Neural Network-Based Stress Detection in Crop Multispectral Imagery for Precision Agriculture

This study explores the application of object detection techniques using neural networks, specifically YOLOv7 and YOLOv8, to classify stress in potato crops using multispectral images obtained by drones. The results indicate that YOLOv8 excels in stress detection using RGB images, while YOLOv7 shows higher accuracy with monochrome images, suggesting its suitability for specialized applications. The combination of RGB and monochromatic images significantly improved accuracy values for healthy and stressed plants, with figures of 0.917 and 0.914, respectively, and F1 score values of 0.902 for healthy plants and 0.881 for stressed plants. In addition, the importance of non-visible bands, such as NIR and RED EDGE, is highlighted to achieve complete and accurate evaluations. This work highlights the effectiveness of object detection techniques with neural networks in stress classification in agricultural images. It proposes future research to validate models in various crops and environmental conditions, thus improving precision agriculture practices.