Content uploaded by Varun Pande
Author content
All content in this area was uploaded by Varun Pande
Content may be subject to copyright.
Content uploaded by Khaled Elleithy
Author content
All content in this area was uploaded by Khaled Elleithy
Content may be subject to copyright.
Content uploaded by Khaled Elleithy
Author content
All content in this area was uploaded by Khaled Elleithy
Content may be subject to copyright.
Optimized Algorithm for Fire Detection over WSN using Micaz Motes
Varun Pande, Wafa Elmannai, Khaled Elleithy
Department of Computer Science and Engineering
University of Bridgeport
Bridgeport, CT 06604, USA
{welmanna, elleithy, vpande}@bridgeport.edu
ABSTRACT
Environment degradation around the world has motivated
many researchers to deal with an important yet
endangering aspect of rural and forest fires. Most of the
current developed technologies are based on detection of
the fire rather than verifying it. Detecting the fire can be
useful in many cases but it is still not efficiently
implemented in real time systems. However, detecting
fire systems can help many modern cities improve their
smoke detection systems. Therefore, in this paper we
introduce and implement a fire detecting algorithm built
on measuring the temperature of a certain area and
detecting the fire. We have also used a mathematical
model called the “Acoustic ranging Technique” to detect
the location and set the alarm in case of a fire. In our
implementation we have used multiple MTS 300 sensor
boards mounted on MICA2 motes,in order to sense the
temperature of the fire with respect to the energy
consumption. Hence, with use of the implemented
algorithm, we can verify the size of the fire from
temperature recorded and analyzed based on the color
temperature. Finally, in this paper we could prove that the
relation between the type of fire and its colors can be used
in detecting the size of the fire efficiently.
1. INTRODUCTION
Several years ago, forest fires were a major issue that was
a threat to the environment. Lately even civilized cities
have started facing this problem. Not only the
environment is harmed by the fires but also human beings
suffer due to this. Especially during the summer seasons
instates like California, each year thousands of houses are
burned down and hundreds of people die of fire disasters.
Therefore, pollution of water and environmental damage
is resulted from the fire remnants. [8].
Furthermore, some natural resources can cause fire. In
2003, 45,000 people were evacuated and 239 houses were
burned in Okanagan Mountain Park, Canada. The
lightning and winds caused the spreading of fire around
that forest. According to reports, this firestorm burned
25,912 hectares even though 1,000 fire fighters within 60
departments worked nonstop for long periods of time
trying to stop it [2]. This was not the last firestorm that
was in that province.In 2006 more than $156 million was
spent on such disasters [3].
Fire detecting techniques based on wireless sensors has
motivated many researches to improve and built early
warning and better detection system. Sensors proved their
efficiency in many monitoring systems. Sensors are used
to detect and monitor a lot of natural and non-natural
factors. With their ability of sending and receiving signals
and data, sensors can determine and evaluate the type and
speed of natural elements like fire, humidity, smoke,
wind,etc., for example sensors in fire detectors [8]. These
sensors send electrical signals between each other and
notify the base station about the identified risk. Wireless
sensor network (WSN) are defined as a network that
include thousands of distributed nodes ‘sensors’ over the
wireless links for the detection systems [1].
WSNs are used in many fields such as in industry of
commerce and agriculture. Two factors are to be
considered in WSNs. They are the limited in energy and
have small memory [10]. However, in our implementation
we use an efficient communication strategy to allow the
closest sensor to receive the signal while others are
asleep. Then send it to thecentral base station in order to
reduce the power consumption. Therefore, WSNs can
help the fire detecting systems by using these sensors to
sense the fire type and its temperature.
The use of these wireless sensors in fire detecting systems
help these sensors to send the alarm to the base station
over the wireless links whenever the fire is detected.
Based on that, the base station runs the application to
determine the location, temperature and the size of the
fire. This application can help the fire fighters or fire
department to determine the size of the fire by the degree
of the color based on the color temperature based
database.
This paper organized as follows: the related work is
described in section 2, the Ad Hoc Network design in
section 3, overview of MICAZ-MTS 310 sensor board
and simulation setup in section 4, simulation results, and
analysis in section 5 and conclusion in section 6.
2. RELATED WORK
There are limited fire detection systems that have been
implemented and presented in literature. Most of these
systems are only theoretical models rather than
implementation. There is other multiple systems that are
based on a single technique or on mixed techniques (e.g.
mixed of multisensory and wireless cameras).However, in
this section we have studied several proposed works
which are related to our work.
The Forest Fire Surveillance system (FFSS) was
improved by Son et. al.in order to overview South Korea
Mountain [12]. This paper combined a web application,
transceiver, WSN and middleware. The nodes used
TinyOS as their operating system. The nodes collect the
dimensions of the environment requirements such as
humidity and temperature. The received node (Sink-node)
sends these measurements to the software. Then the risk is
determined by the middle ware software. Finally, the
alarm is activated by FFSS for early warning of the fire.
In [5], the fire behavior is detected by Fire Wx Net. Fir
Wx Net targets to measure the condition of the current
weather and detect the visual data for both the base station
as well as around the forest. These zone images are taken
continuously. Hence, the natural factors such as
temperature and wind speed should be measured by the
system from time to time to determine the improvement
of the risk. Also, the beauty of the system was that the
team could prove that their system could be a realistic
system for fire behavior analysis.
A new algorithm was introduced by Li et. al. in order to
implement fire detection system for boreal zone in
Canada [9]. The authors did not use only the satellite
remote sensor in their system, but also they used the
Advanced Very High Resolution Radiometer (AVHRR)
to provide a high quality images around the fire. The
system showed good features such as quick response and
high quality of images for long distances.
Different from the systems presented earlier in this
section, in [7] Wireless Local Area Network (WLAN)
was used to determine the fire. The design in [7] is based
on the verification of the fire rather than detect it. The
improvement of this work is using the cameras and multi-
sensor nodes. Hence, this system follows several steps to
detect and verify the fire. Whenever a multi-sensor node
detects the fire, the server receives the warning alarm over
the network. The end user invokes the proposed
application on the server to determine the closest camera
easily. Lastly, the taken images are sent to the sink nodes.
This system is used in real time.
A reliable framework is proposed in [4] for general
reportage over the WSNs. The experiment shows that this
general frame work can be used for fire detection. The
developing team’s focus is on the importance of
circulation of detection systems. In order to reroute the
data to the sink rapidly, the team introduces a new
algorithm for collecting the data. The system filters the
delayed - aware data as well as the significant data.
The presented techniques in this section have different
features but lack the ability to specify the relation of the
fire type and its temperature to its colors. In our new
implementation presented in this paper, we show and
elaborate how these factors can determine the size of the
fire. Further, we are using MICA2 motes to sense the
temperature of the fire and define its location using
acoustic ranging.
3. PROPOSED WORK
3.1. Over view and experimental setup:
In this section we demonstrate the setup of our approach
and the hardware features that support our deployment.
Also, in this section we discuss several devices used in
the proposed architecture. In our experimental tests, we
use MTS300 sensor board. This board includes important
modules such assounder, thermometer, a Dual-Axis
accelerometer (ADXL202), photon cell and Dual-Axis
magnetometer [6]. However, the main components we use
are the microphone, the sounder, the photon cell and
thermometer to measure the temperature. Each
component has a unique feature for our experimental tests
and is explained in the following sub-sections.
A. Microphone circuit
Acoustic ranging and a general acoustic recording
measurement are the two main functions in this circuit.
This circuit includes both a pre- amplifier (U1A-1) and
second-stage amplifier with a digital-pot control (U1A-
PT2). In acoustic recording and measurement principle
the output of the low level of microphone is increased by
preamplifier. With the help of an output selector (MX),
the analog digital converter (ADC) can serve the output of
microphone’s low level. That can be done when the
output selector links both ADC2 and mic_out signal. The
storage devices of motes store the audio files. In second
output, after the Tone Detector received the mic_out
signal from filter (U), the analog microphone signal will
be turned to the low or high level. The converted stage
occurs usually when the tone becomes 4 kHz. This Tone
will be generated on the board later by the sounder [11].
B. The Sounder
The sounder is a simple frequency with a range of 4KHZ.
It includes both frequency and drive control circuitry. The
power switch is the only device that can control this
sounder’s operation.
C. The Light and Temperature Sensors
A/D converter channel (ADC) is the shared channel for
the light and the temperature sensors as shown in Figure
1. It has a circuitry which allows only one sensor to work
at a time [11]. The Cdse Photo is the light sensor. There
are two conditions for this sensor that is a dark and a light
condition. The digital control signal has to be on to use
this sensor. ADC is always connected to the light sensor
output in order to know the condition of the sensor.
Figure1:The power controlled signal of MTS300 for Temperature and
Light sensors.
The main focus of our application is to apply the acoustic
ranging principle for tone detector and sounder sensor.
Our experiment shows that both RF packet and radio
signal are sent at the same time. By setting a counter on
the timer on mote’s process, it can receive RF packet and
start to listen for the tone at the same time. Also, the
beauty of this application is that, we can detect the fire
location easily. This is done by increasing the counter
until the microphone on the base station detects the tone
produced by the sensor that detected the fire. This way we
can calculate the time of flight of the sound and thus
comparatively localize the location of the fire even
though the wireless nodes are distributed randomly.
3.2. The Work Process and System Architecture
In this section we explain the procedure and the flow of
our experiment as shown as in Figure 2.
Before initializing the master node, all the nodes are
randomly distributed over the simulation area including
the master node as shown in Figure 3. Figure 3 shows the
architecture of the system and how the communication is
done between the nodes and the master node as well as
between the master node and the base station. Since the
distribution is random, there is no known location of the
nodes. However, an experimental area can include close
to 1000 nodes and 200 master nodes. During the
simulation and placement of the nodes, some nodes are
specific for sending the RF packet to the master nodes by
the activation of thelight sensor.
Start
Initialize Master
Node
Initialize Light
Sensors
Communicate
With Base
Station
If Night
Temperature
Readings
Yes
No
Initialize tone
detector on
Spectral
Readings
Send Data To
Basestation
Data On
Spectral
Readings
Data On
Temperature
Readings
Analyze Data
Display
Output
Stop
Activate
Sounder
If Threshold
Broken
Yes
No
Send RF
Packet
Figure 2: MTS300 Fire Detection system flow chart
Other nodes are specific toactivate the sounder whenever
fire is detectedor the temperature threshold is broken.
3Mts
3mts
3mts
3mts
Base Station
Master Node
Acoustic Signal
Acoustic Signal
Acoustic Signal
Acoustic Signal
Figure 3: The System Architecture
Each node has its own module for sensing,(each MTS 300
panels committed to 2 separate MICAZ panels). Each
sensor components has a separate sensor boards (each of
the MTS 300 boards accomplish a separate task), and a
light sensor is used by one board to detect the fire while
the other measures the temperature and sets up the
sounder (the fire is detected by the cdse photo cell).
Whenever, any light sensor of the node detects the fire,
then its parallel sensor has to activate the sounder by
sending an acoustic signal if the threshold of the
temperature is broken. The detected node will send the RF
packet to the master node.
According to the acoustic ranging principle, the counter
of master node has to be reset as soon as the node receives
the RF packet. To allocate the fire place we have to
calculate the time of flight of the sound signal by the
sounder. It is calculated based on the time that took of the
master node’s microphone to receive the sound signal.
This approach is one of the easiest and non-expensive
ways to allocate the fire without having a prior knowledge
of the location of the node or acknowledgement from the
detected node about the fire’s location.
3.3 Methodology
In our experimental tests we use MICA2 motes. The
model we use is MPR400CB and the data radio is
CC1000900mhz. Hence, the MTS300 CA is used in order
to extent the sound, temperature and light sensors. The
proposed frame work involves programming the motes
for permitting them to communicate with the master node
and allow the master node to communicate with base
station. The sensors send the readable data to the base
station. The data describes the increase in the temperature
of the fire as shown in our implementation in Figure 4.
Based on our study we found that there is a relation
between the fire’s temperature and their colors. It is
shown below whenever the temperature increases the
color changes in several gradients. That can help us in
detecting the size of the fire based on its colors.
1. Red: as the temperature increases, the redness changes
until it reach the cherry color before it changes to next
case.
a. Observable: 980 °F (525 °C).
b. Bright: 1,300 °F (700 °C).
c. Cherry, bright: 1,500 °F (800 °C).
d. Cherry, full: 1,700 °F (900 °C).
e. Cherry, pure: 1,800 °F (1,000 °C).
2. Orange:
a. Deep: 2,000 °F (1,100 °C).
b. Pure: 2,200 °F (1,200 °C).
3. White
a. Whitish: 2,400 °F (1,300 °C).
b. Cheerful: 2,600 °F (1,400 °C).
c. Impressive: 2,700 °F (1,500 °C).
In section 4 we discuss the results of our implementation
and how with higher temperature and lower resistance of
CKT we can get correct data.
Figure 4:measured data of temperature of the fire is giving by several
detected sensors
4. ANALYSIS THE RESULTS
As shown in section 3, we have used the MTS300CA
sensor. MTS300CA helps in monitoring the energy of the
fire and detects it as it moves around. The more fire
spread around, the more nodes become active to detect the
fire.
A. The relation between the measured temperature and
the accuracy data scenario:
The complete Fire Detection system is implemented in
NES C in UNIX Tiny OS platform or its equivalent
Cygwin on a windows platform. A standard MEMSIC
sensor board the MTS 300 mounted on top of the MICAZ
was used for communication. An MIB 510 Program board
isused to program the individual motes. The experimental
results are conducted on a Pentium 2.4GHz computer
system.
Figure 5 shows the relation between the resistances in the
board to the temperature. By examining Figure 5, we can
conclude that the temperature increases as the resistance
decreases which lead to the accuracy of the measurements
in temperature. The Cdse Photon sensors of the MTS 300
boards works as the primary detector in the initial
detection of fire because factors like extreme sunlight or
artificial light can confuse the photo sensor the thermistor
of MTS300 takes care of the secondary measurement of
temperature. Although the motes work perfectly
individually, in order to cover a large area it is required to
have multiple motes communicating with each other. As
the distribution of motes is random we use the acoustic
ranging technique which allows the motes to define its
location to the base station.
Figure 5: Resistance VS Temperature Graph
To evaluate the performance of our proposed Fire
Detection system, a number of scenarios were created
using up to 4 pairs of motes (8 modules) to conduct our
experiment. We measure the intensity of light that allows
the data base to confirm the type of fire especially the size
based on the temperature inputs. The detection rate is the
ratio of successful attempts (cases where the best
detection is the correct detection) to the total attempts.
The results show that the detection rate is 92 %( increase
in number of motes might provide betterresults). This
makes our technique very suitable for many fire detection
applications.
B. Power Consumption Scenario:
Figure 6 shows the behavior of power consumption of the
sensors. Since the biggest issue in the wireless sensor
networks is the energy consumption especially when there
is random distribution of the sensors, it will be difficult to
detect the sensor location to recharge in case the nodes
run out of power. However, our results show that we can
save in energy consumption of the sensor with our new
algorithm. Since the experimental testsare based on
activating several sensors which in and around the fire
while the sensors areasleep,only the closest node will be
active to detect the fire and inform the base station. When
the detected node informs the base station, base station
will send a signal back to all the nodes in the base stations
vicinity for accepting all the readable data.
We also use the thermistor to measure the temperature.
The measured temperature is retransmitted to the base
station. With the use of the radio channel of the MTS300
we can send warning alarm to other nodes. Finally in
Figure 6 it is clear to see more temperature data was
recorded when more power was consumed.
Figure 6: the behavior of the power consumption Vs the temperature
5. CONCLUSION
Fire detection has been usedin many applications such as
Forest fire Prevention systems, surveillance of volatile
labs, etc. In this paper, we have proposed a new technique
that implements a new fire detection algorithm. We have
used MTS300CA that shows a high ability of monitoring
in the MANET areas. The approach is fast, easy, and it
provides a practical solution to the detection problem for
either accuracy of data or power consumption. The
system has a success rate of 92% for a large set of
scenarios. Our implementation could prove high result not
only in detecting the fire but also in verifying the fire
location specifically.
This work is extended in the by using the IMB400
multimedia boards and using filtering mechanisms and
computer vision technology to improve the accuracy of
detection as well as the exact location of the fire.
REFRENCES
[1] Ahmad Abed Alhameed Alkhatib and Gurvinder
Singh Baicher, “An Overview of Wireless Sensor
Networks”, 2012 International Conference on Computer
Networks and Communication Systems (CNCS 2012),
2012.
[2] B.C. Ministry of Forests and Range.Fire Review
Summary for Okanagan Mountain Fire (K50628).
[3] B.C. Ministry of Forests and Range Web Page,
http://www.for.gov.bc.ca .
[4] E. Ngai, Y. Zhou, M. Lyu, J. Liu, “A Delay-aware
Reliable Event Reporting Framework for Wireless
Sensor-actuator Networks, Ad Hoc Networks,” Vol. 8,
Issue 7, pp. 694-707, 2010Sdf.
[5] Hartung, C.; Han, R.; Seielstad, C.; Holbrook, S. “Fire
WxNet: a multi-tiered portable wireless system for
monitoring weather conditions in wild land fire
environments,” In ACM, 4th International Conference on
Mobile Systems, Applications and Services, Uppsala,
Sweden, June 19-22, 2006.
[6]http://bullseye.xbow.com:81/Products/productdetails.a
spx?sid=177.
[7] J. Lloret, M. Garcia, D. Bri, S. Sendra, A Wireless
Sensor Network Deployment for Rural and Forest Fire
detection and verification, sensor Nodes, Vol. 9 , Issue
Detection and Verification, Sensor Nodes, Vol. 9, Issue
11, pp. 8722-8747, 2009.
[8] Jaime Lloret, Miguel Garcia, Diana Bri and Sandra
Sendra, “A Wireless Sensor Network Deployment for
Rural and Forest Fire Detection and Verification”, sensors
,2009.
[9] Li, Z.; Nadon, S.; Cihlar, J. Satellite-based detection
of Canadian boreal forest fires: development and
application of the algorithm. Int. J. Remote Sens. 2000,
21, 3057–3069.
[10] Mohamed Hefeeda and MajidBagheri, “Wireless
Sensor Networks for Early Detection of Forest Fires
”, Mobile Adhoc and Sensor Systems, 2007.MASS
2007.IEEE Internatonal Conference on, 2007.
[11] MTS/MDA Sensor Board User’sManual, June 2006,
PN: 7430-0020-04.
[12] Son, B.; Her, Y.; Kim, J. “A design and
implementation of forest-fires surveillance system based
on wireless sensor networks for South Korea Mountains,”
IJCSNS 2006, 6, 124-130.
