Sugar beet farming goes high-tech: a method for automated weed detection using machine learning and deep learning in precision agriculture

Abstract

The main objective of this study is to develop a method for the automated detection and classification of weeds and sugar beets. Precision agriculture is an essential area of research that aims to optimize farming practices and reduce the use of harmful chemicals. For this purpose, the Faster RCNN and Federating Learning (FL)-based ensemble models were utilized to classify a specific dataset. In the first stage of the study, feature extraction is performed from the images in the dataset and classified by machine learning algorithms. Then, classification is carried out with the help of FL based deep learning ensemble models. Within the scope of the study, grid search is used for hyperparameter optimization and the results are obtained by a tenfold cross-validation method. Among all tested algorithms, the FL-based ensemble model constructed using the ResNet50 model exhibited the highest accuracy rate of 99%. This system has the potential to significantly reduce the use of herbicides and other chemicals in agricultural practices, promoting a more sustainable form of agriculture.