A Multistage Ai Framework For Prenatal Crop Monitoring, Risk Detection And Storage Management

Agriculture faces challenges in productivity and eco- nomic issues because there is no real-time information available about soil, crops, climate, and market conditions. The current smart farming systems tackle these issues individually and do not provide a complete, lifecycle-based decision support system. This paper introduces a Multi-Stage AI Framework for Prenatal Crop Surveillance, Risk Assessment, and Storage Solutions that follows a systematic Sense-Analyze-Act approach. The proposed framework combines IoT-based field sensing with ESP32 modules and environmental sensors. It also uses data analysis through machine learning and deep learning algorithms. Random Forest analyzes soil and assesses crop suitability. YOLOv8 detects pests and diseases in real time from leaf images. Long Short-Term Memory (LSTM) networks predict market prices based on historical data. A Flask-based backend manages data routing and preprocessing, while a Streamlit front end provides real- time alerts, visualizations, and decision support for farmers. This multi-layered architecture promotes modularity, scalability, and smooth integration of hardware and software. The prototype developed demonstrates real-time sensor data collection, image monitoring, and predictive analysis, offering a farmer-focused decision support system to reduce crop loss, improve storage efficiency, and support sustainable farming practices.