Enhancing Agricultural Field Classification of Remote Sensing Images using YOLOv8

Crop monitoring, yield, and precision farming all require proper monitoring of agricultural fields with the help of satellite images of the remote sensors. Conventional ways of analyzing images are normally characterized by a difficult state due to alteration of light, foregrounds of the picture, smearred or obscure field borders and a convoluted backdrop of the picture. This paper proposes a more sophisticated approach to classification of agricultural fields on the basis of the YOLOv8 machine learning model. Being based on the Browwian benefits of YOLOv8 such as improved detection and learning properties, the algorithm makes finding and classifying farm-boxes on large-scale satellite images efficient. The model gets trained on a fairly edited image collection of remote sensing images depicting the aspect of different groups of crops and field configurations. Normalization, data augmentation and annotation optimization are preprocessing functions that are adopted to improve model reliability and generalization. The measures used to evaluate the performance of the system include the standard measures, such as precision, recall, F1-score, and mean Average Precision (mAP). It has been experimentally demonstrated that the YOLOv8-based approach performs better with both classification and fast inference in comparison to the traditional machine learning solutions and earlier versions of YOLO.