International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 3 Issue: 6

ISSN: 2321-8169
3699 - 3703

_______________________________________________________________________________________________

Aquatic Robot Design for Water Pollutants Monitoring

Anuradha A. Maindalkar

Saniya M. Ansari

ME (VLSI & Embedded Sys.) Dept. of Electronics & Tele.

Dr. D. Y. Patil SOE, Pune, India.
[email protected]

Professor, Dept. of Electronics & Tele.

Dr. D. Y. Patil SOE, Pune, India.
[email protected]

Abstract this paper focuses on the design and development of an Aquatic Robot for water pollutants monitoring. An aquatic robot is integrated
with the smartphone for data acquisition. The implemented design contains CV algorithm for image processing on openCV platform. Regularly
monitoring aquatic pollutants is needed for pure aquatic environment and safe aquatic life. The human health and water transport are also main
consideration towards this robot design. The proposed Aquatic robot consists of sensors and camera for sensing hazardous pollutants and
capturing images of surrounding environment respectively. The aquatic robot can accurately detect pollutants and display results on smartphone
in the presence of various conditions.
Keywords- Aquatic Robot, Raspberry-pi, Arduino, OpenCV, Smartphone.

_________________________________________________*****_________________________________________________
monitoring an environmental process in a large region by a
I.
INTRODUCTION
small number of robotic sensors [2]. The aquatic environment
Monitoring aquatic debris or pollutants is of great interest to
presents an almost ideal test-bed for the evaluation and
the ecosystems, marine life, human health, and water transport
development of robotic technologies. The environment is
[1].Water resources and aquatic ecosystems are facing various
highly dynamic and three dimensional. Extremely limited offphysical, chemical, and biological threats from climate
board communication underwater requires that robots must
change, industrial pollution, improper waste disposal, diseaseoperate fully autonomously or under operator control through a
causing microbes, and even terrorist attacks. Sensing of
tether [3].
aquatic environments is of increasing importance to public
Recently, there is an increasing interest in the development
health, port and river security, ecosystem sustainability,
of autonomous, in situ, real-time sensing platforms for aquatic
marine biology, and aquaculture industry. Manual sampling,
environments [1-4]. Moreover, sensing platforms equipped
via boat/ship or with handheld devices, is still a common
with embedded camera sensors provide a promising solution
practice in the monitoring of aquatic environments. This
for efficiently and intuitively observing aquatic environments,
approach is labor-intensive and has difficulty capturing
so we have developed an inexpensive, multipurpose aquatic
dynamic and spatially distributed phenomena of interest.
sensor node enabling long- term aquatic monitoring. Figure 1
shows a completed aquatic robot. The node includes several
kinds of aquatic sensors submerged in water and an onboard
camera sensor on top. The former periodically collect the
water-quality data and the latter captures images for pollutants
detection. However, there are some specific problems awaiting
solutions for the vision-based aquatic pollutants detection.
First, due to the impact of uncontrollable water surface
disturbances such as waves, camera shaking and swaying
reflections on the water, it is quite difficult to reliably identify
pollutants objects. Second, embedded platforms are not
Figure 1. Aquatic Robot
competent for continuous and real-time image processing
because of the constraints on computational power and memory
Mobile sensor networks can be greatly used to monitor
[5]. This means the pollutants detection algorithm should have
environmental variables such as temperature, pH, salinity,
low computation complexity at the same time as achieving
toxins, and chemical plumes. Signicant advances have been
desirable detection accuracy. Third, due to the restricted
made in the area of mobile sensor networks and their
accessibility for replacing batteries of aquatic sensor nodes, the
applications to environmental sciences. Decentralized
limited energy supply remains a challenge for long-term
environmental modeling by mobile sensor networks was
monitoring. Apart from the acquisition, processing and
presented in in which control laws were developed for mobile
transmission of water-quality data, both pollutants detection
sensors to maximize their sensory information. The Mahdi
and image transmission incur high energy consumption.
Jadaliha et.al provides practical solution to the problem of
3699
IJRITCC | June 2015, Available @ http://www.ijritcc.org

_______________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 3 Issue: 6

ISSN: 2321-8169
3699 - 3703

_______________________________________________________________________________________________
Due to space constraints, this paper is restricted to discuss
the vision-based pollutants detection using our embedded
sensing platform. The main contributions of this paper are as
follows:
(1) We present the design of aquatic robot and implement
the pollutants detection algorithm on the embedded platform.
(2) We propose an accurate and computationally efficient
approach to pollutants detection.
II.

PROPOSED METHODOLOGY

requirement. There are simple and easy-used open source

peripheral driver libraries.
B. Mechanical Design
The robot is approximately 65 cm long, 50 cm wide (at the
fins), and 13cm high. It is encased within an acrylic
waterproof shell and displaces about 18 kg of water. An
Aquatic Robot weighted about 4.3 kg. The robot is powered
by two onboard batteries providing over two hours of
continuous operation. Camera, sensor and control signals are
sent to a floating platform at the surface via fiber-optic tether.

A. Hardware Design
Robotic unit is consisting of array of sensors and camera
which is movable around its axis and also vertically.
Raspberry Pi is used for image processing and sending the
image to the user through the Wi-Fi is used for communication
between Arduino and Raspberry Pi (ARM Processor). Motor
driving circuits are used for operating motors. An Aquatic
robot is capable of moving in water by a DC motor. The motor
is manipulated by a programmable control board, which can
communicate with the smartphone through short-range
wireless links such as Bluetooth.

Figure 3.

Mechanical Design of Aquatic robot.

As interaction between the robots body and water occurs, fluid

mechanics are also involved, particular the effects of drag on
the robot. Drag force can be calculated through applying the
equation,
FD = 0.5*PV2*Cd A
Besides drag, hydrostatic pressure is also an important aspect to
take into account. Looking plainly at a flat plate being
submersed underwater, the pressure acting on the plate is
directly proportional to its depth underwater, as illustrated in
Figure 3. By applying the equation,
P

Figure 2.

Block diagram for Robotic Unit

The Raspberry Pi here used for image processing on Open

CV platform. The Raspberry Pi is a credit-card sized computer
that plugs into your TV and a keyboard. It is a capable little
computer which can be used in electronics projects, and for
many of the things that your desktop PC does, like
spreadsheets, word- processing and games. It also plays highdefinition image. Raspberry Pi has a strong processing capacity
because of using the ARM11 architecture and Linux-based
system. In terms of control and interface, it has 8 GPIO, 1
UART, 1 I2C and 1 SPI, which are basically meet the control

Hydrostatic

=hg

Where
p = pressure (N/m2, Pa)
h = height of fluid column, or depth in the fluid
at which the pressure is measured (m, in)
= density of liquid (kg/m3)
g = the gravitational constant (9.81 m/s2, 32.17405 ft/s2)
Under
2m
in
seawater
maximum
around=2*999.97*9.81= 19619.411 Pa

pressure

Weight Distribution As the overall weight of the robot is an

important aspect for it to overcome the buoyancy force and
operate on the seabed, the detailed breakdown of the weight of
the various components of the robot.

3700
IJRITCC | June 2015, Available @ http://www.ijritcc.org

_______________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 3 Issue: 6

ISSN: 2321-8169
3699 - 3703

_______________________________________________________________________________________________
C. Sensors Data Processing
The developed robotic system has embedded several kinds
of sensors for autonomous navigation and quality detection of
water. Sensors to be used for autonomous navigation are
ultrasonic sensors and accelerometer. The sensors used for
quality detection are temperature sensor and gas sensor. In this
paper interfacing of all the sensors, DC motors, Bluetooth with
Arduino (ATmema328) processor is explained.
An interfacing diagram is shown in fig.4. Ultrasonic sensor
is used to calculate distance of obstacle from real position of
robot. Accelerometer measures the three axial positions
according to the movement or shaking of waves in water
environment. Gas sensor MQ2 is used here to measure the
amount of gases dissolve in water. If water body is
contaminated with large number of harmful gases then gas
sensor indicates its presence. The data collected by all the
sensors are then sending to the smartphone via Bluetooth. With
these features, the aquatic robot can move autonomously on
surface of water.

Figure 4.

of gradient magnitude. Hough transform then nds the horizon

line through a voting process based on the number of edge
points that each candidate line.

Sensor Interfacing of Aquatic

D. Image Processing
In this paper we present a smartphone-based sensing
platform that utilizes CMOS camera, inertial sensor, and other
resources. Different from aforementioned sensing platforms,
we aim to combine camera sensors with other types of aquatic
sensors into a sensor node for multipurpose aquatic
monitoring. The key to aquatic pollutants detection is to
extract the foreground objects from image sequences captured
by camera sensors. A common way to foreground extraction is
background subtraction. One such means, Gaussian mixture
model (GMM) [1-6] has been widely adopted because of its
ability to deal with subtle illumination changes.
Here we employ CV algorithm for image processing on the
OpenCV platform. As the camera is an interface with the
Raspberry-pi processor the libraries of algorithm are uploaded
in processor. The flowchart of CV Algorithm is shown in
Figure 6.The images taken by camera processed inside the
Processor. For each frame, we rst convert it to a gray scale
image and detect edges using a Sobel operator. The Sobel
operator detects an edge point according to the local intensity

Figure 5.

Flowchart of CV Algorithm

We employ Hough transform to extract the horizon line.

Hough transform has been widely used to identify the
positions of lines in an image [8]. To reduce the energy
consumption in image processing, we adopt a lightweight
background subtraction approach to detecting the foreground
pollutants objects. We rst convert the representation of an
image to HSV (Hue, Saturation, and Value) model. In HSV,
hue represents the color, saturation is the dominance of hue,
and value indicates the lightness of the color.
In our approach, the background of a pixel is represented
by a Gaussian mixture model (GMM) [7]. The GMM
comprises K three-dimensional Gaussians, where each
Gaussian characterizes the three channels of HSV color space.
When a new frame is available, each pixel is compared with
its background model. If the HSV vector of a pixel in the new
frame does not fall within a certain range from any mean
vector of the K Gaussians, this pixel is considered a
foreground pixel candidate; otherwise, it is classied as
3701

IJRITCC | June 2015, Available @ http://www.ijritcc.org

_______________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 3 Issue: 6

ISSN: 2321-8169
3699 - 3703

_______________________________________________________________________________________________
background. Therefore, the color difference between the
foreground and background affects the classication accuracy.
The binarized foreground image contains randomly distributed noise pixels. Because the background subtraction is
conducted in a pixel-wise manner, the labeling of foreground
and background can be affected by camera noise, resulting in
false foreground pixels. To deal with these noise pixels, we
adopt the opening operation in image morphology. The
opening operation, which consists of erosion followed by
dilation, eliminates the noise pixels through erosion while
preserving the true foreground by dilation.
After the noise removal, we employ region growing to
identify the pollutants objects from the foreground image. It
uses the foreground pixels as the initial seed points and forms
connected regions that represent the candidate pollutants
objects by merging nearby foreground pixels.
III.

EXPERIMENTAL RESULTS

Here Aquatic robot evaluated through experiment done in

swimming tank. The experimental results are obtained on
smartphone validate the feasibility of Aquatic robot by
evaluating the computation overhead, effectiveness of each
module, and the overall performance of a fully integrated
Aquatic robot. The simulations extensively evaluate the
performance of Aquatic robot working under wide ranges for
parameter settings.Fig.6 shows the simulation results of all the
sensors implemented in hardware of aquatic robot.

Figure 6.

a.
Figure 7.

b.

c.

Image processing output a. foreground image, b.

background image, c. output image.

The output of image processing and all sensors are

obtained on smartphone via Wi-Fi network and bluetooth
respectively.The main advantage of this is that if Aquatic
robot is placed at remote location we are able to access data of
surroundig aquatic environment accurately.
Since the aquatic robot is designed at a low cost, it
can be used in various applications like sampling lakes,
monitoring aqua farms and safeguarding water reservoirs.
Aquatic robot could give researchers far more precise data on
aquatic conditions, deepening the knowledge of critical water
supplies and habitats. Aquatic robot will carry sensors
recording things such as temperature, dissolved oxygen,
pollutants and toxic algal blooms. Image processing algorithm
improves the accuracy for detecting pollutants.

Simulation Result of Sensors of Aquatic Robot

Fig.7 shows the results of image processing done in

openCV which is directly obtained on smartphone.The base of
image processing is the CV algorithm. The following figure
shows a sample of background subtraction. Specically,
Figure 7(a) is the original frame where the red dashed lines
represent the extracted horizon lines for this frame and the
registered predecessor frame, respectively. Figure 7(b) shows
the background model, where each pixel is the mean vector of
the Gaussian with the largest weight in the GMM. Figure 7(c)
is the result of background subtraction with image registration.

Figure 8.

Results obtained on Smartphone a) output of

sensors on Bluetooth b) output of camera on
Bluetooth

CONCLUSION
This paper presents design of aquatic robot for aquatic
pollutants monitoring. The android smartphone is integrated
with aquatic robot to capture images and to acquire data of
different sensors. The real time pollutants detection is done
with the CV algorithm efficiently. In our future work, this can
3702

IJRITCC | June 2015, Available @ http://www.ijritcc.org

_______________________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 3 Issue: 6

ISSN: 2321-8169
3699 - 3703

_______________________________________________________________________________________________
be extended to, use aquatic robot in an inland lake and evaluate
it under various conditions such as debris flow speed and
brightness/lightening.
ACKNOWLEDGMENT
We take this opportunity to thank all the individuals for
their guidance, help and timely support which made us to carry
out the project work in stipulated time. We also thank all
teaching and non-teaching staff members for their patience and
special consideration for the completion of our work. Lastly,
we offer our regard for all those, in particular our parents and
who supported us in any respect during this project work.

[3]
[4]

[5]

[6]

REFERENCES
Yu Wang, Rui Tan, Guoliang Xing, Jianxun Wang, Xiaobo Tan,
Xiaoming Liu, and Xiangmao Chang, Aquatic Debris
Monitoring Using
Smartphone-Based Robotic Sensors IEEE CONFERENCE
PUBLICATION, APRIL 2014.
[2] Mahdi Jadaliha and Jongeun Choi, Environmental Monitoring
Using Autonomous Aquatic Robots: Sampling Algorithms and
[1]

[7]
[8]

Experiments, IEEE TRANSACTIONS ON CONTROL

SYSTEMS TECHNOLOGY, VOL. 21, NO. 3, MAY 2013.
Gregory Dudek, Michael Jenkin, Chris Prahacs, Andrew Hogue,
Junaed Sattar, A Visually Guided Swimming Robot.
Yong Wang ,, Dianhong Wang , Qian Lu , Dapeng Luo and Wu
Fang , Aquatic Debris Detection Using Embedded Camera
Sensors,
sensors
ISSN
1424-8220
www.mdpi.com/journal/sensors, 30 January 2015.
Aghdasi, H.S.; Abbaspour, M.; Moghadam, M.E.; Samei, Y.
An energy-efficient and high-quality video transmission
architecture in wireless video-based sensor networks. Sensors
2008, 8, 45294559.
Stauffer, C.; Grimson, W.E.L. Adaptive background mixture
models for real-time tracking In Proceedings of the IEEE
Computer Society Conference on Computer Vision and Pattern
Recognition (CVPR 1999), Fort Collins, CO, USA, 2325 June
1999; pp. 246252.
A. K. Jain, Fundamentals of Digital Image Processing.
Prentice Hall, 1989, vol. 3.
J. Illingworth and J. Kittler, A survey of the Hough transform,
Computer Vision, Graphics, and Image Processing, vol. 44, no.
1, pp. 87116, 1988..

3703
IJRITCC | June 2015, Available @ http://www.ijritcc.org

_______________________________________________________________________________________________