Professional Documents
Culture Documents
VISI KOMPUTER
DISUSUN OLEH:
SURABAYA
2022
Turkish Journal of Computer and Mathematics Education Vol.12 No.2 (2021), 1672-1686
Research Article
Article History: Received: 10 November 2020; Revised: 12 January 2021; Accepted: 27 January 2021;
Published online: 05 April 2021
Abstract: In the field of agriculture, especially paddy plants, there is a demand for research to classify the paddy
diseases at early stages. This is feasible if there are automated systems that can assist the farmers to recognize the
paddy diseases from the paddy leaf images of the plants. The recognition of agricultural plant diseases by
utilizing the image-processing and machine learning techniques can certainly minimize the reliance on the
farmers to protect the yield of paddy crops. In this paper, an attempt has been made to pre-process the images to
prepare the feature-set for Classifiers and then feature extraction algorithms are used to extract the relevant
features from the processed images. The feature-set is then supplied to the classifiers for identification of Paddy
Leaf diseases. The usage of cascaded classifiers has been explored to detect the diseases of paddy leaves. An
attempt has also been made to use genetic algorithm with nearest neighbour algorithm to identify the diseases of
paddy leaves. The proposed automated system can be used on Android , Windows platform and Apple platform
for quickly identifying the paddy leaf diseases as the entire implementation has been performed using MATLAB.
The proposed automated system can certainly help the farmers to classify the diseased paddy leaves at early
stage to protect the crops from further damage.
Keywords: Paddy disease detection, Classification, Adaboost algorithm, Bagging classifier.
1. Introduction
The machine learning based automated systems are the need of the hour for the Indian fields that produce
great amount of rice to identify the diseases as soon as initial symptoms of the diseases appear on the paddy
leaves and to save the yield from further damage (Qing & Zexin, 2009). Machine learning techniques help to
extract the features from the images and then redundant features can also be reduced using machine learning
methods (Pugoy & Mariano 2011). Once the feature set is ready, again machine learning and/or deep learning
methods can be used to classify the images in appropriate classes. The plant images can be easily identified using
machine learning as well as deep learning approaches with high accuracy (Phadikar & Sil, 2008). In Figure. 1, a
generic approach to identify the plant diseases using machine learning approaches is demonstrated in a pictorial
representation.
In generic approach, the images of diseased plants are captured and processed, so that the features from the
images can be extracted for further processing. After pre-processing of images, the features of the images are
extracted to prepare a feature set. The parameters of feature set can directly be supplied to the classifiers or
feature reduction algorithms can be used if feature set contains redundant features that do not play any role in
classification of the images. The classifiers are trained using the feature set and tested after attaining an
acceptable accuracy. The model gets approval after gaining good accuracy in both trained as well as tested data/
feature sets. However, machine learning approaches are providing acceptable accuracy in classification of
images of diseased plants but deep learning techniques are gaining more attention because of delivering better
accuracy in identification of plant diseases from the images of the diseased plants.
1672
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
Agriculture is India's largest source of revenue. The vast majority of the people in India rely on agriculture
for their income. Agriculture is the most important sector of the Indian economy (Agriculture Sector in
India,2015). Over 58% of the rural residents rely on farming for their primary source of income relates (Verma,
2017). Rice is a basic food for the majority of rural populations, and is the second most-produced cereal. Rice is
a nutritional staple in India. Rice disease causes 10 to 15% of crop losses in Asia (Papademetriou, 2000). An
agricultural crop is grown in five continents, namely Asia, Africa, America, Europe, and Oceania. Deliver and
Devour 91.05% of the world's rice according to the Food and Agriculture Organisation of the United
Nations (FAOSTAT) The rice production as per the population in different areas across the world like Africa-
2.95% of , America-5.19% ,Asia-91%, Europe-0.67% and Oceania-0.15% of the population as seen in Figure. 2
America 5.19 %
Afirca 2.95 %
Oceania 0.15 %
Europe 0.67 %
Asia 91%
Rice consumption is projected to increase more rapidly than supply in most countries. In this case, some kind
of harm to the crop is unacceptable (Khoenkaw ,2016). Prevalence of rice disease has often been difficult to
ascertain. a naked-eye study was the only way to diagnose the rice disease For disease detection to be effective,
constant observation of the field is needed (Prakashet & Saraswathy, 2017). Visual research relies on the use of
continuous human interpretation, which greatly increases its expense, effort, and (Jaskaran & Harpreet, 2018).
1673
Nilam Sachin Patil, E. Kannan
When the unprecedented growth of the number of people occurs, so does the steady increase in the market for
food products. The strain places the whole population under the necessity of utilizing new technologies for early
detection and successful care, so there is little space for error. Image analysis techniques are one of the cost-
effective and reliable approaches used for differentiating plant diseases (Jitesh & Harshad, 2016); (Iswarya &
Maheswari, 2019). Plants are susceptible to infection when fungi and bacteria cause illness in them. leaf blight,
brown blot, sheath blight, and leaf scorch (Rice Production , 2015). These diseases will place severe economic
pressure on the rice farmers across the board. farmers may ignore diseases or struggle to recognize them, which
lead to the loss of the crop Per illness has its own solution. when a disease appears on a plant, farmers have to
monitor its spread (Basavaraj & Surendra 2020). This disease detection process needs some due diligence during
the selection of pesticides. seemingly-infected cattle and letting themselves be captured by an automated device
may be a potential alternative for farmers. With such mechanism the farmers can be kept updated on diseases
immediately, many of them can save money and time from the major economic losses (Usha & Priyadharshini,
2019).
One of most popular problems from agricultural science is the identification of diseases from the images of
the plant leaves and to suggest the remedies to get rid of these diseases. The remedies can be suggested by the
automated systems if the identification of the plant diseases can be made accurately. The usage of the image
processing techniques along with machine learning based image identification techniques minimize the need of
manual ways for the farmers to safeguard their crops from the diseases. The automated strategies are very useful
for classifying paddy leaf disease from the images of the diseased plants. Rice diseases are identified from the
images of the diseased leaves which enables accurate detection and classification of the paddy leaf diseases
without investing time on manual inspection. Machine learning based methods assist in detecting the paddy leaf
diseases in a faster and accurate manner. The outcome of these automated systems can eventually improve the
agricultural production and reduces economic losses. Hence, this paper proposes image processing, feature
extraction and classification techniques to identify the paddy leaf diseases through an automated system.
An attempt has been made to use machine learning algorithms for identification of the paddy leaf disease.
1. The images of paddy leaves have been processed to reduce the unwanted distortions, enhances image
features for further processing. It involves resizing of images, brightness correction, filtering, illumination
corrections, noise removal, geometric transformations, and grey scale transformations.
2. After pre-processing of images, feature extraction algorithms are applied, so that the relevant feature
dataset can be attained from the processed images.
3. Next, cascaded classifiers with modifications with respect to the problem statement are applied to obtain
better accuracy.
4. The proposed automated system can be used on Android and Windows platform for quickly identifying
the paddy leaf diseases.
In this paper is structured mainly into five sections. The paper begins with background study, need for the
work and contribution of the research work, followed by literature survey in section II, this section discusses the
existing methods used in identification of paddy leaf diseases. Section III describes about the proposed
methodology for identification of the paddy leaf diseases. Next section discusses the obtained results. The last
section explains conclusion of the research study presented in this paper.
2. Literature review
Geraldin B. Dela Cruz(2019) introduced a smartphone application that assisted the rice farmers to determine
nitrogen deficiency depending on the plant coloration. The tool may be used instead of or in conjunction with
conventional nitrogen use. An easy to use technology for farmers was proposed without any training to the
model. This paper introduced automated image processing methods using numbers to obtain high accuracy
results. The target outcomes were computed by using the z-score statistical approach.
Anthon G. and Wickarchi N.(2009) discussed the crucial role of Paddy disease classification and recognition
in terms of financial growth of the agricultural sector. To find a reliable image-based diagnostic device for paddy
diseases, digital cameras were used to take images under experimental conditions for this research study. Three
1674
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
tropical diseases were chosen for this study: rice blast (Magnaphelgrisea), rice sheath blight (Rhizoctoniasolani)
and brown dots (Cochiobotriamiyabeanus). Digital picture processing began with a green paddy bacterial
infection leaf. These images were partitioned into the segments using the tools of geometry and trigonometry.
There was a great degree of precision was achieved for classifying the images of diseased paddy leaves.
Jitesh P. Shah et al. (2016) had provided a survey report on the infected rice samples by using the image
analysis and machine learning techniques. This paper surveyed several image processing and machine learning
methods that were applied for plant disease diagnosis and classification. This study provided detailed insights
into 19 research works based on rice plant diseases. This research also offered a survey on critical parameters
including dataset size, classifier number, preprocessing, classifier forms, and accuracy etc. J. Yang et al. (2016)
suggested an accurate and efficient Nitrogen dosing with a laser-induced fluorescence (LIF). The LIF approach
was used to determine the vitamin B2 levels in rice using ultraviolet light(355 nm) (excitation light source). This
described the differences in the fluorescence spectra observed in Nitrogen-supplemented and Nitrogen-free rice
leaves. The numbers of relevant features were then extracted from the fluorescence spectra, and tailored for N-
fertilizer dose. Binary SVM method has been used for classification. The accuracy of their system was around
95%.
Archana K.S. & Sahayadhas A. (2018) discussed a disease detection technique in the agriculture sector by
using advanced picture analysis to extract features and classification algorithms. The major challenge in
detection of plant leave diseases are feature recognition/feature extraction and accurate classification. This paper
proposed an algorithm for accurately predicting the bacterial infection(Oryza sativa)in rice plants at early stages.
Numerous picture segmentation and classification techniques were attempted for identifying the paddy leaf
diseases. The detailed analysis was made to assess the performance of the proposed work.
Latte M V & Shidna S. (2016) described the methods to detect and analyze the Paddy leaf deficiencies
including nitrogen, phosphorus, and potassium with pattern detection techniques. Using the nutrient elements
technology, the aim was to identify several leaf elements that could become potential reason behind deficiency in
paddy leaves. In this paper, a pattern detection method was proposed to analyze the color pattern of the paddy
leaves such as red, green, blue, and gold light brown color. The rice plant colors could also depict the dead
leaves. A database of phosphorus, potassium, and nitrogen-deficient paddy leaves was prepared before applying
the pattern detection techniques. A comparative study was made to analyze the performance of proposed study
over the existing methods. The color patterns were used to detect the deficiency in paddy leaves more accurately
in comparison to other nutritional components.
Prabira K.S et al. (2020) had examined the use of image processing methods for diagnosing the rice plant
diseases. The authors discussed image processing methods, disease recognition methods, and feature extraction
methods on the basis of state-of-the-art work. This paper also provided highlights on the recent developments in
automated rice plant disease detection techniques, and also discussed the potential challenges and opportunities
in the field of paddy leaf diseases. Prab Kumar et al. (2017) discussed the techniques to identify the stages of the
rice plant diseases. The color abnormalities and dead zones on leaves were visualized. The experimental results
proved that the procedure for identification of faulty components on the leaves of the rice crop was highly
accurate. S.Phadke and J. Coval (2008) demonstrated the simple prototype for identifying the rice plant diseases
along with the identification of stages of therice plant diseases based on the images of the rice plants. Digital
cameras were used to identify the soiled parts of the rice plants with high accuracy.
Ramesh S. & Vydeki D. (2018) discussed a disease detection technique for rice plants. The images of
healthy and infected leaves were collected for supplying the inputs to the system. For the training process, the
simulation produced a reliable estimation of 90% of the infected images and an additional 86% for the healthy
ones. S. Sengupta & A. K. Das (2017) proposed rule-based classification method using the principles of Particle
Swarm Optimization (PSO) and Association Theory (AT) to present an incremental classification scheme for
classifying the diseases. The characteristics of rice diseases are constantly shifting with temperature, biological,
and geographic influences, and thus the incremental classifier is appropriate for the prediction of
diseases. Accuracy of the suggested technique has been assessed using all rice disease models and many times
and using state-of-art classification methods. Statistical methods are often used to examine the results of the
process which are used to determine their meaning and performance. Y. Zhou et al. (2020) conducted a
systematic study on over 1883 rice varieties in China in order to classify blast-resistance alleles. This research
work included the range of viable blast resistance alleles to include thirteen novel alleles (resistance strains).
1675
Nilam Sachin Patil, E. Kannan
However, many research approaches have been presented by the researchers but still there is a need of
simple and viable automated system that can quickly recognize the rice plant diseases and help the farmers to
take corrective measures to protect their crops at early stages,
3. Proposed Methodology
Classification output
Accuracy Assessment
Dataset: The collection of data and preparation of dataset is very important for developing any machine
learning based application, so for developing the Paddy leaf diseases diagnostic system. We have collected the
images/videos for paddy leaves manually from the fields of Maharashtra, and we have also used online
repositories of images such as Kaggle for the training of machine learning algorithms in initial phases.
Images Pre-processing: Before using the dataset to train our model, the series of pre-processing steps are to
be applied to the data to enhance the images for further usage. During pre-processing, the image data reduces the
unwanted distortions, enhances image features for further processing. It involves resizing of images into uniform
sizes, brightness correction, filtering, illumination corrections, focus corrections, noise removal, thresholding,
geometric transformations, and grey scale transformations. This process prepares the images of diseased paddy
leaves for further analysis.
Following are the raw images from different categories like Hispa, Brownspot, Leafblast and Healthy leaves.
1676
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
The dimension of the image is 1051 x 1051 and the format of the image is JPEG. We have reduced a
dimension when the images are transformed from the BGR color space to gray scaled images. Grayscale is a
range of monochromatic shades from black to white. Many image editing programs allow us to convert a color
1677
Nilam Sachin Patil, E. Kannan
image to black and white, or grayscale. This process removes all color information, leaving only the luminance
of each pixel. The luminance of a pixel value of a grayscale image ranges from 0 to 255. The conversion of a
color image into a grayscale image is converting the RGB values (24 bit) into grayscale value (8 bit). Gray scale
is the most preferred pre-processing technique in image processing as it is one layer image from 0-255 whereas
the RGB technique has three different layered images. Therefore we have preferred grey scale image instead of
RGB. RGB color increases the complexity of the model and hence converting the images to a grayscale is
beneficial in order to reduce the complexity of the ML based classification model.
Figure 8. Original Image Shape (1051, 1051, 3) Figure 9. Grey Scaled image (1051, 1051)
Then we have taken analyzed the distribution of the intensity values of all the pixels presented in the images.
It was required to flatten the image matrix and put the intensity values of all the pixels in one single row vector
to find the prominent peaks.
Figure 10. Presenting the distribution of intensity values Figure11. Presenting the distribution of Zoomed
intensity values
1678
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
It can be observed that there are 2 prominent peaks. The count of pixels with intensity values around 0 is
extremely high (30000). It is expected that the leaf should cover a smaller portion of the picture as compared to
the background color which is primarily black. The next thing to do is to separate the two, that is, the leaf from
the background. The optimal separation value is somewhere around 20 but rather than relying on such
descriptive statistics, we have used a more formal approach known as Otsu's method. Otsu's method assumes that
the image contains two classes of pixels following bi-modal histogram (foreground pixels and background
pixels), it then calculates the optimum threshold separating the two classes so that their combined spread (intra-
class variance) is minimal, or equivalently, so that their inter-class variance is maximal. Otsu’s method exhibits
relatively good performance if the histogram can be assumed to have bimodal distribution and assumed to
possess a deep and sharp valley between two peaks .Next masking has been done. Masking helps in locating the
whole leaf from the image.
We have also applied K-means for segmentation of the leaf from the background. The comparison has been
made between the labels of Otsu and K-Means at a pixel level, summing over the booleans and dividing them by
the total number of pixels in the image. If the result is 1, it means there is no difference at all.
Feature Extraction: For a given image of diseased paddy leaves, the feature extraction begins from an initial
set of data and builds features, so that the more informative feature dataset can be prepared. Feature extraction is
a kind of process, where an initial set of raw variables is processed to more manageable groups.
In the proposed research methodology, we have used HOG (Histogram of oriented gradients) feature
extraction method. HOG is a feature descriptor used to detect objects in computer vision and image processing.
The HOG descriptor technique counts occurrences of gradient orientation in localized portions of an image -
detection window, or region of interest (ROI). It divides the image into small connected regions called cells, and
for each cell compute a histogram of gradient directions. It thendiscretizes each cell into angular bins according
to the gradient orientation.Each cell's pixel contributes weighted gradient to its corresponding angular bin.The
grouping of cells into a block is the basis for grouping and normalization of histograms.Normalized group of
histograms represents the block histogram. The set of these block histograms represents the descriptor. HOG is a
proven feature descriptor used to detect objects in computer vision and image processing. Hence, we haveused
HOG for feature extraction.
Feature Reduction:The feature reduction technique known as PCA(Principal component analysis) has been
applied to remove redundant features which do not impact the decision making of classifiers. The reduced
feature-set has been prepared and used further by the classifiers to identify the diseases of paddy leaves.
Figure 12. PCA is presenting extracted features from images of diseased plants, where each paddy leaf image is
represented by a distinguished colour.
1679
Nilam Sachin Patil, E. Kannan
Classification using Cascaded Classifiers:We have then made use of cascaded classifiers (such as
AdaBoost, and Bagging classifiers) with modifications in the algorithms for identification of diseases of paddy
leavesby using the reduced feature set to achieve better accuracy. Cascading is a particular case of ensemble
learning based on the concatenation of several classifiers, using all information collected from the output from a
given classifier as additional information for the next classifier in the cascade.
a. Adaboost Algorithm
Boostingis a general ensemble method that creates a strong classifier from a number of weak classifiers. This
is done by building a model from the training data, then creating a second model that attempts to correct the
errors from the first model. Models are added until the training set is predicted perfectly or a maximum number
of models are added. AdaBoost is a kind of gradient boosting with built-in functionality of cross-validation. It
allows the user to run a cross-validation at each iteration of the boosting process and thus it makes the process of
getting the exact optimum number of boosting iterations in a single run quite easy. Hence, we are using
AdaBoost in our research work for promising results.
b. Bagging Classifier
Bagging Classifier is one of the most powerful and popular machine learning algorithms. It is an ensembling
algorithm that is used for classification. Bagging stands for bootstrap aggregation and is a powerful statistical
method for estimating a quantity from the given dataset. Bagging tries to employ similar learners on small
samples of the datasets and then takes a mean of all the results. Bagging uses bootstrap sampling to obtain the
data subsets for training the base learners. An ensemble method in Bagging classifier combines the results from
multiple machine learning algorithms altogether to make better and accurate predictions than any single
algorithm. For aggregating, the outputs of base learners, bagging uses voting for classification. We have made
use of bagging with diverse algorithms for voting and we have attempted to improve the output by experimenting
with multiple algorithms using bagging concept. The results obtained are shown in the next section.
A. Initialization of NNA
A portion of the TP is used to initialize the vectors that are used by NNA. During this procedure, weighted
distances to a given individual are determined between the image vectors and initialized vectors.
B. Training of GA
The training phase of the GA needs the problem formulation as an individual or a chromosome, and the fine
tuning of evolutionary operators such as mutation, crossover, and elitism. The replace mutation and alternating-
position crossover are used with the rate of 0.03 and 0.8 respectively.
1680
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
This section provides insights into the results obtained by using the three machine learning algorithms.
Assessment of Accuracy: The results obtained from machine learning based classifiers will be assessed to
measure the performance in terms of accuracy of the detected paddy leaf diseases. We will be using Confusion
matrix, Area under Curve, F1 score, Precision and Recall matrices to compare the performance of the cascaded
algorithms used for classification.
Following are the observations for the above confusion matrix as shown in Figure 13.
• In the testset, 120 images belong to "Hispa" disease. 85 images are correctly identified and 35 are
misidentified.
• In the testset,123images belong to "LeafBlast" rice plant disease. 59 images are correctly identified and
64 are misidentified.
• In the testset, 138images belong to "BrownSpot" rice plant disease. 53 images are correctly identified
and 85 images are misidentified.
• In the testset, 142 images belong to "Healthy" rice plants. 65 images are correctly identified and 77
images are misidentified.
•
1681
Nilam Sachin Patil, E. Kannan
Figure 14. depicts area under curve or RoC for AdaBoost classifier. ROC curve measures the performance
for the classification of paddy leaf diseases s at threshold settings. It plots the True Positive Rate against False
Positive Rate at different classification thresholds. AdaBoost is showing good performance for detection of rice
plant diseases.
Figure 15. Performance evaluation of AdaBoost using Precision, F1-Score and Recall
Figure15. depicts the performance evaluation parameters such as Precision, Recall and F1-Score for
AdaBoost classifier and the results can be further improved by fine tuning of parameters in the proposed
technique.
1682
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
Following are the observations for the above confusion matrix as shown in Figure 16.
• In the testset, 120 images belong to "Hispa" disease. 81 images are correctly identified and 39are
misidentified.
• In the testset,123 mages belong to "LeafBlast" rice plant disease. 19 images are correctly identified and
104 are misidentified.
• In the testset, 138images belong to "BrownSpot" rice plant disease. 58 images are correctly identified
and 80 images are misidentified.
• In the testset, 142 images belong to "Healthy" rice plants. 100 images are correctly identified and 42
images are misidentified.
Figure 17. provides AUC for Bagging classifier and shows the aggregate measure of performance across all
possible classification thresholds.
1683
Nilam Sachin Patil, E. Kannan
Figure 18 shows precision score, recall score and F1-score with respect to Bagging Classifier.
Following are the observations for the above confusion matrix as shown in Figure 19.
• In the testset, 120 images belong to "Hispa" disease. 90 images are correctly identified and 30 images
are misidentified.
• In the testset,123 mages belong to "LeafBlast" rice plant disease. 69 images are correctly identified and
54 are misidentified.
• In the testset, 138 images belong to "BrownSpot" rice plant disease. 78 images are correctly identified
and 50 images are misidentified.
• In the testset, 142 images belong to "Healthy" rice plants. 114 images are correctly identified and 28
images are misidentified.
Figure 20 shows performance of GA with respect to evaluation parameters such as F1-score, precision and
recall. The results show that GA based classifier outperform other two classifiers. Next is AdaBoost which
performs more accurately than Bagging Classifier and then comes Bagging Classifier. All the classifiers can be
1684
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning Methods
used to prepare an automated classifier for the detection of paddy leaf diseases. None of the discussed methods
give 100% accuracy but still these methods can be used for saving the time of farmers for identifying the
diseases of paddy leaves. The accuracy attained by GA classifier is 96% at training feature-set and 91% at testing
feature-set. The accuracy attained by AdaBoost is 88% at training feature-set and 84% at testing feature-set
whereas the accuracy attained by Bagging algorithm is 86% at training feature-set and 81% at testing feature-set.
5. Conclusion
An attempt has been made to use machine learning algorithms such as AdaBoost and Bagging Classifier
along with evolutionary algorithm such as GA for the identification of the paddy leaf diseases. The images of
paddy leaves have been processed toreduce the unwanted distortions, to enhance the image features for further
processing. It involves resizing of images, brightness correction, filtering, illumination corrections, noise
removal, geometric transformations, and grey scale transformations. After pre-processing of images, feature
extraction algorithms are applied, so that the relevant feature dataset can be attained from the processed
images.Next, the cascaded classifiers with modifications with respect to the problem statement are applied to
obtain better accuracy. Genetic algorithm has also tried with NNA for identifying the images of paddy leaf
diseases. The proposed automated system can be used on Android and Windows platform for quickly
identifying the paddy leaf diseases. It can certainly help the farmers to classify the diseased paddy leaves at early
stage to protect the crops from further damage.
References
1685
Nilam Sachin Patil, E. Kannan
16. Prabira K.S, Nalini K.B, Amiya, K.R& Santi, K.B. (2020).Image Processing Techniques for
Diagnosing Rice Plant Disease: A Survey, Procedia Computer Science, 167, 516-
530,doi.org/10.1016/j.procs.2020.03.308
17. Prabira, K., Baishalee N. & Nilamani, B (2017). Detection of Healthy and Defected Diseased Leaf
of Rice Crop using K-Means Clustering Technique, International Journal of Computer
Applications, 157(1), 24 -27, doi.org/10.5120/ijca2017912601
18. Prakash, R. M., Saraswathy, G. P., Ramalakshmi, G. Mangaleswari K. H. & Kaviya, T. (2017),
“Detection of leaf diseases and classification using digital image processing,” International
Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS), pp.
1-4
19. Pugoy R, & Mariano V.Y., (2011). Automated rice leaf disease detection using color image
analysis. 3rd international conference on digital image processing, 8009. Chengdu: SPIE; F1-F7.
20. Qing, Y., Zexin,G., (2009). Application of support vector machine for detecting rice diseases using
shape and color texture features," International Conference on Engineering Computation, pp. 79-
83.
21. Ramesh, S. & Vydeki, D. (2018). Rice Blast Disease Detection and Classification Using Machine
Learning Algorithm, 2nd International Conference on Micro-Electronics and Telecommunication
Engineering (ICMETE), IEEE, doi.org/10.1109/icmete.2018.00063
22. Rice Production (Peace Corps): Chapter 14 - Diseases of rice (2015),[Online]. Available:
http://www.nzdl.org.
23. Santanu P. & Jaya S. (2008). Rice disease identification using pattern recognition techniques, 11th
International Conference on Computer and Information Technology, IEEE,
doi.org/10.1109/iccitechn.2008.4803079
24. Shampa, S. &Asit K. D (2017). Particle Swarm Optimization based incremental classifier design
for rice disease prediction, Computers and Electronics in Agriculture, 140, 443 – 451,
doi.org/10.1016/j.compag.2017.06.024
25. Usha, K., Kanagasuba,R. S, Jaichandran R &Priyadharshini C (2019) Detection and Classification
of Paddy Crop Disease using Deep Learning Techniques, International Journal of Recent
Technology and Engineering ,8(3), 4353 -4359
26. Verma, P., (2017). Rice Productivity and Food Security in India, Centre for Management in
Agriculture, Indian Institute of Management Ahmedabad.
27. Yang, J., Gong, W. & Shi, S. (2016). Accurate identification of nitrogen fertilizer application of
paddy rice using laser-induced fluorescence combined with support vector machine, Plant, Soil
and Environment, 61(11), 501-506,doi.org/10.17221/496/2015-pse
28. Ying, Z., Fang, L., Qiong ,W., Weicong He, Bin Yuan &Wenya Yuan. (2020). Identification of
Novel Alleles of the Rice Blast-Resistance Gene Pi9 through Sequence-Based Allele Mining, Rice,
springer, 13(1), doi.org/10.1186/s12284-020-00442-z
1686
Identification of Paddy Leaf Diseases using Evolutionary and Machine Learning
Methods
Pada bidang pertanian khususnya padi, perlu adanya klasifikasi sejak dini penyakit yang
disebabkan oleh padi. Dalam hal ini, dimungkinkan jika petani memiliki sistem otomatis yang
dapat mengidentifikasi penyakit padi dari citra daun padi. Tentu saja, menggunakan teknik
pemrosesan gambar dan pembelajaran mesin untuk mendeteksi penyakit tanaman dapat
meminimalkan ketergantungan pada petani untuk melindungi hasil panen padi. Pada artikel ini,
saya mencoba untuk melakukan pre-process gambar untuk menyiapkan set fitur untuk classifier.
Algoritma ekstraksi fitur kemudian digunakan untuk mengekstrak fitur yang relevan dari gambar
yang diproses. Kumpulan fitur tersebut kemudian dikirim ke classifier untuk mengidentifikasi
penyakit daun pada padi. Penggunaan kaskade pengklasifikasi telah dipelajari untuk mendeteksi
penyakit daun pada padi. Upaya juga telah dilakukan untuk menggunakan algoritma genetik dan
algoritma tetangga terdekat untuk mengidentifikasi penyakit daun pada padi. Sistem otomatis
yang diusulkan dapat digunakan pada sebuah platform Android, Windows dan Apple untuk
mengidentifikasi penyakit daun padi dengan cepat karena semua implementasi dilakukan
menggunakan aplikasi matlab. Sistem otomatis yang diusulkan tentu akan membantu petani
memilah daun padi yang sakit lebih awal untuk melindungi tanaman mereka dari kerusakan lebih
lanjut.
Sektor Pertanian adalah sumber pendapatan terbesar India. Sebagian besar orang di India
bergantung pada pertanian untuk pendapatan mereka. Pertanian merupakan sektor terpenting
dalam perekonomian India. Lebih dari 58% penduduk pedesaan bergantung pada pertanian untuk
sumber pendapatan utama mereka. Beras adalah makanan pokok bagi sebagian besar penduduk
pedesaan, dan merupakan sereal kedua yang paling banyak diproduksi. Nasi adalah makanan
pokok di India. Penyakit padi menyebabkan 10 sampai 15% kehilangan panen di Asia. Tanaman
pertanian ditanam di lima benua, yaitu Asia, Afrika, Amerika, Eropa, dan Oseania. Kirim dan
Melahap 91,05% beras dunia menurut Organisasi Pangan dan Pertanian Amerika Serikat Bangsa
(FAOSTAT) Produksi beras per populasi di berbagai wilayah di dunia seperti Afrika 2,95%,
Amerika 5,19%, Asia 91%, Eropa 0,67% dan Oseania 0,15% dari populasi. Penyakit-penyakit ini
berdampak serius pada perekonomian. Tekanan pada petani padi secara menyeluruh. Petani
mungkin mengabaikan penyakit atau berjuang untuk mengenali apa yang menyebabkannya.
Setiap penyakit yang menyebabkan gagal panen memiliki solusi tersendiri. Ketika penyakit
terjadi pada tanaman. Proses pendeteksian penyakit ini memerlukan perhatian pada periode-
periode seperti berikut: Pemilihan pestisida. Sapi yang tampak terinfeksi dan membiarkan diri
mereka ditangkap oleh peralatan otomatis menjadi pilihan bagi para petani. Mekanisme seperti
hal tersebut dapat memungkinkan petani untuk menjaga informasi penyakit tetap berjalan dengan
teratur, banyak dari mereka akan dapat menghemat uang dan waktu dari kerugian finansial yang
sangat besar.
Metode Penelitian
1. Pra-pemrosesan gambar dari penyakit daun padi untuk diproses.
2. Proses ekstraksi fitur dari gambar penyakit daun padi.
3. Penyaringan terhadap fitur yang tidak diinginkan menggunakan teknik pengurangan fitur.
4. Klasifikasi penyakit daun padi menggunakan pengklasifikasi bertingkat berdasarkan
machine learning.
Dataset
Akuisisi data dan persiapan kumpulan data sangat penting untuk pengembangan mesin
apa pun Aplikasi berbasis pembelajaran untuk mengembangkan sistem diagnosis penyakit daun
padi. Repositori gambar seperti Kaggle untuk melatih algoritma pembelajaran mesin pada tahap
awal.
Pra-pemrosesan
Berikut adalah langkah-langkah Pra-pemrosesan sebelum melatih model dengan dataset:
Diterapkan pada data untuk mengoptimalkan gambar untuk penggunaan lebih lanjut. Pra-
pemrosesan mengurangi data gambar distorsi yang tidak diinginkan, meningkatkan fitur gambar
untuk diproses lebih lanjut. Ini termasuk mengubah ukuran gambar agar konsisten
Ubah Ukuran, Koreksi Kecerahan, Pemfilteran, Koreksi Pencahayaan, Koreksi Fokus,
Pengurangan Kebisingan, Ambang Batas, Transformasi Geometris dan Transformasi Skala Abu-
Abu. Proses ini menyiapkan gambar beras yang sakit Daun untuk analisis lebih lanjut.
Dibawah ini adalah gambar mentah dari berbagai kategori seperti Hispa, Brownspot, Leafblast,
Healthy leaf. Ukuran dimensi gambar yang digunakan adalah 1051 x 1051 dan format gambar
JPEG. Dikurangi Dimensi saat gambar dikonversi dari ruang warna BGR ke gambar skala abu-
abu. Skala abu-abu adalah Berbagai nuansa monokromatik dari hitam hingga putih. Banyak
program pengeditan gambar memungkinkan untuk mengonversi warna untuk mengubah gambar
menjadi hitam, putih atau skala abu-abu. Proses ini menghapus semua informasi warna, hanya
menyisakan pencahayaan dalam setiap piksel. Nilai piksel luminance untuk gambar skala abu-
abu berkisar dari 0 hingga 255. Gambar berwarna menjadi gambar skala abu-abu yang mengubah
nilai RGB (24-bit) menjadi nilai skala abu-abu (8-bit). skala abu-abu Karena ini adalah gambar
lapisan tunggal dari 0 hingga 255, ini adalah teknik pra pemrosesan yang paling disukai untuk
pemrosesan gambar. Metode RGB memiliki tiga lapisan citra yang berbeda. Warna RGB
meningkatkan kompleksitas model, jadi berfungsi untuk mengubah gambar menjadi skala abu-
abu yang dapat membantu mengurangi kompleksitas model klasifikasi berbasis machine
learning.
dapat melihat bahwa ada dua puncak yang menonjol. jumlah piksel dengan nilai
intensitas sebesar 0 Sangatlah tinggi (30000). Sebagai perbandingan, daun diharapkan menutupi
sebagian kecil gambar dengan warna latar belakang sebagian besar hitam. Hal selanjutnya yang
perlu memisahkan keduanya. Nilai split terbaik adalah sekitar 20, tetapi jangan terlalu mengacu
pada hal itu. Untuk statistik deskriptif, digunakan pendekatan yang lebih formal yang dikenal
sebagai metode Otsu. Metode Otsu dapat mengasumsikan gambar berisi dua kelas piksel yang
mengikuti histogram bimodal (piksel latar depan dan latar belakang). kemudian hitung ambang
batas optimal yang memisahkan dua kelas sedemikian rupa sehingga distribusi gabungan (varian
antarkelas) diminimalkan, atau secara ekuivalen varians antar kelas dimaksimalkan. Metode Otsu
dipamerkan kinerja yang relatif baik jika histogram dapat diasumsikan memiliki distribusi
bimodal dan diasumsikan memiliki lembah yang dalam dan tajam di antara dua puncak.
Penyembunyian berikutnya telah dilakukan. Masking membantu dalam menemukan seluruh
daun dari gambar. Kami juga telah menerapkan K-means untuk segmentasi daun dari latar
belakang. Perbandingannya adalah dibuat antara label Otsu dan K-Means pada tingkat piksel,
menjumlahkan boolean dan membaginya dengan jumlah total piksel dalam gambar. Jika hasilnya
1 berarti tidak ada perbedaan sama sekali.
Ekstraksi Fitur
Ekstraksi fitur adalah semacam proses, di mana satu set awal variabel mentah diproses ke
grup yang lebih mudah dikelola. Pada citra tertentu dari penyakit daun padi, ekstraksi ciri
dimulai dari inisial kumpulan data dan membangun fitur, sehingga kumpulan data fitur yang
lebih informatif dapat disiapkan. Dalam metodologi penelitian yang diusulkan, digunakan fitur
HOG (Histogram of oriented gradients) metode ekstraksi. HOG adalah deskripsi fitur yang
digunakan untuk mendeteksi objek dalam visi komputer dan pemrosesan gambar. Teknik
deskriptor HOG menghitung kemunculan orientasi gradien di bagian gambar yang dilokalkan
jendela deteksi, atau region of interest (ROI). Ini membagi gambar menjadi daerah terhubung
kecil yang disebut sel, dan untuk setiap sel menghitung histogram arah gradien. Ini kemudian
mendiskritkan setiap sel menjadi tempat sampah bersudut sesuai ke orientasi gradien. Setiap
piksel sel memberikan kontribusi gradien tertimbang ke bin sudut yang sesuai pengelompokan
sel menjadi blok adalah dasar untuk pengelompokan dan normalisasi histogram. histogram
mewakili blok histogram. Himpunan histogram blok ini mewakili deskriptor. HOG adalah
deskriptor fitur terbukti digunakan untuk mendeteksi objek dalam visi komputer dan pemrosesan
gambar. Oleh karena itu, dapat menggunakan HOG untuk ekstraksi fitur.
S. Ramesh *, D. Vydeki
Electronics and Communication Engineering Department, VIT University, Chennai 600127, India
A R T I C L E I N F O A B S T R A C T
Article history: In the agriculture field, one of the recent research topics is recognition and classification of
Received 25 March 2019 diseases from the leaf images of a plant. The recognition of agricultural plant diseases by
Received in revised form utilizing the image processing techniques will minimize the reliance on the farmers to pro-
28 August 2019 tect the agricultural products. In this paper, Recognition and Classification of Paddy Leaf
Accepted 6 September 2019 Diseases using Optimized Deep Neural Network with Jaya Algorithm is proposed. For the
Available online 6 September 2019 image acquisition the images of rice plant leaves are directly captured from the farm field
for normal, bacterial blight, brown spot, sheath rot and blast diseases. In pre-processing,
Keywords: for the background removal the RGB images are converted into HSV images and based
Paddy leaf diseases on the hue and saturation parts binary images are extracted to split the diseased and
Optimized Deep Neural Network non-diseased part. For the segmentation of diseased portion, normal portion and back-
Jaya optimization algorithm ground a clustering method is used. Classification of diseases is carried out by using Opti-
K-means clustering mized Deep Neural Network with Jaya Optimization Algorithm (DNN_JOA). In order to
Color features precise the stability of this approach a feedback loop is generated in the post processing
Texture features step. The experimental results are evaluated and compared with ANN, DAE and DNN.
The proposed method achieved high accuracy of 98.9% for the blast affected, 95.78% for
the bacterial blight, 92% for the sheath rot, 94% for the brown spot and 90.57% for the nor-
mal leaf image.
Ó 2019 China Agricultural University. Production and hosting by Elsevier B.V. on behalf of
KeAi. This is an open access article under the CC BY-NC-ND license (http://creativecommons.
org/licenses/by-nc-nd/4.0/).
tion and health monitoring on plants is very harmful. The Lu at al. [16]. They used a dataset which contains 500 images
related studies of the recognition of plant diseases mean that include diseased and non-diseased paddy stems and leaves.
the diseases are the visible patterns observed on the plants Classification was carried out with ten common rice diseases.
[6]. Manually, the plant diseases are more difficult for the They showed that their approach attained higher accuracy
monitoring process. For the manual process it needs more than the conventional machine learning method. The exper-
processing time, large amount of work and expertize in the imental outcomes represented the effectiveness and feasibil-
diseases of plant. So that for the plant disease recognition, ity of their proposed model.
image processing techniques are utilized [7]. For the evaluation of ROI, a segmentation technique based
The image processing steps for the recognition of diseases on neutrosophic logic extended from the fuzzy set was intro-
comprise acquisition of images, pre-processing of images, duced by Gittaly Dhingra et al. [17]. They used three member-
segmenting the images, extracting the features and finally ship functions for the segmentation. To detect the plant leaf
classification [8]. These techniques can be performed only as diseased or not feature subsets were considered on the
on the external appearances of the infected plants [9]. Com- basis of segmented regions. Various classifiers were employed
monly, in most of the plants to detect the plant diseases, for the demonstration and the random forest method over-
leaves are the important source. In rice plants, sheath rot, leaf come the other approaches. They used a dataset with 400 leaf
blast, leaf smut, brown spot and bacterial blight are the most images which included 200 diseased leaf image and 200 non-
common diseases [10]. However for different plants, the diseased leaf images.
symptoms of the plant diseases are varied. The plant diseases Using image processing techniques, D. Nidhis et al. [18]
are different in color, size and shape and every disease has developed a method for detecting the disease type affected
individual features. Some diseases have yellow color and by the paddy leaves. By evaluating the percentage of diseased
some have brown color [11]. Some diseases are same in their area, the severity of the disease infection was calculated.
shapes but differ in colors whereas, some are same in color Based on the severity of diseases the pesticides were utilized
but differ in their shapes. After the segmentation of diseased for the bacterial blight, brown spot and rice blast which are
and normal portion the features related to disease can be the main diseases affect the paddy crop and their
extracted [12]. productivity.
Normally, the manual detection of plant diseases is naked For the identification and the classification of rice plant
eye observation of experts which consumes more time, diseases a new method was presented by Taohidul Islam
expensive on large farms [13]. It is difficult to process and also et al. [19]. In their work, on the basis of percentage of RGB
sometimes it produces an error when identifying the disease value of the diseased part, they detected and identified the
type [14]. Because of the unawareness of suitable manage- diseases by employing image processing techniques. They
ment to rectify rice plant leaf diseases, the rice production utilized Naı̈ve Bayes classifier which is a simple classifier to
is being reduced in recent years [15]. To overcome this a suit- classify the disease into various classes. Their approach suc-
able and fast recognition system on rice leaf disease is cessfully recognized and classified three main types of rice
needed. Hence this paper proposes a novel method for recog- plant diseases by using only one feature. Thus it was a faster
nizing the diseases of rice plants using their images. This method which required less time for computation.
research mainly focuses on four most common rice plant dis- Using the image processing techniques to automatically
eases named as Brown spot, Leaf blast, Bacterial blight and recognize the diseases in paddy leaves Gayathri Devi and Nee-
Sheath rot. Our contribution in this work includes as follows: lamegam [20] developed an approach. They used hybridized
From the farm field in real world circumstances we captured gray scale co-occurrence matrix, DWT and SIFT for the extrac-
the rice plant leaves images and prepared the dataset. Back- tion of features. After extracting the features they were given
ground elimination is carried out by the pre-processing. Seg- to various classifiers include multiclass SVM, Naı̈ve Bayesian,
mentation is performed for clustering the diseased portion back propagation neural network and KNN for the classifica-
and the normal portion of the leaves. For the classification, tion of diseased and normal plants.
optimized Deep Neural Network with Jaya algorithm is pro- Aydin Kaya et al. [21] investigated the results of the effect
posed. JOA is used for the best weight selection of DNN. of four different transfer learning models for deep neural
Rest of the paper is organized as follows: Section 2 con- network-based plant classification on four public datasets.
tains the recent works related to rice plant disease classifica- Their experimental study demonstrated that transfer learn-
tion. Section 3 defines the problem statement. Section 4 ing can provide important benefits for automated plant iden-
comprises our proposed methodology. Section 5 shows tification and can improve low-performance plant
the experimental analysis. Section 6 represents the classification models.
conclusion. Alessandro dos Santos Ferreira et al. [22] used the Convo-
lutional Neural Networks to perform weed detection in soy-
2. Related works bean crop images and classified the weeds among grass and
broadleaf. An image database was created containing over fif-
Some of the recent researches related to the recognition and teen thousand images of the soil, soybean, broadleaf and
classification of rice plant diseases are given as below. grass weeds. The Convolutional Neural Networks used in this
On the basis of deep convolutional neural network, a novel work represented a Deep Learning architecture that has
rice plant disease detection approach was developed by Yang achieved remarkable success in image recognition.
Information Processing in Agriculture 7 ( 2 0 2 0 ) 2 4 9 –2 6 0 251
For detecting the leaf diseases, the conventional methods are Acquisition of images is the process of collecting the images
human vision based approaches. In these cases seeking the which are used for this research. From the farm field in real
expert advice is time consuming and very expensive. The world circumstances we captured the rice plant leaves images
human vision based methods suffer many drawbacks. The by using high resolution digital camera. Then for the recogni-
accuracy and precision of human vision approach is depen- tion of diseases all the captured images are moved to the
dent on the eyesight of the person or expert hired. Machine computer where the implementation process will be carried
learning based method enables to identify the types of dis- out. The dataset contains the images having the leaves with
eases, make the right decision and to select proper treatment. various degree of disease spread. The data set are collected
One of the advantages of using machine learning based from rural area of ayikudi and panpoli, Tirunelveli District,
method is that it performs tasks more consistently than Tamilnadu. The images are captured and a dataset is pre-
human experts. Therefore, to overcome the drawbacks of pared totally with 650 images which include 95 normal
conventional methods there is a need for a new machine images, 125 bacterial blight images, 170 blast images, 110
learning based classification approach. Very few recent devel- sheath rot images and 150 brown spot images. Some of the
opments were recorded in the field of plant leaf disease detec- sample images are given in (See Fig. 2).
tion using machine learning approach and that too for the
paddy leaf disease detection and classification is the rarest. 4.2. Pre-processing
To select the centroid of each cluster, the highest value are selected. The value of black color and the selected cen-
from the hue values of the normal and the diseased portion troid values are fed in the clustering process. After clustering
Information Processing in Agriculture 7 ( 2 0 2 0 ) 2 4 9 –2 6 0 253
X
n
Cty ¼ Pyx ðy xÞ2 ð4Þ
x¼0
X
n
ðy MÞðx MÞ
Cny ¼ Pyx ð5Þ
Fig 4 – (a) Histogram of hue part (b) Histogram of background x¼1
Sy
removed blast affected portion (c) Histogram of background
removed normal portion. X
n
2
Ey ¼ Pyx ð6Þ
x¼0
hidden layers. The proposed architecture of DNN is shown in updates the weight of the nodes in the hidden layers. Due
Fig. 6. to the increase in the training iterations, this neural network
By considering the effort of preference weight fitness, the continually fits the labelled training data’s decision boundary.
DNN is designed with two hidden layers for perfectly learning To enhance the training speed of the DNN and the classifica-
the mapping relation between the input and output data. In tion accuracy, two hidden layers are constructed. In the hid-
the training phase, by using the JOA the DNN iteratively den layer the total number of nodes is evaluated by using
Eqn. (7).
pffiffiffiffiffiffiffiffiffiffiffiffi
n¼ aþbþc ð7Þ
data sample ðx; lÞ for a hidden layer, the loss form can be com- convergence speed. The implementation of this algorithm
puted as can be briefly summarized as follows:
Initialize the population size, the number of iterations and
1 Xm
SðWs ; bs ; x; lÞ ¼ jjhj ðWs ; bs ; xÞ lj jj22 ð10Þ the termination criteria. From the population, the best and
2m j¼1
worst solutions should be determined with respect to the
where Ws and bs are the subsets of biases and m is the num- objective function. The current solution based on the best
ber of neurons in the hidden layer. and worst solutions should be adjusted by using Eq. (12).
Cross entropy is used as the loss function of DNN as the x0j;i;G ¼ xj;i;Gr1;j;G ðxj;best;G jxj;i;G jÞr2;j;G ðxj;worst;G jxj;i;G jÞ ð12Þ
preparation for training and testing. The use of cross-
entropy losses greatly improved the performance of the sig- where xj;best;G and xj;worst;G are the values of the jth variable
moid and softmax output models. The cross entropy loss is for the best candidate and the worst candidate, respectively;
evaluated by the Eq. (11). r1;j;G and r2;j;G are the random numbers in the range [0, 1].
n h i The adjusted solution is compared with the previous solution.
1X ^ k þ ð1 Yk Þlog 1 Y
^k
CE ¼ Y k log Y ð11Þ If the previous one is improved it will replace the previous
n k¼1
solution else it will retain the previous solution [23]. The pro-
where, n represents the training sample quantity, Y k indicates cess is repeated from until the termination criteria are
the kth actual output of training set, Y^ k is the kth expected achieved. The flow chart for the best weight selection process
output of testing set. We are using JOA algorithm for the opti- is shown in Fig. 7.
mal weight selection of DNN network.
In the population improving the fitness value of every 4.6. Post-Processing
solution is the main intention of JOA. By updating the values,
this algorithm efforts the fitness value to shift towards the After performing the classification process, in order to precise
best solution. After that, the new solutions and the old solu- the stability of the proposed method a feedback loop is gener-
tions are compared and for the next iteration only the best ated in which the feedback is sent to the segmentation phase.
solutions are considered. Moreover, it only requires tuning In the segmentation phase a post-processing method, mor-
of the population size and number of iterations which leads phological opening is applied on the diseased segment to
to the implementation of the algorithm is simple as its solu- remove the green portion if there any exists. So that the per-
tion is updated through only one phase using a single equa- formance of classification will be improved.
tion. It0 s dominant over other optimization techniques with
respect to low computational complexity and time and faster 5. Experimental results
Value (NPV), precision, True Positive Rate (TPR), True Negative posed method DNN_JOA the accuracy of normal image is
Rate (TNR) and loss function. 90.57%, bacterial blight is 95.78%, blast is 98.9%, brown spot
Fig. 9 represents the comparison graph of accuracy for the is 94% and sheath rot is 92%. When using ANN classifier the
five classes with respect to the four classifiers which includes accuracy of normal image is 77.24%, bacterial blight is 78%,
the proposed and existing classifiers. When using our pro- blast is 85%, brown spot is 81.5% and sheath rot is 78.3%.
When using DAE classifier the accuracy of normal image is
81.4%, bacterial blight is 86.2%, blast is 91.5%, brown spot is
87.7% and sheath rot is 83.4%. When using DNN classifier
the accuracy of normal image is 83%, bacterial blight is
91.7%, blast is 96.2%, brown spot is 90.6% and sheath rot is
88.5%.
Fig. 10 represents the comparison graph of F1-score for the
five classes with respect to the four classifiers which includes
the proposed and existing classifiers. When using our pro-
posed method DNN_JOA the F1-score of normal image is
81.25%, bacterial blight is 88.75%, blast is 96.86%, brown spot
is 85% and sheath rot is 91.86%. When using ANN classifier
the F1-score of normal image is 61.45%, bacterial blight is
68.67%, blast is 74.89%, brown spot is 65% and sheath rot is
71.56%. When using DAE classifier the F1-score of normal
image is 69.23%, bacterial blight is 74.68%, blast is 84.79%,
brown spot is 76.45% and sheath rot is 79.98%. When using
DNN classifier the F1-score of normal image is 72.68%, bacte-
rial blight is 81%, blast is 88.58%, brown spot is 80.6% and
sheath rot is 84.67%.
Fig. 11 represents the comparison graph of FPR for the five
classes with respect to the four classifiers which includes the
Fig 8 – Confusion Matrix of DNN-JOA. proposed and existing classifiers. When using our proposed
method DNN_JOA the value of FPR of normal image is 6.2, bac- proposed and existing classifiers. When using our proposed
terial blight is 4.7, blast is 1.8, brown spot is 4.4 and sheath rot method DNN_JOA the value of FDR of normal image is 17.1,
is 5. When using ANN classifier the FPR of normal image is bacterial blight is 19.5, blast is 10, brown spot is 26.6 and
14.78, bacterial blight is 12.7, blast is 8.45, brown spot is sheath rot is 25. When using ANN classifier the FDR of normal
13.56 and sheath rot is 11.78. When using DAE classifier the image is 32.82, bacterial blight is 33.6, blast is 30.1, brown spot
value of FPR of normal image is 12.3, bacterial blight is 9.3, is 36.5 and sheath rot is 33. When using DAE classifier the
blast is 5.4, brown spot is 10.3 and sheath rot is 9.9. When value of FDR of normal image is 27.9, bacterial blight is 29.1,
using DNN classifier the FPR of normal image is 10.6, bacterial blast is 25, brown spot is 31.4 and sheath rot is 34.5. When
blight is 6.6, blast is 4.2, brown spot is 8 and sheath rot is 8.9. using DNN classifier the FDR of normal image is 22.67, bacte-
Fig. 12 represents the comparison graph of FNR for the five rial blight is 22.6, blast is 21.2, brown spot is 29.4 and sheath
classes with respect to the four classifiers which includes the rot is 27.3.
proposed and existing classifiers. When using our proposed Fig. 14 represents the comparison graph of NPV for the five
method DNN_JOA the value of FNR of normal image is 10, bac- classes with respect to the four classifiers which includes the
terial blight is 14, blast is 5.78, brown spot is 8 and sheath rot is proposed and existing classifiers. When using our proposed
7. When using ANN classifier the FNR of normal image is 22.6, method DNN_JOA the value of NPV of normal image is 91.2,
bacterial blight is 23, blast is 14, brown spot is 22.5 and sheath bacterial blight is 95, blast is 97.4, brown spot is 93.6 and
rot is 20.1. When using DAE classifier the value of FNR of nor- sheath rot is 96.1. When using ANN classifier the NPV of nor-
mal image is 19, bacterial blight is 21, blast is 10.4, brown spot mal image is 75.6, bacterial blight is 74.89, blast is 82.56,
is 17.8 and sheath rot is 18.3. When using DNN classifier the brown spot is 71.54 and sheath rot is 77.9. When using DAE
FNR of normal image is 17.8, bacterial blight is 17, blast is classifier the value of NPV of normal image is 79.2, bacterial
9.18, brown spot is 14.5 and sheath rot is 14.6. blight is 87.45, blast is 90.56, brown spot is 83.78 and sheath
Fig. 13 represents the comparison graph of FDR for the five rot is 81.54. When using DNN classifier the NPV of normal
classes with respect to the four classifiers which includes the image is 88.5, bacterial blight is 91.45, blast is 94.5, brown spot
is 89.78 and sheath rot is 87.
R E F E R E N C E S
[15] Pinki F, Tazmim N, Khatun SMM Islam. Content based paddy area in agro-field. In: In: Proc. Second International
leaf disease recognition and remedy prediction using support Conference on Inventive Communication and Computational
vector machine. In: In: Proc. In Computer and Information Technologies. p. 62–6.
Technology (ICCIT). 20th International Conference; 2017. p. [20] Devi T, Gayathri PN. Image processing based rice plant leaves
1–5. diseases in Thanjavur Tamilnadu. Cluster Comput 2018:1–14.
[16] Lu Y, Yi S, Zeng N, Liu Y, Zhang Y. Identification of rice [21] Kaya A, Keceli AS, Catal C, Yalic HY, Temucin H,
diseases using deep convolutional neural networks. Tekinerdogan B. Analysis of transfer learning for deep neural
Neurocomputing 2017;267:378–84. network based plant classification models. Comput Electron
[17] Dhingra G, Kumar V, Joshi HD. A novel computer vision based Agric 2019;158:20–9.
neutrosophic approach for leaf disease identification and [22] dos Santos FA, Freitas DM, da Silva GG, Pistori H, Folhes MT.
classification. Measurement 2019;135:782–94. Weed detection in soybean crops using ConvNets. Comput
[18] Nidhis AD, Pardhu CNV, Reddy KC, Deepa K. Cluster based Electron Agricul 2017;143:314–24.
paddy leaf disease detection, classification and diagnosis in [23] Rao RV. Jaya: A simple and new optimization algorithm for
crop health monitoring unit. Comput Aided Interven solving constrained and unconstrained optimization
Diagnost Clinic Med Images 2019;31:281–91. problems. International Journal of Industrial Engineering
[19] Islam T, Sah M, Baral S, Choudhury RR. A faster technique on Computations 2016;7:19–34.
rice disease detection using image processing of affected
Recognition and classification of paddy leaf diseases using Optimized Deep Neural
network with Jaya algorithm
Latar Belakang penelitian
Banyak negara yang mana pertanian adalah sumber daya utama mereka. namun , tidak sedikit juga dari
negara negara tersebut mengalami gagal panen yang disebabkan oleh bencana alam, kekurangna air,
penyakit tanaman, dll. Dalam riset pertanian salah satu hal yang banyak diperbincangkan adalah
bagaimana mengenali penyakit tanaman. Penyakit tanaman sendiri selalu selalu mempunyai pola tertentu
namun, tidak semua orang bisa mengetahui bagaimana pola penyakit tanaman tersebut hanya dengan
mata telanjang. Selain itu pengenalan pola penyakit dengan mata telanjang merupakan tugas yang cukup
sulit dan diperlukan ahli dalam pengenalan ini, karena itulah pengenalan penyakit dapat memakan waktu
yang lama dan juga biaya yang besar.Untuk mengidentifikasi penyakit tanaman,disarankan menggunakan
machine learning dan pengenalan gambar karena machine learning ini mempunyai beberapa fitur yaitu,
mengidentifikasi penyakit, membuat keputusan yang benar, dan bisa memilih cara pengobatan yang benar
untuk penyakit tersebut.Salah satu keuntungan ketika memakai machine learning adalah machine learning
bisa melakukan tugas secara konsisten dari pada metode konvensional.
Sistem yang ada menggunakan lima fase yang meliputi akuisisi citra, pra-pemrosesan, segmentasi citra,
fitur ekstraksi dan klasifikasi. Gambar padi yang didapatkan dari ladang pertanian dan kumpulan datanya
dibuat.Pada fase pra pemrosesan dimensi gambarnya diperkecil latar belakang pada gambar juga
dihilangkan. Kemudian langkah yang dilalui adalah segmentasi citra di mana k-means clustering metode
ini diterapkan untuk membagi bagian normal dan bagian yang sakit. Kemudian klasifikasi penyakit
dilakukan dengan metode DNN_JOA. hasil tidak sesuai maka dikembalikan ke fase segmentasi untuk
presisi stabilitas metode yang dikembangkan.
Akuisisi gambar adalah proses mengumpulkan gambar-gambar yang digunakan untuk penelitian. Dari
ladang pertanian secara nyata keadaan dunia kami menangkap gambar daun tanaman padi.
Fase Pra-pemrosesan bertujuan meminimalisir kebutuhan memori dan daya gambar dalam dataset diubah
ukurannya dan dipotong menjadi dimensi 300 450 piksel.Dalam fase ini yang paling penting adalah
menghilangkan gambar background dengan menerapkan fusi berbasis nilai rona. Mengubah RGB model
diubah menjadi HSV. Dari model HSV terlebih dahulu S nilai dipertimbangkan untuk proses karena
melebihi putihnya.Proses fusi menghilangkan latar belakang dengan menetapkan nilai piksel sebagai 00 s.
Nilai piksel 0 menunjukkan warna hitam dalam model RGB. Hanya bagian daunnya dengan bagian yang
sakit hadir di latar belakang ini dihapus gambar
K-means clustering berdasarkan segmentasi Untuk segmentasi citra metode K-means clustering adalah
dipekerjakan dalam pekerjaan ini. Clustering adalah proses untuk mengelompokkan gambar menjadi
cluster. Bagian yang sakit diambil dari gambar daun dengan pengelompokan ini. Dalam gambar daun saat
menerapkan pengelompokan ini, cluster diharapkan dapat membagi bagian. Teknik ini diterapkan pada
bagian rona model HSV dari latar belakang dihapus gambar.
Fitur ekstrasi
Fitur yang diekstrak dalam fitur karya ini adalah tekstur dan fitur warna. Fitur warna termasuk
mengekstraksi nilai rata-rata dan nilai standar deviasi.fitur tekstur mencakup fitur GLCM seperti
homogenitas, kontras, korelasi, dan energi.
Kerangka kerja DNN memiliki tiga komponen utama : lapisan input, lapisan output dan lapisan
tersembunyi. DNN dirancang dengan dua lapisan tersembunyi untuk pembelajaran yang sempurna
hubungan pemetaan antara data input dan output. Di fase pelatihan, dengan menggunakan JOA DNN
secara iteratif. memperbarui bobot node di lapisan tersembunyi. Jatuh tempo untuk peningkatan iterasi
pelatihan, jaringan saraf ini terus sesuai dengan batas keputusan data pelatihan berlabel.
Pasca Pengolahan
Setelah dikelolah untuk memastikan stabilitas metode yang diusulkan, loop umpan balik dihasilkan di
mana umpan balik dikirim ke fase segmentasi. Pada fase segmentasi metode pasca-pemrosesan,
pembukaan morfologi diterapkan pada segmen yang sakit untuk menghilangkan bagian hijau jika ada.
Sehingga kinerja klasifikasi akan meningkat
metodologi yang diterapkan usulkan menggunakan DNN_JOA dalam platform Python versi 3.6. Kinerja
metode DNN-JOA diperkirakan dan dibandingkan dengan kinerja pengklasifikasi yang ada seperti ANN,
DAE dan DNN. Hasilnya dibandingkan berdasarkan penyakitnya kelas yang meliputi normal, hawar
bakteri, bercak coklat, penyakit busuk pelepah dan blas. Dari kumpulan data 70% gambar digunakan
untuk pelatihan, 20% digunakan untuk pengujian dan sisanya 10% digunakan untuk validasi.
Hasil percobaan telah diterapkan dengan Metode DNN_JOA. Menggunakan pengklasifikasi DNN_JOA
yang tertinggi akurasi dicapai untuk gambar daun yang terkena ledakan yang adalah 98,9%. Menunjukkan
matriks kebingungan yang diperoleh untuk metode yang kami usulkan. Dari matriks kebingungan ini,
True Positive (TP), Negatif Benar (TN), Positif Palsu (FP) dan Negatif Palsu (FN) nilai diprediksi.
Nilai TP, TN, FP dan FN dari matriks kebingungan di atas adalah 19, 109, 1 dan 1 masing-masing untuk
gambar normal; nilai TP, TN, FP dan FN adalah 22, 103, 2 dan 3 masing-masing untuk busuk pelepah;
nilai dari TP, TN, FP dan FN masing-masing adalah 25, 102, 1 dan 2 untuk bintik coklat; nilai TP, TN,
FP dan FN adalah 26, 99, 3 dan 2 masing-masing untuk penyakit hawar bakteri; nilai TP, TN, FP dan FN
masing-masing adalah 29, 98, 2 dan 1 untuk ledakan yang terkena gambar.
Perbandingan akurasi untuk lima kelas sehubungan dengan empat pengklasifikasi yang meliputi:
pengklasifikasi yang diusulkan dan yang ada. Saat menggunakan metode yang kami usulkan DNN_JOA
akurasi gambar normal adalah 90,57%, hawar bakteri 95,78%, blas 98,9%, bercak coklat adalah 94% dan
busuk pelepah adalah 92%. Saat menggunakan pengklasifikasi ANN, akurasi citra normal 77,24%, hawar
bakteri 78%, blas 85%, bercak coklat 81,5% dan busuk pelepah 78,3%. Saat menggunakan
pengklasifikasi DAE, keakuratan gambar normal adalah 81,4%, hawar bakteri 86,2%, blas 91,5%, bercak
coklat 87,7% dan busuk pelepah sebesar 83,4%. Saat menggunakan pengklasifikasi DNN akurasi gambar
normal adalah 83%, hawar bakteri adalah 91,7%, blas 96,2%, bercak coklat 90,6% dan busuk pelepah
88,5%.
Perbandingan FNR untuk lima kelas terhadap empat pengklasifikasi yang mencakup pengklasifikasi yang
diusulkan dan yang ada. Saat menggunakan metode yang kami usulkan DNN_JOA nilai FNR citra
normal adalah 10, hawar bakteri 14, blas 5,78, bercak coklat 8 dan busuk pelepah adalah 7. Saat
menggunakan pengklasifikasi JST, FNR citra normal adalah 22,6, hawar bakteri adalah 23, blas 14,
bercak coklat 22,5 dan busuk pelepah 20,1. Saat menggunakan pengklasifikasi DAE nilai FNR citra
normal adalah 19, hawar bakteri 21, blas 10,4, bercak coklat 17,8 dan busuk pelepah 18,3. Saat
menggunakan pengklasifikasi DNN, FNR citra normal adalah 17,8, hawar bakteri adalah 17, blas adalah
9,18, bercak coklat adalah 14,5 dan busuk pelepah adalah 14,6.
menunjukkan grafik perbandingan akurasi pelatihan dan pengujian. Untuk proses pelatihan jumlah sampel
yang dipertimbangkan adalah 450 dan untuk proses pengujian 125 sampel dipertimbangkan. Baik akurasi
pelatihan dan pengujian meningkat dengan meningkatnya jumlah sampel yang dipertimbangkan. Pada
perbandingan kedua akurasi pelatihan sangat dicapai sebagai 99% dengan 450 sampel dan akurasi
pengujian sangat mencapai 97% dengan 125 sampel dengan menggunakan metode DNN_JOA.
Link Referensi
Recognition and classification of paddy leaf diseases using Optimized Deep Neural network with
Jaya algorithm
https://www.sciencedirect.com/science/article/pii/S2214317319300769
TUGAS KELOMPOK SESI 2
VISI KOMPUTER
DISUSUN OLEH:
SURABAYA
2022
JURNAL 1
Metode Segmentasi yang digunakan pada jurnal pertama yang telah dirangkum adalah
metode segmentasi thresholding.
Pengertian
Thresholding adalah metode segmentasi citra yang memisahkan objek pada citra dari
latar belakang berdasarkan perbedaan terang dan gelap. Area gambar yang cenderung gelap akan
digelapkan penuh (nilai intensitas 0), dan area gambar cenderung terang akan diterangkan penuh
(nilai intensitas 1). Sehingga, keluaran dari proses segmentasi threshold adalah citra biner dengan
nilai intensitas 0 atau 1. Setelah citra tersegmentasi atau objek berhasil dipisahkan dari latar
belakang, citra biner yang dihasilkan dapat digunakan sebagai mask pada gambar aslinya.
Tahapan
3. Penyaringan terhadap fitur yang tidak diinginkan menggunakan teknik pengurangan fitur.
Pra-pemrosesan
Berikut adalah langkah-langkah Pra-pemrosesan sebelum melatih model dengan dataset:
Diterapkan pada data untuk mengoptimalkan gambar untuk penggunaan lebih lanjut.
Pra-pemrosesan mengurangi data gambar distorsi yang tidak diinginkan, meningkatkan fitur
gambar untuk diproses lebih lanjut. Ini termasuk mengubah ukuran gambar agar konsisten Ubah
Ukuran, Koreksi Kecerahan, Pemfilteran, Koreksi Pencahayaan, Koreksi Fokus, Pengurangan
Kebisingan, Ambang Batas, Transformasi Geometris dan Transformasi Skala Abu-Abu. Proses
ini menyiapkan gambar beras yang sakit Daun untuk analisis lebih lanjut.
Ekstraksi Fitur
Pada citra tertentu dari penyakit daun padi, ekstraksi ciri dimulai dari inisial kumpulan data dan
membangun fitur, sehingga kumpulan data fitur yang lebih informatif dapat disiapkan. Dalam
metodologi penelitian yang diusulkan, digunakan fitur HOG (Histogram of oriented gradients)
metode ekstraksi. HOG adalah deskripsi fitur yang digunakan untuk mendeteksi objek dalam visi
komputer dan pemrosesan gambar. Teknik deskriptor HOG menghitung kemunculan orientasi
gradien di bagian gambar yang dilokalkan jendela deteksi, atau region of interest (ROI). Ini
membagi gambar menjadi daerah terhubung kecil yang disebut sel, dan untuk setiap sel
menghitung histogram arah gradien. Ini kemudian mendiskritkan setiap sel menjadi tempat
sampah bersudut sesuai ke orientasi gradien. Setiap piksel sel memberikan kontribusi gradien
tertimbang ke bin sudut yang sesuai pengelompokan sel menjadi blok adalah dasar untuk
pengelompokan dan normalisasi histogram. histogram mewakili blok histogram. Himpunan
histogram blok ini mewakili deskriptor. HOG adalah deskriptor fitur terbukti digunakan untuk
mendeteksi objek dalam visi komputer dan pemrosesan gambar. Oleh karena itu, dapat
menggunakan HOG untuk ekstraksi fitur.
Ekstraksi Reduksi
Teknik pengurangan fitur sebagai PCA (Analisis komponen utama) telah diterapkan untuk
menghapus fitur redundan yang tidak memengaruhi pengambilan keputusan pengklasifikasi.
yang dikurangi set fitur telah disiapkan dan digunakan lebih lanjut oleh pengklasifikasi untuk
mengidentifikasi penyakit daun padi.
a. Adaboost Algorithms
Boosting adalah metode ensemble umum yang menciptakan classifier kuat dari sejumlah
classifier lemah. Ini dilakukan dengan membangun model dari data pelatihan, kemudian
membuat model kedua yang mencoba untuk memperbaiki kesalahan dari model pertama. Model
ditambahkan sampai set pelatihan diprediksi sempurna atau jumlah maksimum dari model
ditambahkan. AdaBoost adalah semacam peningkatan gradien dengan fungsionalitas validasi
silang bawaan. Dia memungkinkan pengguna untuk menjalankan validasi silang pada setiap
iterasi dari proses peningkatan dan dengan demikian membuat proses mendapatkan jumlah
optimal yang tepat untuk meningkatkan iterasi dalam sekali percobaan.
b. Bagging Classifier
Bagging Classifier adalah salah satu algoritma pembelajaran mesin yang paling kuat dan populer.
Ini adalah ansambel algoritma yang digunakan untuk klasifikasi. Bagging adalah singkatan dari
bootstrap aggregation dan merupakan statistik yang kuat, metode ini berfungsi untuk
memperkirakan kuantitas dari dataset yang diberikan. Bagging mencoba untuk mempekerjakan
pelajar yang sama di small sampel dataset dan kemudian mengambil rata-rata dari semua hasil.
Bagging sendiri menggunakan bootstrap sampling untuk mendapatkan subset data untuk melatih
pelajar dasar. Metode ensemble dalam pengklasifikasi Bagging menggabungkan hasil dari
beberapa algoritme pembelajaran mesin sekaligus untuk membuat prediksi yang lebih baik dan
akurat daripada algoritma yang lain. Untuk agregasi, output dari base learner, bagging
menggunakan voting untuk klasifikasi
Diklasifikasikan Berdasarkan komponen yang menjadi acuan untuk pemisahan objek. Acuan
umum yang sering digunakan adalah warna objek. Warna dapat dinilai sebagai komponen utama
karena memiliki dua faktor.
Pertama, warna dapat menyederhanakan proses identifikasi objek dengan menguatkan perbedaan
deskripsi objek.
Kedua, visualisasi bentuk warna dan intensitasnya dapat dilihat oleh manusia secara jelas.
Segmentasi HSV merupakan proses pemisahan objek dengan seleksi warna berdasarkan nilai
hue,saturation dan value.
● Hue : merupakan atribut yang mempresentasikan warna murni
● Saturation : merupakan atribut yang menunjukkan efek cahaya putih yang mempengaruhi
tingkat Dominasi warna
● Value : merupakan atribut yang menunjukkan perbedaan kecerahan pada warna murni
Pra-pemprosesan
Gambar asli berbasis RGB dikonversi menjadi basis HSV dengan proses mengubah
Saturasi pada gambar dan menjadi binary image atau gambar hitam putih, setelahnya
digabungkan dengan Gambar asli Berbasis RGB awal maka background akan menghilang.
Dalam perhitungan fitur, piksel hijau memberikan kontribusi yang merugikan. Jadi
mungkin ada peluang untuk mempengaruhi akurasi klasifikasi. Pada model hue, warna hijau
berada di antara 17,2 derajat dan 45 derajat yang dipetakan sebagai nilai minimum 0,048 dan
nilai maksimum 0,125. Berdasarkan nilai maksimum dan minimum ini, topeng biner dibuat
untuk menghilangkan warna hijau dari bagian yang sakit. Gambar 5 menunjukkan citra clustered
dari citra bagian hue. Keluaran dari clustering adalah dua cluster yaitu bagian normal dan bagian
yang sakit.
Sistem yang ada menggunakan lima fase yang meliputi akuisisi citra, pra-pemprosesan,
segmentasi citra, fitur ekstraksi dan klasifikasi. Gambar padi yang didapatkan dari ladang
pertanian dan kumpulan datanya dibuat.Pada fase pra pemrosesan dimensi gambarnya diperkecil
latar belakang pada gambar juga dihilangkan. Kemudian langkah yang dilalui adalah segmentasi
citra di mana k-means clustering metode ini diterapkan untuk membagi bagian normal dan
bagian yang sakit. Kemudian klasifikasi penyakit dilakukan dengan metode DNN JOA. hasil
tidak sesuai maka dikembalikan ke fase segmentasi untuk presisi stabilitas metode yang
dikembangkan.
Fitur ekstraksi
Fitur yang diekstrak dalam fitur karya ini adalah tekstur dan fitur warna. Fitur warna
termasuk mengekstraksi nilai rata-rata dan nilai standar deviasi.fitur tekstur mencakup fitur
GLCM seperti homogenitas, kontras, korelasi, dan energi.
Kerangka kerja DNN memiliki tiga komponen utama : lapisan input, lapisan output dan
lapisan tersembunyi.DNN dirancang dengan dua lapisan tersembunyi untuk pembelajaran yang
sempurna hubungan pemetaan antara data input dan output.Fase pelatihan, dengan menggunakan
JOA DNN secara iteratif. memperbarui bobot node di lapisan tersembunyi. Jatuh tempo untuk
peningkatan iterasi pelatihan, jaringan saraf ini terus sesuai dengan batas keputusan data
pelatihan berlabel. Untuk meningkatkan kecepatan pelatihan DNN dan akurasi klasifikasi, dua
lapisan tersembunyi dibangun.pada lapisan tersembunyi, jumlah total node dievaluasi dengan
menggunakan persamaan
Penggunaan algoritma JOA untuk pemilihan bobot jaringan DNN yang optimal. Dalam
populasi meningkatkan nilai fitness dari setiap solusi adalah tujuan utama dari JOA.Algoritma
ini berupaya agar nilai fitness bergeser ke arah solusi terbaik sehubungan dengan kompleksitas
komputasi yang rendah dan waktu dan cepat.
Penamaan ukuran populasi, jumlah iterasi dan kriteria terminasi. Dari populasi, solusi
terbaik dan terburuk harus ditentukan dengan fungsi tujuan. Solusi saat ini berdasarkan solusi
terbaik dan terburuk harus disesuaikan dengan menggunakan Persamaan.
Pasca Pengolahan
Setelah dikelola untuk memastikan stabilitas metode yang diusulkan, loop umpan balik
dihasilkan di mana umpan balik dikirim ke fase segmentasi. Pada fase segmentasi metode
pasca-pemrosesan, pembukaan morfologi diterapkan pada segmen yang sakit untuk
menghilangkan bagian hijau jika ada. Sehingga kinerja klasifikasi akan meningkat
DAFTAR PUSTAKA
[2] Adi Pamungkas, “Segmentasi Citra dengan Metode Thresholding,” Pemrograman Matlab,
Apr. 08, 2017.
https://pemrogramanmatlab.com/2017/04/08/segmentasi-citra-dengan-metode-thresholding/
(accessed Sep. 12, 2022).