A novel hand-eye calibration method of picking robot based on TOF camera

Area of Science:

• Robotics
• Computer Vision
• Calibration Techniques

Background:

• Hand-eye calibration is crucial for robotic manipulation, especially in unstructured environments like fruit picking.
• Existing methods can lack stability and accuracy, particularly with varying lighting and poses.
• Time-of-Flight (TOF) cameras offer depth information that can improve calibration robustness.

Purpose of the Study:

• To propose a simple and stable hand-eye calibration method for fruit-picking robots.
• To improve the accuracy and robustness of calibration using TOF camera data.
• To validate the method's effectiveness in a simulated fruit-picking environment.

Main Methods:

• A TOF depth camera is mounted on a robot end-effector to capture calibration board images from multiple poses.
• Circle center extraction and sorting algorithms are used to obtain accurate calibration point data.
• Iterative optimization using Singular Value Decomposition (SVD) and weighted point residuals refines hand-eye parameters.
• Deep learning and 3D vision are combined for peach identification and positioning to verify the calibration.

Main Results:

• The proposed method achieves improved accuracy and stability in hand-eye calibration.
• The technique demonstrates a strong ability to identify and mitigate gross errors.
• Experimental simulations with a JAKA robot and TuYang camera confirmed the method's feasibility and effectiveness.
• The calibration process is simple to operate, with low-cost, easy-to-manufacture calibration boards.

Conclusions:

• The developed hand-eye calibration method is simple, stable, and cost-effective for fruit-picking robots.
• The integration of TOF cameras and optimization techniques enhances calibration performance.
• This approach provides a reliable solution for precise robotic manipulation in agricultural applications.