ML-Based Soil Analysis for Crop Suggestion and Fertilizer Recommendation

Contemporary agriculture requires intelligent responses to the issues of resource utilization and unforeseeable environmental conditions. We propose a 6-wheel autonomous agriculture vehicle system, through in situ sensing, machine learning (ML) and cloud-based visualization system, to achieve more efficient and environmentally friendly agriculture activities. It uses an ESP32 WROOM microcontroller for obtaining information from a series of environmental sensors like DHT22 (temperature & humidity), LDR (light intensity), along with individual soil moisture and NPK (nitrogen, phosphorus, potassium) sensors. The NPK sensor data is handled by an RS485 module. This extensive network of sensors forms an integrated view on the subsurface and atmospheric conditions. In order to facilitate intelligent decision making, the proposed system makes use of ML models for rainfall prediction (Linear Regression) and crop recommendation (Random Forest) and fertilizer recommendation (Random Forest). These models are trained on useful datasets to give precise and context-based information. The system is cloud-connected and includes a local web server for easy remote monitoring and control. It is built on PHP, HTML, CSS, Javascript and JSON, and acts as a server that runs on “http” and has elaborate visualizations of sensor data and ML suggestions in a more digestible manner. The server has its own IP address, so it can be accessed from any Web-connected computer.