Deep Learning-based Image Processing for Early Detection of Stripe Rust in Wheat Crops using CNN Model

This research proposes a novel approach utilizing Convolutional Neural Networks (CNNs) for the early detection of stripe rust (Puccinia striiformis) diseases in wheat crops. The study focuses on leveraging image processing techniques and deep learning concepts to analyze leaf images, identifying characteristic yellow stripes of pustules caused by the fungus Puccinia striiformis f.sp. tritici. The CNN model, a key component of the research, is intricately designed to capture and learn intricate patterns associated with stripe rust, offering a proactive solution to mitigate its impact. The methodology encompasses preprocessing steps, convolutional and pooling layers, and fully connected layers to enhance the model’s predictive capabilities. Evaluation metrics, including accuracy, sensitivity, precision, recall, and $F 1$ score, demonstrate promising results with an overall accuracy of $86 \%$. The research emphasizes the importance of suitable variety selection, seed and fertilizer treatments, and timely fungicide application as part of an integrated disease management strategy. Limitations, future enhancements, and practical recommendations for farmers are also discussed, emphasizing the significance of removing volunteer wheat plants and selecting resistant varieties. The proposed CNN model stands as a promising tool for precision agriculture, aiding in the timely and efficient management of stripe rust in wheat crops.