抄録
This study proposed a novel calibration method for stereo vision using the shape of the field lines on an image, particularly focusing on a curved path on an athletic track. The proposed method involves calculating the transformation coefficient between the 3D and 2D image coordinates to minimize the gap between the observed curved paths on an image and projected curved paths with the transformation coefficients. This is based on determining the radius of curvature of an athletic track from each track drawing. We conducted an experiment on an approximately 40.317-m-long region along the curved path on the athletic track, capturing three cameras (resolution = 3840 × 2160). We compared the 3D coordinates reconstructed using transformation coefficients calibrated by both the proposed method and the directory linear transformation (DLT) method. The intraclass correlation coefficient [2.1] of the proposed method and DLT method were 0.999 (p < 0.0001) for all components. The differences between the upper and lower limits of the agreement were 1.783%, 0.187%, and 0.700% for the vertical, radial, and azimuthal components, respectively, relative to the calibration region. Employing the proposed method, we can undertake 3D motion analysis during competitions without establishing reference points in the field.
