| 摘要 |
PURPOSE: A method is provided to prevent an elipolar geometry issued at recognizing a three-dimension object by use of two or more images by recognizing the three-dimension object using a two-dimension image. CONSTITUTION: The method for recognizing a three-dimension object using a two-dimension image comprises the steps of: generating a model base of the three-dimension object; obtaining specific points included in an inputted two-dimension image (505); generating a virtual coordinate of six points (1, 2, 3, 4, 5, 6) comprising four points (1, 2, 3, 4) which consists of a plane among the specific points (515); expressing a virtual plane formed by the virtual coordinate by &thgr; and Φ (520) and increasing weights of items which search &thgr; and Φ as an index in a hash table included in the model base of the three-dimension object (525); and searching items exceeding a predetermined threshold value among the hash table (530).
|
