3-D Reconstruction by Using Arrayd Cameras
Grant-in-Aid for Scientific Research (C)
|Allocation Type||Single-year Grants|
|Research Institution||Tokyo Institute of Technology|
OKUTOMI Masatoshi Tokyo Institute of Technology, Graduate School of Information Science and Engineering, Associate Professor, 大学院・情報理工学研究科, 助教授 (00262303)
|Project Period (FY)
1996 – 1997
Completed(Fiscal Year 1997)
|Budget Amount *help
¥2,200,000 (Direct Cost : ¥2,200,000)
Fiscal Year 1997 : ¥1,100,000 (Direct Cost : ¥1,100,000)
Fiscal Year 1996 : ¥1,100,000 (Direct Cost : ¥1,100,000)
|Keywords||stereo vision / stereo DOC space / surface model / multiple baseline / genetic algorithm / Hough transform / virtual image generation / road region extraction / 対応度画像 / 3次元構造復元|
(a) We introduced a new data structure, "stereo DOC (Degree of Correspondence) space", in order to integrate the image information obtained from multiple cameras and to extract 3-D structure robustly.
(b) We investigated the characteristics of the stereo DOC space and the constraints which an actual object surface must satisfy.
(c) We introduced the following two methods to search the object surface in the stereo DOC space.
・A method to obtain the 3-D surface model directly which consists of arbitrary-shaped trianglar patches using a genetic algorithm (GA)
・A method to identify the planes which correspond to the object surfaces and the borders of the planes using Hough transform
(d) We showed the effectiveness and the current problems of the obove methods by the experiments using synthesized and real images.
(e) Based on the obove results, we proposed a method to generate a virtual camera image with an arbitrary point of view using multiple real images.
(f) As another application, we proposed a method to detect road regions in the observed image for visual navigation of an autonomous land vehicle.
(g) Also, we made a technical survey for stereo methods.
Research Output (27results)