Calibration of the omnidirectional vision system for robotics sorting system

Area of Science:

• Robotics
• Computer Vision
• Machine Learning

Background:

• Object sorting in robotics requires accurate obstacle distance perception.
• Omnidirectional vision systems offer comprehensive environmental data.
• Structured light provides simple detection for vision systems.

Purpose of the Study:

• To propose an improved extrinsic calibration method for omnidirectional vision systems.
• To enhance measurement accuracy and robustness in robot vision systems.
• To develop a simulator for testing and validating calibration methods.

Main Methods:

• Utilized an omnidirectional vision system with a fisheye camera and structured light.
• Developed and applied an improved extrinsic calibration technique.
• Created a Unity-based simulator for generating synthetic datasets and ground truth annotations.

Main Results:

• Achieved measurement errors below 0.5% for synthetic data and 1.6% for real data.
• Reduced calibration computation time by 12%.
• Demonstrated superior robustness and accuracy compared to existing methods.

Conclusions:

• The proposed extrinsic calibration method significantly improves vision system performance for robotics.
• The developed simulator facilitates efficient testing and validation of robotic vision calibration techniques.
• The findings contribute to more accurate and reliable robot navigation and operation.