Prabu, International Journal of Advanced Engineering Technology E-ISSN 0976-3945
Research Paper
AN EFFICIENT SURVEILLANCE SYSTEM TO DETECT
ELEPHANT INTRUSION INTO FOREST BORDERS USING
SEISMIC SENSORS
Dr. M. Prabu
Address for Correspondence
1
Department of Computer Science and Engineering, Adhiyamaan College of Engineering, Dr. M. G. R. Nagar,
Hosur, Tamil Nadu, India - 635109 India
ABSTRACT
Human-Elephant conflict is the major problem in the forest borders with large elephant herds. In this paper, an automated
system to detect the intrusion of elephants into the human habitat in forest borders is proposed. The seismic signals
generated by the movement and other sounds made by the elephants are received by the system and is transmitted to the
central processor along with the GPS position of the receiver node. The received signals contains some noise and it is should
be removed using algorithm. After removing the noise signal is matched with the database patterns. On match of a stored
pattern, the system generates a SMS to forest authorities. The proposed is very efficient with good computation power and
low cost.
KEYWORDS: Human – Elephant Conflict Seismic Sensors
INTRODUCTION d) Degradation of climate in the forest region.
Indian sub-continent is diverse in biological and non- e) Artificial water resources constructed near
biological aspects. The diversity in the flora of the the habitat of the elephants.
forest areas in the country attracts a wide range of f) Occupying traditional migration paths of
animals to have it as the habitat. One of the most elephants by human constructions.
prominent species in the Indian forest is the elephant. g) Trenches constructed improperly, damaged
India is the major habitat for 60 percent of the Asian or not maintained.
elephants (Elephas Maximus) of the total estimated h) Periodic migration of the herd to other
population of 40,000-50,000. Asian elephants lose comfortable habitats.
their habitat as the forest areas are rehabilitated to i) Forest fire causing all the wildlife to lose its
human settlements, industries and agricultural lands. habitat.
These conversions lead to the increased shortage of With these factors contributing the intrusion there are
the natural food and water resources needed by the several systems and initiatives developed to reduce
herd of elephants. Shortage and fragmentation of human-elephant conflict around the globe. Some of
habitat makes the herd of wild elephants to enter into the traditional, conventional and experimental
the human habitat in the forest borders and roadways. methods proposed by the humans are discussed
There exists a severe damage to crops, human lives below:
and also the elephants. The crop damage is costlier as 1) Air guns: Air guns produce sudden shock
the herd damages the larger portion of the vegetation waves on air producing burst sounds that
when it moves into the farm lands. This loss by annoys elephant herd.
elephants can be mitigated by taking precaution 2) Non - electric fences: Simple fencing
measures on intrusion of the herd. But the detection technique used to block the path of the
and tracking of the herds is hard due to their size and expected intrusion.
complex nature of the movement. The elephants are 3) Electric fences: Electric fences uses low
considered endangered species as it faced heavy loss voltage AC power supplies to avoid the
of lives by poachers. The poaching for ivory of the intruding animals into habitat.
elephants, attacks made by humans to protect their 4) Chilli rope fences: Two stringed fences are
farm lands, captivation of baby elephants and constructed around the vegetation or habitat
accidents are some common reasons of the reduction with the mixture of dry chilli powder and
of elephant population. This paper is structured as engine grease applied on the strings.
follows: In section 2 of this paper, an analysis of the 5) Loud alarms: Alarms producing loud noise
previous works on elephant detection with various triggered by a trip wire is used to defer the
techniques has been elaborated. Section 3 discusses elephants back into forest.
the study area of the system. Section 4 formulates the 6) Chilli smokes: Animal dung’s and/or red hot
problem statement and motivation for the proposed chillies mixed and burnt to produce pungent
system. Section 5 describes the proposed smoke clouds blown in the direction of the
methodology of the problem statement. Section 6 elephants to raid them.
discusses the experimental results of the system. 7) Watch towers: Watch towers of certain
Section 7 gives the conclusion of the proposed work. height is constructed and used for
RELATED WORK surveillance of elephant’s intrusion to avoid
Based on the study of the previous research works the damage caused by them.
and the field research the factors contributing for the 8) Solar powered torches: Solar torches are
intrusion of elephants along the forest borders are specially designed to produce powerful light
summarized below: and are used to raid elephants in the forest
a) Vegetation on farm lands with no protection borders.
fences attracts the elephants. 9) Trenches: Trenches of considerable width
b) Damage of fences constructed in the forest and depth is constructed in the forest borders
borders either by humans or natural ailment. to prevent the intrusion of the elephants.
c) Shortage of food resources in the forest
perimeter.
Int J Adv Engg Tech/Vol. VII/Issue I/Jan.-March.,2016/166-171
Prabu, International Journal of Advanced Engineering Technology E-ISSN 0976-3945
10) Fire: A considerable amount of fire is burnt simulation model the results of the real time
on the expected intrusion path of elephants implementation model is not elaborately discussed.
to defer them back into forest. Ranjit Manakandan et.al. presented a case study on
11) Fire crackers: Crackers producing loud noise the dispersal and movement of the elephants between
is used on intrusion of elephants to raid the Hosur – Dharmapuri Forest Division and
them. Koundinya Wildlife Sanctuary7. The migration of the
12) Throwing arrows / stones: On intrusion of elephants is due to the various parameters like
elephants, sharp objects on considerable vegetation, scarcity of shade, scarcity of grass, water
mass like arrows or stones are thrown on scarcity, forest fires and human elephant conflicts.
them to raid the herd. The study also concluded that the aforementioned
With these manual intervention methods, there are conservation issues there is a decline in population of
several automated system as a result of scientific the elephants in KWS as most of the elephants have
advances are developed to detect the elephants. Some translocated to the Hosur – Dharmapuri Forest
of them are discussed below: Divisions. The authors also suggested evaluation and
The authors, C. Arivazhagan and B. Ramakrishnan restoration of the forest division to improve the
presented a paper on the conservation of elephants elephants count in the forest region.
with the focus on connectivity of elephant’s habitat1. James D. Wood et. al. presented a technique to
The authors aimed to increase the population of estimate the population of the elephants in Central
elephant and protect the habitat of elephants from Africa where it is hard to enter the dense forest8.
degradation and fragmentation. The study was done Since the forest is very denser aerial surveys are not
in southern part of the Indian sub-continent. possible so they developed a system using seismic
Pieter I. Oliver et. al. discussed a method to detect sensors to estimate the total population of large
the elephants using the dung2. With the dung decay mammals in a specific region. The seismic detection
rates and distance sampling techniques, they have system has successfully detected large mammals with
detected and estimated the population size, age group higher accuracy.
in Southern Mozambique region. S J Sugumar and R Jayaparvathy presented a system
Prithviraj Fernando et. al. proposed a solution to to detect elephant intrusion along the forest borders
track elephant movement patterns in their habitat by using real time imaging9. The proposed system is an
direct observations to conserve elephants and to automated unsupervised elephant image detection
avoid human-elephant conflict3. They have used system that reduces human elephant conflict in the
radio collars to get GPS locations of elephants every context of elephant conservation. The elephant’s
4 hours. Data received from the individuals are image captured in the forest border areas and is sent
interpreted in an Excel sheet with statisticXL 1.8 to a base station for processing the image using Haar
add-in. The data is then mapped on the toposheets to wavelet and image vision algorithm. The system also
visualize the collected data. proposed an optimized distance metric for lesser
Tucker Balch has proposed a methodology for animal retrieval time compared to Euclidean and Manhattan
tracking using RFID system4. The RFID tags are algorithms. A paper on a low cost infrasonic
detected and interpreted by a control hub and the recording system is proposed in the year 201310. The
locations are determined. It has a drawback of short system is built to record the infrasound calls made by
range detection and it has low update rate of the elephants. With the recorded infrasonic calls the
locations. elephants can be detected efficiently with low cost
S J Sugumar and R Jayaparvathy have proposed an compared to other systems. The system also records
analytical model for surveillance and tracking of infrasounds generated by other sources of the
elephant herds using a three - state Markov chain5. environment which can be used for research
The design gives the migration pattern of elephants purposes.
and behavior of the elephants over the whole year in STUDY AREA
different climatic periods of three villages near forest The Krishnagiri and Dharmapuri districts of Tamil
borders. The study is conducted in the Coimbatore Nadu hold Hosur Forest Division and Dharmapuri
Forest Division which falls in the dense forests of the Forest Division. The forest divisions lies very much
Western Ghats of India. The intrusion detection closer to the electronic hub of India- Bangalore.
system is developed based on pattern of movement of Developmental activities of these regions have severe
the herd to detect the intrusion of the elephant herds impact on the forest area which includes the
into the locality or habitat of the humans. On construction of roads, railways, human settlements
intrusion of the elephants the system warns the forest and commercial constructions like special economic
officials with an alert message. The system is zones11.
hardware implemented as a prototype to detect
intrusion of elephants into the forest borders.
Ramkumar R et. al. has developed a simulation
model called ASRET to reduce the Human-Elephant
Conflict (HEC)6. This system generates early
warning to prevent conflict between humans and
elephants. The system is a simulation model
simulated in MATLAB Simulink with the central
processing unit, primary unit and secondary unit. The
signals from the sensor nodes are processed by the
central processing unit and the secondary unit is
responsible for the transmission of signals to and Figure1. Study area of Krishnagiri and Dharmapuri
from the sensor nodes. Since the system is a District of Tamil Nadu, India
Int J Adv Engg Tech/Vol. VII/Issue I/Jan.-March.,2016/166-171
Prabu, International Journal of Advanced Engineering Technology E-ISSN 0976-3945
forest borders. It is reported that around 20 people are
killed by elephants in the span of three years from
2013 to 201513. Also there are unreported deaths
which show the fact that above 40 people loses their
lives by elephant attack.
As the transport ways of roads and rail lines
connecting major cities to the villages lies in the
crucial elephant crossing regions there exists frequent
accidents when the elephants try to cross these paths.
The railway track from Hosur to Dharmapuri of
length 75 km is the major section of the Southern
Railways experiencing elephant accidents. The
Figure 2. Enlarged map of the study area showing number of elephant’s death by train hits in this
detailed range of forest divisions and elephant intrusion section in the year of 2015 is reported as 4. Also the
areas. people in the forest borders go into forest to collect
The two forest divisions of Hosur and Dharmapuri logs, fire sticks, medicinal plants, fruits, cattle
have most of the reserve forests of the state, grazing and to collect natural honey where there is a
including Sanamavu Reserve Forest, Melumalai higher risk of elephant’s attack. There is a yet another
Reserve Forest, Anusonai and Udedurgam Reserve risk of elephant attack while the farmers try to ride
Forest, Veppanahalli Reserve Forest, Rayakottai elephants from their farm lands. There are unreported
Reserve Forest depicted in figure 1 and figure 2. deaths by both of the above mentioned situations.
The forest area of the Krishnagiri district is 2, 02,409 PROBLEM FORMULATION
ha which constitutes 9.6% of the state and the forest The objective is to detect the elephants trying to
area of the Dharmapuri district is 1, 64,177 ha which intrude the human habitat in the forest borders. The
constitutes 7.8% of the state. The forest division is an system uses seismic sensors to detect the seismic
extension of the Bannerghatta National Park which signals generated by the movement of the elephants.
lies to its north and also Cauvery Wildlife Sanctuary The signals are processed; features are extracted and
to its south. The Forest division also covers the are matched with stored patterns of elephants in the
Sathyamangalam forest range and extends well to the database. Also the signals are filtered to an extent
Nilgiris. The Forest division serves as a meeting using the FastICA algorithm. The signals are
point of the Western Ghats and Eastern Ghats visualized to see the processed and filtered seismic
forming a vital link to the elephant’s movement signals using the spectrogram. From the processed
between the greatest forests of South India. signals, only a segment of the seismic data detected is
The Krishnagiri and the Dharmapuri districts have used for match. The segment is called as window. On
forest cover of 3,034 km2 of the total geographic area successful match of the recorded pattern, the central
9,622 km2, which constitutes 241 km2 of very dense processor will generate an alert message with the
forest, 1,078 km2 of moderately dense forest and GPS position of the sensing node.
1,715 km2 of open forest12. The forest range lies at an Let Sn be the segmented seismic record sample where
elevation of 300 m to 1400 m above the mean sea n = 0, 1, 2… N – 1, and recorded seismic segments
level and most of the forest region lies between 11º ranging from x= 1, 2, 3… X. Then the signal
12' N to 12º 49' N latitude and 77º 27' E to 78º 38' E produced along with the noise can be given as
longitude. There is a greater diversity in the climatic follows:
conditions due to the varied altitude of the various
regions of the forest division (It is recorded that the
eastern and western part of the forest regions
experiences completely contrast climatic conditions). where be the noise signal added to the original
The forest division falls in the chain of Eastern Ghats signal. The is the signal generated by the recorded
and the areas in and around Denkanikottai has been seismic segment and the corresponding seismic
declared elephant reserve due to large number of record sample. Since the noise signal is not easily
elephant’s death in the past decade. The crucial predictable it is kept separately.
elephant crossings over human habitations lie at MATERIALS AND METHODS
different regions of the forest divisions. These forest Independent Component Analysis for Noise
divisions hold most of the roads and rail lines to Removal:
connect various towns, cities and the district The Independent Component Analysis is an effective
headquarters through the forest, also most of the and efficient signal processing technique that can
villages are positioned near the forest boundaries and separate noise signals from the original source signal.
the forest ranges. The major elephant crossings in the The ICA splits set of independent sources based on
roadways and rail lines are Hosur to Denkanikottai of their statistical independency from a noisy signal. In
length 29.7 km, Hosur to Kelamangalam of length 20 the mathematical model, it can be represented as:
km, Kelamangalam to Uddanapalli of length 14 km,
Denkanikottai to Anchetty of length 24.8 km, where x=[x1, x2,….,xn]T is observed signal,
Jawalagiri to Anchetty of length 40.2 km, s=[s1,s2,…sn]T is the source signal, A defines the m x
Hogenakkal to Urigam of length 58.2 km. These n mixing matrix and n is the noise signals mixed up
regions are crucial to get affected by the elephants same as intrinsic components.
and there exists loss of lives of both humans as well The ICA algorithm takes linear transformation of the
as elephants. The vegetation of the forest borders and equation to estimate the source components based on
the natural elephant habitat acquired by humans assumption of independency.
attracts elephants to enter into habitat of humans the
Int J Adv Engg Tech/Vol. VII/Issue I/Jan.-March.,2016/166-171
Prabu, International Journal of Advanced Engineering Technology E-ISSN 0976-3945
The figure 3 shows the migration logic based on the
T - availability of the food resources and comfortable
where z=A w, w is the estimator of the row matrix A
1 environment. For the logical variable term Very Less:
and y is the best estimation of the source signals.
With the efficient processing power of ICA, the Hot is 1, it shows that the rainfall is very less and
FastICA algorithm is based on the fixed-point climate is very hot: on this state the elephant has
algorithm that can process independent components. higher probability of migration. For the logical
The FastICA algorithm is shown in the following variable terms less: Dry is 1, it shows that the rainfall
steps: is less and climate is dry: on this state the elephant
1) Center the data to make its mean zero: has higher probability of migration. . For the logical
variable terms Average: Moderate, High: Cool, Very
High: Cold is 1, it shows that the rainfall is normal
2) Calculate the covariance matrix
and has good climate: on this state the elephant has
zero probability of migration.
Where E is the orthogonal matrix of the eigen vectors
of and D is the diagonal matrix of the eigen
values,
3) Whiten the data to produce the whitening
vector
Where V is the whitening matrix, Figure 3. Migration analysis of elephants with respect to
the available resources and climate.
4) Choose an initial vector of unit norm The Table 1 depicts the possibilities of the migration
5) Update situation on existence of variable combinations of
climate and rainfall. As rainfall and climate are major
6) Normalize parameters for the availability of food for the herd.
Table 1: Analysis of Expected Migration State
7) Check convergence, if not converged, go back to
step 5
8) If converged, calculate the independent component
Fuzzy Logic to analyze Elephant Migration based
on Resources and Climate:
Availability of comfortable environment and
resources demands the elephant herd to decide
whether to migrate to other region or not. It is a
complicated analysis to predict the availability of Analysis of Elephant Migration on the whole year
resources and environment as the habitat perimeter of using Mobility Markov Chain:
the herd is very large. To analyze the availability of The analysis of migration of the elephant herd during
the resources and good environment fuzzy logic is various periods of year will help to alleviate the
used. Although the Hosur – Dharmapuri forest Human Elephant Conflict by taking counter measures
divisions have better climate and food resources for precisely. In general, the whole year can be divided
the elephants there is a shortage of food resources at as Hot, Dry, Moderate, Wet and Cold Seasons. To
various periods of the year. analyze the movement of the herd previous records of
The availability of the resources is based on the two movement of herds to various regions of the Hosur –
factors: rainfall and climate. The logic takes these Dharmapuri Forest Divisions are collected.
two factors as input parameters and the resultant or Mobility Markov Chain (MMC), a probabilistic
the output variable is the decision of migration. The automaton is used for analyzing the migration
rainfall and climate is chosen as parameter as these regions called Point of Interest (POI) 14. The Mobility
factors influences the availability of food resources in Markov Chain is build with:
the forest like greens, trees, fruits. A set of states, representing the POIs where
The input variables, rainfall and climate are grouped frequent intrusions are recorded in the
to form the variable terms: Very Less: Hot, Less: villages or regions along the forest borders.
Dry, Average: Moderate, High: Cool, Very High: Here the set of states is P={DRF, SRF,
Cold. KRF, JRF, URF, KORF, ARF, URRF,
The Migration will be based on the variable terms PRF, OTHER} where DRF represents the
formed above. E.g. For the variable terms Very Less: Dharmapuri Reserved Forest, SRF
Hot, Less: Dry the migration probability is very high represents the Sanamavu Reserved Forest,
and for the other variables there is a least probability KRF represents the Kelamangalam
of migration. i.e., Reserved Forest, JRF represents the
IF (Very Less: Hot) is 1 THEN Migration is TRUE Jawalagiri Reserved Forest, URF
IF (Less: Dry) is 1 THEN Migration is TRUE represents the Uddanapalli Reserved
IF (Average: Moderate) is 1 THEN Migration is Forest, KORF represents the Kollatti
FALSE Reserved Forest, ARF represents the
IF (High: Cool) is 1 THEN Migration is FALSE Anchetty Reserved Forest, URRF
IF (Very High: Cold) is 1 THEN Migration is represents the Urigam Reserved Forest and
FALSE
Int J Adv Engg Tech/Vol. VII/Issue I/Jan.-March.,2016/166-171
Prabu, International Journal of Advanced Engineering Technology E-ISSN 0976-3945
PRF represents the Pennagaram Reserved shows the transition probabilities of the herd
Forest, the last state represents the other migration throughout the year. Here the probability
infrequent POIs of migration. of moving from DRF to SRF is 0.33, but the
A set of transitions, representing the probability of the vice- versa will not be the same, as
movement of herd from a region to another there exists different migration counts from and to
region. The set of transitions is T={t1,1, t1,2, the POIs. The Denkanikottai RF is the major habitat
t1,3, …,ti,j}, where ti,j corresponds to the of the elephants as they roam in and around the forest
probability of moving from a state of POI regions. The herd moves to the other forest regions
to the other. For a path with no transition, with the probability of 0.33 for Sanamavu, 0.34 for
zero probability is marked and is not shown Kelamangalam, 0.09 for Jawalagiri, 0.03 for other
in the graph. For the transitions from a state forest regions, and 0.21 for Denkanikottai RF itself.
to other state the sum of probability of all The sum of probabilities leaving the DRF will be
the transitions should be equal to one. 1.00, as the probability should be exactly equal to
i.e., . one for any event. The probability that a herd
The weights are associated with the states migrates from SRF to KRF is maximum compared to
in the form of the integer. The weights are all other regions which is marked as the frequently
assigned to a state based on the duration used path. The probability of movement from KRF to
and total counts of migration events DRF might be larger but during dry seasons the herd
happened in the POI. The last state moves from Hosur Forest Division to Dharmapuri
OTHER has no weight assigned as it Forest Division.
contains only the infrequent regions other The path extends from KRF to PRF through JRF,
than the frequent POIs. The probability of ARF and URRF. The herd moves and stays in ARF
the migration is computed on the basis of for longer durations if there appears a comfortable
the number of migrations occurred on climate. The elephant herd rolls around the forest
average in the last five years. With the ranges of ARF, URRF and PRF often as these forest
transition probabilities computed the figure ranges lies closer to each other. The probability of the
4 is drawn. elephant herd moving from the DRF to other forest
regions is 0.03 which is comparatively lesser
compared to all other migration probabilities showing
that the elephants have lesser chances of migration to
those POIs. The transition matrix in table 2 depicts
the transition probabilities depicted in the Mobility
Markov Chain State Diagram. With the analysis of
the data, it is found that there are different paths of
migration of the herd between Hosur and Dharmapuri
Forest Regions. There are multiple migration paths:
(i) DRF > SRF > URF > KRF > JRF > ARF > URRF
> PRF (ii) DRF > KRF > KORF (iii) DRF > JRF >
ARF > URRF > PRF (iv) DRF > SRF > KRF > JRF
> ARF > URRF > PRF (v) DRF > KRF > JRF >
ARF > PRF. There are other paths that a herd can use
Figure 4. Mobility Markov Chain for Elephant
Migration analysis in Hosur – Dharmapuri Forest to migrate between the forest ranges, but they are
Division used in very lesser frequencies. The figure 5 shows
The state diagram is designed in the decreasing order the block diagram of the proposed detection system.
of the weights assigned. The higher weights represent
the frequent POIs of the elephant herd. The figure 4
Table 2: Transition Matrix for Migration Analysis of the Elephant Herd in
Hosur – Dharmapuri Forest Division
Figure 5. Block diagram of the detection system
Int J Adv Engg Tech/Vol. VII/Issue I/Jan.-March.,2016/166-171
Prabu, International Journal of Advanced Engineering Technology E-ISSN 0976-3945
the implemented algorithms mainly use climate and
EXPERIMENTAL RESULTS AND DISCUSSION environment to track the herd. With this system the
A simulation model is developed to detect the detection rate of the elephant is higher than any other
elephant intrusion along forest borders. The system techniques with the detection rate of 93.75 percent.
takes input from the seismic sensor in the form of Further study will involve feature normalization, and
raw seismic waves produced as a result of movement feature classification with the help of Principal
in surface of the land. The received waves are Component Analysis algorithm. The algorithm will
processed using a signal processing module lead to classification of different species on move.
developed as a part of the simulation model. The REFERENCES
module covers various levels of processing to detect 1. Arivazhagan, C. and Ramakrishnan, B., Conservation
the target pattern in the recorded waves. The target perspective of Asian Elephants (Elephas maximus) in
Tamil Nadu, Southern India, International Journal of
pattern is the seismic signals of the elephant herd. Biological Technology (IJBT) An International Journal
The received signals are processed with Fast Fourier of Biological Technology (Gayathri Teknological
Transform (FFT) and Fast Independent Component Publication), 1(Special Issue): 17, 18.
Analysis (FastICA). The FastICA algorithm mitigates 2. Olivier, P.I., Ferreira, S.M. and Van Aarde R.J., Dung
survey bias and elephant population estimates in
noise signals in an efficient manner. With the use of southern Mozambique, African Journal of Ecology,
FastICA algorithm, the raw signals received from the 2009,47, 202–213.
seismic sensor can be filtered. The filtered signals are 3. Fernando, P., Leimgruber, P., Prasad, T. and Pastorini,
passed to the next level to extract the features in the J., Problem-elephant translocation: Translocating the
problem and the elephant?, PLoS ONE 7:e5091,2012.
signal. The extracted features will have the original 4. Balch, T., Dellaert, F., Feldman, A., Guillory, A., Isbell,
patterns of the footfalls of an elephant or the herd. C., Khan, Z., Pratt, S., Stein, A. and Wilde.H, How
The signals are segmented to form records of variable multi-robot systems research will accelerate our
length. The variable length records are given as input understanding of social animal behavior. Proc. of the
IEEE, 94, 1445-1463.
to the pattern recognizer that matches stored patterns 5. Sugumar, S. J. and Jayaparvathy, R., An early warning
of the elephants in the database with the Dynamic system for elephant intrusion along the forest border
Time Warping Algorithm (DTW) (300 samples of areas, Current Science, 2013,104, No.11, 1515-1526.
preprocessed seismic signals of footfalls of the 6. Ramkumar, R., Sanjay Deb., Rajanna .K .M., An
automated system for remote elephant tracking to
elephant herd and elephants are stored in database). If reduce human elephant conflict, International Journal of
the pattern recognizer produces positive interrupt Computer Applications, 2014,24-27.
then the alert message generator generates alert 7. Ranjit Manakandan., Swaminathan, S., Daniel, J. C.
message to the stored GSM numbers. On negative and Ajay Desai, A case history of colonization in the
Asian elephant: Koundinya Wildlife Sanctuary (Andhra
interrupt from the pattern recognizer the system flush Pradesh, India), Gajah, 2010, 33, 17-25.
the received signals and wait for the next seismic 8. Jason Wood, D., Caitline O’Connel-Rodwell, E. and
wave to be processed. The figure 6 depicts the Simon Klemperer, .L, Using seismic sensors to detect
seismic signal produced by the footfall of the elephants and other large mammals: a potential census
technique, Journal of Applied Ecology, 2005,42, 587-
elephant. Table 3 presents the true positive rates 594.
evaluated after 150 tests. 9. Sugumar, S. J. and Jayaparvathy, R., An improved real
time image detection system for elephant intrusion
along the forest border areas, The Scientific World
Journal , 2014.
10. Girisha Durrel De Silva and Kasun De Zoyza, A low
cost infrasonic recording system, Wireless Sensor
Network Research Laboratory, University of Colombo
School of Computing, Sri Lanka, 2013.
11. Hosur – Dharmapuri Biodiversity Survey,
Figure 6. Time series of single foot fall of the elephant at www.asiannature.org/our-programmes/ environmental-
a sampling rate of 200Hz survey-and-assessment/hosur-%E2%80%93-
dharmapuri-biodiversity-survey, retrieved 24-09-2014.
Table 3: Test for True Positive Rate
12. Forestry an overview, http://agritech.tnau.
ac.in/forestry/forestry_introduction.html, retrieved 24-
09-2015.
13. Deccan Chronicle, http://www.deccanchro
nicle.com/150730/nation-current-affairs/ article/20-
killed-elephant-attacks-3-years, retrieved 12-09-2015
14. Gambs, S., Killijian, M.O. and Del Prado Cortez, M.N.,
Show me how you move and i will tell you who you
are, Transactions on Data Privacy, 4, no. 2, 103–
CONCLUSIONS 126,2011.
A seismic surveillance system was designed for the
recognition of large mammals like elephants. The
focus of the study was to analyze the usability of
seismic features for effective detection of the seismic
waves generated by the movement of the
pachyderms. Several sampling of the records were
carried out in order to find suitable geophone
configurations and to establish good samples of
seismic record pattern in the database, which will
afford reliable results. An analysis algorithm based
on Mobility Markov Model (MMM) was
implemented, which is able to locate a herd. Further
performance increase is achieved by exploring
differences in position and the climatic conditions, as
Int J Adv Engg Tech/Vol. VII/Issue I/Jan.-March.,2016/166-171
