Image Processing Technique for the Detection of Al-Berseem Leaves Diseases Based on Soft Computing

Detecting plant diseases using the traditional method such as the naked eye can sometimes lead to incorrect identification and classification of the diseases. Consequently, this traditional method can strongly contribute to the losses of the crop. Image processing techniques have been used as an approach to detect and classify plant diseases. This study aims to focus on the diseases affecting the leaves of al-berseem and how to use image processing techniques to detect al-berseem diseases. Early detection of diseases important for finding appropriate treatment quickly and avoid economic losses. Detect the plant disease is based on the symptoms and signs that appear on the leaves. The detection steps include image preprocessing, segmentation, and identification. The image noise is removed in the preprocessing stage by using the MATLAB features energy, mean, homogeneity, and others. The k-mean-clustering is used to detect the affected area in leaves. Finally, KNN will be used to recognize unhealthy leaves and determines disease types (fungal diseases, pest diseases (shall), leaf minor (red spider), and deficiency of nutrient (yellow leaf)); these four types of diseases will detect in this thesis. Identification is the last step in which the disease will identify and classified.


Introduction
Student Agriculture is a source of income for many countries in the world, especially countries that have a fertile climate and soil to produce various agricultural crops. These countries seek to achieve quality in agricultural production, which leads to paying more attention to agricultural diseases and pests affecting plants. Technology can help to find quick and effective solutions in detecting diseases before causing huge losses in crops and productivity.
Image processing can play an important role in diagnosing the diseases through analyzing the image, classifying the symptoms of diseases and then identifying the specific disease.
Early discovery of diseases will reduce the spreads of diseases and find treatment without affecting productivity.
Image processing can help farmers to save time and effort by early diseases discovery. Farmers determine the type of disease through symptoms that appear on the plant surface. However, similarities in the symptoms can cause a confusion in determining the accurate disease. Image processing techniques analyzes the symptom color and shape to determine the type of disease accurately and clearly. Image processing is an algorithm that studies image input produces an output of specific features (Nikkhouy, & Abusham, 2011). There are many algorithms that are used to improve images quality. Redistributing colors, the brightness and the contrast in the image all of these help in In 2011 Patil & Bodhe studied the image processing technique to detect diseases affect sugarcane. Fungi the most important diseases affect sugarcane where produce large losses occur on crops. They merged two methods, first one a simple threshold, and other one triangle thresholding to divide the plant leaves to parts and identify the injured part.
That gives a more accurate result and reduce the bad diagnosis (S. B. Patil & Bodhe, 2011).
In 2011 Mythili study the image noise which affects the image process. There are many filter techniques uses to remove noise from the picture, the noise quantization, and non-symmetric noise. In his paper, compare the filtering techniques use in removing noise and gave a result, using a fuzzy filter technique. This technique work with color image reduces the noise that affects the color component in the image. Remove the noise from the picture help to reduce time to detect diseases (if picture clear, detecting operation will be easy and fast) (Mythili, 2011;Hasoon et al., 2011).
In 2012 Chaudhary & Godara studied the first step of image classification, where spot color of diseases in plant leaves. The disease type is determined by the color and the area of the affected part. Color segmentation improves detect diseases. They implemented the Otsu method in "Monocot" and "Dicot" to study the background of color space and how to make it an affected in classify the image. They Improved the algorithm to remove background noise and make the image smooth from impurities. Then they study the color model (CIELAB, YCbCr and HSI) compare between them and used each one in image process part to remove noise from the image. These methods contributed in to reduce noise from pictures (Chaudhary, Chaudhari, & Godara, 2012).
In 2016 Chatterjee & Rao study the need to find a solution to losses in crops, they used the artificial intelligence In 2017 Raut and Fulsunge the writer focuses on improving the productivity of crops from monitoring the disease that affects plants. The traditional way to determine the disease not an effective, to get a perfect solution for it, the writer implements a modern technique to discover the disease in leaves and fruit. The MATLAB program has the algorithm to classify and identify diseases such as the k-means clustering and SVM. The technique of MATLAB program collects a picture of health leaf and not healthy and save it in a database. Then will preprocess the image and enhance a picture then apply K-mean to cluster image and use the SVM algorithm to classify and training image. In 2018 Tlhobogang & Wannous studied the ICT how it is important if used in the agricultural field. This technology makes a big change if implemented in agricultural. The writers focused on the causes make detection disease late, and some of the reason for spread diseases, are the lost communication between the farmers, and there is no data source contain all disease information. They focus on the input of the ICT technology in agricultural will solve many problems. Writers work in image analysis and neural networks to build data source contain disease information to improve the machine learning and training it to get fast result. They study also the problem make ICT not public, the differentiation between farmer level and agricultural area make cusses to stop spread ICT in the agricultural field This work aims to accurately classify and cluster diseases that affect al-berseem leaves. Also, to modify images through using local graph structure and histogram to improve their quality. The importance in this study derives from the importance of early detection of disease in reducing huge losses of crops and economy. Al-berseem farmers in Oman from different regions have been facing problems with diagnosing disease of their crop. This research utilizes image processing technique based on soft computing using in MATLAB program to classify pictures and discover the disease to help farmers select the correct medicine. One of the challenges is that image processing technique needs time to build a good database in order to be able to classify and detect untested images (Shaaban, 2021). Also, the study may take more time since it is my first time to use MATLAB program. There is a lack of studies and information concerning diseases that affect al-berseem leaf in Oman. This fact along with collecting data from the field can also be seen as a challenge that will take time in completing of the work.

Problem Statement & Objectives
The first symptoms affect the plant are similar and common in many diseases like changing in the color of leaves.
Farmers may diagnose the disease wrongly with the old way of diagnosing by using the naked eye. The need to find a way to detect diseases fast and choose a correct medicine is a very important step because it will save the crops and money. Image processing technique provides features through classifying pictures and modifying them to get a quick result. Al-berseem is one of the popular crops that has been cultivated in Oman since a long time. The problem of study focuses on diseases that affecting al-berseem leaves. The following questions make the study problem clear.

•
What are the diseases affecting al-berseem leaves?
• What are the causes of the disease?
• How can image processing help farmers in detecting diseases?
The main aim of the work is to accurately classify and cluster diseases that affect al-berseem leaves. A subobjective of this study is to modify images through using local graph structure and histogram to improve their quality. Other objectives of the study are: • To review related literature and studies.
• To detect unhealthy leaves of al-berseem.
• To classify al-berseem leaf diseases using texture features.
• To analyze the infected leaf and know the specific type of disease on time.

Study Area
The study is conducted in Mahdha city. Mahdha is in Al-Buraimi Governorate in Sultanate of Oman. It is characterized as a border area with United Arab Emirates. Farmers export al-berseem to it. Mahdha city has good soil and climate that contribute to have high quality of al-berseem. Figure 1. shows the study area.

Research Methodology
This study is multistage approach that classify the images farmers using different types of smart devices. The Figure 2 shows the steps of the methodology that is followed in this study, which explain how image processing techniques will enhance the images quality (Hassin & Abbood, 2021) and detect diseases.
Step 2: Pre-processing image to improve the quality and remove noise from it.
Step 3: Cluster the image based on the affected area.
Step 4: Obtain the segmented parts.
Step 5: Compare the features from one cluster to the other clusters.
Step 6: Extract and detect of disease based on neural network classifier.

Image Processing Types Using Local Binary Pattern (LBP)
Local binary pattern (LBP) is a texture descriptor and a common technique that is used to extract texture and describe detail of the texture. In 1996, first algorithm of LBP was reported where image operation was used to detect features and classify texture. LBP is classified as a strong tool that can extract texture features through images analysis

Research Experimental Setup
The Pictures are the main data used in this research, called primary data, which was collected to study a specific problem (al-berseem leaves diseases). Use iPhone7 to capture al-berseem leaves, use digital cam and saved in computer to help reduce and control image size. First, select the disease wants to study. Then took photos of healthy leaves and unhealthy leaves (Radha, 2017). They found four types of disease affect al-berseem leaves in the period of collecting data (fungal infections, pest diseases (red spider), leaf minor (shall), and deficiency of nutrient (yellow leaf)). Take more than 100 pictures of each illness, spent three months collecting all data. Contact with the agricultural engineers in Mahdha and Al-buraimi to classify the type of diseases. They classify and give correcting diseases names. Figure 4 shows a sample of collected data from Mahdha farms. and what are the changes made in leaves, help to classify diseases (Khirade, 2015). The classification method is an important step, it identifies differentiation between symptoms of the disease with different conditions and several features (Arnal Barbedo, 2013). By using the histogram to reduce the size and minimize the size of pictures and focus only in the affected area. A histogram is a tool used to discover an important statically shape for continuous data. Here will fetch the disease in the matrix with specific features, then the program will start to compare the affected leaves with healthy leaves. The program will be training all newly entered data, with the data already saved in the program.
It keeps the data features in the program and when the new data enter, will read and save the features. Finally, for the new data entered to the program will go through all previous sections, then will get the accurate result in a short time.
Lastly, collect all information come from the upper section to get the result. Here will show which disease affect alberseem leaves, from the features of the picture, after clustering it by affected area. It will compare disease features saved previously in the program with new features collected from the new picture entered. Finally, will detect disease types and show the result.

Results and Discussion
The

Conclusion
This research focuses on diseases that affected al-berseem leaves and how detecting diseases will change the productivity of crops. This research used MATLAB 2016 to identify diseases and classify their types. Early detection of diseases makes a considerable difference in productivity and economic income because most losses in this field come from misdiagnosing the disease or late detection of the disease. This work showed that modern photo techniques contribute to saving a lot of time. It helps determine immediately appropriate solutions and, therefore, to control the spread of diseases among crops.
The future work of this research looks forward to creating an application of a plant disease detection system on mobile phones. So, farmers can use it easily. Also, the study seems to complete all pests that affect the al-berseem plant clover, not only the problems that affect the leaves. A different perspective is that studying other al-berseem planting regions in Oman would provide more information on the topic.