Enhanced Image Processing and CNN Approaches for Accurate Detection of Wheat Stripe Rust

Stripe rust, induced by the fungus *Puccinia striiformis f.sp. tritici*, poses a significant risk to wheat fields, dispersing via spores carried by the wind and resulting in broad outbreaks. Proper disease management includes applying fungicide treatments on seeds, fertilizers, and timely sprays, especially if the disease is identified early. Due to the fluctuating timing and location of outbreaks every season, decisions on fungicide application should rely on data from the current season. This research centers on accurately identifying wheat stripe rust disease by utilizing image-processing and deep learning methods. Numerous techniques were assessed, such as transfer learning and different convolutional neural network (CNN) structures. Preprocessing images to eliminate background noise and optimizing the deep learning model with fewer training epochs lead to highly precise classification of wheat diseases, reaching a 97 % accuracy rate on the test set. Types like healthy, septoria, and stripe rust were effectively recognized. The research also examines various advanced methods in image processing and deep learning, offering valuable guidance for future studies. This review is focused on helping researchers discover efficient ways to detect wheat stripe rust, which will help improve disease control.