طبقه‌بندی بیماری‌های قارچی گیاه جو به کمک روش‌های پردازش تصویر و هوش مصنوعی

Barley is one of the four major crops produced worldwide. Plant pathogens are considered the most significant limitation in global barley production. These pathogens can destroy the green areas of the leaves and limit the ability of the plant to form and fill seeds. Rhynchosporium commune and Pyrenophora teres are among the most important pathogens of barley. Recent advances in agricultural technology have increased the demand for the development of non-destructive disease detection methods. Today, computer tools based on image processing and artificial intelligence have been developed in agriculture to monitor crop growth. Some applications of these tools include nutrient deficiency assessment during plant growth stages, water stress monitoring, growth process tracking, pest detection, environmental impact analysis, and plant disease diagnosis. In this study, image processing and artificial intelligence were used to diagnose and classify two pathogens R. commune and P. teres. For this purpose, a dataset consisting of 1000 images of healthy barley leaves and 2000 images of plants infected with fungal disease was used. To implement the disease detection algorithm, a program was prepared in the MATLAB programming environment. In the disease identification algorithm, after separating the diseased area from the images of barley leaves, image texture, shape, and color features were extracted. These features are considered essential for extracting information from images of agricultural products. Then, by combining the random forest classification method with the meta-heuristic gray wolf algorithm to optimize the hyperparameters of the random forest, diseases were diagnosed and classified and the most effective image features were identified. The accuracy of diagnosing and classifying diseases with the proposed algorithm was equal to 91.33 and 90.50% for the two studied fungal diseases, respectively. Also, the feature selection results showed that image texture features have a much greater effect than other extracted features in the diagnosis of the two fungal diseases.