GAN-Based Super-Resolution for Disease Detection in Aerial Images: A Case Study of Potato Crop

Precision agriculture requires high-resolution images to monitor crop growth and health effects. However, capturing such images from aerial platforms faces challenges due to altitude, motion blur, and limited resolution. Early detection and identification of these diseases can help prevent their spread and minimize the impact on crop yields. To address the challenge of capturing high-quality images, this paper relies on the new degradation model BSRGAN to enhance low-resolution aerial images of potato crops, initially sized at \(750 \times 750\) , to a higher resolution of \(3000 \times 3000\) . The experimental results indicate that the proposed BSRGAN method surpasses existing super-resolution techniques in terms of visual quality. The study utilizes the Potato Multispectral Images Dataset, which has dimensions of \(750 \times 750\) . Interpolation methods are then applied to enhance the resolution, resulting in a high super-resolution dataset with dimensions of \(3000 \times 3000\) . Subsequently, the dataset is labeled using the publicly available tool makesense.ai (makesense.ai: https://www.makesense.ai/index.html ), based on three categories of crop health: (a) Healthy, (b) Potato Leafroll Virus (PLRV), and (c) Verticillium wilt, as specified by the Potato Disease Identification, Agriculture, and Horticulture Development Board 2023 (AHDB). After labeling the data, we use You Only Look Once version 8 (YOLOv8) for potato crop health detection on the high-resolution dataset with dimensions of \(3000 \times 3000\) . The YOLOv8 algorithm has been trained on a high super-resolution dataset for object detection and classification. It achieves an impressive mAP better than without super-resolution of over 73%, specifically for healthy potato leaves, with an overall mAP of 56% across all three categories. These findings demonstrate the potential of deep learning-based approaches to accurately and efficiently identify potato leaf diseases, empowering farmers to protect their crops.