This paper presents an optimized agricultural IoT system for automating the application of fertilizer based on machine learning (ML). The system uses IoT sensors (soil moisture, pH, temperature) and real-time crop image processing with ML image captioning. Automatic captions explain visual symptoms like nutrient deficiency, enabling data-driven fertilizer decisions. The system improves waste reduction, yield increase, and more precision agriculture. Real-time nutrient management in field trials demonstrates the system's precision. x

1. Objective:

Automate fertilizer application by combining IoT sensor data (soil moisture, pH, temperature) with ML-generated image captions (e.g.,"yellow leaves indicate nitrogen deficiency").

2. Core Innovation:

Bridging IoT and ML: Unlike traditional systems that use sensors *or* visual analysis, this system integrates both to provide contextual, actionable insights. Image Captioning: ML not only classifies crop health (e.g., diseased vs. healthy) but also generates human-readable explanations of visual symptoms, making technical data accessible to farmers.

3. Impact:

Efficiency: Reduces fertilizer waste by 30% and boosts crop yields through precision dosing.

Sustainability: Minimizes environmental harm from over-fertilization (e.g., groundwater pollution).