Professional Documents
Culture Documents
Design of Plant Leaf Diseases Detection System Employing IoT
Design of Plant Leaf Diseases Detection System Employing IoT
Department of ECE
Jaypee University of Information Technology, Solan, Himachal Pradesh, India
*jain.shruti15@gmail.com
Abstract: Plant diseases can affect an extensive range of fewer workers and equipment. Therefore, monoculture
commercial crops, and result in a momentous loss in yield. farming aids in boosting a farm industry's production and
Most diseases are caught by the farmer walking through the efficiency. Plants have an autoimmune disease. Additionally,
field and checking the crop. In this research article, the it ensures harvesting and planting efficiency, which denotes
authors have designed a model for the detection of plant an increase in production and automates the profit [6, 7].
diseases using the Internet of Things (IoT) and Machine Plants can be affected by a varied range of diseases, with
Learning (ML) models. Plant diseases constitute a major bacterial, fungal, nematode, and viral infections. Fungal
production hazard in agriculture and major crop failures in diseases are the basis of fungi that infect plants & interfere
the history of man have been due to these diseases. The main with their normal growth and development. It includes
goal is to guarantee farmers' early detection of plant diseases, powdery mildew, rust, and root rot. Bacterial diseases are
which will reduce the need for pesticides. A more stable and originated from bacteria that infected plants and produce
wealthy society is made possible by eliminating the problem toxins that can damage plant tissue. It includes bacterial wilt,
of plant diseases. Farmers click the image of the leaf and fire blight, and bacterial leaf spot. Viral diseases are caused
sensors with Arduino send the real-time data to the cloud by viruses that infect plants and replicate inside their cells. It
where think-speak will visualize the data which was tested includes tomato and tobacco mosaic virus etc. Nematode
and validated using ML models. Support Vector Machine diseases are caused by worms called nematodes that infect
(SVM) and k-NN are used as the ML models implemented by plants and feed on their cells, disrupting their normal growth
the existing online database. The emergence of accurate and development like soybean cyst nematode, and crown
techniques in the field of leaf-based image classification has gall. In addition to infectious diseases, plants can also be
shown impressive results. SVM results in 94.65% accuracy. affected by physiological disorders, which are caused by
environmental stresses or nutritional imbalances. Examples
Keywords: Plant leaf detection, Arduino, Internet of things, of physiological disorders include nutrient deficiencies,
Classification transplant shock, and sunscald.
Authorized licensed use limited to: National Inst of Science Edu & Research -NISER. Downloaded on December 08,2023 at 05:52:09 UTC from IEEE Xplore. Restrictions apply.
essence, it is a science that seeks to teach computers to solve plants for their living. Plant diseases constitute a major
challenging and complicated issues[12]. AI production hazard in agriculture and major crop failures in
isthetermusedtodescribetheintelligenceexhibitedbyrobotsorot the history of man have been due to these diseases. The
herdevices.It pertains to the study of computers [13]. It has agrochemical industry has risen to the challenges by
greatly enhanced human life, and it is now a prominent and providing antidotes for the treatment of all kinds of plant
practical area of study in computer science. A kind of ML diseases as they break out. There are ample numbers of
known as DL uses artificial neural networks to adapt to and chemicals available in the market that was manufactured to
learn from massive volumes of data. Additionally, high-level control plant leaf diseases. This is done by inhibiting the
abstractions are learned from data via hierarchical growth or by killing the disease-causing pathogens. An effort
architectures. It is a novel method that has previously been to improve plants by breeding for disease resistance is the
applied in well-known AI applications. Three key causes, other way to curb plant diseases. This approach exploits the
including the development of chip processing, the availability resistance that is available in nature through genetic transfer
of expensive computer hardware, and machine learning from source to desired plants as reproductive boundaries
techniques, have contributed to the current boom in deep permit. Image processing is a technique that examines and
learning. Numerous deep-learning algorithms have been manipulates images using computer algorithms. The
thoroughly examined and debated in recent years. Machine detection of plant diseases is one of the applications of AI in
learning is an application of AI that relies on information and the world of agriculture. Fungi, bacteria, and viruses are the
algorithms to assess situations and make judgments devoid of most prevalent illnesses that can harm leaves, but insects can
human input. It talks about how computers carry out also have an influence that manifests as spots on plant leaves.
activities on their own depending on their past knowledge In this research article, an application for farmers was
[14]. In the realm of agriculture, image processing is mostly designed that will be used to upload real-time images of
used to identify plant illnesses and to pinpoint the specific leaves that classify from sick plants and receive information
portion of a plant's leaf that the disease has impacted [15]. about the disease. The main methodology used in this paper
Due to the effectiveness and precision of its algorithms, is illustrated in Fig 1.
machine learning is utilized as a tool for disease diagnosis in
agriculture [16].To address several issues with agriculture
and plant diseases, computer application in the field of
agriculture was established. Due to its great performance and
accuracy, AI is the most often employed technique for the
detection of plant diseases [17].
Authorized licensed use limited to: National Inst of Science Edu & Research -NISER. Downloaded on December 08,2023 at 05:52:09 UTC from IEEE Xplore. Restrictions apply.
of the earliest signs is the black & yellow patches may also determines each pixel’s particular intensity. Since the chosen
cause necrotic blotches as shown in Fig 2 (b). High image was created in RGB format, it will be converted to
temperatures and humidity promote the emergence and grayscale. The classification of the illness was done using the
spread of Cercospora Leaf Spot of sickness. On the plant's machine learning approaches namely k nearest neighbor
leaf, there are tiny grey specks that identify it as shown in Fig (kNN) and Support Vector Machine (SVM) approaches [19].
2 (c).
The model is trained which was validated by real-time
images using IoT. IoT touches every facet of our lives. IoT
helps to interconnect physical objects equipped with sensing,
actuating, and computing power and thus lends them the
capability to collaborate on a task in unison remaining
connected to the internet. The lifeline of IoT lies in the ability
of IoT devices to transmit and receive information. Wireless
communication is the ideal form of communication for IoT.
(a) (b) However, IoT devices do not send and receive data in the same
way hence different connectivity options have been developed
to cater to the different applications. These connectivity options
have different coverage, and power requirements and offer
different bit rates. It is prudent to consider these variations
when deciding which option to utilize for a given IoT device.
IoT can contain a myriad of devices according to its
architecture and application. When talking about the protocol
used only for IoT is MQTT. Other protocols used in an IoT
ecosystem are XMPP and HTTP. They also use other protocols
for managing the network such as SNMP, NETCONF, and
(c) CLI. The farmers checked the real-time images which were
processed and implemented by Arduino and different sensors.
Fig 2: Different plat leaf diseases (a) Alternaria Alternate, (b) Arduino Uno is a microcontroller development board based on
Bacterial Blight, (c) Cercospora Leaf the Microchip ATmega328P microcontroller and made by the
company Arduino. Arduino is like the spinal cord. It connects
to a large number of peripherals and provides spontaneous
The detection accuracy not only depends on the ML
communication. It is good for low-intelligence tasks. Arduino
algorithm but also depends on the type of images used. For
is more rugged and has a much wider operating voltage range.
plant leaf detection, a multispectral image provides better
More importantly, it has got a pretty neat IDE to easily
information for plant disease detection. These multispectral
program making it suitable for controlling your sensors &
images will allow the algorithm to be less computationally
actuators. The images were stored in the cloud using Arduino
expensive and with better accuracy than by using RGB
and were visualized by Thing Speak.
images. The unwanted distortion, like spores and dust, is
eliminated to enhance the image quality which is done in a
III. RESULTS AND DISCUSSION
pre-processing step. In this step hue of the photo, and scan is
changed. Mainly images are transformed from RGB to
Our economy is highly dependent on Agriculture and address
L*a*b*colour space (where L*stands for the luminance layer
several issues with agriculture and plant diseases, computer
and a*and b* for chromaticity layers). After this stage
application in the field of agriculture was established. Plants
segmentation is done. Segmentation serves to distinguish the
have an autoimmune disease. Hyperactivity of plant innate
non-disease zone from the area that the illness allows on the
immune receptors causes abnormal defense activation due to
leaf picture. There are different techniques for segmentation,
which the growth of the plant is retarded or it may have
where the Region of Interest (RoI) is marked. The RoI has a
spontaneous lesions that are an indicator of plant disease.
lower overall size than the original picture. In this paper, K-
Due to its great performance and accuracy, AI is the most
mean clustering is used as a segmentation algorithm. With
often employed technique for the detection of plant diseases
each observation belonging to the cluster with the closest
[7]. If proper care is not taken then it can cause serious
mean, this method's main goal is to split n observations into
effects on plants which affect productivity and quantity. In
k clusters. Each pixel in the uploaded image is labeled based
this article detection of unhealthy regions or analysis the leaf
on the k-means results, and the clustering results are saved in
infection like bacterial blight, and Alternaria alternata is done
a blank cell array. The selection of a practical cluster is a
using Arduino and ML algorithms. Arduino is the
crucial component of this process. The cluster was chosen
microcontroller where all the data sets and codes are stored.
because it represents the largest area that the sickness has
The hardware design of our model is shown in Fig 3.
impacted. During the subsequent step of feature extraction,
the characteristics of the selected clusters are retrieved.
There are other techniques for extracting features, but in this
article, the Gray Level Co-occurrence Matrix
(GLCM)features are employed [18]. The GLCM is renowned
for its capacity to extract textural characteristics that will
simplify any further analysis. Using the image, this approach
Authorized licensed use limited to: National Inst of Science Edu & Research -NISER. Downloaded on December 08,2023 at 05:52:09 UTC from IEEE Xplore. Restrictions apply.
by being given an integer value. T he two illnesses: Alternaria
alternata and Bacterial Blight for our model are illustrated in
Fig5&Fig 6.
Fig 4 shows the designed model using IoT via Arduino and
the data is stored in the cloud. Readings from the sensors
were also sent to a Thingspeak channel to visualize the data.
ThingSpeak is an open data platform and API which allows SVM and kNN classifiers are used to classify the features
one to collect, store, analyze, visualize, and act on data from retrieved from the dataset. The classification accuracies are
sensors or actuators like Arduino. Several groups of illnesses tabulated in Table 2.
affect a lot of leaves. Each illness in a class is given a label
Authorized licensed use limited to: National Inst of Science Edu & Research -NISER. Downloaded on December 08,2023 at 05:52:09 UTC from IEEE Xplore. Restrictions apply.
Table 2: Classification Accuracy [3] Monoculture Farming Explained: WhatAre the Pros and
Classifiers Accuracy Cons?” EARTH OBSERVINGSYSTEM,29 Dec.2021,
https://eos.com/blog/monoculture-farming/.
SVM 94.65% [4] Otani,Saria,etal.“Disease-FreeMonocultureFarmingbyFungus-
GrowingTermites.”ScientificReports, vol. 9, no.
kNN 93.45% 1,2019,https://doi.org/10.1038/s41598-019-45364-z.
[5] Power, J. F., and R. F. Follett. “Monoculture.” Scientific
94.65% and 93.45% accuracy are obtained using SVM and American, vol. 256, no. 3,
ScientificAmerican,adivisionofNatureAmerica,Inc.,1987,pp.7
kNN classifiers respectively. Real-time images of various
8–87,http://www.jstor.org/stable/24979342.
illnesses and their derived characteristics are used to train the [6] Kumar,Surender,andRupinderKaur."Plantdiseasedetectionusin
model. Finally, based on this comparison, the system will be gimageprocessing-areview." InternationalJournal of
able to determine the kind of sickness, which means that it ComputerApplications 124.16(2015).
contrasts the values of the vector properties in the database. [7] Singh, Vijai, and A.K. Misra. “Detection of Plant Leaf
Therefore, it could predict the disease depending on the Diseases Using Image Segmentation andSoft Computing
values that are matching. The data we obtained includes Techniques.” Information Processing in Agriculture, vol. 4,
various images of leaf illnesses that were downloaded from no. 1, 2017, pp.41–
the internet; these images are saved in a database on the 49.,https://doi.org/10.1016/j.inpa.2016.10.005.
[8] Chamasemani,F.F., and Y.P.Singh. Malaysia, Bio
computer and are utilized, together with images of healthy
InspiredComputing:TheoriesandApplications(BIC-
leaves, to train our s ys t e m . Additionally, creating the TA),2011SixthInternationalConference.FacultyofInformation
standalone application will increase the technology's usability Technology,MultiMediaUniversity. Cyberjaya, Malaysia
and accessibility for farmers. As a consequence, a standalone [9] ChihWei,H.,andC.Lin.2002.Acomparisonofmethodsformulticl
application for distinguishing between healthy plants and asssupportvectormachines.IEEE Transactions on
plants afflicted by illness has been created. Detection of plant NeuralNetworks, 13(2):415-425
diseases using technology is beneficial as it reduces the work [10] Helly, M. E., A. Rafea, and Salwa-El-Gammal. 2003. An
of monitoring big farms. integrated image processing systemforleaf diseasedetectionand
diagnosis.in Proc. IICAI, pp.1182-1195
[11] Shruti Jain, Meenakshi Sood , “ SVM Classification of Cell
IV. CONCLUSION AND FUTURE WORK
Survival/ Apoptotic Death for Color Texture Images of
Survival Receptor Proteins”, International Journal on Emerging
The easiest way to maintain a high yield is through the Technologies 10(2): 23-28(2019).
identification of plant diseases using IP. The major objective [12] Jean,W.2009.Extensionplantpathologist,georgiaplantdiseaselo
of this paper is to demonstrate how image-processing tools ssestimates,www.caes.uga.edu/publications
help farmers to increase yields while also ensuring precise [13] ManuBN(2022).Plant LeafDisease
findings for the diagnosis of plant diseases. We can design a DetectionandClassificationusingMulticlassSVM
model using an existing dataset where different IP steps were [14] Classifier(https://www.mathworks.com/matlabcentral/fileexch
employed. 94.65% accuracy is obtained using SVM which ange/55098-plant-leaf-disease-detection-and-classification-
using-multiclass-svm- classifier),MATLAB Central File
results in a 1.2% accuracy improvement in comparison to
Exchange.RetrievedApril29, 2022.
kNN. The model is validated by the real-time images that [15] ,W.,
farmers took from their farms. The proposed model helps in andS.Wang.2008.Imagethresholdingusingweightedparzenwind
detection of the plant leaves disease. In the future, we will try owestimation.JournalofAppliedSciences, 8(5):772-779
to develop a mobile application that will greatly simplify the [16] Kim, D. G., T. F. Burks, J. Qin, and D. M. Bulanon. 2009.
procedure for farmers and to employ drones to increase the Classification of grapefruit
dataset of training photographs and boost the precision of our peeldiseasesusingcolortexturefeatureanalysis.InternationalJour
suggested method. nalonAgricultureandBiologicalEngineering, 2(3): 41-50
[17] Gonzalez,R.,R.E.
Woods.2008Digitalimageprocessing.Thirdedition,PearsonEdu
References
cation,Prentice-Hall,Inc
[18] Bhusri S , Jain S, Virmani J , “Classification of breast lesions
[1] "AdvantagesandDisadvantagesofMonoculture."Gardenerdy, using the difference of statistical features” Research Journal of
22 Jan.2015,https://gardenerdy.com/advantages- Pharmaceutical , Biological and Chemical Sciences (RJPBCS)
disadvantages-of-monoculture/. ,7 (4), 1365-1372: July- Aug 2016
[2] Khirade, Sachin D., and A.B. Patil. “Plant Disease Detection [19] Jain S, “ Computer Aided Detection system for the
Using Image Processing.” Classification of Non Small Cell Lung Lesions using SVM”,
2015InternationalConferenceonComputingCommunicationCo Current Computer-Aided Drug Design, 16(6), 2021 , pp
ntrolandAutomation,2015,https://doi.org/10.1109/iccubea.201 833-840.
5.153.
Authorized licensed use limited to: National Inst of Science Edu & Research -NISER. Downloaded on December 08,2023 at 05:52:09 UTC from IEEE Xplore. Restrictions apply.