摘要 |
PROBLEM TO BE SOLVED: To obtain a device and a method which produce a recognition model capable of detecting an object and detect an object to be detected on the basis of the recognition model when an object to be detected is photographed in an unknown image and even when a few apparent changes such as a direction and occlusion exist. SOLUTION: This device inputs plural images in which an object to be detected and the other things except the object are photographed (101 and 102), extracts colors, edges, texture or a characteristic vector obtained by connecting them from the respective inputted images (111), classifies the space of the characteristic vector into an image characterizing area and areas other than it hierarchically (112) and produces a recognition model in the characteristic space (113). Also, in the detection of a local area, it selects a local area to be a collation object by scanning an unknown image in a search window, extracts the characteristic vector from each local area and decides whether or not the object to be detected is photographed in the unknown image with the collation calculation of each extracted characteristic vector with the recognition model.
|