As urbanization expands and food security becomes critical, remote sensing and machine learning (ML) have emerged as essential tools for large-scale vegetation mapping. This paper reviews the advances, challenges, and perspectives in vegetation classification, focusing on horticultural and agricultural crops. It explores how ML techniques handle complex datasets to create accurate predictive models for sustainable land management.
Research indicates that classification accuracy depends on image type, study area, and the algorithm used. Approximately 30% of reviewed studies prioritize features such as spectral indices, textures, and elevation data. Methodologically, supervised learning dominates at 94%—comprising hierarchical (46%), mathematical (24%), and layer-based (24%) models—while unsupervised methods account for only 6%. Regarding focus, 31% of studies target horticultural products, 27% agricultural crops, and 42% other vegetation types.
These findings underscore the necessity of integrating diverse satellite data, features, and ML methods for scalable and precise classification. The study emphasizes that the future of precision agriculture lies in interdisciplinary collaboration between agricultural sciences, data mining, and remote sensing. By providing a comprehensive review with temporal and graphical analyses, this article establishes a new classification framework for the field. It serves as a vital resource for researchers and policymakers aiming to optimize resource utilization and address global food security and climate crises.