استفاده از هوش مصنوعی عمیق برای ردیابی خودکار بیماری زنگ جارویی زرشک: مطالعه موردی، شهرستان قاین، استان خراسان جنوبی، ایران

Barberry is a significant and highly valued commodity in Iran, with more than 98% of the country's seedless barberries coming from South Khorasan province. Around 40,000 families in this province depend on the production and sale of this product as their primary source of income, both directly and indirectly. The primary and detrimental disease that impacts seedless barberry shrubs in South Khorasan province is known as barberry witches' broom rust, which is caused by the fungal pathogen Puccinia arrhenatheri. This disease has been spreading in the barberry gardens of the province for almost twenty years. Recently, the spread of the disease has increased significantly due to climate changes and consecutive droughts in the province. As a result, new areas of the province are being reported as infected with the disease every year. The development of new technologies has led to advancements in plant disease diagnosis methods. Traditional methods have evolved into more sophisticated techniques, including deep learning and machine learning. In 2016, a group of researchers from the University of Bonn in Germany introduced an innovative technique that employs deep learning to monitor and detect plant diseases. They devised convolutional neural networks (CNNs) to construct a profound learning model, which achieved an accuracy of over 99% in distinguishing between healthy and diseased leaf images. Timely detection of barberry broom rust disease is essential for effective management and decision-making. It allows for the implementation of an integrated disease management approach to protect the quality and quantity of this valuable product. In barberry orchards that have a history of infection with this disease, the buds of infected shrubs swell earlier than those of other shrubs. However, the conventional methods used to identify plant diseases often result in plant destruction, have limited access to experts, and are less accurate in detection. To address these issues, it is necessary to design automated methods that can quickly and accurately identify the disease and categorize it. Therefore, the primary objective of this investigation was to detect the barberry witches' broom rust disease efficiently and accurately. To achieve this, a novel automated detection approach was developed, employing artificial intelligence. This approach utilized deep convolutional neural networks and incorporated a generalized gradient descent optimizer. To execute this methodology, an advanced smartphone camera was utilized to capture precise images of both diseased and healthy foliage. Subsequently, these images underwent analysis through this technique. The resultant framework attained a remarkable recognition accuracy of 99.23% within a mere 3 milliseconds, outperforming conventional deep structures with comparable characteristics. These discoveries underscore the substantial benefits of our approach.