Markerless Hand-Eye Calibration by Flange Ellipse Detection
Ruoyu Jia, Ruomeng Fan, Qitong Guo, Xiaohang Shi, Masahiro Hirano, Yuji Yamakawa
AI summary
Problem
Current markerless hand-eye calibration methods struggle with poor generalization to new robots, high computational demands, or reliance on expensive fiducial markers that complicate deployment in non-laboratory settings.
Approach
The framework detects the robot's standardized circular flange using a lightweight object detector and classical ellipse fitting, then cross-validates the outputs via an IoU filter to robustly solve the Perspective-n-Point problem for pose estimation.
Key results
- Surpasses existing methods in calibration accuracy while running efficiently on CPU-only hardware
- Maintains consistent precision on unseen robot platforms without retraining
- Eliminates manual annotation needs through an automated dataset generation pipeline
- Provides real-time geometric feedback to optimize data collection efficiency
Why it matters
Democratizes high-precision vision-guided robotics by providing a scalable, low-cost calibration solution that works out-of-the-box across diverse industrial robot arms.
Abstract
This paper proposes a simple yet effective mark- erless hand-eye calibration method that achieves low cost, high accuracy, and strong generalization across different types of robots. The method utilizes a circular flange, a stan- dardized structure in industrial robots, for calibration via the perspective-n-point (PnP) algorithm, achieving superior performance with a simpler pipeline. The entire system is built using mature, off-the-shelf components, avoiding complex architectures. By combining a lightweight object detection net- work (e.g., Faster R-CNN) with classical geometric techniques, we construct a flange detector that is both accurate and robust. The training process requires no manual annotations, and the resulting model generalizes well across various robot platforms. Experiments demonstrate that our method achieves higher calibration accuracy than more complex existing ap- proaches. Notably, the method maintains consistent precision even when applied to previously unseen robots. All develop- ments are available at: https://github.com/Ruoyu-Jia/ Markerless_hand_eye.