Low-Cost Plant-Protection Unmanned Ground Vehicle System for Variable Weeding Using Machine Vision

Area of Science:

• Agricultural Engineering
• Robotics
• Computer Vision

Background:

• Traditional agricultural spraying machinery contributes to pesticide waste and environmental pollution.
• There is a need for precision agriculture technologies to optimize resource use and minimize environmental impact.

Purpose of the Study:

• To develop and evaluate a machine vision-based variable weeding system for unmanned ground vehicles (UGVs).
• To reduce pesticide waste and environmental pollution by enabling adaptive spraying based on vegetation distribution.

Main Methods:

• Utilized a machine vision system with the normalized super green (2G-R-B) algorithm and fast iterative threshold segmentation for vegetation detection.
• Implemented a PID control algorithm with fuzzy rule-based adaptive parameter adjustment (Kp, Ki, Kd) and an interleaved period PWM controller.
• Conducted orthogonal testing with varying spraying duty cycles (25%, 50%, 75%, 100%) to assess system performance.

Main Results:

• The chosen image processing combination effectively distinguished vegetation from the background, improving pixel extraction accuracy.
• Orthogonal testing demonstrated minimal pressure variation (<0.05 MPa) and low spraying errors (average <2%, max <5%).
• Field trials confirmed the system's ability to adjust spraying volume in real-time according to vegetation distribution.

Conclusions:

• The developed machine vision-based variable weeding system is a low-cost and effective solution for precision agriculture.
• The system enhances control accuracy and stability, significantly reducing pesticide waste and environmental impact.