Professional Documents
Culture Documents
CHAPTER 1
INTRODUCTION
Precision Agriculture addresses these issues by saving water resources utilized for
cultivation, increasing the crop yield by continuously sensing the soil moisture, soil
humidity, soil temperature, soil pH, and other supporting tasks. In many developed
countries, precision agriculture is entirely IoT based. It successfully moves towards
the increase of the crop yields noted by Shifeng et al. (2011).
The two significant challenges in precision agriculture addressed by Yuxiao et al.
(2010) are
1. To bring the awareness of utilizing the procedure of the IoT-based framework
into its economical implementation.
2. Balancing the challenges and make effective utilization of sensors.
Technological advancements have shown their improvements in various fields such
as business management, medicine recommendation, weather prediction models,
education system, and communication systems. However, its contribution to
agriculture is less. Most of the people whose livelihood depends on agriculture are in
rural areas where they do not have facilities of technology in the agriculture sector to
yield high production, as pointed by Yuxiao et al. (2010).
1.2 WSN-BASEDPRECISIONAGRICULTURE
In Wireless Sensor Network (WSN), numerous sensor nodes are
interconnected, limiting the processing capability. A considerable number of sensor
observations get transmitted to the sink node, increasing network complexity. Zhu G
(2009) has stated the significant drawback in everyday internet usage, as the Internet
is a virtual network of devices, plays a significant role in any application considered.
The author Prabha et al. (2021) pointed that IoT makes a network of internet-
connected objects capable of collecting or changing information that uses embedded
sensors or devices. It primarily connects a machine with an IP address somewhere
into the world. It can be linked in conformity with any other device, anywhere within
the world known as the Internet. IoT requires a unique paradigm to provide optimal
network performance in sharing accurate data in the wireless sensor network.
Since the WSN is data-centric, data processing is a critical problem in the WSN
network, which determines the performance of the WSN. Usually, the sensor node is
battery powered with limited energy stated by the author's Pierce and Elliott (2008).
In precision agriculture, an enormous amount of data is acquired and processed. This
idea of precision agriculture using WSN will help the farmers increase the crop yield
by applying modern technologies as a replacement for conventional farming
practices. Lea-Cox et al. (2008) and green. O et al. (2009) has stated that agricultural
yield is incredible. Smart Irrigation using IoT with machine learning techniques
plays a significant role in optimum water-resource utilization.
It causes as ever effect on plants if the plant is not adequately cared for
frequently. As a result of this, respective product quantity, productivity, or quality
will be affected. The recommended technique proves to be favorable in monitoring
huge crop fields. This system helps in predicting the irrigation planning based on the
irrigation requirements with the various sensing parameters considered as pointed by
Gutiérrez. J et al. (2013).
SMARTFARMING
CropStorage
Advanced agricultural system aid in IoT due to their efficient and easy maintenance of
the farm. Also, creating this type of automation system helps to water the plants without
human assistance. Moreover, it will have the option of planting seeds, measuring soil
moisture. The various functional elements of precision agriculture presented in Figure 1.2.
The aspects such as sensing, evaluating, and treatment are primarily involved in monitoring
the pests causing crop diseases. Recently image processing is the critical technology in
which raw images captured from the agriculture field, analyzed to characterize the insect
types for pest control. These IoT based devices for pest control has reduced the overall
expenditures by supporting the restoration of the natural climate.
The agricultural industry focuses on IoT technology is to gain benefits within a short
period, with low investment making farming innovative and high-tech utilizing IoT devices
in the farming field. Gubbi et al. (2013) has stated that due to the IoT utilization in the
irrigation field, there is a significant decrease in manual and physical work. The sensors
implemented in the irrigation fields monitor and store the data for its use in the future
process. Advanced sensors connected to the cloud network used for monitoring. The authors
TongKe (2013) and Roopaei et al. (2017) stated that the data obtained through the sensors
are more accurate and precise, supporting the IoT cloud environment for further prediction
process.
The IoT can bring out lots of benefits to the farmers like water monitoring, water
usage limitation, crop monitoring, pre-prediction of the environmental impact, less human
resources, increased crop productivity, improved harvesting yield, proper field management.
The author Bhowmick et al. (2019) noted that from sowing the seeds to selecting crops
based on the environment, IoT could support them. It helps the farmer to choose the right
plant which can uphold in a particular situation. The complete IoT setup monitors the whole
field and takes a survey on the environmental factors, finding and reporting abnormalities by
triggers and alerts.
Smart decision support system using IoT, as shown in Figure 1.3and machine
learning algorithms are becoming an essential factor for agricultural practices, as Viani et al.
(2017) and Balducci et al. (2021) stated. Machine learning helps enhance and modernize the
irrigation system to meet the increasing demand for the irrigation system. The intelligent
irrigation planning system can monitor water conservation for irrigation. (ACRIS
Multiparameter Optimization System for Precision Agriculture)
a) Weather Condition
The authors Mizunuma et al. (2003) point those changes in climate due to weather have
a severe impact on agriculture. From the observation of climate studies, several experts in
the agriculture sectors have suggested that there will be a decrease in agricultural production
of 30% by the year 2050 due to the adverse climatic changes. Climatic changes influence the
identified agriculture parameters, decreasing the quality of the yield. A quick relief to this
problem is using IoT to adapt and address climate change.
c) Usage of Fertilizer
The quantity of the fertilizer applied during the farming activity has a significant
responsibility in the harvest. Each farmer should make accurate decisions on the usage of
chemicals based on crop specifications. The impact of fertilizer usage kills the nutrient
content of the plant and remains a significant reason for many acute diseases.
d) Soil Variations
The main success of agriculture is the soil, which treated as a significant component in
farming. The soil moisture is sensed and monitored automatically under precision
agriculture. The modern devices for agriculture monitoring are available to watch soil
statuses, such as gamma-radiometric soil sensors, a soil moisture sensor and electrical
conductivity sensors. The other sensors used for monitoring the climatic condition
enumerate the physiological status of the crops.
earlier by the farmer. The water is not a single factor-dependent; it depends on various
factors like climate, season, soil type, crop types, and growth stage. In general, during
cultivation, the crop loses water through transpiration and evaporation.
1.7 YIELDMONITORINGANDFORECAST
Yield monitoring analyzes the range of factors corresponding to the agricultural yield,
as shown in Figure 1.4, such as moisture content and the quantity of harvested crops. It
accurately confirms and verifies the crop recordings to generate the wet level estimations
and identify how properly the plant has done is working and what has to do next. Crop
forecasting is a work possible based on the yield harvest, which supports the farmers in
planning their future requirements and making appropriate decisions.
Therefore, it is not surprising that the quality of human capital endowment of the
economic agent and the type of enterprise operated greatly influence the choice and use of
information. Endogeneity is a pervasive problem in precision agriculture technology.
Besides, the motive of enhancing activities like access to capital, off-farm income, and farm
income by potential end-users of innovation can positively affect their adoption of a novelty,
pointed by Nikkila et al. (2010).
Developing countries focus on a cost-effective smart irrigation system and need for
countries like India, where 4% of freshwater is serving17%oftheworldpopulation.
Henceforth, the smart irrigation system with an efficient strategy for water resource
utilization is modeled based on the latest technologies. Most of the irrigation system depends
on a few essential parameters such as precipitation, evaporation, and soil moisture for
designing the smart irrigation system described by the authors Shang et al. (2007) and Davis
et al. (2010). By imposing the brand new sensing of IoT in agriculture, the standard farming
elements substantially changed. Currently, the integration of wi-fi sensors unimaginably
promotes agriculture with smart farming. Its limitations on dense farming issues, yield
optimization, drought response, land suitability, irrigation, and pest control are the
challenges that still need to address.
1.9 ADVANCEDAGRICULTURALPRACTICES
Food quantity and quality enhanced by adopting the novel methods. The methods
practiced for centuries. To increase crop production, focusing on fertilizers, seeds, etc are
required. This methodology is not possible with conventional strategies, and there exists an
adequate demand in meeting this gap. The scientists think alternate approaches on food
supply such as Bioengineered, Genetically Engineered foods challenge DNA and genetic
issues.
Several studies highlight that these foods cause severe human effects such as
infertility, human health disruptions, ageing, fault insulin regulations, etc. Similar
technologies acceptance in society leads to bio-organic food products. Research conducted
for decades helps in conventional agriculture processes improving the yield with minimum
effect. Sophisticated and controlled techniques are required to tackle these issues and move
towards cultural and urban farming, for which the usage of advanced techniques based on
sensor technologies is the mandatory requirement today.
1.10 MOTIVATIONOFTHESTUDY
Today agriculturalists are involved in maximizing production with minimum costs.
This method requires a novel approach to address the issue and increase yield and
profitability, reducing the environmental impacts of food production. To incorporate
precision agriculture in the cultivation field requires various ecological parameters that in
directly as sess the above issue. However, the intra-field variability in sugar beet yield and
quality is generally unknown; therefore, it needs to quantified concerning soil properties and
micro-climate conditions. The observed variability in the crop environment can then be
managed by customizing the inputs to the areas where they are needed. Therefore the
precision agriculture model has been utilized and advanced here following the crop growth
accounting the field variations such as air temperature, soil moisture, and soil type.
precision agriculture improve the Indian agricultural sector's. The study proposed focuses on
The identification of limiting factors that affect the crop production.
Chapter 1 gives an overview of the research problem, itsneed, listing the use of various
machine learning techniques in agriculture, the issues and challenges faced in the agriculture
field, and the research study's objective and contribution.
Chapter 2 offers a background for the rest of the thesis. This chapter presents a
comprehensive literature survey on the use of machine learning techniques in precision
agriculture.
Chapter 3 presents the overview of the proposed research, named "ACRIS: Agriculture
Cultivation Recommender and Smart Irrigation System," constructed in 3 modules. This
chapter describes the first module of ACRIS, "ARV Recommendation System: Agriculture
Factor-Based Relevance Vector Analysis Model for Identifying Accurate Farming
Recommendations", in concern with providing favorable recommendations for farmers who
utilize IoT in Precision Agriculture for increasing their cultivation yield.
Chapter 4 discusses the second module of ACRIS, "AISM System: Advanced Irrigation
Planner based on Soil Moisture Prediction for Farmers in Precision Agriculture", the module
predicts the soil moisture and plans the irrigation for the farmers involved in precision
agriculture to reduce the water consumption utilized for cultivation and increase the
production yield.
Chapter 5 presents the final module of ACRIS, "AMOP System: ACRIS Multi parameter
Optimization System for Precision Agriculture", the module optimizes the water level for
future irrigation decision-making to maintain good crop growth and water stability in
precision agriculture by comparing the utilization of water content during the various stages
of plant growth.
Chapter 6 concludes the thesis by providing a summary of each contribution and describing
the entire architecture of the study ACRIS: Agriculture Cultivation Recommender and Smart
Irrigation System," for farmers in Precision Agriculture to support them for decision-making
based on the observed information processing concerning the smart irrigation planning and
high yield crop production reducing the utilization of water.
1.12 SUMMARY
In this chapter, the need for agriculture, an overview of precision agriculture, how
IoT used in precision agriculture, significant issues related to agriculture in IoT, the
advanced agriculture practices, the motivation for the research, and the thesis contribution
dealt with in the introduction.
CHAPTER 2
LITERATURESURVEY
2.1 INTRODUCTION
Agriculture, a required field, needs to be balanced with the population growth. Alan L.
Olmstead and Rhode (2009) observed that the modern agricultural system is commercialized
and labour-intensive. The input resources used in the early year 1920 and 1970 were 30 %,
producing a 180% yield. Moreover, the yield increase was not expanding the utilized data
sources, rather than the innovation on productive farming. Researchers have recently
identified that the utilization of sifting machines, mechanical innovation, and synthetic
manures calculate agricultural profitability. The agriculturists began utilizing information
and communication technology to system at their money-related information and monitor
their business interchanges with outsiders from the past decade. Nowadays, data assumed to
be an essential part of an individual's life. In this manner, the horticulture industry allows the
farmers to gather and assess farming measures using the observed data. For example,
sensors, cultivating hardware, meteorological sensors are proficient in imparting the correct
data.
The endeavors cover all means in the generation chain involving the day by day
agricultural tasks. Wang et al. (2006) and Sorensen et al. (2010) reported that farmers often
experience an overload of information, which creates a different form of the data source.
Wang et al. (2007) applied the sensors to monitor temperature, soil moisture, and humidity
and communicated the sensed information to the farmers by informing them through third
parties like meteorological stations. At that point, farmers combined the information
effortlessly and had clear choices to deliver individual items, enhancing their salary and
legislative standards (Lan (2012) and Razi and Nath (2019). McCown R.L additionally
suggests a similar thought et al. (2012), data gathered with the farmer's interior arrangement
creates knowledge to learn and build a genuine intellectual framework.
Allen and Wolfert (2011) proposed numerous proprietary solutions to help farmers
effectively monitor their farms. Nikkila Retal. (2010) found more complex frameworks that
track to zoographical regions and climate changes. In later days, Farm Management
The sensor technology in the agricultural field solves issues related to the yield and
proposed proper monitoring. The authors, M. R.
M. Kassim et al. (2014) and Sahota et al. (2010) elucidated the Network layers' sensor
technology. Mampentzidou et al. (2012) provided the fundamental role of sensor technology
and the essential components in agriculture. Precision Agriculture Monitor System (PAMS)
monitors agricultural work using a sensor proposed by Shining Li et al. (2011). IFarm
Framework system suggested for managing the water consumption for improving the
production by enhancing the socioeconomic factors. Wang N et al. (2006) offered an outline
for the latest sensor technology designing for ecological nursing. Anisi M.H et al. (2015)
classified the sensor technology based on its performance metrics. Million
innovations. Using this system, the farmer can manage the irrigation shortcomings of today's
agriculture as proposed by Batte (2005) and Csoto (2010).
2.2 PRECISIONAGRICULTURE: AREVIEW
Precision agriculture considered an improved input management strategy, which can help
farmers to increase the farm's profits (through the efficient use of farm inputs), and at the
same time, enhance the environment-adoption of soil sampling technology described on the
intertemporal soil sampling. However, there is no information on farmers' behaviour
concerning the extent of sampling. The farmers ' choices and insights must be carefully
understood to ensure the promotion and use of these efficient irrigation technologies
(Riquelme et al., 2009). This study contributes to the understanding of the farmer in concern
to the adoption and use of efficient irrigation technologies.
According to Allen and Wolfert (2011), farmers attracted to use precision agriculture
practices because of their potential to improve on-farm profits and the environment.
Currently, the leading technologies that are associated with this agricultural practice are
Geographical Information Systems (GIS), Global Positioning Systems (GPS), Sensors,
Variable Rate Technology (VRT) and Yield Monitoring (YM). The adoption of precision
agricultural practices was first reported in the Midwest's Corn Belt, which helped increase
corn, wheat, and soybean yields, as well as reduce production costs, an upward adoption
trend of precision farming in 14 Southern states, from about 63 percent in 2009 to about 73
percent in 2013.
The most widely adopted precision agricultural technologies in the southern U.S states are
grid and zone soil sampling, variable rate application of lime, phosphorus, and potassium,
and soil survey maps (Costo 2010). The southern U.S, including Alabama, Georgia,
Mississippi, North Carolina, South Carolina, and Texas, is the most critical geographical
cotton production region in the U.S (Zwart and Bastiaanssen2004). For instance, in some
Texas counties, cotton production exceeds about 50 percent of harvested acres. Most
southern U.S states perceive other farmers, farm equipment sellers, crop consultants,
university extension, news media, and government agencies as essential sources of precision
agricultural practices (Zwart and Bastiaanssen 2004). So far, no technique exists, which may
help in figuring out irrigation practices over massive areas.
EoverallEtEm (2.1)
The maximum temperature (°C), and Tmin is the minimum temperature (°C) and T is the
extra-terrestrial radiation (MJm−2day−1). Temperature
R
a
and solar radiation are utilized for estimation, which is represented by Equations (2.3) and
3
(2.4).E .3.8710 .S.(0.6T 0.4T 29) (2.3)
To 1 R max min
∘C
5T 35 1.1
max 1
max ∘C 1 max
T 35 1.10.05(T 35) (2.4)
∘C
T 5 0.01exp[0.18(T 20)]
max 1 max
The ultimate goal of the proposed approach is to make effective irrigation decisions
based on optimum water utilization. Prediction of soil moisture is crucial for the powerful
irrigation management system. Evapotranspiration plays an vital in the estimation of soil
moisture.
requirements and essential strategies concerning water, save from rain and plan irrigation.
Conservation of water from rain-fed and performing irrigation in practical has to raise in
saving water resource efficiently. As the water use in each manufacturing unit is growing
day by day. The interim climatic changes and the parameters in concern, such as
evapotranspiration, decide the rainwater existence (Rawal 2017). Thus the evaluation of
climatic parameters concerning crop production is essential to avoid water crisis and food
shortage globally.
The farmer no longer accesses the actual equipment flow, the theoretical device flow
rate based totally on limiting beneficial resource flow fees for 24 h according to today
observed.TIC calculated as in Equation(2.5).
TIC(mm/day)=IrrigatedArea(ha) 240.1
More precise values on crop water intake provided utilizing the lysimeters. This
method is the same old technique for the direct measuring EvapoTranspiration (ET) by
Nikkila et al. (2010). The devices notice the variations in a plot for a decided crop through
all experiments. The device data with additional statistics received from meteorological
station used for the water management process. These records have periodically gathered the
use of numerous devices, and after those records are analyzed, an irrigation control choice
made. Using ICT, this executed quickly and in real-time stated by Wang et al. (2007).
Additionally, technologies related to automation and far away from management, such as
electronics, automation, sensing technologies, and robotics, have advanced during the last
decade, primarily due to the decrease in the production fees of the fundamental components.
As described by the authors Razi and Nath (2019), various gadgets widely implemented in
enormously technical sectors.
Reference crop EvapoTranspiration (ETo) estimation is essential in irrigation water
manipulation to calculate crop water requirements and its scheduling, including the rainfall-
runoff modelling and several exceptional water resources media. Based on its significance,
several direct and indirect techniques hired to estimate the reference crop evapotranspiration.
However, its achievement has limited since direct measurement strategies lack precision and
lack of accuracy due to scale troubles related to more accurate indirect techniques. For
instance, the Penmane Monteith benchmark model is highly non-linear and requires weather
input records that not mechanically monitored. Artificial Intelligence (AI), Neural
Computing strategies that may efficiently map complex, non-linear input-output
relationships provide a helpful alternative in such situations.
For a Wireless Sensor Network (WSN) (Ojha et al., 2015) scenario with ZigBee
(Razi Qet al. 2019) and (Gutierrez et al. 2014), the technology uses multiple sensor nodes in
a specified area. González Perea R et al. (2016) proposed a hybrid heuristic methodology
that combines Decision Trees with the Genetic Algorithm to find the optimal decision tree to
model farmer's behavior, predicting the occurrence of irrigation events. The method tested in
an entire irrigation district. Results showed that the optimal models developed have been
able to predict between 68% and 100% of the actual irrigation events and between 93% and
100% of the adverse irrigation events.
knowledge regarding the character of these complicated relationships is not required. They
excel in applications that call for non-linear and multidimensional mapping of inputs to
outputs. ANNs are tolerant of noisy facts and generalize properly within a problem region
wherein the information can be poorly allotted or spotty. Once learned, neural networks can
compute predictions rapidly with a high diploma of accuracy and integrated smooth
algebraic equations into large programs. The main criticism of ANNs is that the reasoning in
the back of the enter-output courting mapping is opaque, leaving us primarily unable to
comply with and interpret how choices made. Some techniques exist to analyze
knowledgeable neural networks, but they are commonly involved, no longer robust, and can
prove challenging to interpret even the relative importance of entering features.
Finally, locating a most effective answer that reaches the minimal global errors for
the trouble is not always assured with a neural network, even though they often outperform
more excellent conventional strategies. Methods for changing step size encompass steadily
decreasing it over time, the usage of momentum phrases that allow past steps to contribute
through helping keep away from community minima, or permitting adaptive sizing based
totally on whether the error fee is enhancing or getting worse at every step noted byPuertoa
et al. (2013). The outstanding venture of the agricultural area is to produce greater food from
much less water, which done by growing Crop Water Productivity (CWP) by the author Lan
(2012). Globally measured common CWP values in keeping with unit water depletion are
1.09, 1.09, 0.65, zero.23 and 1.80 kgm−3 for wheat, rice, cottonseed, cotton lint, and maize,
respectively as referred by Zwart and Bastiaanssen (2004). The range of CWP is extensive
for vegetation like wheat, rice, cottonseed, cottonlint, and maizeand consequently offers
impressive opportunities for maintaining or increasing agricultural manufacturing with 20–
40% much much fewer water resources. The variability of CWP can ascribed to: (i) climate;
(ii) irrigation water manipulation and (iii) soil (nutrient) manipulation, among others. The
vapour pressure deficit is inversely associated with CWP. Vapour strain deficit decreases
with latitude, and for this reason, favorable areas for water irrigated agriculture located at the
better latitudes. The most excellent stop is that CWP can be increased notably if irrigation is
decreased and crop water deficit is independently induced as stated by Zwart and
Bastiaanssen (2004).Figure2.1depicts the representation of the single artificial
neuron.Figure2.2provides the network structure of ANN.
Figure 2.2 Representation of the Artificial Neural Network (ANN) (Zwart and Bastiaanssen
(2004))
S.No. Factors
1 MoistureContent
2 Density
3 PlasticityCharacteristics
4 AmountofFineParticles
5 Gradation
6 OptimumMoistureContent
7 MaximumDryDensity
8 ParticleShape
In current years many packages had been proposed for Wireless Sensor Networks
(WSN) stated by Riquelme et al. (2009). One of the precision agriculture-based suggestion
pointed by Anisi et al. (2015), where WSN can play an essential part in dealing with the
control of water assets for irrigation. It provides information about the adjustment within the
crops to evaluate the optimum point for harvesting and estimating fertilizer necessities to
predict crop performance very accurately.
Abrisqueta et al. (2015), focused on a model for a district-level irrigation plan and the
identification related to this. The proposed model was the advanced version of I2 improving
Perea et al. (2019), showing an R2 variation of 93%. The main reasons behind these results
were the usage of the Bayesian framework with the optimized parameters of the genetic
algorithm. It proves that Bayesian changes have shown variations in previous versions and
the rules set. The rule's input validation no longer requires a separate statistics set isolated
from the training set when the sampled records or databases are minor in size. These data
contain statistical information on irrigation relevant to various seasonal weather and crop
conditions. Thus the usage of the Bayesian framework has provided an overall performance
improvement in the early prediction of weather situations, as noted by Perea et al. (2019).
2.6 SUMMARY
In this chapter, the endeavors of all generation agricultural tasks along with various
survey conducted on precision agriculture, its empirical estimations in decision-making are
presented. Also, review on various neural computing techniques and various crop growth
influencing factors are discussed here.
CHAPTER3
AGRICULTURE FACTOR BASED RELEVANCE
VECTORANALYSISMODELFORIDENTIFYING ACCURATE
FARMING RECOMMENDATIONS
3.1 INTRODUCTION
The Internet of Things Role in Precision Agriculture has significantly developed novel
techniques that can feed nine billion people by 2050. Nowadays, farmers collect real-time
data from sensors and monitor the farming fields to effectively harvest higher yields and
gain profits, as stated by Christine Dewi and Rung Ching Chen (2019). In most developed
countries, farmers deploy sensors in farming fields to automate the farming task by utilizing
real-time environmental conditions.Considering India,diverse climatic conditions yield
varieties of vegetables and fruits around the entire year, making India stand the second-
highest in agricultural production.
The author Firasath Nabi et al. (2020) suggested that farmers’ requirements on solutions to
reduce crop loss, plant disease prevention, etc. to produce higher yield remains a challenge.
Recent technological advancements such as precision-based agriculture exploit the use of
IoT, leading to sustainable agriculture growth. IoT in precision farming collects and transmits
the huge amount of field data, analyzes and makes accurate farming decisions as stated by
Bauer and Aschenbruck (2021). Agronomist suggestions on the suitable environmental
factors that affect crop growth a represented inTable3.1.
The system identifies environmental factor variations and recommends its integration
in the field, meeting the site-specific requirements of modern agriculture practices described
by the authors Christine Dewi and Rung Ching Chen (2019). The functioning of the system
is classified into three phases, as depicted in Figure 3.1
The primary objective in any cultivation is the rapid germination from the weed-free
seedbed. For which the soil has to be prepared and loosened, this significantly requires time.
Manyresearchers havedeveloped thebest sensor design oflowpower and low cost to acquire
field data.The author Firasath Nabi et al. (2019) suggestion of (VH400) Soil Moisture sensor
and (DS1822) Soil Temperature shown the best achievements in the single-tier
heterogeneous network with (XBee-PRO S2) ZigBee module and (MTSMC-G2-SP) GPRS
module working on solar-powered in the agriculture field irrigation attracted as one of the
primary consideration in this research. Thus the environmental factors measured were the
soil temperature, soil humidity, and water level.
Algorithm3.1–ARVRecommendationAlgorithm
Input: Sensor Data (Soil Temperature(Degree Celsius) andRelative
Humidity)
Output:CultivationRecommendation Begin
TrainARV
BuildtrainedARVmodel
{
Get the current sensor dataPerformtherelevantvectoranalysis
{
𝒚(𝒙,𝒘)=∑𝒏 𝝎𝒎𝝋𝒎(x)+𝝎𝟎=𝛚.𝛗
𝒊=𝟎
}
if(Predictedvalue==Estimatedvalue) Provide irrigation recommendations else
GOTOARV
}
Display(AnalyzedData);
Display(RecommendationforFarmer); End
The data collected were processed and stored on a centralized database. Initially, the
crops may depend on high radiations. Later as stages get evolved, high temperatures may
harm the yield. During such times, farmers opt ventilation strategy to control the
temperature. High humidity shall expand the ailment. Similarly, yield influenced by the soil
water. Hence the above motivation on the impact of temperature, humidity, and water level
in cultivation created insights in considering these factors as important data information in
this research. The data collected through the wireless channels for its analysis by
agriculturalists provides real-time decision support to the farmers.
0 Perfect Fit
Negative Invalid
p
predicted value. The MSE calculation expressed in Equation (3.1)
yˆ
p1 y p
p
MSE
N
2
(3.1)
N ation
for the
where observ
1 N Is the ed and
error predict
calcul ed
N p1
vector
RMSE
N
p1
y y
pp
2
(3.2)
N ˆ
N 2
yy ˆ
p p
NSE1 p1
(3.3)
N 2
p1
ypyp
where,
yˆp, the forecasted agricultural parameter
y
,the mean of the observed agricultural parameter
1 PerfectFit
0 Accuratelybasedonthemeanoftheobserveddata
The ARV Recommendation is applied to find the land position where all the climatic
factors are in good condition and preferable to promote cultivation (Wilk Sampaio de
Almeida et al. 2021). The analyzed parameter prediction is evaluated and compared with the
well-known precision agriculture level predictor APAL (Australian Precision Agricultural
Laboratory) for its standard. APAL delivers high-quality, independent soil, plant, and water
analytical service benchmarks to enable better agronomic decisions. (Bramley, R. G. V.,&
Janik, L. J. 2005) Considering the achievements of the applications benchmarked the APAL
parametric values, the APAL level set as a benchmark in evaluating the performance of the
algorithm ARV Recommendation. The validation of the proposed algorithm concern with
the RMSE compared with the existing ANN RMSE concerning data reliability and accuracy
3.3 EXPERIMENTALRESULTSANDOBSERVATIONS
Data Collection, Experimental Setup, and Sensor Deployment
The environmental factors considered for measurement are soil temperature, soil
humidity, and water level. The ARV Recommendation algorithm used the real-time sparse
dataset collected from the agriculture field, considered the training samples. Plants of 100
varieties arbitrarily picked as testing samples. The implementation carried out in the
agriculture field in a tropical climate. The sensor observations recorded over one month at 8
am, 10 am, 12 pm, 2 pm, and 4 pm each day. The humidity sensors utilized to catch
dampness information at a consistent interval of every 60 minutes.
In the agricultural field, the sensors placed at a distance of 8m for sensing the
microclimate throughout the entire growing season. The author Firasath Nabi et al. (2019)
suggestion of Soil Moisture Sensor (VH400) and Soil Temperature (DS1822) utilized in the
agriculture field irrigation is attracted as one of the primary consideration. Sensors utilized
to measure soil temperature, soil humidity, and pH value sensor. The sensed parameter used
a control unit. Here the considered agriculture field is divided into equal tubs, each tub of
one unit area of length 8m and breadth 7m as depicted in Figure 3.2
Each tub consists of two sensor nodes. The specific location fixed for each sensor
node will enable detecting the fault nodes. The procedure applies the fault tolerance and
check pointing based IoT technique. For a considered agricultural field, the number of
1 1
1(N d fw)1
fw fw
Where Ns represent the number of the required sensor nodes,Nd the number of
divisions in the agricultural field andfwrepresents the field width. The sensor nodes
distributed according to the field width. The sample deployment of sensors in the
agricultural field of the total area 105.8m X 39.7m depicted in Figure 3.3.
Figure 3.3 depicts the data acquisition boards, such as MDA300 sensing the moisture
level of the soil. These boards built with the processor unit, which performs processing and
storage. Figure 3.4 depicts the hardware components on a single sensor.
Each sensor consists of i)SensingUnit, ii)ProcessingUnit, iii) Transceiver for
communication between the WSN node and the base station, and iv) Power Unit. Figure 3.4
depicts the architecture of the sensor, along with its communication module. Finally, the
ARV Recommendation algorithm is implemented in MATLAB (2021), setting 5-fold cross-
validation to evaluate the training and testing samples. The crop yield calculated using the
grain flow rate and the area covered as expressed in Equation (3.6) (Rohit Kumar Rajak et
al. 2017)
46.29
47.22
46.89
46.26
46.58
47.68
45.86
45.43
46.56
45.17
45.06
45.93
45.02
43.83
44.05
44.34
45.58
43.77
26.56
25.67
26.98
25.14
26.78
26.33
25.04
25.77
25.87
24.87
25.02
24.56
24.79
24.98
24.06
25.23
24.59
23.89
24.62
24.73
23.43
60
50
40
JITS College of Engineering, Kalagampudi Page 42
30
20
10
Relative Humidity
0
RMSEValidation
0.9
0.8 0.77 0.769
0.74 0.737 0.739
0.693
0.67
0.7
0.576 0.597
0.6 0.55 0.563 0.554
RMSE
0.469 0.474
0.5
0.4
0.3
0.2
0.1
0 1 2 3 4 5 6 7
RMSE(ExistingANN)
Days
RMSE(ProposedARV)
Figure 3.5 Environmental Factor-based Condition Analysis
RMSEValidation
0.9
0.4
0.3
0.2
0.1
0
1
[Type
2
the
3
document
4
title]
5 6 7
Days RMSE(ExistingANN)
RMSE(ProposedARV)
Cultivation
Mode of Water Consumption Total Water
(No. of Days
Cultivation (In ml/Day) Consumption(ml)
Considered)
Thus the ARV Recommendation System can support both the farmers and plant
pathologists. These include
Provide accurate information on temperature, humidity, and water factors.
3.5 SUMMARY
This chapter discussed the framework and experimental results of the module “ARV
Recommendation System: Agriculture Factor-Based Relevance Vector Analysis Model for
Identifying Accurate Farming Recommendations”. The parameter which holds the smaller
RMSE considered as the sufficient parametric conditions for cultivation. Figure 3.6 shows a
better ARV RMSE value of 0.042, 0.038, and 0.0.35 for the first 3 days compared to the
RMSE of the existing ANN model.
Further, the volume of water required for irrigation predicted in this module. The
irrigation planning and monitoring of the proposed “ACRIS: Agriculture Cultivation
Recommender and Smart Irrigation System” is explained in Chapter 4.
CHAPTER4
ADVANCED IRRIGATION PLANNER BASED ON SOIL MOISTURE
PREDICTION FOR FARMERS INPRECISION AGRICULTURE
This chapter presents the second module of ACRIS, “AISM System: Advanced
Irrigation Planner based on Soil Moisture Prediction for Farmers in Precision Agriculture”,
the module predicts the soil moisture and plans the irrigation for the farmers involved in
precision agriculture to reduce the water consumption utilized for cultivation and increase
the production yield.
4.1 INTRODUCTION
The primary objective of the proposed module, "AISM System", is to significantly
optimize the utilization of the water resource and support farmers involved in precision
farming. Irrigation plays a major role in crop cultivation. Ignorance of this will cause a
rigorous effect on plants affecting the quality and productivity of the yield. The developed
AISM System is favorable in monitoring larger area crop fields. Prediction of soil moisture
can support irrigation management in agriculture. Irrigation planning will support the farmer
in evaluating the water storage, refilling the estimation rate, and fulfilling the economic
demands of water by maintaining sustainable soil water resources (Rahim Barzegaret al.,
2017).
The emergence of technology has a significant role in the agriculture domain. Today
people prefer automation in all processes, avoiding human assistance. Artificially applying
water to the agriculture field is termed irrigation (El Mezouari et al. 2019 and Masseroni et
al., 2020). Many researchers have focused on an intelligent irrigation system to automate
and schedule watering needs. Soil moisture is an important factor in this activity. The
proposed system AISM aims at minimizing water usage. Sharmin Akter et al. (2021)
statement in the previous chapter describes that there will be an increased nine billion
population in the year 2050. Also, the results in Chapter 3 show that during the daytime,
high temperatures and low humidity are reducing the soil moisture level. Soil moisture is
influenced by evaporation and precipitation (Amarendra Goap et al., 2021).
The primary objective of the module in this research is to predict the soil moisture
utilizing the machine learning algorithm and plan irrigation in the agriculture field, as shown
in Figure 4.1. The carried out work takes two input datasets i) dataset observed from the
agricultural field and ii) web dataset aggregated with weat her forecasting statistics such as
temperature and humidity from the webs forecasting portals such as Open Weather and Accu
Weather. The AISM System performs two functions. The primary function utilizes the novel
SLR (Sparse Linear Regression) algorithm for soil moisture prediction. The secondary
function utilizes the novel K_SLR (K_ Sparse Linear Regression) algorithm for smart
irrigation planning in the agriculture field. The original scheme of the SLR (Sparse Linear
Regression) algorithm is modified from the Linear Regression (LR) algorithm to suit the
needs of the agriculture domain. The K_SLR (K_ Sparse Linear Regression ) the hybridized
collaboration of the SLR (Sparse Linear Regression) and K-Mean Clustering algorithm. The
developed novel SLR (Sparse Linear Regression) algorithm is compared with the best
achievement shown by Support Vector Machine (SVM). SVM is a popular supervised
machine learning algorithm utilized for various challenges in classification and regression.
The individual observations are simply coordinated as support vectors. The SVM features
outliers and finds the hyper plane of exhibiting the maximum margin. The technique kernel
function takes low input space and converts it into high dimensional. That is, it considers the
complex data and transforms them into labels of required definitions. The classification is
validated by implementing the K-Mean Clustering algorithm in collaboration with the
Support Vector Machine (SVM) algorithm (KSVM). The results are compared with the
proposed K_SLR (K_ Sparse Linear Regression) algorithm fulfilling the objective of the
research described. Thus the proposed System plan the irrigation concerning the reduction of
water consumption to maintain healthy plants. The optimization of the proposed system is
described in the forthcoming Chapter 5.
4.2 PROPOSEDMODEL
The proposed “AISM System: Advanced Irrigation Planner based on Soil Moisture
Prediction for Farmers in Precision Agriculture" consists of two main functions, as depicted
in Figure4.1, the framework of the AISM System. Primary Function: Prediction of the Soil
Moisture Secondary Function: Irrigation Planning.
Algorithm4.1–K_SparseLinearRegression(K_SLR)
Input: Observed Sensor Data (Soil Temperature, Soil Humidity, Soil
Moisture), Web Data(Temperature, Humidity)
Output: Soil Moisture Prediction Begin
Initialization: Collection of field and web datasets; Begin//Training
Train SLR; Predict PSM; Train SVM; Predict PSM; End //Training
Begin//Prediction
Calculate the Predicted Soil Moisture Difference (PSMD);
SetP0,P1,P2,……Pn,predictiondifferencePSMDfortheday1,2, 3…n;
Set Centroid value==Pi; Threshold T==Tmax; For (i=0 to n; i++)
Apply K_SLR;
Predict GSM and MSE; For (Tmax>GSM)
Smart Irrigation() End For
End For End//Prediction
End
Algorithm4.2–K_SLR//SmartIrrigation()//
{
SetIrrigationdateID If (CD >= ID)
{
While(Tmax>GSM)
{
Relay==1;//irrigationstart
}
}
Else
{
Relay==0;//irrigationstop
}
}
End
Sparse Linear Regression (SLR) algorithm, transform the observed raw sensor data
to the vector representation. The kernel considers a pair of data points as a similarity
function. The kernel function is modified in accepting sequential data, images, and vectors.
For k, the similarity function, Zi training sample, and Z', the unlabeled input the weighted
similarity measure is expressed in Equation (4.1).
The novel approach to the usage of weather forecasting datasets with k means clustering in
the K_SLR algorithms 4.1 and 4.2 depicted in Figure 4.1.
Wi
z =sgn
n ¿
∑
k (z , z )
❑ i= i (4.1)
❑ Zi
The AISM System predicts the upcoming days' soil moisture from the given input
datasets. For irrigation control, a relay switch is connected to the water pump, enabled with
Wi-Fi. The interface of the web triggers the signal for real-time field monitoring. Next,
theirrigationplanningfunctionoperatesineitherautomaticmodeormanual mode, remotely
controlled. The observed level of ground soil moisture on the current date with predicted
precipitation of the previous date considered advanced suggestions for the irrigation
proposed.
The farmer schedules the irrigation at a specified soil moisture threshold. The Threshold T
considered with the predicted pattern of ground soil moisture, and previous precipitation
information schedules the irrigation task. The water motor controlset to ON and OFF
conditions. The R-Pi python module alerts the Arduino-Uno andcontrols the relay to switch
between the ON and OFF states of the water motor.
The minimum threshold„ Tmin‟and maximum threshold„ Tmax‟ described in the above
algorithm will initiate and end the irrigation, considering the recorded observations. Ground
Soil Moisture (GSM) predicted with MSE, and its comparison with T maxutilized to decide the
mode of irrigation, either as manual or auto mode. The signal value “1” for “ON” sent to the
motor. For irrigation to stop it sends “0”. Thus the AISM System predicts the desired
volume of water required for cultivation. The System has great scope for its application
support to farmers who depend on the terrestrial tank due to water scarcity. (Shramin Akter
et al. 2021).
Figure 4.2 Sensor Deployed in the Agriculture Field (Amarendra Goap 2021)
The standalone sensors, specifications presented in the table are deployed in the form, as
shown in the sample agricultural field, as depicted in Figure 4.2.
DHT22 Sensor-SoilTemperatureandHumidity
Initially, to execute the primary function, the datasets are collected and transmitted for
processing. The intelligence-basedmachine learning algorithms are proposed (SLR &
K_SLR),processes the sensed data, and predicts the required information for future
prediction in consideration of the two dataset categories i) the online web information and ii)
sensed information. The figure 4.3 shows the overview of End –To -End Smart farming
Moisture Sensor
The humidity sensor measures the water content (moisture) of the soil. The sensor
recommends plant irrigation in the agriculture field, botanical garden (Shahidul Islam et al.,
2015). The ground moisture operating voltage is 5 V, the current required is less than 20
mA, the interface is analog-type sensors and operates between 10 and 20℃. The soil
moisture sensor uses capacitance to measure the dielectric permittivity of the surrounding
soil. In the soil, dielectric permittivity is a function of water content. The sensor creates a
voltage proportional to the dielectric permittivity and, therefore, to the soil's water content.
The sensor calculates the average water content over the entire length of the sensor.
(Vaishali et al. 2017) The soil moisture sensor measures moisture loss due to evaporation.
The moisture content and its irrigation in control mostly use the humidity sensor software
and the Arduino Uno board. The soil moisture connects to the VCC% v of Arduino UNO,
GND soil moisture, and interconnected sensor Arduino UNO and the last door of the A0
sensor connected to the 0 Arduino analog board (Baraka et al., 2013).
Relay
Relay has on and off conditions to control the motor(Senpinar, 2019). There are
many types of modules, such as single channels, double channels, four channels, and eight
channels (Vaishaliet al. 2017). The relay can handle the high power electric motor and needs
to be calibrated to withstand overloads or failures. Concerning the mains voltage, the relays
have three possible connections: common, usual, and normal pins. There is no contact
between the common pinand the normally open (NO) pin. On load conditions, the common
pin is activated. (Reche et al., 2014). All the pins of the forwarding. The connection between
the relay module and the Arduino is really simple. The GND of the relay goes to the ground.
IN1 relay port connected to the Arduino digital pin. Check the first channel of the relay.
Water Pump
(DC, 12V) For this study, a water pump is required, which must be DC, 12V. The
DC motor is a commonly used motor with DC power distribution systems. Some rotors carry
magnets, and the stator grabs the conductors. The supports are used to allow the rotor to
rotate continuously towards its axis (Reshma and Babu, 2016).
Table 4.3 Soil Moisture based on sensor data and predicted algorithms
SMDbasedon SMusingSMDby SLR SMusingSMD by
Date sensor data prediction K_SLR prediction
The SLR algorithm predicts and generates the Prediction Soil Moisture (PSM) value,
comparing it with the web-based interface for real-time monitoring gives the Ground Soil
Moisture (GSM).The result of PMD shown in Table 4.4 and Table 4.5 presents the final
predicted soil moisture difference used in the smart irrigation "AISM System," showing the
effective utilization of the water from the rain precipitation enabling productivity increase in
the agriculture field. The amount of water required for automatic irrigation (1250ml) is less
compared to manual irrigation (1550ml). Manual irrigation has a water leak. It is always
better to adopt automatic irrigation (Bouderbala et al. 2019 and Chaudhry, S. and Garg.,
2019).
Farmers in developing countries have a lack of knowledge that proper irrigation can
gain yield. And the optimum water level is sufficient for good crop growth. This irrigation
planning system, "AISM System", will support the farmers in identifying the optimal water
level point and increasing their farming harvest. The AISM System will support more on
irrigation planning of the farmers during seasons making water level usage efficient in large
cultivation areas. Figure4.3 Specifies the proportion ality of the yield in terms of utilized
water level.
Thus, the sensor recording from the web portal supports visualizes the upcoming day‟s
precipitation enabling well-planned irrigation.
Table 4.4 Final Predicted Soil Moisture using K_SLR algorithm with observed sensor data
1.4
SMD based on sensor data and prediction algorithm
1.2 1.14361.1334 1.1664
1.0954 1.0543
0.9853 0.9607.89934
1 0.9489 0.9564
0.8456 0.8703.490120.894
0.8543
0.7943 0.789 6 0.7643
SMD(%)
0.8 0.7432
0.6754
0.5678
0.6
0.4
0.2
0
SM(vmc)
Date
SMDbasedonsensordata SMDusingSLRprediction SMDusingK_SLR
18.0000
17.0000
4/1/20215/1/20216/1/2021 7/1/202 8/1/20219/1/202110/1/2021
1
Date
SMbasedonsensordataSMusingSMDbySLRpredictionSMusing SMDby K_SLR
Figure 4.6 Soil Moisture based on sensor data and predicted algorithms
The accuracy of the proposed algorithm is anticipated to plan the smart irrigation
based on the estimated soil moisture predicted by the existing methods and provides optimal
irrigation. Table 4.4 shows the experimental analysis with a good estimation of the soil
moisture with the aidfield data and the forecast information. This observation could be
utilized to forecast the water requirement for optimum irrigation and the effective utilization
of natural rain. Statistical measures such as R squared and Mean Squared Error is used to
estimate the proposed algorithm's estimation error rate. The farmer set the soil moisture
threshold level under the auto mode. Based on the threshold level, the system will
automatically schedule the irrigation chart based on the weather forecast (precipitation)
information and predicted soil moisture. Whereas under the manual mode, the user will
make the scheduling decisions concerning the precipitation information and predicted soil
moisture.
ComparisionofcorrelationandMSEbetweenSMDS based on
sensor data and prediction
0.7 0.6592
0.6
0.5
0.4
0.2413
0.3 0.1403
0.1035
0.2
0.1
0 PredictedSMDusingSLRPredictedSMDusnigK_SLR
PredictionAlgorithms
R(Correlationcoefficient) MSE(MeanSquaredError)
Figure 4.6 shows the graphical result of the moisture level observed during the few
days of the considered month. It observed from the obtained results of the prediction made
using SMD with the K_SLR, K_SLR shows a higher accuracy with 99.75%. The proposed
novel K_SLR algorithm is better when compared to the SLR algorithm, confirming the
efficiency of the irrigation planning module.
4.4 CONCLUSION
The proposed module, "AISM System," reads the sensed data to predict the soil
moisture and plan an efficient irrigation scheduling. The previous observations from the
weather forecast data enable future requirements based on the soil moisture level. Thus the
research focuses to brings the efficient utilization of water at a low cost. The information
used by the system is the soil moisture preventing insufficient and excessive irrigation. The
module addresses manual irrigation problems, proving that sensor-based irrigation requires
minimum water and less time than manual irrigation. The MSE and R calculation proves the
efficiency and accuracy of the AISM System. The AISM System sensor-based irrigation
automatically works without human assistance and is the most reliable. The research has
considered a small scale agriculture field; testing on a large scale is the future requirement of
the system.
4.5 SUMMARY
This chapter discussed the framework and experimental results of the module AISM
System: Advanced Irrigation Planner is based on Soil Moisture Prediction for Farmers in
Precision Agriculture. The volume of water required for irrigation is predicted in this
module, showing that the proposed K_SLR algorithm shows a higher accuracy of
99.75 % than the SLR algorithm, confirming the efficiency of the irrigation planning
module. The parameter optimization for precision agriculture discussed in Chapter 5.
CHAPTER5
ACRIS MULTI PARAMETER OPTIMIZATION SYSTEM
FOR PRECISION AGRICULTURE
This chapter presents the last module of ACRIS, "AMOP System: ACRIS Multi
parameter Optimization System for Precision Agriculture." The module optimizes the water
level for future irrigation decision-making to maintain good crop growth and water stability
in precision agriculture by comparing the utilization of water content during the various
stages of plant growth.
5.1 INTRODUCTION
The primary objective of the proposed module, "AMOP System: ACRIS Multi
parameter Optimization System for Precision Agriculture" is to optimize the water-resource
level for farmers to enable future decision-making in precision farming. The module
developed to ensure the efficiency of the proposed research "ACRIS: Agriculture Cultivation
Recommender and Smart Irrigation System" for its usage in larger area crop fields for
sustainable water management and high yield harvest. In the year 2050, there will be a great
demand for water globally due to climate change and population growth. There exists a
challenge to identify a specific measure for optimizing water usage in agriculture to meet the
food and water demand.
The proposed system AMOP aims at optimizing the water resource considering the
crop growth characteristics such as rainfall, evapotranspiration, and moisture for precision
agriculture. The results of the previous chapter confirm that AISM System is efficient for
irrigation planning. Optimization of the system is necessary as the moisture influenced by
evaporation and precipitation (Amarendra Goap et al. 2021 and Wu et al., 2020).
The primary objective of this module in this research is to optimize the ACRIS System
utilizing the machine learning algorithm and suitable future decision-making to farmers on
sustainable cultivation, as shown in Figure 5.1.The carried out work takes crop growth
characteristics as input datasets, the dataset measured from the cultivation field after the
entire plant growth gets completed. TheAMOP utilizes the novel AAGA (Agro
ANN_Genetic Algorithm) algorithm for analysis and decision-making. The original scheme
of the ANN(Artificial Neural Network) and Genetic Algorithm(GA) hybridized to develop
the Agro ANN_ Genetic Algorithm to suit the optimization needs of the agriculture domain.
In particular, the developed novel Agro ANN_Genetic Algorithm compared with the best
achievement shown Artificial Neural Network (ANN) algorithm for its validation, fulfilling
the objective of the research described. Thus the proposed system provides future irrigation
decision-making to maintain good crop growth and water stability in precision agriculture.
farmers to identify the irrigation-based decision making evaluating the existing irrigation
schedule utilized for the entire crop growth.
CropGrowthParameters
AMOPSystem
AAGA
(AgroANN_NeuralAlgorithm)
MeasurementofIrrigationLevel (Plant
Growth Stages)
ResultValidation (Existing–ANN)
ACRIS-RECOMMENDATION
SustainablePrecisionAgriculture–Decision-Making
Figure5.1FramewoerkoftheProposedAMOPSystem
The module utilizes crop growth characteristics as input datasets, predicting the
precipitation, rainfall and evapotranspiration ranges that can support the crop productivity
with future required soil water stability to plan and schedule irrigation (Sorensen et al. 2010
and Togneri et al. 2019). The AMOP System implemented utilizing the novel AAGA (Agro
ANN_ Genetic Algorithm), the hybridized combination of the Artificial Neural Network
(ANN), and the Genetic Algorithm (GA).
Algorithm5.1–AgroANN_GeneticAlgorithm(AAGA)
Input:CropGrowthParameters(SoilTemperature,SoilMoisture,Relative
Humidity, Crop growth stages)
Output:fitnessvalue Begin
ReceivesInput; Population()
{
TrainANN
EvaluateGAfitnessfunction Fitnessevaluation()
{
1 n
MSE (yyˆ)2
(5.1)
n i1 i i
The coefficient of determination R, soil parameters exploit by the rainfall and runoff
expressed as R factor. R factor changes based on daily precipitation. The methods are
available for R factor calculation on a monthly basis as expressed in Equations (5.2) to (5.4)
(YY
2
1 (5.2)
2
)
m
R 2
(YY ) o
Z( Z (5.3)
N 2
RMSE[ fi 1/2
i1 oi)/N]
VIF i
1 (5.4)
2
1R i
2
R the coefficient of determination
where i
The novel approach using of MSE, R2, RMSE, and VIF in the Agro ANN_ Genetic
Algorithm depicted in Figure 4.1.
The AMOP system implemented in the novel Agro ANN_ Genetic Algorithm
(AAGA) allows the users to evaluate and carry out the irrigation schedules that are most
desirable to the farming conditions. The complexity is moderated and integrated into the
AMOP system, calculating the precipitation and evapotranspiration ranges predictions and
initiating the soil water stability by controlled irrigation, increasing the production yield.
Therefore, the system's novelty is the hybridized combination of the Artificial Neural
Network (ANN) and the Genetic Algorithm(GA). The AAGA utilizes the flow level of the
water pushed for irrigation. The precipitation event had been pretty predicted, although
rainfall depths needed further adjustment. Soil water content tiers computed using the crop
increase version observed the trend of soil humidity measurements, while the canopy cover
patterns and the seed cotton yield had been correctly predicted, specifically on the drip-
irrigated plots.
The working principle of the AMOP Systemis as follows: The system implemented
in the AAGA algorithm. Here the ANN-based feature of the system finds the best parameter
creating a classification system (Diego Inacio Patricio and Raafael Rieder 2021). The
decision unit function in the neural network process the input information and initializes the
population () function (Mekonnen et al., 2019). The population () is the measure of the yield
in the Senescence (Maturity) stage, where the crop reaches the maximum height of the plant.
At this stage, the plant crop is ready for harvest. Once the population is initialized, based on
the original Genetic Algorithm, the mutation starts to measure the irrigation levels in the
field - these levels based on the plant growth stages. Usually, plant growth categorized into
four stages: emergence, Rapid Growth, Lag Growth, and senescence. Finally, the fitness
function evaluated calculating the MSE error rate of the system. The variations identified
and performance evaluated with the forecast measures such as R2, RMSE, and VIF. The
water level in the field is analyzed. Thus the crop recommendation performed by the AMOP
system providing the best farming decisions to the farmers in precision agriculture.
The optimization of irrigation through the proposed AMOP system assessed for its R2,
RMSE, and VIF. Table 5.1 presents the various stages of crop growth. The crop growth
development stagetakes around 170 days, and the growth categorized into four stagesi)
Emergence, ii) Rapid Growth, iii)Lag, and iv) Senescence. Their rogation constraints
identified are here with defined inTable.5.1.The performance of the proposed approach
assessed for its R2 of (0.974), low RMSE of (0.051), and VIF (98.789). Thus the AMOP
system evaluation for the considered performance metrics was better than the existing
system Artificial Neural Network (ANN) shown in Figure5.2.
Duration of
StagesofCrop Growth
Crop
Development CropIrrigationConstraints
Growth(days)
1-14
Emergence – 90% of Uppersoillayercontentretained close
thecropseedemerged from to FC levels.
the ground.
15-70 Rapid Growing Stage- Rootzonethesoilwatercontent level
Greenstems with an reaches the maximum canopy
increasing green canopy. growth.
LagGrowingStage- Rootzonethecontentretainedat
71-113 Triggeringofflowers. lowerthanthemaximumcanopy
Senescence (Maturity)- growth.
Maximumheight of Thewaterlevelshouldbeless than the
theplant. maximum canopy growth.
114-170
AAGAsimulatesthewaterlevelinthecropgrowingfields.
ThedailytranspirationTsexpressedinEquation(5.5).
T S xMAxk T xET
s k m s,x 0 (5.5)
Where ET0, the reference evapotranspiration KmTs,xthe midseason crop coefficient
Sk,the crop coefficient
MA micro-advective adjustments
The MA expressed in Equation (5.6)
MA1.72MAMA 0.30CC
2 3
(5.6)
Everyday utilized Ts and ET0the water productivity (WP) , the biomass calculation expressed
in (5.7)
Tr
B=W p ∑ (5.7)
x
ET c
Cropreference
evapotranspiratio
Observations-ETVSR
0 n (ETo),114-170
days,830
Cropreferenceeva
ObservedParameter(mm)
potranspiration(ET
o),71-113
days,650
RainfallMeasurement
Cropreferenceevapotr (R)in mm, 114-170
Cropreferenceevapot anspiration(ETo),15- days,423
ranspiration(ETo),1- 70days,400
14days,350
RainfallMeasurementRainfallMeasurement(R)in
mm, 71-113
(R)inmm,15-70days,238
days,204
RainfallMeasurement(
R)in mm, 1-14
days,104
ObservedDuration
Cropreferenceevapotranspiration(ETo)RainfallMeasurement(R)inmm
ComparisonofANNwithAAGA
PredictionbyImprove Predictionby Predictionby
Predictionby
d Neuro- ImprovedNeuro- ImprovedNeuro- ImprovedNeuro-
GeneticModel,1- GeneticModel,15- GeneticModel,71- GeneticModel,114-
14days,0.9741 70days,0.8965 113 days, 0.9026 170days, 0.9632
Predictionb PredictionbyANNModel,15-70
yANN days,0.84 5376
ObservedParameter(MU)
ObservedDuration
PredictionbyANNModel PredictionbyImprovedNeuro-GeneticModel
Table5.3 Comparison of Existing ANN with the proposed AAGA for the various performance metrics
The performance indices for the entire dataset are observed. The ratings of the
performance indices for both training and testing datasetsarecarriedoutbyR 2, RMSE, and
VIF. The relationships among the AAGA and ANN were measured with the testing and
training datasets. Table 5.3 presents the performance evaluation comparison of the proposed
AAGA with the existing ANN model considering the stages of plant growth towards the
measurement metrics R2, RMSE, and VIF. The proposed AAGA system performs better in
forecasting every day irrigation water demand, while a huge volume of records is evolved.
The AAGA shows better represent ability and Accuracy than the existing ANN model with
very low MSE.
Figure5.4 Performance evaluation R2
The results were compared with the existing ANN model, and observed that the
proposed AAGA technique has high R2 on all the stages of plant growth, proving it to be
better than the other, as shown in Figure5.4.HeretheR2forthedays(1-14)-0.961,(15-70)-0.969,
(71-113)-0.962, (114-170)-0.974. Thus the AMOP system efficiently optimizes the water
level for future irrigation decision-making to maintain good crop growth and water stability
in precision agriculture.
Compared with the existing ANN model, the results observed that the proposed
AAGA technique has a low RMSE error rate on the first three stages of plant growth,
proving it to be better than the other as showninFigure5.5.HeretheRMSEforthedays(1-14)-
0.042,(15-70) -0.038, (71-113)-0.035 is presented. Thus during the early stages of plant
The results compared with the existing ANN model and observed that the proposed
AAGA technique has high VIF on all the stages of plant growth, proving it to be better than
the other, as shown in Figure5.6Here the VIF values for the days (1-14) -95.21, (15-70) -
96.84,(71-113)-97.364,(114-170)-98.789.Thus,theproposedAAGA has a high correlation in
predicting and planning the irrigation for the considered parameters.
5.4 CONCLUSION
The results were compared with the existing ANN model, and it was observed that the
proposed AAGA technique has high R2, VIF with low RMSE thus making the AMOP
system efficiently optimizing the water level for future irrigation decision-making for the
farmers to maintain good crop growth and water stability in precision agriculture fulfilling
the primary objective of the research.
5.5 SUMMARY
This chapter discussed the framework and experimental results of the module, “AMOP
System: ACRIS Multi parameter Optimization System for Precision Agriculture”. The
module proposed concludes the enhanced version of the developed ACRIS System showing
a better R2 value of 97.4%. The overall conclusion of the proposed research work, "ACRIS:
Agriculture Cultivation Recommender and Smart Irrigation System", is concluded further
with the evidence produced in its three implementation stages is discussed and described in
the forthcoming Chapter 6.
CHAPTER 6
CONCLUSION AND FUTURE ENHANCEMENT
6.1 CONCLUSION
Digital technologies are rapidly transforming the efficiency, effectiveness, and
transparency of agricultural production, the natureand labour requirements of
agricultural work, the structure of agricultural value chains, the range of agricultural
business models, and rural living lifestyle opportunities. The thesis describes the
"ACRIS: Agriculture Cultivation Recommender and Smart Irrigation System" for
farmers in precision agriculture, as shown in Figure 6.1 for the welfare of humans to
maintain a sustainable environment. Precision agriculture is a new technology that
requires some new advanced and expensive equipment such as yield monitoring
sensors, GPS receiver, etc. Such facilities are very costly for precision agriculture
farmers. Therefore, many try to avoid precision agriculture systems, especially in
situations where their productions have low commodity prices.
make predictions for the future requirement and the soil moisture level analysis.
Thus, research focus brings an auto and effective mode of water utilization and helps
in bringing water-saving systems based on the proposed algorithm with cost-
effective analysis. By increasing the model strength and quality of performance,
some more advancement in the agricultural field for lifetime usage enhanced for
promoting agriculture. Thus the modules proposed concludes the improved version
of the developed ACRIS System shows a better R2 value of 97.4%. The overall
conclusion of the proposed research work, "ACRIS: Agriculture Cultivation
Recommender and Smart Irrigation System," provides favourable recommendations
for farmers who utilize IoT in Precision Agriculture for increasing their cultivation
yield.
6.2 FUTUREENHANCEMENT
The recommender system can be connected with farmers on the desktop
application or mobile phones to exchange information in the future. Furthermore,
intelligent decisions for planting, fertilizing, and harvesting crops can be considered
future research directions to effectively use the information to achieve higher yields
and earn higher profits. Further, the system developed consideringthe soil
characteristics as parameters for precision agriculture. The system can utilize various
other algorithms such as random forest, K-nearest neighbour and Naive Bayes with
REFERENCES
1. Abrisqueta, I., Conejero, W., Valdés-Vela, M., Vera, J., Ortuño, M.F. and Ruiz-
Sánchez, M.C. (2015), ―Stemwater potential estimation of drip-irrigatedearly-
maturingpeachtreesunder Mediterranean conditions‖, Computers and Electronics
in Agriculture, Vol. 114, pp.7-13.
2. Adeloye, A.J., Rustum, R. and Kariyama, I.D. (2012), ―Neuralcomputing
modeling of the reference crop evapotranspiration‖, Environmental Modelling
& Software, Vol. 29, No. 1, pp.61-73.
3. Ahmad, N., Hussain, A., Ullah, I., & Zaidi, B. H. (2019), ―IOT based Wireless
Sensor Network for Precision Agriculture‖, In 2019 7th International Electrical
Engineering Congress (iEECON), IEEE, pp. 1-4.
4. Akter, S. and Mahata, P. (2021), ―Developinga Smart Irrigation System Using
Arduino‖.
5. Alan L. Olmstead and Paul W. Rhode. (2009), "Conceptual Issues for the
Comparative Study of Agricultural Development", in Pedro Lains and Vicente
Pinilla, (eds.),Agriculture and Economic Developement in Europe Since 1870
(London: Rouledge, 2009), pp. 27-51.
6. Ali, S.M.A.M. (2021), ―A Survey on Agriculture Crop Recommender
System‖,International Journal of Advanced Research in Computer Engineering &
Technology (IJARCET), Vol. 7, No. 2.
7. Allen, J. and Wolfert, J. (2011), ―Farmingfor the Future: towards better
information-based decision-making and Communication-Phase I: An
Australasian stocktake of farm management tools used by farmers and rural
professionals‖, New Zealand Centre of Excellence in Farm Business
Management, Palmerston North. AgFirst Consultancy/Wageningen University
and Research Centre.
8. Anisi, M.H., Abdul-Salaam, G. and Abdullah, A.H. (2015), ―A survey of
wireless sensor network approaches and their energy consumption for monitoring
farm fields in precision agriculture‖, Precision Agriculture, Vol. 16, No. 2,
pp.216-238.