Machine Vision and Network RTK-Based Miss-Seeding Detection and Compensation System for Potato Planting: Design and Integration

To address the limitations of existing potato miss-seeding detection, including vulnerability to dust and vibration, lack of whole-process information management, and delayed reseeding response, this paper develops and verifies a miss-seeding detection and reseeding system for electric-driven potato planters based on machine vision and network Real-Time Kinematic (RTK). The system integrates an industrial camera, a network RTK rover station, and an STM32 microcontroller. A lightweight vision model enables efficient seed spoon identification. Miss-seeding events are real-time bound to high-precision RTK coordinates, with spatial distribution visualized through empirical GCJ-02 offset conversion and a map engine. PWM technology is applied to dynamically adjust the seed-metering belt motor speed for rapid compensation. Bench and field tests indicate that the miss-seeding detection rate exceeded 90% only in the workshop-calibrated, well-lit, no-seed specificity test; under normal field seeding conditions, it was approximately 83–85%, decreasing to about 64% under plant shading with an uncalibrated camera. The reseeding success rate reached approximately 87% at 0.3 m/s. This system achieves full-link closed-loop control and provides a technical solution for intelligent potato seeding.This study should be regarded as a proof-of-concept rather than a field-ready commercial solution.