Proceedings of the 2nd International Conference on Trends in Electronics and Informatics (ICOEI 2018)

IEEE Conference Record: # 42666; IEEE Xplore ISBN:978-1-5386-3570-4

Digital Monitoring of River Pollution Using Satellite

Images in Frequency Domain

Sethu Manickam, V. M. Priyadharshini, Dr. V .M. Anitha Rajathi

Dept. of Information Technology and Management Studies
Anna University – BIT Campus, Trichy
sethu.sxcs@gmail.com, priyadharshinivm@gmail.com, anithahrm@gmail.com

Abstract - The method proposed in this paper is can cause drastic repercussions, such as decimation
associated with the digitizing of river monitoring of aquatic species, and propagation pf epidemics.
through the application of Google Maps API and Water pollution assumes different forms, such as oil
the concepts of Background Subtraction as used spills, acid rain, heightened toxicity of fresh water,
in Digital Image Processing (DIP). The proposed and inordinate siltation. It is a significant menace,
scheme obtains roadmap and satellite images of a that can have a grave impact on our lives, owing to
its toxic nature.
target latitude and longitude and executes a
This paper focuses on digitizing the manner
simplified RGB based image subtraction in which a river body can be supervised via a novel
succeeded by the analysis of water colour in the scheme that utilizes the easily accessible Google
frequency domain following the application of Maps, as opposed to relying on high resolution
Discrete Cosine Transform (DCT) so as to satellite images that have been the staple of previous
monitor minor variations among the highly models for the purpose of extracting information
correlated pixel values. This novel scheme that can edify the user with relevant data regarding
consists of a four-step process for identification of the assigned river, such as the current pollution level
the differences and variations in obscured the and the changes that have been prevalent since the
images obtained in real time through Google last monitored time period. The method incorporates
Maps API and the assimilated images extracted a potent yet uncomplicated scheme, that accepts the
from the database over a profound period of latitude and longitude values as inputs and provides
statistical data calculated by the algorithm in the
time. The output is computed on the basis of the
form of a structured table that is lucid in
percentage of change as indicated by the colour
implementation, thus making the entire process
variations in the two images after the user-friendly.
implementation of the proposed method. 30
different co-ordinate values across the expanse of II. RELATED WORK
the river Kaveri were taken as prototypes on
which the aforementioned method was applied Numerous efforts have been exerted in the
and the RGB mean was calculated. The method field of satellite monitoring of water bodies for
can be scaled to different parts of the country by mitigation of water pollution. Ding et. al had
calibrating their respective databases with fashioned a very efficient model of satellite
historical imagery of maps. In this method, a monitoring with the aid of Landsat remote sensing
formula has also been devised for the image data [2]. It was utilized to produce essential
computation of the Water Quality Index (WQI) water quality parameters such as ammonia nitrogen,
of a given region, based on a physical laboratory and COD content correlation analysis. It reversed
ammonia nitrogen and COD concentration by
test conducted there, employing samples of water
utilizing a remote sensing satellite to execute a
from the given region. sweeping supervision of river ammonia, nitrogen
and COD pollution distribution. In another paper by
Keywords: DIP, Image Subtraction, DCT, Water Prochazka et. al, the use of real data processing is
Quality Index presented [1]. They are taken by satellites operated
by NOAA (The National Oceanic and Atmospheric
Administration). Their work is grounded on the
I. INTRODUCTION relation between environmental sensing and digital
Urbanization can be the root of evil when it processing of observed graphs and images. The
comes to ecological imbalance, especially pertaining entire paper is devoted to the analysis of
to the contamination of water bodies, a ramification mathematical methods sanctioning a discernment of
of industrial and residential runoff. Water pollution the concentration of aerosol particles. Landsat
poses an unparalleled threat to our existence, as it images have also been employed in the detection of

978-1-5386-3570-4/18/$31.00 ©2018 IEEE 142

Proceedings of the 2nd International Conference on Trends in Electronics and Informatics (ICOEI 2018)
IEEE Conference Record: # 42666; IEEE Xplore ISBN:978-1-5386-3570-4

pollution in juxtaposition to ground measurement pixel βij considered as “River Region”. A new image
data [6]. This work was published by Yusop et. al. is created such that the RGB value of pixel
The application of satellite images for monitoring βij(Satellite) is taken as is if the following conditions
water area pollution is the basic kernel of the are true:
research work presented by Mityagina et. al [4]. It is
Red
posited that satellite monitoring would permit the βij(Roadmap) = 163
diagnosis of the source of the pollution, thus helping
Green
to conduct a quantitative assessment of its scale and βij(Roadmap) = 204
predict its drift parameters. Turbidity is used as a
Blue
major parameter in the work presented by Lim et. al βij(Roadmap) = 255
[3]. It is also phrased as a turbidity meter for the
evaluation of the pollution level in adjacent region. In all other cases, the Redβij(Satellite),
Green
In another novel work presented by Notarnicola et. βij(Satellite), Blueβij(Satellite) value is stored as 0.
al, an original measuring parameter is propounded This is to indicate that the pixel βijis not a “River
[5]. Visible and thermal measurements are provided Region”. Region Subtraction provides scope for
by MODIS (MODerate-resolution Imaging Spectro- further calculations to be made on the resulting
radiometer) sensors on board of Terra and Aqua intermediary image. The intermediate image can be
satellites. In most of the apposite extant work, the further classified into “Dry Region” and “Water
major guiding component has been high resolution Region” based on another RGB based comparison.
satellite image processing. We have suggested an Based on our experimental study, we conclude that
auxiliary mechanism with the help of promptly sand pixels have a particular characteristic, wherein
available Google Maps to generate a quality output. the RGB values of the pixel βij(Satellite) can be
easily deduced by the subsequent conditions:
III.PROPOSED WORK
Red
111 ≤ βij(Satellite) ≤ 159
The proposed method entailed the
implementation of four major steps in order to 90 ≤ Green
βij(Satellite) ≤ 200
acquire necessary data from the given input images.
The process constitutes: A. Image Acquisition, B. 90 ≤ Red
βij(Satellite) ≤ 160
Region Based Subtraction, C. Monitoring Variations
Among Pixel Blocks in Frequency Domain, and D. Such a βij(Satellite) pixel is classified as “Sand
Assimilation of Data Pattern. One of the cardinal Region”. For facilitating easier access, such a pixel
benefits of utilizing our proposed scheme is that it is also equated to 0 for the corresponding Red,
involves the application of the readily available Green, and Blue values. The resulting image is then
Google Maps which can be effortlessly used in an subjected to a Discrete Cosine Transformation for
extensive range of areas. Our recommended analysing the colour of the river water.
technique is expounded in detail as follows:
C. Monitoring Variations Among Pixel Blocks in
A. Image Acquisition Frequency Domain
The input data comprises only of the The image of the “Water Region” from the
latitude and longitude values. Each of the inputs is previous step is taken and a 8 x 8 block-by-block
taken as a string of maximum length 8 (including 1 Discrete Cosine Transform is applied on it. The
for decimal point). The proposed system employs colour variations are too minute to be noted in the
the services of Google Maps API to extract 2 map Spatial Domain, thus it is converted to the
images in JPEG format. The input images consist of Frequency Domain as it gives us uncorrelated values
Satellite Image and Roadmap. The Roadmap is used to work with. The average DC value of each 8 x 8
to segregate the expanse of the river from the stretch block is used for drawing conclusions. The βi=0, j=0
of land and road. The technique of subtraction is pixel value in the Frequency Domain gives the co-
explained further in Region Based Subtraction. efficient of the smooth region for the entire block
Unlike the methods employed by preceding models, which precisely depicts the colour range of the given
this technique does not obstinately necessitate the block. The average of all such 8 x 8 blocks are
use of high resolution images. attributed together and it is compared with the same
value obtained from a similar image from an older
B. RGB Based Region Subtraction time period. The average β00 obtained using the
The area covered by the river is subtracted following formula:
from the rest of the satellite image by using RGB i=0 j=0
based comparison. The image in the Roadmap is Avg
β00 = ∑ ∑ β(i/8) (j/8)
used for collation. The pixel value at βij(Roadmap)is n n

such that the Red value equals 163, the Green value
equals 204, and the Blue value equals 255 then the

978-1-5386-3570-4/18/$31.00 ©2018 IEEE 143

Proceedings of the 2nd International Conference on Trends in Electronics and Informatics (ICOEI 2018)
IEEE Conference Record: # 42666; IEEE Xplore ISBN:978-1-5386-3570-4

The Avgβ00 is then compared with the AvgβOld

and the percentage of increase in the colour value V. EXPERIMENTAL RESULTS
can be obtained. The increase in colour value is
directly proportional to the increase in the pollution The propounded methodology was applied
level of the water in that region. Thus, we can on multiple images using Satellite and Roadmaps
deduce the percentage of increase in pollution level appropriated from Google Maps. The system
by extrapolating the available information by using succeeded in eliminating a majority of the hitches
the formula: encountered by the older procedures by
incorporating a more robust algorithm which is
(y – y1) / (x – x1) = (y2 – y1) / (x2 – x1) efficient in identifying the river region immaculately
and predicting the pollution level in the given
D. Assimilation of Data Patterns region. The algorithm was assessed on numerous
locations across rivers in India and the results were
Certain regions have a characteristic colour impeccable as far as the prediction mechanism of
pattern. This may happen due to the presence of colour of water is concerned. The experimental
specific industries or other natural factors. Data results achieved by pixel comparison based on
Patterns are assimilated to eliminate such an error. archived maps have been tabulated as follows:
Persistently occurring values are noted and our
proposed system takes into account the values from Table: Computed RGB Mean values for different
an older time period for comparison. If the values do co-ordinates
not seem to change, the region is considered as “not
polluted”. The pollution level is given in terms of Month
‘Water Quality Index’ (WQI) which is expressed as Sl. & RGB
a ratio as follows: No Latitude Longitude Year mean
March
WQI =0.3 * ( RedβAvg/ Redβdb) + 0.3 * ( 1 12.38293 75.538116 2017 #425F61
Green
βAvg/ Green
βdb) +0.3 * ( BlueβAvg/ Blueβdb) March
2 12.374798 75.549413 2017 #3C565B
March
3 12.369074 75.56298 2017 #425A61
IV. ALGORITHM March
4 12.363057 75.572617 2017 #415762
Step 1: Obtain the Latitude and Longitude for March
requested region. 5 12.337313 75.618695 2017 #3D5866
Step 2: Extract the Satellite and Roadmap of the March
region using Google Maps. 6 12.333709 75.654969 2017 #486373
Step 3: Perform RGB based background March
subtraction using the following conditions: 7 12.329648 75.654969 2017 #4D6E77
Red
βij(Roadmap) = 163 March
Green
βij(Roadmap) = 204 8 12.28245 75.760447 2017 #3F5C75
Blue March
βij(Roadmap) = 255
9 12.431614 75.924048 2017 #314758
Step 4: Subtract the sand pixels using the following
March
conditions: 10 12.560545 75.99182 2017 #103454
Red
111 ≤ βij(Satellite) ≤ 159 March
Green
90 ≤ βij(Satellite)≤ 200 11 12.520999 76.229957 2017 #566A76
Red
90 ≤ βij(Satellite) ≤ 160 March
Step 5: Apply a 8 x 8 block-by-block Discrete 12 12.525691 76.413849 2017 #557481
Cosine Transform (DCT) on the final intermediary March
image. 13 12.486559 76.457494 2017 #1D3D4F
Step 6: Obtain the average DC value, Avgβ00, using March
the following formula 14 12.42609 76.571305 2017 #374554
March
i=0 j=0
Avg 15 12.422223 76.597881 2017 #374554
β00 = ∑ ∑ β(i/8) (j/8)
n n March
16 12.425916 76.65651 2017 #2D404E
Step 7: Compare the Avgβ00 with the AvgβOld obtained March
from archived map of the same latitude longitude 17 12.221065 76.95636 2017 #234962
and extrapolate the pollution level in the zone March
and compute the WQI value as given in Step 8. 18 12.253791 77.150803 2017 #293D59
Step 8: Compute WQI = 0.3 * ( RedβAvg/ Redβdb) + 0.3 March
* ( GreenβAvg/ Greenβdb) +0.3 * ( BlueβAvg/ Blueβdb) 19 12.286256 77.432408 2017 #345971

978-1-5386-3570-4/18/$31.00 ©2018 IEEE 144

Proceedings of the 2nd International Conference on Trends in Electronics and Informatics (ICOEI 2018)
IEEE Conference Record: # 42666; IEEE Xplore ISBN:978-1-5386-3570-4

March
20 12.12183 77.774229 2017 #476379 VI. CONCLUSION
March
21 11.901169 77.840237 2017 #0A3F4C
March The remote sensing data that is computed
22 11.788715 77.805244 2017 #2D4A5D
within the global spectrum is utilized in this paper.
March
Despite the image resolution being compromised,
23 11.417728 77.684749 2017 #26343E
March the output value is fundamentally precise. It has
24 11.075971 78.061226 2017 #475458 merits like periodical continuous monitoring and
March being free of charge. It has a reasonably substantial
25 10.962133 78.221931 2017 #635B54 significance in the research on remote monitoring
March for the whole water quality. It also assumes
26 10.96365 78.387799 2017 #4A606C important practical relevance and application value
March for administering our proposed technique for the
27 10.838261 78.726811 2017 #3B3E47 augmentation of water quality in rivers of all types.
The innovation in this paper is that we have used
Photos: For sample latitude value 10.838261 and readily accessible maps and have utilized feasible
longitude value 78.726811 resources to produce a superlative output which is
accurate to a large extent. Certain progress has been
accomplished by the research on the quality of water
in colour-dyed rivers that are polluted due to textile
industries. Thus, there is vast scope for future
amelioration in the designing of a system of superior
efficiency for the digital monitoring of restricted
rivers. Owing to the impact of insufficient
experimental data, the building model falls short of
the quintessential value. In addition, due to restricted
condition, the model inversion result cannot be
verified by adequate data. Therefore, these problems
need to be remedied and improved further in the
A. B.
future research.

VII REFERENCES

[1]Satellite image processing and air pollution dete

ction, A. Prochazka; M. Kolinova; J. Fiala; P.
Hampl; K. Hlavaty, 2000 IEEE International
Conference on Acoustics, Speech, and
Signal Processing. Proceedings (Cat.
No.00CH37100), Year: 2000, Volume: 6,
Pages: 2282 - 2285 vol.4
C. D.

[2]Monitoring and evaluation on water quality of

Hun River based on landsat satellite data Hua
Ding; Ru Ren Li; Hao Lin; Xin Wang 2016 Progress
in Electromagnetic Research Symposium (PIERS),
Year: 2016, Pages: 1532 - 1537

[3]Turbidity measurement from ALOS satellite ima

gery, H. S. Lim; M. Z. MatJafri; K. Abdullah
OCEANS 2009-EUROPE, Year: 2009, Pages: 1 - 5
E. F.
Fig A shows the obtained Roadmap of given co-ordinates,
Fig B shows the Satellite image of given co-ordinates, Fig [4]Satellite monitoring ofthe Black Sea surface poll
C is the image obtained after RGB based image ution, M. Mityagina; O. Lavrova, 2015 IEEE
subtraction, Fig D shows the water region, Fig E shows International Geoscience and Remote Sensing
the sand region, Fig F depicts the DCT transformed image
of Fig C.

978-1-5386-3570-4/18/$31.00 ©2018 IEEE 145

Proceedings of the 2nd International Conference on Trends in Electronics and Informatics (ICOEI 2018)
IEEE Conference Record: # 42666; IEEE Xplore ISBN:978-1-5386-3570-4

Symposium (IGARSS), Year: 2015,

Pages: 2291 - 2294

[5]Using remote sensing satellites for water quality

monitoring in the UAE, Ammar Al Muhairi; Hosni
Ghedira; Hussain Al-Ahmad; Ali Dawood 2011.
IEEE GCC Conference and Exhibition (GCC)
Year: 2011, Pages: 67 - 68

[6]Urban Pollution Produced by Industrial Settleme

ntsAnalyzed ThroughHigh Resolution Satellite Ima
ges, C. Notarnicola; M. Angiulli; F. Posa, 2006
IEEE International Symposium on Geoscience and
Remote Sensing, Year: 2006, Pages: 3423 - 3425

[7]Monitoring water quality from Landsat TM ima

gery in Penang, Malaysia, Syazwani Mohd
Yusop; Khiruddin Abdullah; Lim Hwee San; Md
Noordin Abu Bakar, Proceeding of the 2011 IEEE
International Conference on Space Science and
Communication (IconSpace), Year: 2011,
Pages: 249 - 253

978-1-5386-3570-4/18/$31.00 ©2018 IEEE 146