords AT89C51 and ISD1420a, a voice chip .The purpose
Image processing, Internet of Things, of this system is to detect fire. When the sensor
Heat signatures, Sensors detects smoke, a voice message will be send to the
relevant department. And the disadvantage of this is
false alarm will be submitted.RakeshV S et.al [4] they
I. INTRODUCTION
designed the system y using Zigbee and FTP Web
Fire is dangerous that could bring the great server. purpose of this system is to avoid false alarm
loss for human life. To prevent these losses various when smoke or intruder movement are detected , the
alarm system has been developed. As technologies system sends warning message through SMS.And the
involved the various automatic fire alarm system is disadvantage in this method is does not take any
used In existing method, sensors are used to find the action to stop the fire. Moreover, the single board
fire. But the major disadvantage in sensor method is computer is used, and it is expensive and has lower
sensing fire only when it reaches the programmed technical specification compare to raspberry pi
level of the temperature and also it cannot generate
any report for analysis process. To obtain a cost III. ISSUES TO BE ADDRESSED
effective fire alert solution ,we use image processing
system and raspberry pi to detect the fire .By using An important point in smoke and fire
raspberry pi ,it consumes low power, low cost and detectors are typically used to detect the particular
execute faster to detect the fire.The main advantage particles is generated by smoke and fire by using
of the System is the early warning. This system can ionization or photometry. And then sensors are used
be installed anywhere for fire detection. So we do not to sense particles. The main weakness of point
need any other sensor. Here, the camera will capture detectors is that they are in limited distance and used
the video and separate the image into frames. And in open or large spaces. By using sensor many of
then the frames are compared to the original them meets the false alarm.And then sensors not
image. Which is already booted in the rasspberry pi cover large space while detecting the fire.Many
sensor are costly when compare to the raspberry pi.

ISSN: 2348 – 8387 www.internationaljournalssrg.org Page 18

 SSRG International Journal of Computer Science and Engineering - (JCSE) – Volume 4 Issue4 April 2017

IV. PROPOSED WORK based on light in the image. Here

The proposed system uses Image processing; value 125 is used.
strength of using image processing in fire detection is B)cieLAB filter[7]:
the ability to serve large and open spaces. Proposed
system consists of three stages: In the first stage, Here the LAB color model is used .cieLAB
camera will capture the image and it will send that color model are Highlights red, yellow and related
image to controller for further evaluation. And then colors like orange. For all pixels in the frame the
the process of further detection has been started. mean value of L, A and B components are identified.

In the second stage, the images are converted into  For every pixel four filters are used.
frames and it will compare those images into already o If L>L mean
booted images. In the third stage, MMS will be sent o If A>A mean
to the user . o If B>B mean
o If B>A mean
V. MODULE DESCRIPTION  Whose values run from 0 (black) to 100 (white).
We segmented our system design into five modules.  The central vertical axis represents lightness
Video recording and transferring to controller is first (signified as L*)
module of our system design. In this module the
camera captures continuous frames from the area of  The color axes are based on the fact that a color
its coverage. All the captured images or frames can't be both red and green, or both blue and
transfer to the controller for applying image yellow, because these colors oppose each other.
processing function. Second module is Color based
 On each axis the values run from positive to
segmentation in this module after separating the
negative.
frames as single image it is subjected to standard
color base segmentation. The segments are further  On the a-a' axis, positive values indicate amounts
divided into constant sized blocks. of red while negative values indicate amounts of
green.
Third module is Fire pattern recognition in this
module the blocks of the segmented image will be  On the b-b' axis, yellow is positive and blue is
examined for the presence of heat signature or fire negative. For both axes, zero is neutral gray
patterns.Emergency trigger is our four module here, if Both:
any peculiar pattern is identified in any of the blocks
for a particular period of time, it will switch on the To satisfy various lighting conditions both the RGB
emergency mode for the purpose of MMS send to the and cieLAB filters are used. If any one of the filter
remote user. Multimedia message transmission is our passes a pixel, it is a fire signature.
fifth module in this module the multimedia message C)RGB filter2
is created with its content using messaging API and
alert message with image will be send to the recipient Another filter that uses RGB components. It will
works well at night mode. In this method the R, G,
and B components are compared with threshold
VI. METHODLOGY
values.
Fire is detected using fire patterns with heat
Signature. Heat signature is color patterns to  rt=140, gt=100, bt=100
represent the fire. There are three filters are used to  Three conditions are checked:
find the heat signature. They are:
o R>rt
o RGB filter
o G>gt
o cieLAB filter
o B<bt
o Both
o RGB filter 2
VII. ADVANTAGES
A)RGB filter[6]:
1) It will send the MMS to the remote user
RGB filter is used to Extract Red (R) Green (G) and when it reaches the emergency mode.
Blue (B) component of each pixel. And then in every
pixel two conditions are verified. They are 2) Image Processing in fire detection is the
ability to serve large and open spaces.
o If R >G>B
3) Raspberry pi has higher specification and
o If R> Rt (Rt is the red threshold low cost
value between (0,255). This is

ISSN: 2348 – 8387 www.internationaljournalssrg.org Page 19

 SSRG International Journal of Computer Science and Engineering - (JCSE) – Volume 4 Issue4 April 2017

VIII. RESULTS
The proposed method was implemented using
Raspberry Pi 2[5], OpenCV 3 and python3.4. Figure
1. shows output of the filter where no fire is found.

Figure 3. MMS received in phone

IX. CONCLUSION

We implement the algorithm to detect the fire and

then it will send the image to remote user. In the form
of enhanced an image in different enhancement
degree using raspberry pi. It was found that algorithm
developed for raspberry pi executed successfully.

X. REFERENCES

[1] Md Saifudaullah Bin , Rosni Abu ”Development of Fire

alarm system”, Published in Electrical, Electronics and
system Engineering ICEESE 2013.
[2] C.Shunxia and C. Yanda, “Design Of Wireless Intelligent
Figure 1. Output of filter without fire Home Alarm System”, Industrial Control and Electronics
Engineering (ICICEE), 2012.
Figure 2 shows output of filter where fire occurs. [3] V.Rakesh, P. Sreesh and S. N. George ,”Improved real-time
Figure 3 shows the multimedia message received by surveillance system for home security system”, 2012 Annual
IEEE India Conference (INDICON).
through yowsup package. [4] J.San-Miguel-Ayanz and N. Ravail, “Active fire detection for
fire emergency management: Potential and limitations for the
operational use of remote sensing,” Natural Hazards , vol.
35,no. 3,pp.361-376,2005.
[5] Raspberry pi –www.raspberrypi.org.
[6] Turgay Celik, Hasan Demirel, Huseyin Ozkaramanli, and
Mustafa Uyguroglu, &quot;Fire detection using statistical
color model in video sequences,&quot; J. Vis. Comun. Image
Represent. , vol. 18, pp. 176–185, April 2007.
[7] Turgay Celik , Fast and Efficient Method for Fire Detection
Using Image Processing , ETRI Journal , December 2010,
pp. 881-890.

Figure 2. Detected fire and frame where fire

occurs.

ISSN: 2348 – 8387 www.internationaljournalssrg.org Page 20