| 岸本 泰海 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 加藤 博一, | 舩冨 卓哉, | 久保 尋之, | 田中 賢一郎 |
|
title: Predicting Appearance of Metals using Microgeometry
abstract: In the design process of metal parts, the three-dimensional shape after grinding can be predicted by numerical control machining simulation. However, it is difficult to determine the visual defects of scratches, streaks, and irregularities based on the three-dimensional shape alone. Microfacet-based BRDF is used to render metals, but spatial aliasing occurs when materials that exhibit high-frequency spatial variation are used. In this study, I propose a method to reduce aliasing while maintaining the appearance of machining defects while maintaining the appearance of defects by determining the normal distribution function (NDF) for each surface patch. In this presentation, I show the results of calculating the NDF of a local region and future issues. language of the presentation: Japanese 発表題目: マイクロジオメトリーを用いた金属の外観予測 発表概要: 金属部品の設計工程では，削った後の3次元形状をNCフライスシミュレーションによって予測することができる．しかし，3次元形状だけでは，傷・筋・ムラの見た目の不良を判断するが困難である．金属のレンダリングでは，マイクロファセットベースのBRDFが用いられるが，高周波数の空間変動を示す材料を用いると，空間的なエイリアシングが発生する．本研究では，表面パッチごとに法線分布関数(NDF)を求めることで，加工不良の見た目を維持しつつエイリアシングを低減する手法を提案する．本発表では，局所領域のNDFを計算した結果と今後の課題について紹介する． | |||||||
| 辻 茉佑香 | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 加藤 博一, | 舩冨 卓哉, | 久保 尋之, | 田中 賢一郎 |
|
title: The Touch Sensing Method with Gated Image Using Projector-Camera Synchronization
abstract: We address novel touch-input sensing for the projected screen with a camera image. To realize this touch-input sensing, it is necessary to obtain whether the finger is in touch with the screen surface, and if so, where on the screen surface it is in contact. However a single-camera image can not perform both touch detection and coordinate detection, so the mainstream method is to track the 3D finger position by using such as multiple cameras or ToF cameras. In this research, we capture a limited area slightly above the screen-surface with the gating technique to determine whether the finger touch with the screen surface or not. Our method realized both touch detection and coordinate detection from a single camera image. We also achieved sufficient touch-input sensing with a very simple classical image processing technique since we capture only necessary objects in the scene for sensing. language of the presentation: Japanese 発表題目: プロジェクタ-カメラの同期によるgated imageを用いたプロジェクタ映像とのタッチセンシング手法の開発 発表概要: 本発表では、映像を投影したスクリーンに対する新しいタッチ入力センシング手法について説明します． タッチ入力センシングを実現するためには，指がスクリーン面に接触しているかどうか，接触している場合はスクリーン面のどこに接触しているかを取得する必要があります． しかし，単一のカメラ画像ではタッチ検出と座標検出の両方を行うことができないため，複数のカメラやToFカメラなどを用いて3次元的に指の位置を追跡する方法が主流となっています． 本研究では，画面表面からわずか上の限られた領域をgated imagingを導入して撮像し，指が画面表面に接触しているかどうかを判定します． 本研究では，1台のカメラ画像からタッチ検出と座標検出の両方を実現しました．また，シーン内の必要な物体のみを撮像することで，非常にシンプルな画像処理技術で十分なタッチ入力センシングを実現しています． | |||||||
| TAO XIANGKUN | M, 2回目発表 | 光メディアインタフェース | 向川 康博, | 加藤 博一, | 舩冨 卓哉, | 久保 尋之, | 田中 賢一郎 |
|
title: Foreign object detection through scattering medium using linear sensor/light source and dedicated deblur algorithm
abstract: Foreign object detection is an important step in a food production line. Existing methods mainly address metal detections using technologies such as the X-ray and electromagnetic induction. However, detecting non-metal foreign objects such as animal bone, insect and plastic fragment is still a challenging task. In this study, we propose a novel method using optical and computer vision techniques to tackle this problem. We built an apparatus which is analogous to the linear sensor/line light source combination used in actual food production line. Using the combination of line sensor/light source, the blur introduced by light scattering can be suppressed. Finally, we aim to farther suppress the scattering blur by employing a dedicated deblur algorithm. language of the presentation: English | |||||||
| 石倉 智貴 | M, 2回目発表 | ロボティクス | 小笠原 司, | 加藤 博一, | 高松 淳, | 趙 崇貴, | 丁 明（客員准教授） |
|
title:
Toward a Touch Care Robot: Generation of Gentle Stroke and Touch Using a Human-Imitation Hand
abstract: Touch care provides positive effects such as mitigation of stress and pain. However, from the shortage of therapists and care workers caused by the aging population, patients cannot receive enough care. To solve the problem, touch care provided by a robot has gained popularity. In this research, we focus on gentle stroking and touch performed by a robot that evokes the same emotions as if it were performed by a human. We propose a robot system to gently stroke the back of a human using a human-imitation hand. To evaluate the emotional effects, we compare the combination of two agents (robot and human) and two stroke speeds (3 cm/s and 30 cm/s). The results of subjective and physiological evaluations show a common effect in the speeds, which suggests that the robot stroke can evoke the emotional responses in the same way a human stroke does. With the purpose of correcting the gentle touch performed by therapists, we propose a quantitative measurement system for touch based on the Frustrated Total Internal Reflection (FTIR) method. The system measures the area and center of gravity of the touch surface. These two pieces of information are expected to be the basis for the generation of gentle touch using a robot. language of the presentation: Japanese | |||||||
| 酒巻 有吾 | M, 2回目発表 | ネットワークシステム学 | 岡田 実, | 林 優一, | 東野 武史, | DUONG QUANG THANG |
|
title: Load-Independent Constant Voltage / Current Design for 2x2 MIMO-IPT
abstract: Inductive Power Transfer(IPT) is a wireless power transfer technique, and various electronic devices, such as electric vehicles and electric toothbrushes, employ it. In practice, these applications require constant voltage or constant current characteristics. Furthermore, Multiple IPTs needs to be operated simultaneously in close locations to accommodate many devices. There are several types of research on multiple-input multiple-output (MIMO) IPT to fulfill the requirements. The previous study proposed a MIMO IPT system composed of two constant voltage IPT pairs operating at the same frequency in close vicinity. However, it is sensitive to the variation in the distance between transmitter and receiver. To solve this problem, we propose a robust compensation circuit for a 2-by-2 MIMO IPT with constant current or voltage characteristics. In this presentation, I will present the proposed compensation circuit design procedure and show that the proposed scheme is robust to the distance variation. language of the presentation: Japanese | ||||||
| 佐山 大智 | M, 2回目発表 | ネットワークシステム学 | 岡田 実, | 林 優一, | 東野 武史, | DUONG QUANG THANG |
|
title: Leaked Magnetic Field Reduction for MIMO
Wireless Power Transmission by Source Voltage Phase Control
abstract:Wireless power transmission (WPT) can be applied to various devices from wearable device to EV(Electric Vehicle) and has been expected further development. WPT system has been recently studied in order to improve the transferable power by delivering multiple power streams simultaneously. In MIMO system, the power streams influence each other because of the effect of cross-coupling, and therefore it is difficult to provide constant voltage to the loads when the load resistances and source voltages vary during the charging process. In order to solve this problem, in the previous study, the constant voltage and isolation scheme of two-stream WPT have been proposed by adding the compensation circuit. When the two streams are isolated, adjusting the voltage phase of one stream might not affect output voltage of the other so much. This open a new possibility to reduce the magnetic field leaked to surrounding space by adjusting the phase difference of the sources in destructive manner. Of course, in such a case, the power transfer efficiency may deteriorate. In this work, we will optimize the phase difference of the sources to significantly suppress the leaked magnetic field while maintaining a higher efficiency. language of the presentation:Japanese | ||||||
| 本多 右京 | D, 中間発表 | 自然言語処理学 | 渡辺 太郎, | 中村 哲, | 進藤 裕之 | |
|
title: General-Purpose Image Captioning with Noisy Resources
abstract: Image captioning is a task to generate the descriptions of a given image in natural languages. The aspects and domains of the captions can be diverse depending on the interests of the users. To train a caption generator that covers the diversity, however, exhaustive manual annotations are required. Our goal is to generate diverse captions from limited resources, alleviating the enormous annotation costs. Currently, we are working on unsupervised image captioning for this purpose. In this setting, one has to generate captions with images and their pseudo-captions, without any manually paired images and captions. Previous work tackled this by closing the gap between the images and pseudo-captions in their feature space. The problem of this approach is that it cannot be sensitive to the noises contained in the pseudo-captions. In our work, we focus on removing the fractions of the pseudo-captions that are irrelevant to a given image. We introduce an effective and efficient filtering mechanism and experimental results show that our work outperforms the previous work, with much less time for training. language of the presentation: Japanese | ||||||