学位論文

3D-Hough法を用いた屋内空間の三次元平面形状の検出

3D shape detection of the indoor space based on 3D-Hough method

jpn

2013-03-24

Not Applicable (or Unknown)

2013-03-24

修士(理学)

情報科学研究科情報科学専攻; 指導教授: 藤田悟

This paper describes methods for detecting the 3D shapes of the indoor space that is represented as a combination of planes such as a wall, desk, or whatnot. Detecting the planes makes it possible to perform calibration of multiple sensors and 3D mapping, and then produces various services such as the acquisition of life logs, AR interaction, and invader detection. This paper proposes and verifies three algorithms. First, it mentions a way to use2D-Hough.The proposed technique converts 3D data into 2D data by slicing. Second, it mentions a way to use 3D-Hough. The proposed technique requires appropriate parameters with which performances are well controlled. Finally, it mentions another way to use3D-Hough in the voxel space. The voxel space expresses more precise surface with cumulative 3D data, and makes it possible to detect the planes more accurately. This paper solves to detect the planes in those methods, detect intersection point formed by three planes in real time, and detect more than ten planes in reconstruction 3D space using 3D-Hough in voxel space.

Thesis