Вы находитесь на странице: 1из 6

Machine Learning Based Cloud Integrated Farming

Jetendra Joshi Siddhanth Polepally Pranith Kumar


Department of Electronics and Department of Electronics and Department of Electronics and
Communications Communications Communications
NIIT University, Rajasthan, India NIIT University, Rajasthan, India NIIT University, Rajasthan, India

Rohith Samineni S R Rahul Kaushal Sumedh


Department of Electronics and Department of Electronics and Department of Electronics and
Communications Communications Communications
NIIT University, Rajasthan, India NIIT University, Rajasthan, India NIIT University, Rajasthan, India

Dandu Geet Kamal Tej Vishal Rajapriya


Department of Electronics and Communications Department of Electronics and Communications
NIIT University, Rajasthan, India NIIT University, Rajasthan, India

ABSTRACT is highly inefficient and produces a low yield. Technological


With the increasing population, the demand for more agricultural advancements in the area of agriculture have shown reliable
production is only surged. This increasing demand can only be results and also help in increasing the competence of farming
attained with progressing technology. Internet of Things (IoT) is activities. But when computerization was implemented and
dramatically advancing the way we live our life by full-scale manual labor had been replaced by automatic machineries, the
control over data with minimal human involvement. Using IoT to yield percentage has improved immensely. We came up with a
meet this high demand for production is achieved. In this paper new innovative way where smart farming can be done on a small
we propose a bot that can be used for small scale farming in areas scale using IoT.
like gardens or backyard. A simple web application is provided IOT (Internet of Things) is any physical system that can be
for the user to decide the plants to be farmed and the bot does rest assigned an IP address and can transfer data over a network. It
of the work. This bot can plant the seeds, water each plant at helps in increasing the efficiency and accuracy of transferring
required intervals and even plot the weed to bury it. A database is data over a network. In simple words, it is a network of
provided with the information about several parameters to be interconnected things and devices which are embedded with
taken care of for each kind of plant. Different sensors are used to sensors, network connectivity, software and necessary electronics
sense the properties of soil and environment which can be used to that enables them to collect and exchange data making them
anticipate the near changes and take necessary steps. Image reactive. The IOT architecture is capable of providing highly
processing is being used for detection and prevention of weed secured encryption and is also developed enough to support an
growth. We adopted Bayesian methods of machine learning to operating system. The agricultural farming scenario appears to be
efficiently estimate the performance parameters by probability one of the most favorable application areas for IoT. Farming
distribution. sector is highly un-organized in few countries; most of the
systems followed are passed down through generations. Using
CCS Concepts IoT to achieve this can be productive.
•Computer systems organization➝Robotic autonomy
1.1 Important components of IOT
Keywords IoT is defined as any physical object that can sense and affect the
physical environment. These physical objects are referred to as
Internet of Things (IoT); smart agriculture; weed removal; “things”. Humans are also included in IoT in various scenarios
shadow removal; Machine Learning; Bayesian Statistics. where they control the environment via mobile applications. The
main components that make internet of things a reality are
1. INTRODUCTION physical objects or “things”, sensors, actuators which are used to
Agriculture is a wide financial sector and it plays a vital role in sense the physical environment such as brake controller in a car,
the overall economic development of a nation. In agricultural different types of platforms, different types of services and all of
sector, growth is important for the enlargement of an economy. these are interconnected over a network. Types of platforms
Unfortunately, farmers still use ancient methods of farming which consist of middleware that are used as access to devices or data
analytics. Types of services include cloud services that can be
Permission to make digital or hard copies of all or part of this work for used to process big data and convert it into valuable information.
personal or classroom use is granted without fee provided that copies are This valuable information can be used by applications to help
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. Copyrights for
control the physical as well as virtual environment using a
components of this work owned by others than ACM must be honored. network. Widely accepted architecture of IoT is depicted in figure
Abstracting with credit is permitted. To copy otherwise, or republish, to 1.
post on servers or to redistribute to lists, requires prior specific permission
and/or a fee. Request permissions from Permissions@acm.org.
ICMLSC '17, January 13-16, 2017, Ho Chi Minh City, Viet Nam
© 2017 ACM. ISBN 978-1-4503-4828-7/17/01…$15.00
DOI: http://dx.doi.org/10.1145/3036290.3036297
DOI: http://dx.doi.org/10.1145/12345.67890
1
directed acyclic graph (DAG). For example, a Bayesian network
could represent the probabilistic relationships between speed and
time. Given speed, the network can be used to compute the
probabilities of the latency of different sensors which we estimate
in this paper. Efficient algorithms exist that perform inference and
learning
The paper has been divided into VI sections. In section II related
works; previously achieved works has been discussed. Section III
mentions problem formulation. In section IV proposed model, the
hardware, software and image processing has been explained
along with results. In section V, an analysis of proposed model is
performed. Section VI is conclusion and finally, ending the paper
with references in section VII.

2. RELATED WORKS
Internet of things is a developing technology that connects
devices with in a network. This technology developed enables the
Figure 1. Architecture of IoT. control and monitoring in various domains including home
An individuals’ life is completely inundated with IOT appliances, industrial production, health monitoring applications,
applications surrounding him. For instance, a simple wrist band smart cities, smart grid, agricultural applications and many more
around a person’s hand measuring the sleep cycles knows when to explained in [1],[2]&[3]. In [4], [5] IoT is discussed in depth
wake him up for his daily routine. Once the person is up, it starts along with cloud computing. How IoT is currently being used in
a chain of events communicating to several other devices like agricultural field and its future scope is well extrapolated along
warming the room by regulating the air conditioner, starting the with detailed flowcharts of its working. The various areas into
coffee machine, notifying the geyser to start water heating, which IoT applications disseminate is also explained. In [7] the
preparing the garage door to open and so on making a persons life author proposes a IoT based agricultural convergence technology
efficient and easy. This is an application of IOT in smart home for high quality increase in agricultural products. He implemented
systems. On the other hand, consider a scenario where a person precision farming as an alternative to the future agriculture using
has medical issues. In this scenario, person suffering from heart this technology. This allows prediction of supply and demand,
issues is prone to suffer from heart attacks. This wrist band can quality management during the entire life cycle of agriculture. An
detect sudden increase in blood pressure, erratic breathing and insight into key technologies of IoT and system architecture is
changes in vital signs and notifies the user for immediate medical given in [9] which included technology of information collection,
attention. In case of a severe situation, it will alert an ambulance network communication, data fusion and computing. Authors
for dispatch and apprise the hospital so that they can be prepared view on monitoring system for Agricultural production based on
beforehand. IoT is also shown. In [13] [19] author proposes the use of wireless
network based solution for Precision Agriculture along with an
Smart agriculture is a concept that was created in order to overview of current agricultural production status, urging the
maintain a sustainable environment as well as get an optimal yield. need for precision farming. In further explains the feasibility of
Previous agriculture methods were developed without considering precision farming in developing countries with solid ground. In
water consumption, climate change, environmental conditions, etc. [14] author discusses different problems faced by the ‘Food and
As agriculture consumes approximately 70% of the world’s Agricultural Organization’ of the UN with the pacing population.
freshwater supply, water management plays a crucial role and The prominence of smart farming in the coming future is
should be used efficiently. This can be done through smart bolstered with government census data. Various barriers and
agriculture which takes into consideration all aforementioned driving factors that lead for the adaption of smart agriculture are
factors. In smart farming, based on already available data related well interpreted.
to the amount of water required for different types of crops and
also taking into consideration the moisture levels in the soil, a Various domain uses IoT and Big-Data analysis as the key
reliable water management schedule can be designed. Available technology for development. Sensor technology has also been
weather reports can be used as a precursor for further precautions. advancing and several fields like cloud computing and mobile
computing are reliable on it. Bringing together these two
1.2 Bayesian Statistics advancements in technology opens a new range of applications
Machine learning Algorithms have been used extensively for a that can make life less complex. This is explained in [15] [17].
wide range of tasks including classification, regression and Smart agriculture is accomplished in this using IoT, sensors,
density estimation in a variety of application areas such speech Cloud computing, Mobile computing, Data mining and Big-Data
recognition, bioinformatics, computer vision, spam detection, analysis. The paper [16] proposes an electronic system based
fraud detection and advertising networks. The algorithms and implementation of cloud computing, Global System for Mobile
techniques come from diverse fields including statistics, communication for monitoring a variety of environmental
mathematics, neuroscience, and computer science and used even parameters in greenhouse. Wide variety of sensors that can
broader, or most areas related to machines nowadays. measure temperature, light intensity, humidity and moisture levels
are used to make the production as prolific as possible. A new
A Bayesian network, belief network or directed acyclic graphical smart GPS based remote controlled robot to perform tasks related
model is a probabilistic graphical model that represents a set of to fields like animal scaring, weeding, keeping track of variables,
random variables and their conditional independencies via a sensing climatic changes has been proposed in [19] .A brief

2
overview of the need for IoT in agriculture and different sensors, Based off of all the inputs given by the user and the type of crops,
hardware and software that is available in the market can be the Farm bot then divides the plot of land assigned for agriculture
studied from it. A detailed technical review on the current sensor use into different segments based on the plant requirements. The
based Automated Irrigation system is given in [20] along with an Decision Support System (DSS) is a cloud service that uses the
overview of the need for automated irrigation. best algorithms to optimize scheduled events based on the data
that is present in the database. The Farm bot is smart enough to
In [24] [25] [26] author explained different methods to decide how much water a certain type of plant requires and based
accomplish background and foreground extraction. Several on weather forecast or moisture content in the soil, it will also
methods of three frame differencing, background subtraction, decide whether to give more or less water.
mixture of Gaussians, template matching, average mean filter,
average median filter are explained with flow charts and A database is provided that has all information required for each
mathematical representations among which normal background and every type of plant that is compatible with the Farm bot.
subtraction is observed to suffice our need. Several complex edge Raspberry Pi Controller is the brain of the Farm bot. The
detection methods are also explained. In[27] different Raspberry Pi is small sized single-board computer. It has several
morphological operations are discussed which include dilation, pieces of software installed that allow the farm bot to
erosion, hole filling ,opening, closing and so on. communicate with the web app, manage sequences and schedules,
as well as update its sensors, etc. It also uses a USB web cam that
can stream as well as take pictures which can be used to check the
3. PROBLEM FORMULATION progress as well as identify weeds which can hinder the plants
Till this day farming is being enforced using the same old ancient progress.
and complex methods. Delay in any one step of farming can lead
The Farm bot also uses Arduino Firmware that issue commands
to immutable loss. With the rapid increase in technology what can
to physically control motor drivers, rotary motors, etc., by
be done to make this process less complex. Man labor in farming
sending electrical impulses to them.
is decreased with the help of newly emerging applications of
Internet of Things. How does the proposed farm bot help in 4.2 Hardware Implementation:
achieving small scale farming without any involvement of human The interactive mobile app that has been provided to the user will
presence is seen. Farm bot could do planting, smart sowing and first ask the user to provide inputs for the app to continue further.
delete the weed from the field. How does a farmer get a good The app will then (based on the type of seeds we select from the
yield in the field with less stress and how we can eliminate weed database) will divide the entire area into different segments. If the
in our farm using Image Processing, is treated as the major goal. same type of seeds are provided, then the entire area will be
divided equally as same seeds will take up similar amount of
4. PROPOSED MODEL space. If there are different seeds, the app will divide the
area/farm into different segments based on what is optimal from
4.1 System Overview the database.
The agricultural sector is facing many challenges in order to feed
the billions of people that inhabit the planet. Food production Each seed needs to be planted at a certain depth into the soil for
needs to increase along with the increasing population. All of this optimal growth. The seed injector takes each seed and plants the
has to be achieved within the limited land available also, taking seed in the soil at that depth. The depth is taken from the database.
into consideration other parameters such as climate changes, Every seed requires certain amount of water to grow optimally.
limitation of water resources, etc. In order to counter all these The amount of water required depending on the type of the seed is
issues, smart agriculture is the best solution. Smart agriculture or given in the database. Water is then passed into the water nozzle
smart farming or precision farming uses sensing technology, to through a garden hose. The water nozzle accepts a concentrated
make farms more intelligent, where they keep a lot of factors in stream of water which is then turned into a gentle shower. The
scrutiny and save a lot of non-renewable resources while helping database also has the information related to amount of water
the crop yield maximum quantity with quality. The Farmbot is required by each seed under various environmental conditions.
one such device that is used for small scale farming. It can be Since the location of the seed can be obtained from the database,
used by anyone to grow a wide variety of crops together in their the water is distributed properly to its designated location. The
backyard. farm bot does this at regular intervals to the all the seeds/plants
An interactive web application is provided to the user where one based on their requirements.
can easily configure and control their Farmbot from a web The amount of light that a plant is exposed to is a very important
browser on their laptops, desktops, tablets, or smartphones. The factor. Light intensity determines the rate of photosynthesis and
application features real-time manual controls and logging, a this can affect the growth of a plant immensely. The higher the
sequence builder for creating custom routines for FarmBot to number of photons that come in contact with the plant, the more
execute, and a drag-and-drop farm designer where someone can number of chlorophyll molecules get ionized. This helps in the
drag and drop crops into their farm on the interface. Using this, plant producing more ATP and NADPH which is essential for the
we can graphically design and manage our farm. plant. The farm field is equipped with different sensors to
The web application takes some input commands from the user measure the light intensity, temperature and humidity so that a
and lets them design the farm. The MQTT gateway which is a notification can be sent to take necessary steps if any required
cloud application acts an intermediary for all messages between environmental conditions are not met. The sensed valued are
the web app and Farm bot devices. This is how the Farm bot displayed as shown in figure 2.
communicates and understand the commands from the web app
given by the user.

3
0 otherwise
Here,
B( x, y) is the corresponding pixel intensity value at
( x, y) coordinate of background image,
Figure 2. Sensor values on LCD Display. f ( x, y) is the image captured pixel value at ( x, y ) and
F ( x, y ) is the resulting binary image pixel value
Water is a very crucial element for the growth of a plant. Plants
which are not given sufficient amount of water will not grow
at ( x, y ) .
properly as they are not able to transport necessary nutrients from Main challenge faced is that the background image (soil color)
one part of the plant to the other. Excessive water is also an issue though captured at the beginning, does not remain the same as
for a plant. If a plant is given too much water, then the roots of that of the soil color after the plant growth. This can happen
the plants might rot, which in turn would result in less mineral because of several factors like when watered, the soil color is dark
intake from the soil. The farm bot has a sensor which is able to or even when plant shadow is on the soil. False detection can take
measure the moisture in the soil and determine whether each type place and proposed simple background subtraction is not reliable.
of plant requires more or less water. As the farm bot takes To avoid this, complex edge detection techniques can be used to
moisture into consideration, the bot provides the plants with the detect the foreground but here, even simple and less complex
calculated amount of water, not more, not less. empirical method is proposed.

An onboard camera is present on farm bot, which will take an When the soil is watered or covered by shadow, soil as a whole or
image of the plants from time to time to check whether any weeds part exhibits following properties.
are growing along with the plants. Using Image Processing, plants
1. This soil region is usually darker.
and weeds in the image are differentiated and the weeds are then
buried into the soil with the help of the weed remover tool. 2. This region represents the same pixel values but under darker
illumination.
4.3 Image Processing: 3. They share the same background texture pattern.
Image processing plays a major role in taking care of the plants
once grown. Weed should be recognized and removed regularly We use a simpler method of comparing the pixel values in plant
to keep the plants healthy. For this a well adaptive image image and the background image, which has proven applicable to
processing techniques should be used that can work under a wide range of factors. After several experiments it can be
different conditions. The proposed image process is as follows. concluded that if 0.25  f ( x, y) / B( x, y)  0.93 then that
pixel can be considered as soil pixel only and made to 0(black).
Once the plant is grown to a stage, farm bot camera moves to the
centre of a plant where seed is planted. An image is captured and
cropped as per the size given to that particular seed. Then from The result obtained is filled with noise and hence, it is passed
here image processing comes into play for weed detection and through several morphological operations which include median
removal. First phase involved is the foreground, background
filter, closing, opening, erosion, dilation etc.
extraction. Plants and weed is the foreground and the ground/soil
is the background that should be removed. Next is object
detection i.e., differentiating the plants and weeds in the image? 4.3.2Convex hull algorithm
Though the result obtained looks fair, every white object detected
Our database also consists of information about what approximate is considered as a plant. Any part of the object with holes and
area (2 dimensional area from top) does each different type of discontinuities has to be closed. So here convex hull is being used.
plant occupy after given number of days of growth. We compare Convex hull algorithm, in simple terms can be explained as two
this area to the area of the objects (plants and weed) detected. pixels of same object are connected with a straight line and any
Object enclosing the location of the seed planted and whose area pixel on the line is considered as a pixel of that object. This can
is the range from database is considered the main plant. Rest be called as a hole filling method and helps convert almost all
objects detected, which are out of range is considered as weed and connected objects to blobs. Once this is achieved, we are left with
buried. Centroid of these objects can be considered as the co- a binary image with blobs of plant and weed.
ordinates for this weed location.
4.3.3 Identification of weed
4.3.1 Background Subtraction: A bounding box is plotted around each blob detected. Area and
For the first phase of foreground and background extraction, centroid of each blob detected is calculated and stored in an array.
background subtraction is used. Background subtraction is the These areas now are compared to the area expected for that seed
core idea for a number of object detection algorithms. In and the blob representing the plant is located. Rest all blobs
background subtraction a fixed image called background image is identified are considered weed and centroid of the same are
used. Once the image with plant is captured, it is then subtracted passed to the bot. The bot considers these centroid co-ordinates as
from this background image and the resulting image is then the location of the weed and burry them inside. The results from
subjected to the thresholding process to get the final output as a every step involved in image processing works are shown in
logical (binary) image. This can be mathematically represented as figures 3 and 4.

if | B( x, y)  f ( x, y) | Th then F ( x, y)  1

4
The hosted web application is already configured to use the
hosted MQTT Gateway service. All you have to do is enter in
your web application credentials using the FarmBot WiFi
Configurator, and then your FarmBot will be able to talk to the
web application over MQTT. The configuration needed for your
FarmBot will be stored in the authentication token that the web
application gives to your FarmBot during setup.

5.3. FarmBot Raspberry Pi Controller:


Raspberry Pi uses the software to maintain a connection and
synchronize with the web application. This allows FarmBot to
download and execute scheduled actions, be controlled in real-
time, and update logs and sensor data. The controller
communicates with the Arduino over USB to send and also
receive data.

6. CONCLUSION
This study proposed a sensible small scale farming using IoT.
Using the proposed model, an individual can maintain his own
farm in small gardens. Most of the hardware utilized is easily
acquired and cost efficient. Incorporating image processing to
avoid weed growth is seen to be reliable and the results shows
Figure 3. Steps involved in Image Processing. above are consistent with this fact. The proposed simple method
to recognize shadow or drenched soil as background is seen to be
effective and obviate the use of complex foreground extraction
techniques. As part of the future work, it can be extended to large
scale farming.

7. REFERENCES
[1] Luigi Atzori, Antonio Iera, Giacomo Morabito, “”smart
Objects” to “social Objects”: The Next Evolutionary Step of
the Internet of Things”, IEEE CommunicationsMagazine,
January 2014.
Figure 4. Final Step in Image Processing.
[2] Ken Cai, ”Internet of Things applied in Field Information
Monitoring”, Advances in information Sciences and Service
The proposed image processing approach is seen to be optimal, Sciences (AISS) Volume 4,Number 12,July 2012.
less complex and with no errors in various physical conditions.
Since, this process is not performed all day and repeated after a [3] White paper on “What the Internet of Things(IoT) Needs to
brief amount of time(few days to weeks), high complex and Become a Reality”/ freescale.com/IoT,arm.com, May 2014.
advanced methods such as Gaussian foreground extraction model [4] V.C.Patil, K.A.A1-Gaadi, D.P.Biradar, M.Rangaswamy,
or adaptable background subtraction show no major advantage “INTERNET OF THINGS (IOT) AND CLOUD
over basic subtraction. In addition, the proposed empirical way of COMPUTING FOR AGRICULTURE: AN OVERVIEW”,
pixel comparisons take care of the remaining minor subtraction Proceedings of AIPA 2012, INDIA
problems if exists.
[5] Sheetal Israni, Harshal Meharkure, Parag Yelore,
“Application of IoT Bases System for Advance Agriculture
5. PERFORMANCE EVALUATION in India”: International Journal of Innovative Research on
Computer nd Communication Engineering November 2015.
5.1. The FarmBot Web Application [6] Dr. V .Vidya Devi,G. Meena Kumari, “Real- Time
The web application allows you to easily configure and control Automation and Monitoring System for Modernized
your FarmBot from a web browser on your laptop, smartphone Agriculture” ,International Journal of Review and Research
and tablet. The web application features real-time manual controls in Applied Sciences and Engineering (IJRRASE) Vol3 No.1.
and logging, which builds a sequence for creating custom actions PP 7-12, 2013
for FarmBot to execute, and a drag and drop farm designer so any [7] Chandini. K., “A Literature Study on Agricultural Production
one can graphically design and manage their farm. System Using IoT as Inclusive Technology”: Internatnal
Journal of Innovative Technology and Research Volume
5.2. MQTT Gateway 4,number 1, December-January 2016.
The MQTT Gateway is a cloud application that acts as an [8] Narayanaswami, Chandra, and Mandayam T. Raghunath.
intermediary for all messages between the FarmBot web "Application design for a smart watch with a high resolution
application and FarmBot device. It handles socket connections, display." iswc. IEEE, 2000.
device identification, and authentication issues.

5
[9] Hong ZHOU, BingWu LIU, PingPing DONG, “The [19] Nikesh Gondchawar, R.S.Kawitkar, “IoT based Smart
Technology System Framework of the Internet of Things and Agriculture”: International Journal of Advanced Research in
its Application Research on Agriculture.”: School of Computer and Communication Engineering june 2016.
Information Science & Technology, Beijing Wuzi University, [20] Karan Kansara, Vishal Zaveri, Shreyans Shah, Sandip
Beijing, China. Delwadhkar, Kaushal Jani, “Sensor based Automated System
[10] Shifeng Fang, Li Da Xu, Yungiang Zhu, Jiaerheng Ahati, with IOT: A Technical Review”: International Journal of
Huan Pei, Jianwu Yan, Zhihui Liu, “ An Integrated System Computer Science and Information Technologies, Vol 6.
for Regional Environment Monitoring and Management [21] J. Wu, Z. Liu, J. Li, C. Gu, M. Si and F. Tan, “An Algorithm
Based on Internet of Things”, IEEE TRANSACTIONS ON for Automatic Vehicle Speed Detection using Video Camera”,
INDUSTRIAL INFORMATICS, VOL. 10, NO.2, MAY 2014. International Conference on Computer Science & Education,
[11] Clement Atzberger, “Advances in Remote Sensing of 2009 IEEE, ICCSE 2009, pp. 193-196, Nanning 25-28 July
Agriculture: Context Description, Existing Operational 2009
Monitoring Systems and Major Information Needs”, Remote [22] D.Rajesh, “Application of Spatial Data Mining for
Sensing 2013, ISSN 2072-4292. Agriculture”, International Journal of Computer Applications
[12] Y. Kim, R. Evans and W. Iversen, “Remote Sensing and (0975-8887) Volume 15-No.2, February 2011.
Control of an Irrigation System Using a Distributed Wireless [23] Venkata Naga RohitGunturi, “Micro Controller Based
Sensor Network”, IEEE Transactions on Instrumentation and Automatic Plant Irrigation System”, International Journal of
Measurement, pp. 1379–1387, 2008. Advancements in Research & Technology, Volume 2, Issue4,
[13] Manikantan Krishnaswamy Ramakrishna, Prof.Jeff Dimaio, April-2013.
“An Economical Wireless Sensor Network Based Solution [24] N. Prabhakar, V. Vaithiyanathan, A. P. Sharma, A. Singh and
for Precision Agriculture” April, 2016. P. Singhal, “Object Tracking Using Frame Differencing and
[14] “Towards SMART FARMING, Agriculture Embracing the Template Matching”, Research Journal of Applied Sciences,
IoT Vision”: Beecham Resarch Ltd. Engineering and Technology, Vol. 4(24) pp. 5497-5501,
[15] Hemlata Channe, Sukhesh Kothari, Diplai Kadam December 2012.
“Multidisciplinary Model for Smart Agriculture using [25] M. Piccardi, "Background subtraction techniques: a
Internet-of-Things(IoT), Sensors,Cloud-Computing,Mobile- review", International Conference on Systems, Man and
Computing & Big-Data Analysis” Int.J.Computer Cybernetics, 2004 IEEE, Vol. 4, pp. 3099-3104, 10-13
Technology & Applications, Vol 6 (3). October 2004.
[16] Keerthi.v, Dr.G.N.Kodandaramaiah, “Cloud IoT Based [26] H. Zhang and K. Wu, “A Vehicle Detection Algorithm Based
Greenhouse Monitoring System” : Int.Journal of Engineering on Three-frame Differencing and Background Subtraction”,
Research and Applications Vol5, Issue 10, October 2015. Fifth International Symposium on Computational Intelligence
[17] Steve Sonka, “Big Data and the Ag Sector: More than Lots of and Design, 2012 IEEE, ISCID 2012, pp. 148-151,
Numbers”, International Food and Agribusiness Hangzhou 28-29 October 2012
Management Review Volume 17 Issue 1, 2014. [27] I. Kartika, and S.S. Mohamed, "Frame differencing with
[18] G.V.Satyanarayana, SD.Mazaruddin, “Wireless Sensor Based post-processing techniques for moving object detection in
Remote Monitoring System for Agriculture Using ZigBee outdoor environment", 7th International Colloquium
and GPS”, Conference on Advances in Communication and on Signal Processing and its Applications, 2011 IEEE, CSPA
Control Systems 2013 (CAC2S 2013) 2011, pp. 172-176, Penang 4-6 March 2011.

Вам также может понравиться