| 青山 尚正 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 清川 清, | 舩冨 卓哉, | 藤村 友貴, | 田中 賢一郎（客員准教授） |
|
title: Depth Separation Using Full-Waveform LiDAR
abstract: LiDAR is a technology that can reconstruct the distance to an object as a point cloud. However, when there are multiple depths including transparent objects, it is difficult to reconstruct the point cloud correctly. In this research, we propose a method for depth separation by using full-waveform LiDAR, which can record not only the distance to an object but also the temporal change in reflectance, to identify the reflection characteristics of objects in scenes with multiple depths. language of the presentation: Japanese 発表題目: 全波形LiDARを用いた奥行き分離 発表概要: iDARは物体までの距離を点群として復元できる技術である。しかし透明物体などを含む複数の奥行きがある場合、正しく点群の復元をすることが難しい。本研究では、物体までの距離だけでなく反射率の時間変化を記録できる全波形LiDARを用いることで、複数の奥行きのあるシーンでの物体の反射特性を特定し、奥行き分離の手法を提案する。 | |||||||
| 生坂 優太 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 清川 清, | 舩冨 卓哉, | 藤村 友貴, | 田中 賢一郎（客員准教授） |
|
title: Estimating reflectance and shape using time-resolved radiosity
abstract: In a scene occurred inter-reflection, a reflectance is estimated by inverse radiosity with a shape. However, a wrong shape leads to an error of reflectance. In addition, there is no clue how to modify the shape. In this study, we proposed time-resolved radiosity, which incorporates an analysis of time variation in picoseconds into radiosity. In estimating the reflectance, we confirmed that the wrong shape causes inconsistencies in picosecond scale observations by single photon measurement. In addition, the shape is modified by optimization using the inconsistencies. language of the presentation: Japanese 発表題目: 時間分解ラジオシティを用いた反射率と形状の推定 発表概要: 拡散相互反射が起きるシーンにおいて，その形状が与えられると，ラジオシティの逆問題として反射率を推定することができる．しかし，与えられた形状が誤っていた場合，従来のラジオシティでは誤った反射率が推定され，加えて形状を修正するための手掛かりがなかった．本研究では，ラジオシティにピコ秒単位での時間変化の解析を取り入れた時間分解ラジオシティを提案する．反射率を推定する際に，誤った形状ではピコ秒単位での観測に不整合をきたすことを明らかにし，これを単一光子計測による実験で確かめた．さらに，その不整合を用いた最適化によって形状を修正する． | |||||||
| 知念 響紀 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 清川 清, | 舩冨 卓哉, | 藤村 友貴, | 田中 賢一郎（客員准教授） |
|
title: Shape Estimation of Specular Object by Highly Time-Resolved Measurement with Spatial Coordinates Embedded in Time Axis
abstract: The purpose of this research is to measure the shape of objects with dominant specular reflection. Previously, a structured illumination method to measure the shape using multiple displays has been proposed based on the principle of triangulation, which is one of the main principles of 3D shape measurement. In this paper, we propose a shape measurement method for specular objects that combines structured illumination with picosecond time-of-flight measurement, which is another major measurement principle. By embedding the two-dimensional spatial coordinates on the diffuser into the picosecond time axis, we can measure the shape of the specular object in a short time. The feasibility of the proposed method is confirmed by actual experiments on a specular object. language of the presentation: Japanese 発表題目: 空間座標を時間軸に埋め込んだ高時間分解計測による鏡面物体の形状推定 発表概要: 本研究では，鏡面反射が支配的な物体の形状計測を目的とする．従来，3次元形状計測の主要な原理の1つである三角測量の原理に基づき，複数のディスプレイを用いて形状計測を行う構造化照明法が提案されている．本稿では，もう1つの主要な計測原理であるピコ秒単位のTime-of-Flight計測に構造化照明を組み合わせた鏡面物体の形状計測手法を提案する．ピコ秒単位の時間軸に拡散板上の2次元空間座標を埋め込むことで，短時間での形状計測が可能となる．実際に鏡面物体を対象とした実験により提案手法の実現性を確認した． | |||||||
| 赤塚 大地 | M, 2回目発表 | ヒューマンロボティクス | 和田 隆広, | 清川 清, | 趙 崇貴, | 佐藤 勇起, | 高松 淳 |
|
title: Development of a Robot Teaching System Using Extraction of Semantic Motion Constraints in Human Motion
abstract: The intermediate representations between human and robot motions may be strongly related to the reusability of the motions taught to the robot. Therefore, we use the constraints that humans impose on their own motions as intermediate representations to teach motions to robots. We can move tools freely, but they can play their roles only after we move them following constraints called Semantic Constraints considering the tools’ semantics. We can construct a robot teaching system that has high reusability with Semantic Constraints because they have minimal information required on performing the tasks. In our research, we focus on the pouring task and use Semantic Hinge that has object’s rotational movement information as Semantic Constraints on pouring task. The proposed method consists of the finding of the semantic hinge from the depth image and reconstruction of the hand behavior to handle the object with the extracted hinge. By using the depth images as input, we can find the semantic hinge without deciding the object coordinate system. Finally, we plan to move the robot arm using this teaching system and evaluate how much liquid can be poured. language of the presentation: Japanese | |||||||