Rice is one of the most significant food sources of the world. More than fifty percent of the world’s population accept it as a staple food because of its nutrients, vitamins and minerals. But the rice plant is often infected by various diseases that contribute to economic, agricultural and communal deficiency in the development of the agricultural field.
Researchers are keen on the diagnosis of rice plant diseases using the image processing techniques approach since the last 10 years. Various methods have been applied and developed to detect, identify and quantify the rice plant diseases.
This is a paper that reviews some of the research papers that focus on the related subject. This paper aims to categorize and compare the related research papers based on the image processing techniques.
Rice (Oryza sativa) is not only one of the major food sources of the rural population of the world, it is also the source of income for many farmers. Especially in Asian countries, many people’s livelihood depends on the rice crop.
The quality and quantity of the crop is a huge factor for the economic growth of countries, such as India, where agriculture is the main source of income . But every year, around 37% farmers lose their crop to pests and diseases, and these losses can range between 24% and 41%.
There are some common diseases that affect the rice plant. By acquiring images of plants affected with these diseases, image processing techniques are used to identify them.
TYPES OF PADDY DISEASES
Paddy can be infected by diseases at all stages of growth and at all aerial parts of the plant, such as node, leaf, neck etc. The nodal and neck infections are the most severe among them. These can occur due to many conditions, such as nutrient deficiency, pests, unusual environment, bacteria, parasites and so on. Below are given a list of the most commonly occurring paddy diseases and their symptoms.
Sheath blight in rice plant is a fungal disease caused by Rhizoctonia solani Kuhn. It is a fungi that infects the straw of rice. Both leaves and stems are affected by this. If sclerotia falls off the plant straw during harvesting, this can occur.
The Magnaporthe oryzae can cause leaf blast. Soil with low moisture, prolonged and frequent rain showers, and low temperature during daytime can cause the fungi to grow on the leaves and collar. White to green or gray spots and greenish dark borders are the symptoms of this disease.
The fungi named Microdochium oryzae causes the appearance of scalded leaves. Narrow brown-reddish wide bands appear in this disease. It spreads rapidly if the leaf is wounded.
One of the most damaging and most common rice diseases, it has been largely ignored historically. Oval or round shaped dark brown spots on the leaves are the symptoms of this disease.
Bacterial leaf blight
Xanthomonas oryzae causes this bacterial disease that infects the leaves. It causes yellowing and drying of leaves and wilting of seeds. It favors rain fed lowland and irrigated areas in tropical environments .
Using image processing techniques the identification and detection of paddy diseases can become easier. The steps are show in the figure below:
The following purposes are achieved by using the method of image processing with information mining:
- Recognition of leaf and stem infection
- Measurement of the region influenced
- Infected area’s shape detection
- Infected area’s color determination
The steps followed throughout the image processing procedure are as follows:
A. Image acquisition
Image acquisition is the task of acquiring the images needed for the processing tasks. A database was prepared containing images of rice plants with diseases specifically for the purpose of detection. The images are usually captured using scanners, cameras with high resolution or similar gadgets.
B. Image preprocessing:
1. Different preprocessing techniques are used to remove different discrepancies and noises in an image. Dewdrop, dust, shadow effects in an image can cause problems during segmentation and extraction of the images. To prevent this, a number of methods are used to clear up the image.
|Gaussian Filtering ||Noise Reduction||98.63%|
|Regression-Median Filtering ||Cotton plant disease detection||83.26%|
|Weiner filtering ||Healthy & infected leaf detection||80-99%|
T Islam et al  used the Gaussian filtering method for noise reduction. They used green pixel masking with Naïve Bayes’ classifier to detect rice blast, rice brown spot & bacterial blight. They achieve an accuracy of 98.63%. Sarangdhar et al.  used regression median filtering to detect and characterize different diseases of cotton leaves with an accuracy of 83.26%. and Rewar et al.  proposed to detect the infected leaves using Weiner filtering with different window sizes. To remove different background noise, they used adaptive histogram equalization. In order to detect the edges of healthy and infected parts of the leaf, a number of edge detection techniques were used, out of which the canny edge detector gave the best result i.e. 80-99% accuracy.
C. Image segmentation
The process of sectioning a digital image into multiple segments or objects is called image segmentation. In order to find the desired objects this method is used. From the many segments of the image data only the useful and necessary information and features are extracted.
|Otsu method ||rice plant dysfunction classification||78%|
|K-Mean Clustering & Fuzzy C mean ||deficiency of mineral Identification||K-Mean=85%, FCM= 92%,|
|Otsu Method based global and automatic threshold ||diagnosis of paddy disease||87.5%|
Many researchers have used a number of distinct techniques to get the best possible result in the segmentation process. Among them, Phadilkar et al.  have proposed a methodology of leaf complexion based classification of rice leaf diseases. The classification takes two phases to complete, firstly the healthy and unhealthy leaves are classified and separated, then leaves with brown spot and leaf blast are classified. They use the Otsu method to achieve this. Sethy et al.  proposed to identify mineral deficiency of rice plant by the use of SVM with two methods of feature extraction such as K-Mean Clustering & Fuzzy C mean clustering with an average accuracy of 88.5%. Kurniawati et al  have used Otsu Method based global and automatic threshold to diagnose the rice plant disease with an accuracy of 87.5%.
also read – How to Develop a Research Plan for Thesis Paper
E. Feature Extraction
In order to classify an image, the features of that image are one of the primary parameters. The features or inherent characteristics of the objects present in an image, is focused on in this stage of image processing called feature extraction. Once the features are extracted, they can describe the object they were extracted from. Texture, color, edge and shape are the categories features are generally extracted under. Since each disease have their own distinctive color, shape and texture, they are used to classify the paddy diseases.
|Feature extraction-detection of edges ||leaf diseases detection||82%|
|Color extraction ||Soybean leaf defoliation evaluation||96%|
|Texture feature ||Unhealthy leaf detection||94%|
|HOG and LBP ||Detection of of infected area||HOG=94.6% LBP=90.23%|
Gajanan et al.  did research on the recognition and identification of plant disease in 2018. Their primary method in this research was the feature extraction-edge detection. The area, quantity of the spots on leaves and color space is used to determine the ailment of the plant. RGB pictures were utilized in the research of Liang et al.  to evaluate soybean leaf defoliation. This methodology utilized color extraction methods to reach an accuracy of 96% in the appraisal of defoliation. Arivazhagan et al.  proposed a methodology consisting of four phases in the pursuit of infected leaf detection. These phases were color transformation, pixel masking, removing threshold and segmentation. This reached an accuracy of 94%. Pothen et al.  utilized HOG and LBP to extract the features of the infected areas of the leaves, and SVM for classification with an accuracy of 94.6% and 90.23% respectively with polynomial Kernel SVM.
Classification is usually a supervised learning procedure that maps the data into specific groups or classes. This is done in two steps, the first of which is called the training phase. With the specified class labels, the classifier is developed by the classifier algorithm. This is required to generate the classifier model that describes the set of classes that were predefined. The second step uses the generated classifier model for classification. If training data is not available, unsupervised classification is used. Different methods such as Support Vector Machine (SVM), Probabilistic Neural Network (PNN), Convolutional Neural Network (CNN) are used for this step.
|K-Mean Clustering & Fuzzy C Mean ||Leaf disease classification||K-Mean=85.05%, FCM=95%|
|Support Vector Machine ||Rice seed ailment detection||97.2%|
|Back Propagation Neural Network ||Paddy disease detection||100%|
|CNN ||Infected region classification||97.43|
Sethy et al.  proposed a method for identifying mineral deficiencies in rice plants by use of K-mean clustering and FCM. In this research the overall precision reaches 85% for K-Mean and 95% for FCM respectively. A crop health detection method was introduced by Liu et al.  to obtain premium quality production at the end of growth stage. SVC (Support Vector Classification) used to portray empty, corrupted and strong panicles. This reached an accuracy of 97.2%. Orillo et al.  introduced a Matlab based methodology to identify leaf ailments using Back Propagation Neural Network(BPNN). This methodology was used to classify different illnesses of rice plant leaf. This program gave an accuracy of 100% out of 134 images that the database consisted of. T. M. S. Sazzad et al.  proposed a framework where color segmentation, enhancement, color feature and filter for classification process were incorporated for identification. They utilized CNN classifiers in order to increase the accuracy rate of the identified data.
If not given enough attention, agriculture can suffer tremendous losses because of rice plant diseases. Building an automated system using the many techniques available, early notifications of diseases can be provided. In the same direction, this paper briefly discussed some of the most common rice plant diseases that threaten our crops. This paper overviewed segmentation, pre-processing, feature selection, feature extraction, and classification methods. We summarized the different techniques researchers have used for disease detection. Using these approaches, contribution to the economy can be made by addressing and identifying solutions to the problems that plague different domains of our agriculture.
- “Agriculture Economics and Importance of Agriculture in National Economy.”[Online].Available: http://agriinfo.in/?page=topic&superid=9&topicid=185.
- Knowledgebank.irri.org. 2020. Diseases – IRRI Rice Knowledge Bank.[online] Available at: <http://www.knowledgebank.irri.org/step-by-step-production/growth/pests-and-diseases/diseases> [Accessed 20 December 2020].
- Islam, T., et al., 2018. A Faster Technique On Rice Disease Detection Using Image Processing of Affected Area In Agro-Field, Proceedings of the 2nd International Conference on Inventive Communication and Computational Technologies (ICICCT), Coimbatore, 62-66.
- Sarangdhar, A.A. and Pawar, V.R., 2017. April. Machine learning regression technique for cotton leaf disease detection and controlling using IoT. In Electronics, Communication and Aerospace Technology (ICECA), 2017 International conference of IEEE. 2, 449-454.
- Rewar, E., Singh, B.P., Chhipa, M.K., Sharma, O.P. and Kumari, M., 2017. Detection of Infected and Healthy Part of Leaf Using Image Processing Techniques.
- Phadikar, S., Sil, J and Das, A. K., 2012. Classification of Rice Leaf Diseases Based on Morphological Changes, International Journal of Information and Electronics Engineering 2(3), 460-463.
- Sethy, Kumar, P., et al. 2017. Identification of Mineral Deficiency in Rice Crop based on SVM in Approach of K-Means & Fuzzy C-Means Clustering. HELIX 7.5 1970-1983.
- Kurniawati, N. N., Abdullah, S. N. H. S., Abdullah, S. and Abdullah, S., 2009. Texture analysis for diagnosing paddy disease,” 2009 International Conference on Electrical Engineering and Informatics, Selangor,doi: 10.1109/ICEEI.2009.5254824, 23-27.
- Gajanan, D.E., Shankar, G.G. and Keshav, G.V., 2018. Detection of Leaf Disease Using Feature Extraction for Android Based System.
- Liang, W., Kirk, K. R., & Greene, J. K. (2018). Estimation of soybean leaf area, edge, and defoliation using color image analysis. Computers and Electronics in Agriculture, doi:10.1016/j.compag.2018.03.021, 150, 41–51.
- Arivazhagan, S., Shebiah, R.N., Ananthi, S. and Varthini, S.V., 2013. Detection of unhealthy region of plant leaves and classification of plant leaf diseases using texture features. Agricultural Engineering International: CIGR Journal, 15(1), 211-217.
- M. E. Pothen and M. L. Pai, “Detection of Rice Leaf Diseases Using Image Processing,” 2020 Fourth International Conference on Computing Methodologies and Communication (ICCMC), Erode, India, 2020, pp. 424-430, doi: 10.1109/ICCMC48092.2020.ICCMC-00080.
- Sethy, P. K., Negi, B. and Bhoi, N., January 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.
- Liu, Z. Y., Shi, J. J., Zhang, L. W., & Huang, J. F., 2010. Discrimination of rice panicles by hyperspectral reflectance data based on principal component analysis and support vector classification. Journal of Zhejiang University. Science. B, 11(1), 71-8.
- Orillo, J. W., Cruz, J. D., Agapito, L., Satimbre, P. J., & Valenzuela, I., 2014. Identification of Diseases in Rice Plant (Oryza Sativa) using Back Propagation Artificial Neural Network. 7th IEEE International Conference Humanoid, Nanotechnology, Information Technology Communication and Control, Environment and Management (HNICEM) Palawan: IEEE, 1-6.
- T. M. S. Sazzad, A. Anwar, M. Hasan and M. I. Hossain, “An Image Processing Framework To Identify Rice Blast,” 2020 International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA), Ankara, Turkey, 2020, pp. 1-5, doi: 10.1109/HORA49412.2020.9152912.