面心を利用した3次元CADモデルと視覚情報とのパターンマッチング(機械力学,計測,自動制御)

概要

論文の詳細を見る
Much researches on the aspect graph has been done since 1979 when Koenderink and van Doorn first described it in their paper. The main problems of image recognition using the aspect graph are the representation method of aspect and the matching elements between the aspect graph and the image. In this paper, we propose a new solid model, named the Modified coded boundary representation (MCBR) method and the image matching method for model-based recognition using the MCBR method. MCBR method is similar to B-reps, which is well known as a solid model used in the many 3-D CAD systems. However, B-reps has absolute coordinates, the MCBR method does not have it, but has relative coordinates. Because of this feature, it is easy to extract the relationships between the object elements, for examples faces, edges and vertexes. This proposed method is used as a 2-D model to represent the aspect graphs which made from the 3-D models and the objects extracted from the images. And the system uses the relationships between the positions of faces as a matching index. This present method is useful without change of the relationships between faces of the object by changing the gazing line of camera.
一般社団法人日本機械学会の論文
1997-06-25