Article Abstract 2

Agriculture remains a critical contributor to food security and economic stability; however, traditional soil testing practices and intuition-driven crop selection often result in inefficient resource utilization and reduced productivity. To overcome these challenges, this study introduces Agri Vision, a standalone Soil Monitoring and a crop recommendation framework that combines deep learning, machine learning, and IoT enabled sensing deployed on a Raspberry Pi platform. The developed approach identifies soil categories by analyzing images acquired through a USB camera, which are interpreted using a MobileNetV2 driven convolutional neural network, enabling accurate identification of seven soil categories. In parallel, soil parameters including moisture, pH, and temperature are captured through specialized sensing devices designed for accurate environmental measurement, interfaced through the MCP3008 analog-to-digital converter. The predicted soil class and sensor readings are combined and analyzed using a Random Forest model to recommend suitable crops for cultivation. A thkinter-based kiosk interface provides an intuitive user experience by supporting image capture, live sensor visualization, and instant result generation, along with email and quick response-based report delivery. All processing tasks are executed directly on the Raspberry Pi itself, enabling independent operation without internet access and ensuring reliable use in isolated and underserved farming regions. Experimental evaluation demonstrates high accuracy in soil classification and reliable crop prediction, validating the experimental outcomes confirm the practical viability of the developed system. Agri Vision provides an affordable and extensible framework for precision farming while also enabling future upgrades like soil nutrient analysis and intelligent irrigation automation.
Keywords: Precision, Agriculture, Soil Monitoring, Crop Recommendation, Deep Learning, Internet of Things (IoT), MobileNetV2, Random Forest, Raspberry Pi