Deep Learning for Super Resolution of Sugarcane Crop Line Imagery from Unmanned Aerial Vehicles

Improving resolution of sugarcane crop images is crucial for extracting valuable information related to productivity, diseases, and water stress. With the rise of remote sensing technologies like Unmanned Aerial Vehicles (UAVs), the number of images available has grown exponentially. In this study, we aim to enhance image resolution using deep learning techniques, namely MuLUT, LeRF, and Real-ESRGAN, to optimize extraction of sugarcane agronomic characteristics. Although these models were initially designed for landscapes, people, cars, and anime images, our experiments with agricultural images show promising results, outperforming classic upsampling algorithms by an impressive 482.81%. Visually, the image quality improvement is significant, making our approach an attractive alternative for extracting crucial information about the crop. This research has the potential to revolutionize the analysis of sugarcane crops, opening new possibilities for precision agriculture and improved agricultural decision-making.