SAFE SEMI-SUPERVISED FUZZY CLUSTERING AND ITS APPLICATION IN AGRICULTURAL IMAGE SEGMENTATION

Abstract

Image segmentation is an important technique in image analysis and processing in agriculture, aiding in the monitoring of crop health, yield estimation, crop mapping, and resource management. Traditional image segmentation methods face challenges due to the need for large labeled datasets and the complexity of the agricultural environment. To address this issue, the paper proposes combining the advantages of fuzzy clustering and semi-supervised learning, allowing the use of both labeled and unlabeled data to improve segmentation accuracy. The paper also introduces Picture Fuzzy Sets (PFS), an extension of fuzzy sets, which provides a more detailed representation of uncertainty and effectively handles noisy regions in the data. When combined with safe semi-supervised learning techniques, the proposed method ensures high accuracy in agricultural image segmentation, particularly in noisy and uncertain environments. This approach offers significant advantages over existing methods, minimizing the negative impact of labeled data and optimizing the clustering process. The results show that the safe semi-supervised fuzzy clustering method with PFS can be effectively applied to agricultural image segmentation with complex and noisy agricultural data.