Agri-Vision

DOI

https://www.doi.org/10.59256/indjcst.20260501036

Published In

Volume 5, Issue 1 (2026)

Published Online

January-April 2026

Pages

233-238

Article Type

Review Article

Check for updates

The convergence of Artificial Intelligence (AI), the Internet of Things (IoT), and satellite remote sensing is transforming modern agriculture from an empirically driven labor practice into a highly data-centric discipline. While individual machine learning models for crop recommendation, disease detection, and yield forecasting have achieved high isolated accuracy metrics, current literature reveals a critical fragmentation in deployment: these modules rarely interact within a unified decision-support ecosystem. This review systematically analyzes the architectural requirements for an integrated Sense- Analyze-Act agricultural framework. It critically evaluates the efficacy of ensemble learning methods, specifically Random Forest classifiers, for soil-nutrient-to-crop matching; Convolutional Neural Networks (CNNs), like MobileNetV2, for edge- based phytopathological surveillance; and Long Short-Term Memory (LSTM) networks for temporal market price and soil moisture forecasting. Furthermore, the transformative role of Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) in democratizing agronomic advisory services for smallholder farmers is systematically assessed. The review identifies key deployment barriers- specifically domain shift generalization, algorithmic bias attributable to non-representative training corpora, and the absence of robust multimodal sensor fusion architectures. Finally, it articulates a research roadmap emphasizing Federated Learning and autonomous unmanned aerial vehicle (UAV) surveillance to successfully bridge the gap between demonstrated algorithmic potential and smallholder operational reality.