2006 Fiscal Year Final Research Report Summary
3D Modeling of Outdoor Environments by Integrating Multiple Sensor Information
| Project/Area Number |
16300057
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Perception information processing/Intelligent robotics
|
|---|
| Research Institution | NARA INSTITUTE OF SCIENCE AND TECHNOLOGY |
|---|
Principal Investigator |
YOKOYA Naokazu Nara Institute of Science and Technology, Graduate School of Information Science, Professor, 情報科学研究科, 教授 (10252834)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAMAZAWA Kazumasa Nara Institute of Science and Technology, Graduate School of Information Science, Associate Professor, 情報科学研究科, 助教授 (40283931)
KANBARA Masayuki Nara Institute of Science and Technology, Graduate School of Information Science, Assistant Professor, 情報科学研究科, 助手 (10346306)
SATO Tomokazu Nara Institute of Science and Technology, Graduate School of Information Science, Assistant Professor, 情報科学研究科, 助手 (50362835)
|
|---|
| Project Period (FY) |
2004 – 2006
|
| Keywords | Computer Vision / Mixed Reality / 3D Modeling / 3D Environment Model / 3D Reconstruction / Multiple Sensor Fusion / Registration of Multiple Range Data / Digital Archive |
|---|
| Research Abstract |
We have obtained the following research results in this project.
1. 3D Reconstruction from Omnidirectional Videos Captured from the Ground
The objective is to reconstruct outdoor scenes using a shape-from-motion technique from omnidirectional video streams obtained by using the omnidirectional multi-camera system Ladybug. First, we have developed an algorithm for estimating extrinsic camera parameters from omnidirectional videos obtained by using a moving camera system. We have then developed a new multi-baseline stereo algorithm, which is based on counting interest points, for computing a dense depth map. These algorithms have been successfully demonstrated with experiments in real outdoor environments.
2. 3D Reconstruction from HD Videos Captured from the Sky
We have developed a method for acquiring 3D surface data from video streams obtained by using a helicopter. The method here is based on the algorithm mentioned above. The shape-from-motion technique and the multi-baseline stereo alg
… More
orithm are applied to HD video streams for obtaining a dense depth map from the sky.
3. Accurate 3D Measurement Using Laser Rangefinder
The objective is to acquire a textured 3D environment model by accurate 3D measurement using a laser rangefinder. First, we have developed an algorithm for simultaneous registration of multiple range data obtained at different locations and also have developed a method for selecting textures from multiple omnidirectional images. We have then developed a view planning method for efficiently scanning a wide area sequentially. These algorithms and methods have been successfully experimented in our campus.
4. Construction of Textured 3D Scene Model by Integrating Heterogeneous 3D Data
We have developed a 3D active contour model for integrating heterogeneous 3D data obtained above into a unified voxel model representation. This results in the construction of textured 3D outdoor scene models. Experiments have been carried out with an outdoor environment of our campus. Less
|
