摘要 |
PROBLEM TO BE SOLVED: To provide a method that additionally learns various changes included in long-term environment image data without human intervention and a teacher. SOLUTION: A unit 1 extracts face regions from image data and divides off face images on a time/space basis to prepare face image arrangements. A unit 2 computes, for every combination of the face image arrangements, the distance between the face image arrangements on the basis of the distances between images included in the face image arrangements, forms initial subclusters by using the distances between the face image arrangements, combines all the subclusters by using consistent/inconsistent edges, and prepares a moving image dictionary for each combined category.
|