Vision-Based Deep Learning for Crop Classification in Precision Agriculture

As precision agriculture expands, the need for accurate and adaptable crop classification methods becomes critical for improving resource efficiency and supporting sustainable farming. This systematic literature review investigates recent advancements in computer vision (CV) and Deep Learning (DL) for crop classification, synthesizing findings from 87 high-impact journal studies published between 2019 and 2024. The primary methodologies, including (CNNs), Recurrent Neural Networks (RNNs), and transfer learning, are categorized and assessed for their performance across various agricultural applications. Results indicate that CNNs and hybrid models achieve high accuracy (often >90%) in controlled environments, while transfer learning offers improved adaptability across diverse field conditions, albeit with a high computational cost. Key challenges identified include the need for larger, more diverse agricultural datasets and better model generalization to account for variable field conditions. These advancements hold significant promise for enhancing crop classification. Future research should focus on optimizing models for real-time applications and integrating advanced imaging techniques to promote sustainability in agriculture through targeted, data-driven interventions.