International Journal of Soft Computing and Engineering (IJSCE)

ISSN: 2231-2307, Volume-3, Issue-1, March 2013

Multi-Sensor Based Forest Fire Detection

System
Parul Mohindru, Rajdeep Singh

Abstract— Wireless Sensor Networks (WSNs) have become Fuzzy techniques for treating tentative qualitative
hot topic in field of research in recent days. In day to day life we information including fuzzy arithmetic and mathematics,
come across many problems which left unresolved by humans, so fuzzy set theory, fuzzy logic, fuzzy decision making and
at that time we think of collaborating human knowledge with
fuzzy control. Rule based fuzzy operators are a new class of
technology to eradicate the problems. The efficient approaches of
forest fire detection using multi-sensors describes one of the
operators exclusively considered in order to relate the
wireless sensor network applications for detecting a parameter principles of estimated interpretation as shown in Fig 1.
that is fire and reporting it to the base station to save the humans
and wildlife from destruction which is caused by the fire. The
effort offered in this paper conveys the idea of implementing
Fuzzy Logic on the information collected by multiple sensors.
Thus multiple sensors are used for detecting probability of fire
with variations during different time in a day. Each sensor node
senses Temperature, Humidity, Light Intensity, CO Density and
Time for calculating probability of fire. It will improve precision
of the detection system, as well as false alarm rate will be
reduced.

Keywords— Forest Fire Detection, Fuzzy Logic, Sensor

Networks.

I. INTRODUCTION
Wireless sensor network consist of large number of sensor
nodes deployed densely in the environment. It consists of
large number of sensor nodes which collect data and
disseminate the collected data towards sink. This concept
can be used for many crucial applications where manual
involvement of humans is difficult. The summer 2007
showed to the world how big is the devastation caused by Fig 1: Fuzzy Logic
forest fires [1]. So, immediate notification of the fire is the
most critical issue in forest fire detection systems in which II. REVIEW OF LITERATURE
WSN can provide real time fire detection with high
Al-Abbass Y. Al-Habashneh et al. proposed three MAC
accuracy. As forest fires may occur in wild areas due to
protocols for forest fire detection [3]. Two were based on
human carelessness or change in environmental conditions.
carrier sense multiple access and one was based on time
They cause threats to ecosystem, creating economical
division multiple access. Results showed that there is no
damage as well as endangering people’s and animal’s lives.
superior protocol which outperforms others in terms of
Spread features of forest fires show that the fire fighter
power consumption, delay and complexity, however trade-
center should be aware of the threat in at most 6 min after
off does exist.
the start of the fire in order to put out a fire without making
Çağdaş Döner et al. have given a design in which nodes
any permanent damage in the forest [2]. For efficient use of
are randomly deployed with each node having temperature
Wireless Sensor Networks, all nodes are divided into small
sensor attached to it, simple localization technique is used to
groups called clusters. Each cluster has its own cluster head
get coordinates of the nodes and a distributed spanning tree
which communicates with base station. In this way there
for the messaging infrastructure [4].
will be no need of sending data from each sensor node to the
ArnoldoDíaz-Ramírez et al. proposed two algorithms for
base station. Sensor nodes will send their data to respective
forest fire detection [5]. The proposed algorithms are based
cluster head which will save lot of energy that is pointed as
on information fusion techniques. The first algorithm uses a
one of the main drawbacks in sensor networks is battery
threshold method and the other uses Dempster- Shafer
power. Therefore, in this paper an event detection
theory. Both algorithms reported false positives when the
mechanism is projected using fuzzy rule based system.]
motes were exposed to direct sunlight. However, if the
motes are covered to avoid direct sunlight exposure, the
number of false positives may be reduced.
Manuscript received on March 2013. YunusEmreAslan et al. proposed a comprehensive
Parul Mohindru, Department of Computer Science & Engineering, framework for the use of wireless sensor networks for forest
Punjab Technical University, Jalandhar, India. fire detection and monitoring [2]. The framework considers
Rajdeep Singh, Department of Computer Science & Engineering,
Punjab Technical University, Jalandhar, India. all parts of the life cycle of a wireless sensor network system

142
 Multi-Sensor Based Forest Fire Detection System
that is specialized for forest fire detection. While C. Fuzzificaion
considering the early detection of forest fires as the major Fuzzy Logic was inroduced in 1965 by Lotfi A. Zadeh,
goal, the system is constructed that regards the low energy who was professor in computer science at the University of
capacity of sensor nodes and the difficult environmental California in Berkeley. Fuzzy Logic is a multivalued logic
conditions that may adversely affect the network operation that allows intermediate values to be defined between
and performance. conventional evaluations like true/false, yes/no, high/low,
A.K. Singh and Harshit Singh proposed system for fire etc. Fuzzy Logic has emerged as a profitable tool for the
detection using wireless sensor networks and fuzzy type-2 controlling and steering of systems, complex industrial
logic [6]. Fuzzy gives the best result in such cases as the processes, household, entertainment electronics, as well as
gathered data acts as input for the system on which Fuzzy for other expert systems and applications. The aim is to use
logic was implemented to calculate the probability of the fuzzy sets in order to make computers more ’intelligent’,
forest fire. therefore. Fuzziness describes event ambiguity and
impreciseness of linguistic terms. Fuzzy logic fits best in
III. IMPLEMENTATION applications where the variables are continuous and/or
A. Clustering mathematical models do not exist or traditional system
models become overly difficult. WSN is typically used to
Nodes are divided into clusters and each cluster will have supervise some parameters of an environment process. The
their own cluster head. The simulation for clustering of atmospheric events are multifaceted, confusing and
nodes is done in MATLAB. Fig 1 shows random
imprecision embedded in their nature. Consequently, a fuzzy
deployment of nodes is area of interest whereas Fig 2 shows
based approach is a feasible option. The model of fuzzy
each cluster having its own cluster head.
logic system consists of fuzzification, fuzzy rules, fuzzy
inference system and defuzzification process [7].
The Fuzzification is the first step in the fuzzy inferencing
process. This involves a domain transformation where crisp
inputs are transformed into fuzzy inputs. Crisp inputs are the
exact inputs measured by sensors and passed into the control
system for processing, such as temperature, position,
pressure, rpm's, etc. Each crisp input that is to be processed
by the Fuzzy Inferencing Unit has its own group of
Fig 2: Clustering of nodes membership functions or sets to which they are transformed.
The group of membership functions exists within a universe
B. Event Detection Mechanism of discourse that holds all relevant values that the crisp input
Event detection is a popular service in environmental can possess [7]. In our fire detection algorithm temperature,
monitoring and object tracking applications. Ambulatory humidity, CO density, light intensity, time acts as input
medical monitoring, vehicle tracking, military surveillance, fuzzy variables. The Probability of Fire is the output
forest fire detection are some sample applications that event variable. The membership functions LOW, MEDIUM and
detection plays a key role. The popularity of this service is HIGH are defined on temperature, light intensity, humidity
not limited to the application layer. Several wireless senor and CO density whereas BEFORE NOON, NOON, AFTER
network middlewares provide the required primitives, such NOON for Time as shown in Fig 4 to Fig 8. VERY LOW,
as event notification to facilitate event detection tasks in LOW, MEDIUM, HIGH and VERY HIGH are defined on
various applications. Fire detection sensor networks are Probability of fire as shown in Fig 9.
deployed to set an alarm if a fire starts somewhere in the
monitored area. So, whenever fire is detected the nodes will
send information to their respective cluster head. This will
reduce processing cost at each node which can be used for
further computations. This can be simulated in MATLAB
using event detection mechanism as shown in Fig 3.

Fig 4: Membership Function for Temperature

Fig 3: Event Detection Mechanism

Fig 5: Membership Function for Light Intensity

143
 International Journal of Soft Computing and Engineering (IJSCE)
ISSN: 2231-2307, Volume-3, Issue-1, March 2013
inferencing used. AND, OR and NOT are three Zadeh
Operators. The label of the consequent is associated with its
output membership function [7]. There are 5 input variables
that consist of 3 fuzzy linguistic variables. Therefore, the
total 3*3*3*3*3 = 243 rules are used, which are all possible
combinations of the input variables. Thus some of the
example rules in this rule based system are as follows:
IF Temperature is LOW and Light Intensity is LOW and
Humidity is HIGH and CO is LOW and Time is Before
Fig 6: Membership Function for Humidity Noon THEN Fire probability is VeryLow.
IF Temperature is HIGH and Light Intensity is LOW and
Humidity is HIGH and CO is HIGH and Time is Noon
THEN Fire probability is VeryHigh.
E. Defuzzificaion
Defuzzification involves the process of transposing the
fuzzy outputs to crisp outputs. There are a variety of
methods to achieve this such as method of averaging which
is also known as the Center of Gravity method or COG. It is
a method of calculating centroids of sets. In practice,
Fig 7: Membership Function for CO Density
defuzzification is done using centroid method [7].

IV. RESULTS
The results are crisp number from the set [0, 100] which
is the scale of fire probability. Decision making could be
done on the basis of output obtained by making few
conditions.
1. If the output is between 0 and 20, then the probability of
fire is very low.
2. If the output is between 21 and 40, then the probability
Fig 8: Membership Function for Time of fire is low.
3. If the output is between 41 and 60, then the probability
of fire is medium.
4. If the output is between 61 and 80, then the probability
of fire is high.
5. If the output is between 81 and 100, then the probability
of fire is very high.
For example if the input is [89 868 92 76 14], where first,
second, third, fourth and fifth element of input matrix
represents Temperature, Light Intensity, Humidity, Carbon
Monooxide Density and Time respectively, the output is 90.
Fig 9: Membership Function for Fire Probability Thus, the probability of fire is VERY HIGH at Time 2 PM
as shown in Fig 10. Control surface of fire probability based
D. Fuzzy Inference Engine on different input parameters are shown in Fig 11 to Fig 14.
Fuzzy rules are IF-THEN rules introduced by Zadeh.
They may be understood as partial imprecise knowledge on
some crisp function. Gaussian, triangular, and trapezoidal
functions are the most commonly used membership
functions. In the fuzzy rules, triangular and trapezoidal-
shaped membership functions are used for the variables to
simplify the computations. Rule evaluation consists of a
series of IF- Operator-THEN rules. A decision structure to
determine the rules require familiarity with the system and
its desired operation. This knowledge often requires the
assistance of interviewing operators and experts. The most
commonly used fuzzy inference technique is Mamdani
method. Fuzzy rule base drives the inference system to
produce fuzzy outputs, which are defuzzified to get system
outputs. There is a strict syntax to these rules which is:
IF antecedent1 OPERATOR antecedent2, THEN
consequent1 OPERATOR consequent2.
The antecedent consists of input variable IS label, and is
equal to its associated fuzzy input or truth value µ(x). The
consequent consists of output variable IS label, its value
depends on the Operator which determines the type of Fig 10: Fuzzy Inference Rule Based Fire Detection System

144
 Multi-Sensor Based Forest Fire Detection System
V. CONCLUSION
In this paper, we projected an event detection mechanism
for detection of fire and fuzzy approach for fire calculating
probability of fire using multi-sensors. Our proposed forest
fire detection handles the vagueness present in the statistics
successfully and gives the finest results with very low false
alarm rate. The decision based on this approach is more
precise as it gives accurate results with variation of time and
other physical parameters. The membership functions and
the parameters can be changed and modified as required.
Rules also could be altered and adjusted according to
parameters for further work on this model.

REFERENCES
Fig 11: Surface view of fire probability with respect to [1] Bayo, D. Antolín, N. Medrano, B. Calvo, S. Celma,” Early detection
Temperature and Time and monitoring of forest fire with a wireless sensor network system”,
Procedia Engineering, Volume 5, 2010, Pages 248-251, ISSN 1877-
7058.
[2] YunusEmreAslan, IbrahimKorpeoglu, and OzguUluso, “A framework
for use of wireless sensor networks in forest fire detection and
monitoring,” Science direct, vol 36 pp.1-12, Mar 2012.
[3] Al-Abbass Y. Al-Habashneh, Mohamed H. Ahmed, and Taher
Husain, “Adaptive MAC Protocols for Forest Fire Detection Using
Wireless Sensor Networks ,” in proceeding of IEEE electrical and
communication system engineering confrence’, 2009,pp.329-333.
[4] ÇağdaşDöner*, GökhanŞimşek, Kasım Sienna Yıldırım, and
AylinKantarc, “Forest Fire Detection with Wireless Sensor
Networks,”in academia Computer Engineering Department, Ege
University’, 2010, pp.107-109.
[5] ArnoldoDíaz-Ramíreza,*, Luis A. Tafoyaa, Jorge A. Atempa, and
PedroMejía-Alvarezb, “Wireless Sensor Networks and Fusion
Information Methods for Forest Fire Detection✩,” in Science direct
on Electronics Engineering and Computer Science ’,2012, pp.69-79.
[6] A.K. Singh, and Harshit Singh, “Forest Fire Detection through
Wireless Sensor Network using Type-2 Fuzzy System”,International
Fig 12: Surface view of fire probability with respect to Light Journal of Computer Applications,” vol 52– No.9, pp. 19-23,August
Intensity and Time 2012.
[7] M. Ganesh, Introduction to fuzzy sets and fuzzy logic, Prentice Hall
of India Private Limited, New Delhi, 2006.

Parul Mohindru was born in 1989, in Jalandhar, India. She is pursuing

M.Tech in CSE (Networking) from Punjab Technical University,
Kapurthala.. Her research interests include Wireless Sensor Networks and
Mobile Ad Hoc Network (MANET).
Rajdeep Singh was born in India. He is currently pursuing his PhD in
Wireless Sensor Networks from Punjab Technical University. He is
currently working as Head of Department in Punjab Institute of
Technology, Kapurthala.

Fig 13: Surface view of fire probability with respect to

Humidity and Time

Fig 13: Surface view of fire probability with respect to CO

Density and Time

145