Autonomous Reading of Gauges in Unstructured Environments

Area of Science:

• Computer Vision
• Robotics
• Measurement Systems

Background:

• Existing gauge reading software often requires manual input for gauge location and scale.
• Current systems typically only function with a direct, frontal view of the gauge.
• Autonomous gauge reading in unstructured environments remains a challenge.

Purpose of the Study:

• To introduce GAUREAD, an end-to-end computer vision system for autonomous analog gauge reading.
• To develop a system that operates without manual input or predefined gauge parameters.
• To enable gauge measurement in unstructured environments and from varied perspectives.

Main Methods:

• GAUREAD integrates gauge detection, perspective rectification, and scale reconstruction.
• The system processes images to identify and interpret analog gauges.
• The algorithm is optimized for real-time performance on embedded systems like the NVIDIA Jetson Nano.

Main Results:

• GAUREAD achieves autonomous gauge reading with high accuracy.
• Measurement error is within 3% for perspective angles up to 20° and 9% up to 50°.
• The system demonstrates a reading speed of 800 milliseconds per gauge on an NVIDIA Jetson Nano.

Conclusions:

• GAUREAD offers a robust solution for autonomous analog gauge reading in complex environments.
• The system's ability to handle varied perspectives and unstructured data enhances its applicability.
• Potential applications include automating safety and security operations in mobile robotics.