解像度制御を用いたステレオ視による複雑な物体シーンの記述作成 [in Japanese] Making Description of Complicated Object Scenes by Stereo with Resolution Control [in Japanese]
Access this Article
Search this Article
This paper describes a method for making description of complicated scenes including different-sized objects (some objects may have textures on their surfaces) by using stereo with resolution control. In such scenes, some objects may occlude other objects and, thus, surfaces of the occluded objects can be observed as some small surfaces. In case of observing the scenes in a low resolution, the boundaries of small surfaces may not be extracted. In a high resolution, a surface boundary can be fragmented into small pieces if the surface has textures on it and if the brightnesses of some parts of the texture near the boundary are similar to the brightness of the other side of surface separated by the boundary. Thus, it is difficult to obtain scene descriptions by associating object boundaries in single resolution observation because stable boundaries may not be obtained. We propose a method of describing complicated object scenes, in which boundaries of each surface are followed while the resolution of stereo cameras in the observations is selected depending on the complexity near the following boundary in order to obtain stable ones. The description of the object surfaces are constructed from only those stable boundaries. We proved that our method is applicable by some experiments of real scenes.
JRSJ 17(3), 380-388, 1999-04-15
The Robotics Society of Japan