| 摘要 |
There are provided three-dimensional data entering means (10) for entering the three-dimensional data of an object; a reference image storing part (30) for storing a reference image of at least one object; attitude candidate deciding means (20) for producing an attitude candidate that is a candidate of the attitude of the object; comparison image producing means (40) for projecting the three-dimensional data to a two-dimensional image in accordance with the attitude candidate, while producing a comparison image approximate to the reference image; and image collating means (55) for performing a collation based on either a distance value between the reference image and the comparison image or a similarity degree therebetween. This allows collation or retrieval to be performed with a high accuracy even when the reference image of the object is captured under different conditions such as attitude, illumination and the like, or even when no three-dimensional object model of the object can be obtained in advance, or even when only one or few reference images are existent. |
