反想光学环境-冈山大学.PPTVIP

最近の研究紹介仮想光学環境 岡山大学　工学部　情報工学科 （大学院自然科学研究科　計算機科学講座） 尺長　研究室 実光学環境 仮想光学環境 スクリーン物体の形状計測 スクリーン物体（表面）の法線 実サンプルによる反射特性の学習 補間による反射特性の推定 形状と反射特性を用いて生成した投影パターン 仮想光学環境（プロジェクタ１台の系） 仮想光学環境（プロジェクタ４台の系） * 実物体 実照明 観測者 反射特性 Projector PC スクリーン物体 仮想照明 観測者 反射パターン スクリーン物体上に，様々な反射特性?仮想照明を再現 人間が持つ多様な奥行き知覚機能を自然に利用できる Projector Camera Projector Screen object Shape is acquired by active stereovision ３原色（Ｒ，Ｇ，Ｂ）で３次元直交ベクトルを表現‘ 2d BRDF map recovered from one point （black: not sampled） サンプルを装置中央に配置 光源位置を制御しながら回転 固定カメラで反射特性を計測 θi θ’r Sample images Integration Interpolation Reconstruction of 2d BRDF Soup bowl (lacquer ware) Specular reflection is also simulated Location of viewer = location of LCP Light direction ： moving This slide shows an ordinary exhibition of 3d object. In the environment, a human client see a real object under a real lighting condition. This movie shows another example. In the case, a Miroku head is used as a screen object. The left movie shows an overview of the virtual photometric environment. This system consists of four projectors and four cameras. The right movie shows a screen object when imaginary photometric patterns are projected from four projectors. You are welcome to our demonstration in this afternoon, and enjoy yourself in the virtual photometric environment. Seeing is believing !! At first, the geometric data is sampled by a conventional range finder system, which consists of a projector and a camera. After the calibration, 3d depth data is recovered pixel by pixel from a set of images of projected stripes. Once a 3d depth data is provided, a surface normal is estimated by principal component analysis in a local area on the surface. This slide shows color representations of surface normals of two objects, where red, green and blue color correspond to orthogonal 3d directions in 3d space. After the surface normal recovery, we can utilize a geometric surface model of the object for the further processings. A simple measureme