紀要論文

複数のRGBDカメラによるボリュームモデルのリアルタイム生成

Real time volume modeling with multiple RGBD cameras

jpn

24321192

https://doi.org/10.15002/00021533

法政大学大学院情報科学研究科

法政大学大学院紀要. 情報科学研究科編

13

1

6

2017-03-31

Version of Record

With the head mount displays are getting popular, virtual reality technology is put to the practical use. Many devices and applications are introduced to public markets, and consumers are able to use them readily. Most applications, however, use the shape models generated inside the computer, but ones from real objects in real world will become common in near future. It is necessary to develop a technique to generate a shape model from three-dimensional scan of a moving object, this paper proposes a method to create a three-dimensional model in real time using multiple RGBD cameras. The proposed method estimates the shape model using voxels. Simultaneously scan the entire circumference of the subject with multiple cameras. Using depth data obtained from each camera, that express the distance to the surface, we give weights to each voxels at the measured distance. The polygons which are rendered for each voxel are calculated using the Marching Cubes method. In order to reduce the cost of Marching Cube, we introduced Octree to reduce the searching area of rendered polygons. In the existing work using the surface model, the processing speed decreases with the number of cameras, but our proposed method is hardly affected. It gains four times faster than the existing method from the experiment using four RGBD cameras.

Article

AA12746425