A flexible camera calibration method for pose vision measurement system of roadheader

• A new method for on-site camera calibration without manual intervention is proposed. • Construct virtual point and line fusion features to calibrate camera parameters. • A dynamic weight objective function further optimizes the calibration results. • Simulations and experiments show that it has high accuracy and good adaptability. Cameras used for visual measurement in underground roadway construction sites are difficult to calibrate. Therefore, this study proposed an on-site camera calibration method that does not require manual intervention. First, the virtual point and line features on the plane target are constructed using pole-polar constraints. Second, a homography model based on point-and-line features is established. Finally, a combined objective function based on the dynamic weight coefficient is proposed to optimise the calibration results further. Relevant simulations and experiments showed that the maximum internal parameter error obtained using the proposed method was 0.463 %, and the maximum position estimation error of the roadheader was 30 mm. In addition, the results show that this method is superior to other methods in terms of calibration accuracy and robustness, solving the difficulty of camera calibration in underground roadway construction sites and guaranteeing high-precision pose vision measurement.