A REAL-TİME AI-DRİVEN AGRİCULTURAL ROVER INTEGRATİNG PLANT DİSEASE DETECTİON AND GEO-REFERENCED SOİL MOİSTURE ANALYSİS

This paper presents an autonomous agricultural ground robot for real-time monitoring of plant health and soil moisture in large-scale crop fields.The proposed system integrates a deep learning-based perception subsystem with autonomous navigation and a custom ground control station (GCS) to enable continuous and geo-referenced field analysis.Visual data are acquired using an onboard camera and processed in real time on a Raspberry Pi using an object detection model (ODM) to identify disease-related visual symptoms such as discoloration, deformation and leaf degradation.Each detected instance is associated with a confidence score and accurately geo-tagged using GPS data.In parallel, a contact-based soil moisture sensor performs localized measurements at fixed spatial intervals along a grid-based coverage trajectory.All perception outputs are synchronized with navigation data and transmitted via a telemetry link to the GCS where live video with detection overlays, rover trajectory, mission status and sensor telemetry are visualized and logged for postmission analysis.Field experiments conducted under real operating conditions demonstrate stable autonomous operation and achieve an overall plant disease detection accuracy of 85-90%, confirming the effectiveness of the proposed system for precision agriculture applications.