Multi-Sensor Fusion for Environmental Perception and Decision-Making in Smart Fertilization Robots

To address the issues of resource wastage, environmental pollution, and low crop productivity associated with traditional agricultural fertilization methods, we present a smart fertilization robot system based on multi-sensor fusion. This system integrates visual (OpenMV), temperature and humidity (DHT11), and light intensity (BH1750) sensors, leveraging advanced algorithms such as Kalman filtering, Bayesian inference, and Long Short-Term Memory (LSTM) networks to achieve robust environmental perception and precise fertilization decision-making in complex agricultural settings. Experimental results demonstrate significant improvements in fertilization accuracy and resource efficiency, with a 20% reduction in fertilizer waste and an increase in fertilization precision from ±20% to ±2% compared to manual fertilization. The system also exhibits strong adaptability and stability under varying light intensities, soil moistures, and crop growth stages. Our study not only provides an effective technological solution for precision agriculture but also paves a new way for the application of multi-sensor fusion in agricultural robots.