A vision-based automated guided vehicle system with marker recognition for indoor use

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Automated Guided Vehicles (AGVs) are crucial for modern logistics and manufacturing.
• Existing AGV guidance systems can be costly or complex.
• There is a need for low-cost, efficient, and robust indoor navigation solutions.

Purpose of the Study:

• To develop and evaluate an intelligent vision-based AGV system using fiduciary markers.
• To demonstrate a low-cost and efficient vehicle guiding method.
• To achieve a high recognition rate for indoor AGV navigation.

Main Methods:

• Utilized a consumer-grade webcam for image acquisition.
• Employed fiduciary markers with directional cues (letters/triangles).
• Applied hue and saturation values for marker candidate extraction.
• Used bird's eye view and Hough transform for marker detection and pose estimation.
• Implemented distance transform for character recognition within markers.
• Defined four directional signals and 10 alphabet features as markers.

Main Results:

• Achieved a high marker recognition rate of 98.87% during testing.
• Successfully calculated the positional relation between the marker and the vehicle.
• Demonstrated the feasibility of using simple markers for guidance.
• Validated the system's effectiveness in an indoor AGV setting.

Conclusions:

• The proposed vision-based AGV system using fiduciary markers is a viable and effective solution for indoor navigation.
• The method offers a low-cost, efficient, and maintainable alternative to existing AGV guidance systems.
• High recognition accuracy supports the practical application of this approach in industrial environments.