AI-Driven Edge-to-Cloud System for Crop Health Management

Timely detection of plant disease is critical to agricultural productivity, yet existing solutions remain expensive, cloud-dependent, or inaccessible to smallholder farmers. This paper presents a low-cost, real-time Plant Health Monitoring System built on a Raspberry Pi edge-computing platform. The system integrates a camera module for continuous leaf image acquisition, an OpenCV-based HSV colour-space analysis engine for on-device plant health classification, a digital soil moisture sensor for irrigation feedback, and a dual-channel relay module for automated physical response. Leaf health is assessed by segmenting green-pixel regions within detected leaf contours and computing a green density percentage. A rule-based inference engine maps the observed density to one of three disease conditions: Fungal Leaf Spot, Nitrogen Deficiency, or Pest Attack, each accompanied by a probable cause and a recommended treatment. Upon detection of an unhealthy plant, the system activates a disease-control relay, saves the annotated frame locally, and dispatches a structured diagnostic alert including disease name, cause, treatment, green percentage value, and a photograph of the affected leaf to the farmer via the Telegram Bot API. Soil moisture readings independently trigger a pump relay for demand-driven irrigation. All image processing and decision logic execute entirely on-device with no cloud dependency, achieving sub-second detection-to-actuation latency. The complete system is assembled from commodity open-source components, making it financially accessible to smallholder farmers and agricultural extension programs.