A novel hand-eye semi-automatic calibration process for laser profilometers using machine learning.

• A semi-automatic calibration process is proposed to improve the efficiency and accuracy of hand-eye calibration. • A hand-eye calibration postures generation model is developed to automatically drive the robot and machine tool to the target postures for calibration. • A machine-learning method is extended to process the profile data collected by laser profilometers, which is faster, more robust and accurate than existing methods. Hand-eye calibration is a vital step in the laser profilometer integrated measurement system. However, the widely applied standard sphere calibration process requires extensive laborious human work and expertise. To standardize and streamline the hand-eye calibration process for Industry 4.0, a semi-automatic calibration process was proposed, containing a calibration postures generation model and an intelligent circle detection method based on machine learning. With the proposed process, the measurement system can collect required data with minimal human participation and automatically processes them, which can significantly improve hand-eye calibration accuracy and efficiency. Additionally, an exhaustive study of the influence of the number of postures on the calibration accuracy of the industrial robot and automated fibre placement (AFP) machine demonstrates the proposed model's effectiveness. Moreover, the current study's results can also be excellent tools for further hand-eye calibration research. Our code and dataset are made publicly at: https://github.com/tangyipeng100/hand_eye_cali_circle_segmentation. [ABSTRACT FROM AUTHOR]