Hand-Eye Calibration System

A hand-eye calibration system based on FastAPI + Gradio, supporting automatic sampling, image stream, error visualization, and nonlinear optimization.

Hand-eye calibration method was updated to Reprojection-Optimized-based, more detail see HandEyeCalibrator class _geo_error_opt method. This update improve accuracy. (About theory will update later)

For learning and reference purposes. Now, this only support eye on base.

Features.

• Supports ZED camera image acquisition and AprilTag pose estimation

• Supports robot control (e.g., Aubo) and pose sampling

• Supports sample selection based on optical axis angle, distance, and translation constraints

• Supports Spectral/KMeans clustering and nonlinear optimization

• Supports Web UI (Gradio) and API calls (FastAPI)

• Only python

PRPARE

1. INSTALLTION AUBO python PACK, Ref: https://www.bilibili.com/video/BV1aM411x7Ko/?spm_id_from=333.337.search-card.all.click&vd_source=7482ad936e799333160f2fbb0cb32509 or https://www.youtube.com/watch?v=ITNtUZekr7s
2. Note tag families
3. and others ...

Quick Start

bash run.bash

REFERENCES

1. Olga Sorkine-Hornung and Michael Rabinovich, Least-Squares Rigid Motion Using SVD

Accuracy

Accuracy is mm level