The procedure of stitching images of bottle cap sides.

Step 1: The intrinsic parameters of the four cameras are calibrated by Zhang’s calibration method, respectively. Then, the extrinsic parameters of the four cameras are calibrated by the approach designed in this paper. Finally, a four-camera coordinate system is established. Step 2: The center of the four-camera coordinate system is found, and a new coordinate that represents the ideal position of the cap in the four-camera coordinate system is established with the center as the origin. Step 3: A 3D point cloud model of the cap with this new coordinate origin as its center is established. Step 4: A set of images of the cap side is captured by the four-camera system. Step 5: Edge feature extraction is performed for bottle cap side images after image preprocessing. Step 6: The actual position of the cap is determined by exploiting the ideal position and the edge feature information of the cap. Step 7: The best view of cameras for the bottle cap is solved to determine which regions of the cap are seen best from which camera. The best observation regions of cameras for the cap can be obtained. Step 8: According to the best observation regions of cameras, the images belonging to the regions (i.e., region of interest) are cylindrical back-projected and are stitched to generate a full unwrapping image of the cap side.