9 XI November 2021

https://doi.org/10.22214/ijraset.2021.38890
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

Pest Detection in Plants Using Convolutional Neural Network

Ms Savita Sharma1, Rishabh Sharma2, Rishav Kumar3
1
Assistant Professor, Dept of CSE, MAIT Delhi, GGSIPU Delhi
2, 3
UG Students, Dept of CSE, MAIT Delhi, GGSIPU Delhi

Abstract: Agriculture or farming is an imperative occupation since the historical backdrop of humanity is kept up. Artificial
Intelligence is leading to a revolution in the agricultural practices. This revolution has safeguarded the crops from being
affected by distinct factors like climate changes, porosity of the soil, availability of water, etc. The other factors that affect
agriculture includes the increase in population, changes in the economy, issues related to food security, etc. Artificial
Intelligence finds a lot of applications in the agricultural sector also which includes crop monitoring, soil management, pest
detection, weed management and a lot more. Significant problems for sustainable farming include detection of illness and
healthy monitoring of plants. Therefore, plant disease must automatically be detected with higher precision by means of image
processing technology at an early stage. It consists of image capturing, preprocessing images, image segmentation, extraction of
features and disease classification. The digital image processing method is one of those strong techniques used far earlier than
human eyes could see to identify the tough symptoms. Considering different climatic situations in various regions of the world
that impact local weather conditions. These climate changes affect crop yield directly. There is a great demand for such a
platform in the world of today which would enable the farmer market his farm products. We have proposed in this study a system
where farmers can sell their products directly to customers without the intervention of distributors and traders. The predictive
analytics system is necessary for the farmer to get the maximum yield which benefit the farmer. This may be done if the
environment, market conditions and knowledge of the timely planning of farms are known properly.
Keywords: Pest Detection, Artificial Intelligence, Agriculture, Image processing, Convolutional Neural Networks.

I. INTRODUCTION
India is an agricultural country, which depends on agriculture more than 50 percent of the people. The agri-business commitment to
India's national revenue is all the more important, as farming in India is regarded to be a backbone in the Indian economy. Most
Indians are dependent on agriculture explicitly or implicitly. Some are directly connected with agriculture and some others deal with
these commodities [1]. India is able to produce food grains, which in the Indian economy may make a huge impact. In order to reach
a desired Government mark, it is necessary to help small-scale farmers alongside the large farmers in the case of land, banking and
other machinery. For over 58 percent of the population of India, agriculture is the major livelihood source. Gross Value in FY20,
Rs. 19.48 lakh crore was expected to have been added to agriculture, forestry and fisheries (US$ 276.37 billion). At current prices,
Indian share of farming and ally industries in the Indian Gross Value Added (GVA) amounted to 17.8% in FY20. In 2021, after the
pandemic-led downturn, consumer expenditure in India is expected to return to its rise by up to 6.6 percent [1,2]. The Indian food
sector is set to develop rapidly and annually, because of its great value-added potential, especially in the food processing industry, it
is increasingly contributing to global food commerce. The sixth biggest in the world is Indian food and food markets, with retail
sales accounting for 70%. 32 percent of the entire food market in India, one of the major sectors in India, is the Indian food
manufacturing sector, rated fifth in terms of quality, accumulation, trade and anticipated growth. For April 2020 – January 2021,
main exports of agricultural goods amounted to $32.12 billion. AI might provide an advantage for existing practices and procedures
within the rural environment in order to achieve profitability and support. For example, dynamic capabilities such as AI may assist
to identify changes in agricultural products' market value and explicitly offer planting and harvesting instructions to keep away from
major crop losses. In general efficiency and sustainability, early disease detection and changed water system designs might enhance.
Artificial intelligence-enabled weather forecasts constantly provide accurate, remarkable bits of knowledge in day-to-day farming.
Such accurate data may help to reduce crop losses via preventive actions. With AI applications in farming groups, they can enhance
the present strong IT capabilities after some time via learning. Because plant diseases may harm crops, they represent a significant
danger to crop monitoring. Therefore, it is very important to diagnose and manage early plant illnesses. Often, this process requires
the right diagnosis by a professional human competence. Specifically at distant areas and small farms in developing regions this
knowledge is not always available. The creation of efficient image-based prediction techniques, including the use of smartphones, to
capture high quality pictures, may help significantly to the initial diagnosis and decrease in waste.

©IJRASET: All Rights are Reserved 1583

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

In the area of agricultural research, the rapid growth of deep learning technology has effectively implemented Convolutional Neural
Networks (CNNs) that can overcome the disadvantages of machine learning. In the automated detection of pest conditions, the CNN
model works well [3]. The CNN model comprises generally of two major operators, the convolutional stratum and the grouping
stratum. The convolutional layer can extract more complicated and relevant picture characteristics automatically. The pooling layer
lowers the amount of data parameters because of a high calculation of the convolution network. The subject of categorization of pest
images on the basis of CNN models is mostly investigated in current research. However, it is more necessary to identify and locate
every pest in the natural environment than to classify pests.

Figure 1: Farm Share of Agricultural GDP

Figure 1 shows the farm share of Agricultural GDP in India. The agricultural portion in GDP rose from 17.8% in 2019-20 in 2020-
21 to 19.9%. However, GVA growth for farming continued to grow positive by 3.4% for the entire economy during 2020-21, while
it contracts by 7.2% for the whole economy."
Previous yield forecasts were made by looking at the knowledge of the farmer on a given field and crop. However, given the quick
changing conditions, farmers are compelled to grow more and more crops every day. As the existing state of affairs, many of them
are not sufficiently informed of the new crops and of the benefits they obtain from their production. The productivity of agriculture
may also be enhanced under a range of global circumstances via study and provision of crops performance [3,4]. The suggested
system therefore takes the user's position as an input. The soil nutrients such as nitrogen, phosphorous and potassium are derived
from the site, the predicted weather. Machine Learning and Multiple linear regression are used in the suggested system to detect the
data pattern and process it under the input circumstances. In turn, this will offer the finest harvests possible under the conditions of
the environment. The projection will be more accurate if last year's production is also taken into account. This method therefore
proposes profitable crops for the farmer to choose directly.
There is a great demand for such a platform in the world of today which would enable the farmer market his farm products [4]. We
introduced farmers and customers to this system for better and direct contact. The farmer is now transferring his goods to a certain
agent, and he is asking the farmer to attend the market after a certain period to receive the cash from the sold commodity. At the
expense of the market, the agent sells the goods to a different agency or dealer. Each agent attempts to remove his commission from
it. There is no way for farmers to know how much their goods was sold and how much. There's no transparency. There is no facility
for farmers to discover the product rates on various markets where they may sell their stuff for big profit. Farmers are often unaware
of the government's initiatives and compensations. Despite all the possibilities offered by doors, the farmers cannot benefit from
them. The major objective of this online application is to link consumers and farmers directly to farmers and consumers. Since the
majority of farmers do not know about the latest equipment and technology owing to a job loss and time wasted [5]. If a farmer
finds out about what pesticides or fertilisers he has used in his farm, his or her data will be kept so that they may readily find out
what pesticides or fertiliser they have used to farm.

©IJRASET: All Rights are Reserved 1584

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

II. LITERATURE REVIEW

The automatic detection of pests in recent years has been an important subject for study. In most situations, visibility, machine
learning or technology for detecting herbs is picked and employed. However, in the same job there is typically no comparison of the
many available approaches. Many computerized pesticide identification and recognition study focuses on a particular technical
method, although many technological options are not being evaluated. In recent years, computer vision and identification of objects
made enormous progress. Prior to this, the typical method was based on detectors' algorithms for features such as Salient Regions,
SURF, SIFT, MSER, etc. Some learning methods using these attributes are employed when data is retrieved. Depends on specified
functions, the performance of methods. Image classification is indeed a difficult procedure that must be redone if the issue and the
dataset change [6,7]. This difficulty occurs in every effort to detect plant illnesses through the use of computer vision since they
trust hand-made functions and algorithms for improved images. Deep learning techniques may be used to overcome the problem of
manual extraction of features as feature extraction is done automatically. Machine and deep learning advances allow the reliability
of item identification and detection to be dramatically improved. In the case of illness detection, machine learning approaches were
used on the one hand. In agriculture research projects, several of these approaches were employed, such as Artificial Neural
Networks (ANN's), Decision trees, K-means or KNN [7]. One of those technique proposed widely in the field of illness diagnosis is
Vector Support Machines (SVMs). Various techniques have been analysed to identify diseases and classify them using machine
learning in tomato crops. First, tomato yellow leaf curl disease is identified using RGB pictures and various master learning methods
(SVM, linear kernel, quadratic kernel, radial base function, multilayer perceptron, multilayer kernel). The average accuracy of this
method was 90%.
Bakhsh pour and Jafari (2018) developed a method for weed detection. Using an artificial neural network and a support vector
machine, the pattern is identified for each species of plant. SVM has a 96.67 percent accuracy rating, whereas ANN has a 92 percent
accuracy rating [8].
Ray et al. (2017) devised a method for detecting fungal illness at an early stage, and correctly diagnosing the disease aids in its
prevention. It included an overview of current and upcoming disease detection techniques. On-field validation is aided using
biosensors in this sector.
Carranza et al. presented a deep learning-based method for identifying species in herbarium collections. It focuses on how
convolutional neural networks may aid with automated plant species identification. Image- In the convolution neural network
method, net classification is highly effective. For domain-specific training, transfer learning is also employed [8,9]. When taught
and evaluated for a variety of species, the results suggest that it is more accurate. It was demonstrated that transfer learning to
another location may be accomplished using herbarium datasets, even when the species do not match.
Lu et al. (2017) presented a method for diagnosing rice illness. This is accomplished using a deep convolutional neural network.
There are 500 pictures in the dataset utilized for the research. For identifying reasons, ten different kinds of rice disease are
employed. 95.48 percent accuracy was attained during this process.
Gan et al. (2018) have completed an essential task of mapping citrus yield [9]. To determine if the fruit is ripe or not, an image-
based method is utilized. Green fruit is detected using a combination of color and thermal images. For the categorization of ripened
and green fruit, the CTCP algorithm, also known as color thermal combined probability algorithm, is employed.
Liang et al. (2019) used a severity estimation method to determine the degree of illness in a plant. For diagnosing the illness, the
PD2SE-Net method has been developed. During the identification phase, the visualization and augmentation processes are carried
out. As an add-on structure, the ResNet50 architecture is employed. For illness severity, the accuracy is 0.98 and 0.99.
Chapman et al. (2018) used data from an oil palm farm to use a Bayesian network to forecast yield. The accuracy and r2 (0.6 and
0.9) scores for this network are greater [10]. The Bayesian network's parameters were described in detail. Five distinct classes
methods were used to measure the depth of the soil.
A method for detecting wheat leaf rust has been presented by Azad Bakht et al. (2019). It's time to calculate the leaf area index. At
various levels, the canopy-scale is examined. For identification, machine learning techniques such as boosted regression tree,
Gaussian process regression, support vector machine, and random forest approaches are utilized.
Iqbal et al. (2018) did a study focusing on citrus plant disease and the categorization of the many diseases that affect citrus plants. It
also goes through the many techniques utilized in the segmentation, feature extraction, feature selection, image processing, and
classification methods in depth. It also goes into the automated technologies that are used to detect and classify items [10,11]. Citrus
disease is known by several names, including canker, black spot, citrus scab, melanosed, and gearing. For illness extraction, the K-
mean method is employed, and the approaches used for each stage of analysis are compared to the current survey. Color features are
computed and classified using the Back Propagation Neural Network (BPNN) and the Grey Level Co-Occurrences Matrix (GLCM).

©IJRASET: All Rights are Reserved 1585

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

Pre-processing, color-based transformation, picture improvement, noise reduction, resizing, and segmentation techniques are all
covered. Various texture, color, and shape-based feature extraction techniques are used. It contains an overview of several classifier
techniques as well as their applications. According to the findings, the pre-processing approach increases segmentation accuracy.
Kaya et al. (2019) discovered that manually categorizing data has several key drawbacks, including the fact that it is costly, time-
consuming, and requires expertise. During the classification phase, a Deep Neural Network is offered as a solution. Plant
categorization model performance has increased. It contains a comparison of several methods and their best results, such as (DF-
VGG16/LDA = 99.00, DF-Alex net/LDA = 96.20, CNN-RNN = 98.80, CNN = 99.60, (CNN, SVM = 97.47)). Input, 3*3 conv,
ReLU, pool, 3*3 conv, ReLU, pool ReLU, FC-class size, SoftMax are all part of the proposed CNN design. The classification
accuracy for each model in the training dataset and for the pre-trained model is shown in the image.
It takes the image as input and assigns different weights to the various items in the image, making each object distinct from the
others. The pre-processing burden in the convolution neural network is minimal when compared to other classification techniques
[12]. Neuron connection in the human brain is analogous to the neurons linked to CNN. It could learn its filters. The dependencies
existing in the spatial and temporal components are captured with the help of the applicable filter. During the process, it plays a vital
role in reducing the picture to simpler forms, with no image loss. Object detection (R-CNN, Fast R-CNN, and Faster R-CNN) and
semantic segmentation are two major applications of convolution neural networks (Deep parsing network, fully convolution
network). CNN architectures are made up of a series of layers that change one activation volume to another with the aid of
differentiable functions. CNN is primarily built with layers like as convolution, pooling, and fully linked. CNN's essential
component is the convolution layer. It carries the network's computing workload. Dot products are mostly performed between the
kernel and the limited region. The depth is large in the kernel, while the spatial is less. In kernel, just the depth will be increased,
while the height and width will be reduced. Image representation is achieved by sliding the kernel over the image's height and
breadth (Hang et al., 2019). Secondly, support vector machine techniques are utilised with both RGB pictures and spectral reflection
for the detection and quantification of tomato leaf miners. Thirdly, tomato powdery molten fungus Oidium neolycopersici is
identified by utilising SVM algorithms and visual thermal and stereo light. Fourthly, powdery mildew is found in tomatoes with
self-organizing maps and RGB pictures. In this work we just utilise 138 images, a modest number of images to get the data set
variability [13]. Most of the disease detection and classification classificatory were developed using few data sets relying on image
extraction to categorise the leaves. In order to create reliable image classification, a big, labelled and validated collection of pictures
of sick and healthy plants is needed. No dataset with these characteristics was accessible until quite recently. The PlantVillage
initiative has already begun to gather and classify tens of thousands of illustrations of normal and sick plants to address this
problem. The PlantVillage dataset is used for building deep neural networks for the diagnosis of various crop diseases [14]. It is
used with the most recent pest identification and machine learning research projects.

Figure 2: PlantVillage Dataset

Figure 2 depicts the PlantVillage dataset. It consists of 54306 healthy and unhealthy leaf images divided into 38 categories by
species and disease.
Farmers may use several apps to forecast crop yields depending on meteorological variables. In order to anticipate crops, machine
learning techniques have been utilised. For the five climate factors the Random Forest Algorithm is used to train the model,
however additional inputs like as soil quality, pest, chemical materials utilised are not taken into account. The model was taught to
build random forest by 200 decision-making trees.

©IJRASET: All Rights are Reserved 1586

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

This approach is based mostly on weather forecasts, plantations of crops, crop forecasts, and crop costs. For this model, the data set
for economic farming is analysed [14,15]. It is then pre-processed and divided into training and test data. For excellent precision,
Support Vector Machine and random forest methods are employed. The final result is to forecast crop yields and to designate crop
yellow yields as the best bio condition. In the developing country it is hard to achieve smart farming since many farmers do not
know the technologies and are uneducated.
There were four types of agricultural yield predicting methods or combinations: (a) field investigation, (b) plant growing modelling,
(c) remote sensing, and (d) statistical models. The merits and disadvantages of these techniques. Field surveys attempt to detect the
reality of the ground using farmers' reports and objective surveys. Due to sampling mistakes and non-sampling, these studies suffer
from decrease in replies, resource constraints and dependability [16]. Process-oriented crop models simultaneously increase crops
and develop crop by inputs depending on crop characteristics and environmental circumstances. They employ agro-growth and
development concepts, which apply throughout time and space. However, all yield reduction variables are not taken into
consideration and substantial data and validation needs are present. Remote sensing attempts to get current crop information via
satellite pictures. Remote sensing information is available internationally and does not suffer from human mistakes under open data
regulations. Satellite data readings only offer indirect measures of the agricultural yield, specifically measured irradiance, so as to
translate satellite data into yield predictions on physicochemical or analytical frameworks. Statistics models employ weather
indicators and predictors for the results of the three preceding techniques [16,17]. These models assess the yield rate trend for the
development and management of genetics and fit linear models between predictors and residues. They offer high precision, but
cannot be expanded into various space and time settings. Reusability in agricultural system modelling was not a design objective;
the underlying science has been given more attention. Examples of machine applications that learn to forecast agricultural
production are similar in design. Methods have not been concentrated on reusability or transferability. Our machine learning
platform has been developed to focus on flexibility and reusability.

III. CONVOLUTIONAL NEURAL NETWORK FOR PEST DETECTION

Convolutional Neural Networks (CNNs) fall into the deep neural network model category and have regularized multilayer
perception. A completely connected network causes data overfitting. It outperforms hierarchical structures and aids in the resolution
of problems. Using simple patterns to create complicated patterns is a great way to get started. In a neuronal network, the pattern of
neurons is called neuronal connectivity. The biological process is analogous to the convolution neural network [17]. It requires only
tiny amounts of data throughout the image classification process. It takes an image as input and then assigns different weights to the
various items in the image, making each thing unique. The pre-processing task in the convolution neural network is minimal when
compared to other classification algorithms. Neuron connectivity in the human brain is similar to that of CNN's neurons. It is
capable of learning the filters it contains. The dependencies existing in the spatial and temporal components are captured with the
help of the relevant filter. During the process, it plays a significant role in reducing the image to simpler forms, with no image loss.
CNN architectures are made up of a series of layers that translate one activation volume to another using differentiable functions.
CNN is primarily built using layers such as convolution, pooling, and fully connected. On CNN, the convolution layer is the most
important component. The network's computational load is carried in it. Between the kernel and the limited region, it primarily
executes dot products. The kernel has a large depth but a tiny spatial dimension. In kernel, only the depth will be increased, but the
height and breadth will remain tiny. Image representation is achieved by sliding the kernel in the image's height and breadth.

Figure 3: Convolutional Neural Network for Pest Detection

A CNN model is trained with a class label dataset and then finally tweaking it by utilising just a few instances from the target
domain dataset as shown in Figure 3.

©IJRASET: All Rights are Reserved 1587

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

IV. PROPOSED METHOD

The proposed solution is an android application. In Plant Disease detection the image of the plant is simply taken and uploaded to
the mobile device. Then this image is supplied by a Convolutional Neural Network encoding this image in a numerical array, which
is classified in the model with the other numerical arrays. The model is a TensorFlow model, which is built from the huge size of the
conventional TensorFlow model into a TensorFlow model. This model helps to categorize the numerical value of the image supplied
into data sets. When a numerical array matches the trust is calculated and the trust value is displayed.
The created technology is linked with a smartphone to enhance the efficiency of farmers [18]. A CNN Object Detection model is
implemented on a mobile device using the proposed system using the Keras platform to find pests in the picture. Plant disease
detection includes five key steps: image acquisition, image pre-processing, image segmentation, feature extraction and grading.
Digital camera or scanner is used as part of image processing, pre-processing comprises enhancing image processing, dividing
pictures where the afflicted and healthy regions are divided, extracting the feature defines the area of infection and helping classify
illness.

A. Dataset
For Pest Detection we have used The Plant Village dataset. It is split into 18 groups, comprising 54306 pictures of various plant
leaves. This collection includes 13 plant kinds and 26 plant disease categories. The data collection includes both healthy and ill
pictures of crops. Fourteen crop species, including: apple, blueberry, squash, strawberry, orange, peach, pepper, potato, raspberry,
soy and tomato, are shown. The two areas for each class are the plant name and illness name. As shown in Figure 4 all the pictures
are scaled and divided for further categorization and pre-processing.

The Pest Detection module follows the following steps:

1) Image Acquisition: Image acquisition is the collection or collection procedure with and without illness of plant leaf pictures.
The system's accuracy depends mostly on the picture kinds utilised, as training is carried out. Images are taken from or
collected using a digital camera on the farm [19]. The quality of the image relies on the kind and orientation of the digital
camera employed. The first procedure is to acquire the picture data that is utilised as a computational input. Image data entry in
.bmp,.jpg,.png,.gif format should be provided.
2) Image Pre-processing: Pre-processing of the picture follows acquisition of the image. Image pre-processing refines images by
noise, enhance, resize, increase data, cuts, convert colour space, smoothing etc. images. The recorded leaf pictures might reveal
insects, insect faeces, dust and squirrels etc. that are all thought to be eliminated noise as shown in Figure 4. Enhanced distorted
pictures with noise reduction filters that eliminate distortions [20]. Contrast improvement techniques are required if the image
contrast is poor. The job requires just sheet pictures and the rest of the pieces are regarded as the backdrop. Hence, approaches
to remove background leaves from entire pictures are utilised for removing them.
3) Image Segmentation: In the identification of leaf diseases, the segmentation of picture plays an essential role, as pre-processed
images are taken from the area of interest. The division of the picture into distinct portions of a leaf requires the division of the
image. Segmentation may be carried out by utilising several approaches, such as Otsu, k-means, thresholding, region, edge, etc
[21]. Deformation segmentation takes the intensity values into account when splitting photographs and this is an example of
edge detection. Colour variations are seen in infected leaves and such leaf pictures are broken off using the clustering process k-
means to remove sick parts from the leaves.
4) Feature Extraction: The extraction procedure involves the identification and extraction of intrinsic properties known as image
disease descriptive features. Colour, texture and form characteristics are generally extracted. Colour characteristics distinguish
between one illness and another based on colour and are main colour aspects of the histogram and moments [22]. For the
categorization of diseases, textures that indicate how picture textures are dispersed are retrieved. Examples of textural
characteristics include entropy, homogeneity and contrast. The form shows how the symptoms of the disease differ from each
other. For leaf diseases, structural extraction is better than colour and texture.
5) Disease Classification: As shown in Figure 4, the collected characteristics are utilized for the categorization of leaf diseases.
Classification is a monitored approach for mapping pictures from leaves to various disease classifications [23]. The classifier
technique produced describes the predetermined set of illness classes by learning from pictures with disease labels as shown in
Figure 4. This phase of learning is known as the stage of training. The trained classifier is used to test the pictures and the
precision achieved is dependent on the trained classifier.

©IJRASET: All Rights are Reserved 1588

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

Figure 4: Flowchart for Pest Detection

V. COMPARISON BETWEEN DIFFERENT PEST DETECTION TECHNIQUES

Table 1 shows the comparison between different Pest Detection techniques based on the classification methods used, name of the
disease and accuracy obtained.

CLASSIFICATION PLANT DISEASE RESULT

REFERENCE METHODS

[1] 1. Support Grape Scab and Rust disease 80% accuracy

Vector
Machine
[2] 2. Artificial Neural Apple Infected 82% accuracy
Network
[6] 3. Neural Cotton Cercospora, Red Spot 89.56%
Network accuracy
[8] 4. Fuzzy KNN Tomato Nutrient deficiency More than 83.5%
accuracy
[9] 5. Feed Grape Downey Mildew 85% accuracy
Forward
BPNN

Table 1: Comparative Study of different Pest Detection Techniques

©IJRASET: All Rights are Reserved 1589

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

VI. RESULTS AND DISCUSSIONS

The image of the leaf can be taken and sent to the android device in the pest detection module. After this the model will predict the
type of pest in the leaf. As shown in Figure 7, for the pest detection model, the accuracy of the model obtained is 95.16%. Also, the
confidence value is determined and the value of the confidence is presented if a numeric array matches.

In the android application, the user can check pest in the plants via:
A. Camera
B. Gallery
C. Live Detection

The performance metrics that are considered in our proposed work are as follows.
1) Performance Accuracy: The total number of correctly classified images to the total number of images.
2) Loss Function: How well the architecture models the data.
3) Precision: The ratio of the number of correctly predicted observations (true positives) to the total number of positive
predictions (true positives + false positives).
4) Recall: the ratio of correctly predicted observations (true positives) to all observations in that class (true positives + false
negatives).
5) F1 Score: the Harmonic Mean between precision and recall.
6) Time requirement (in sec) per epoch for training each DL model.
The output is as follows:

Figure 8 represents the detection of pest in Tomato plant and the name of the pest is Septoria leaf spot.

Considering four fundamental classification metrics:

 validity,
 precision,
 recall, and
 f1 scores,
the total performance of the system designed for pest recognition is evaluated.

©IJRASET: All Rights are Reserved 1590

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

In addition, a comparison is made with the performance of earlier research, i.e., neural network back propagation (BP) and single-
shot detector (SSD) MobileNet, for the outcome of the pesticide categorization. These findings corroborate the results in order to
detect automated agricultural pests from our suggested Faster CNN. The SSD MobileNet was successful in determining all the
pictures of pests that have been examined.
But the Faster CNN suggested achieves the greatest mean precision value (98.0%) over BP's and SSD's 50.0%, and SSD's 86.0%
respectively. respectively, the Faster CNN offered. Furthermore, our suggested Faster CNN model's prediction accuracy has been
validated for multiple sizes of pest pictures evaluated, by altering the percentage of data divided into 70–30% and 90–10% for
training and tests. Normally, by expanding from 70% to 90% of the complete data set the performance of agricultural pest classifiers
is enhanced. Both BP neural network and SSD MobileNet classification accuracy rose to 43.0% and 85.6%, respectively, for 30%
testing data, but our proposal Faster CNN's accuracy rate remained high at 94.0% at the same test data amount.

VII. CONCLUSION
The aim of this paper is to identify illnesses in crops by means of the deep learning method, which is the Convolutional Neural
Network. The model is essentially evaluated for certain species of plants with certain kinds of plant diseases. The template was built
using tensor flow, Keras and Android. The total system findings indicate that the MobileNet model functions better than other
models and provides improved accuracy in illness detection. The number of plant types and their diseases will expand as an addition
to the project. The model will also be enhanced by increasing the training and testing parameters. Without involving any middlemen
between farmers and consumers and earning profit, farmers/customers may sell/purchase farm products at an ideal cost. Farmers
will find it more helpful to know information about existing farms and feel it is a safer and more valuable website. In order to
propose optimal harvests with greater precision and productivity the framework uses supervised machine training algorithms. The
model is trained to validate the performance of the current model created using ANN with decision-tab classifier. The accuracy
values of the measured values have shown a better 95 percent accuracy suggestion model built using ANN compared with 92
percent accuracy achieved from decision book classification systems. In addition, with bigger datasets, ANN performs well.

REFERENCES
[1] Liu, Jun, and Xuewei Wang. "Plant diseases and pests detection based on deep learning: a review." Plant Methods 17.1 (2021): 1-18.
[2] SANJU, S. KRITHIKHA, and DR BL VELAMMAL. "An Automated Detection and Classification of Plant Diseases from the Leaves Using Image processing
and Machine Learning Techniques: A State-of-the-Art Review." Annals of the Romanian Society for Cell Biology (2021): 15933-15950.
[3] Selvaraj, Michael Gomez, et al. "AI-powered banana diseases and pest detection." Plant Methods 15.1 (2019): 1-11.
[4] Pattnaik, Gayatri, Vimal K. Shrivastava, and K. Parvathi. "Transfer learning-based framework for classification of pest in tomato plants." Applied Artificial
Intelligence 34.13 (2020): 981-993.
[5] Reardon, Thomas, et al. "“Pivoting” by food industry firms to cope with COVID‐19 in developing regions: E‐commerce and “copivoting” delivery
intermediaries." Agricultural Economics 52.3 (2021): 459-475.
[6] Kartikeyan, Punitha, and Gyanesh Shrivastava. "Review on emerging trends in detection of plant diseases using image processing with machine
learning." International Journal of Computer Applications 975 (2021): 8887.
[7] Patel, Deven, and Nirav Bhatt. "Improved accuracy of pest detection using augmentation approach with Faster R-CNN." IOP Conference Series: Materials
Science and Engineering. Vol. 1042. No. 1. IOP Publishing, 2021.
[8] Afifi, Ahmed, Abdulaziz Alhumam, and Amira Abdelwahab. "Convolutional Neural Network for Automatic Identification of Plant Diseases with Limited
Data." Plants 10.1 (2021): 28.
[9] Mique Jr, Eusebio L., and Thelma D. Palaoag. "Rice pest and disease detection using convolutional neural network." Proceedings of the 2018 International
Conference on Information Science and System. 2018.
[10] Jiao, Lin, et al. "AF-RCNN: An anchor-free convolutional neural network for multi-categories agricultural pest detection." Computers and Electronics in
Agriculture 174 (2020): 105522.
[11] Karar, Mohamed Esmail, et al. "A new mobile application of agricultural pests recognition using deep learning in cloud computing system." Alexandria
Engineering Journal 60.5 (2021): 4423-4432.
[12] Gavhale, Kiran R., and Ujwalla Gawande. "An overview of the research on plant leaves disease detection using image processing techniques." IOSR Journal of
Computer Engineering (IOSR-JCE) 16.1 (2014): 10-16.
[13] Chen, Jian-Wen, et al. "A smartphone-based application for scale pest detection using multiple-object detection methods." Electronics 10.4 (2021): 372.
[14] Aquil, Mohammad Amimul Ihsan, and Wan Hussain Wan Ishak. "Evaluation of scratch and pre-trained convolutional neural networks for the classification of
Tomato plant diseases." Int J Artif Intell 10.2 (2021): 467-475.
[15] Thakur, Sahil, et al. "Plant Disease Detection and Solution Using Image Classification." (2021).
[16] Singh, Piyush, Abhishek Verma, and John Sahaya Rani Alex. "Disease and pest infection detection in coconut tree through deep learning
techniques." Computers and Electronics in Agriculture 182 (2021): 105986.
[17] Ozguven, Mehmet Metin, and Kemal Adem. "Automatic detection and classification of leaf spot disease in sugar beet using deep learning algorithms." Physica
A: Statistical Mechanics and its Applications 535 (2019): 122537.
[18] Fuentes, Alvaro, et al. "A robust deep-learning-based detector for real-time tomato plant diseases and pest recognition." Sensors 17.9 (2017): 2022.

©IJRASET: All Rights are Reserved 1591

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue XI Nov 2021- Available at www.ijraset.com

[19] Lu, Jinzhu, Lijuan Tan, and Huanyu Jiang. "Review on Convolutional Neural Network (CNN) Applied to Plant Leaf Disease Classification." Agriculture 11.8
(2021): 707.
[20] Fuentes, Alvaro F., et al. "High-performance deep neural network-based tomato plant diseases and pest diagnosis system with refinement filter bank." Frontiers
in plant science 9 (2018): 1162.
[21] Shruthi, U., V. Nagaveni, and B. K. Raghavendra. "A review on machine learning classification techniques for plant disease detection." 2019 5th International
Conference on Advanced Computing & Communication Systems (ICACCS). IEEE, 2019.
[22] Ngugi, Lawrence C., Moataz Abelwahab, and Mohammed Abo-Zahhad. "Recent advances in image processing techniques for automated leaf pest and disease
recognition–A review." Information processing in agriculture 8.1 (2021): 27-51.
[23] Suma, V., et al. "CNN based leaf disease identification and remedy recommendation system." 2019 3rd International conference on Electronics,
Communication and Aerospace Technology (ICECA). IEEE, 2019.

©IJRASET: All Rights are Reserved 1592