A core position measurement method for measuring, from an end face image of a multi-core optical fiber including a plurality of cores and a common cladding surrounding the plurality of cores, relative positions of the plurality of cores with respect to the common cladding having a different brightness from the plurality of cores in the end face image, the core position measurement method comprising:
- a first step of associating, with each of a plurality of brightness threshold value candidates prepared in advance, one or more contours extracted from the end face image using one of the brightness threshold value candidates, and calculating one or more feature vectors associated one-to-one with the one or more contours corresponding to the one brightness threshold value candidate;
- a second step of generating a first set including one or more first feature vector candidates, each of the one or more first feature vector candidates being a feature vector satisfying a structural condition of the common cladding among the one or more feature vectors calculated in the first step;
- a third step of calculating, for each of the one or more first feature vector candidates belonging to the first set, a distance from one first target feature point whose position is designated by one first feature vector candidate to each of first peripheral feature points whose positions are designated by remaining first feature vector candidates, and selecting, as a first brightness threshold value for extracting a contour of the common cladding, a brightness threshold value candidate corresponding to a first feature vector candidate having a maximum degree of coincidence of the position of the first target feature point with the first peripheral feature points among the plurality of brightness threshold value candidates;
- a fourth step of generating a second set including second feature vector candidates from the one or more feature vectors calculated in the first step, each of the second feature vector candidates being a feature vector satisfying a structural condition of the plurality of cores;
- and
- a fifth step of calculating, for each of the second feature vector candidates belonging to the second set, a distance from one second target feature point whose position is designated by one second feature vector candidate to each of second peripheral feature points whose positions are designated by remaining second feature vector candidates, and selecting, as a second brightness threshold value for extracting contours of the plurality of cores, a brightness threshold value candidate corresponding to a second feature vector candidate having a maximum degree of coincidence of the position of the second target feature point with the second peripheral feature points among the plurality of brightness threshold value candidates, wherein
- each of the one or more feature vectors calculated in the first step is a position vector whose components are represented by a plurality of pieces of structural data converted into numerical form as features of a corresponding contour of the one or more contours,
- each of the first target feature point, the first peripheral feature points, the second target feature point, and the second peripheral feature points is defined as a point in an m-dimensional spatial coordinate system with an origin at each of start points of the one or more feature vectors, where m is an integer of 2 or more and the number of components of the feature vector is m,
- the degree of coincidence of the position of the first target feature point is evaluated based on a coincidence score obtained by adding a decreasing function of the distance to each of the first peripheral feature points calculated for the first target feature point,
- the degree of coincidence of the position of the second target feature point is evaluated based on a coincidence score obtained by adding a decreasing function of the distance to each of the second peripheral feature points calculated for the second target feature point, and
- the relative positions of the plurality of cores with respect to the common cladding are measured based on the structural data of the first feature vector candidate corresponding to the first brightness threshold value selected in the third step and the structural data of the second feature vector candidate corresponding to the second brightness threshold value selected in the fifth step.