公开(公告)号：US10921520B2

公开(公告)日：2021-02-16

申请号：US16529035

申请日：2019-08-01

Applicant: FURUKAWA ELECTRIC CO., LTD.

Inventor： Jun Nishina ,  Hideaki Hosoi

IPC: G02B6/25 ,  G02B6/00 ,  G02B6/255 ,  G01N21/64 ,  G06N20/00 ,  B23K26/03 ,  G06N3/08

Abstract: Brightness profile data having a number of dimensions is extracted based on image data in a radial direction of an optical fiber, the brightness profile data representing features for each rotation angle of the optical fiber. Machine learning uses training data to create a prediction model that based on the brightness profile data determines a rotation angle of each pair of optical fibers is. The pair of optical fibers are rotated to the determined rotation angle and then fusion spliced. The training data indicates a correspondence relationship between the rotation angle of the optical fiber and brightness profile in the radial direction for each rotation angle of the optical fiber. The prediction model can determine a rotation angle of an arbitrary optical fiber based on brightness profile data of the arbitrary optical fiber.