SUBJECT RECOGNITION DEVICE, SUBJECT RECOGNITION METHOD AND SUBJECT RECOGNITION PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide a subject recognition device capable of improving subject recognition accuracy while maintaining smallness in the number of subject observation images and high speed performance of recognition.SOLUTION: A subject recognition device comprises: feature point extraction means which extracts feature points respectively from a subject development image generated on the basis of three-dimensional shape data and texture data of a subject and a plurality of retrieval object images; three-dimensional spatial distance calculation means which calculates a three-dimensional spatial distance with an adjacent feature point for each feature point of the subject development image; two-dimensional spatial distance calculation means which calculates a two-dimensional spatial distance with the adjacent feature point for each feature point of the subject development image; inter-feature vector distance calculation means which calculates an inter-feature vector distance between each feature point of the subject development image and a feature point of each of the retrieval object images; and collation means which calculates a score on the basis of the three-dimensional spatial distance, the two-dimensional spatial distance and the inter-vector distance, votes the calculated score for the corresponding retrieval object image, and selects an image of a subject candidate from the retrieval object image to output it on the basis of the voting result.