Vegetation health assessment system: towards precise agriculture using artificial intelligence

Abstract

This paper introduces a Vegetation Health Assessment System (VHAS) that integrates deep learning regression and large language models to improve precision agriculture by predicting and analyzing vegetation health in Southern Nigeria. The VHAS uses a customdesigned Efficient Vision Transformer (EVT) to reliably anticipate vegetation indices from remotely sensed data. The Google Gemini large language model (LLM) then processes these predictions, producing complete reports that include thorough analysis, highlight potential concerns, and provide context-specific management advice. A case study highlights the system's ability to deliver insightful assessments of vegetation health, going beyond simple index numbers and recommending actionable solutions for improving plant health. The VHAS's capacity to combine many data sources and models while generating human-readable reports considerably improves precision agriculture decision-making, outperforming traditional methods, and many other AI-based alternatives. While the work shows encouraging outcomes, it also exposes limitations and proposes future research approaches, with an emphasis on improving model accuracy, data diversity, and quick engineering techniques. The VHAS provides an important step towards more precise and efficient vegetation health monitoring and management in a variety of fields, including agriculture, forestry, and environmental conservation.