| 豊田 真行 | M, 2回目発表 | ヒューマンロボティクス | 和田 隆広, | 加藤 博一, | 趙 崇貴, | 佐藤 勇起, | 高松 淳 |
|
title: Development of End-effector Reproduce Contact Surface and Gentle Touching Motion of Humanitude Touch
The technique of touch in Humanitude has a positive effect on relieving stress and making people relaxing. If robots can reproduce the touch motion, it is expected to spread high-quality care to more people. We think it is especially important to develop an end-effector that directly touches to people and a planning motion to reproduce the touching motion of Humanitude. In this study, we focus on the contact surface during the touch motion, and develop an end-effector and a motion planning method that can reproduce contact surface same as an expert of Humanitude. By verifying whether the developed end-effector and touch motion can actually provide a relaxing mental state, we can confirm the importance of the contact surface in reproducing the touch of Humanitude. So far, we have conducted contact surface analysis using an FTIR touch panel. As a result, we were able to confirm that 1) the touches of experts reflect the know-how in the touch of Humanitude, which is to open the fingers and touch from the fingertips, and 2) the specific characteristics show that even non-experts can approach the touch motion of experts by touching based on this know-how. In addition, some required elements to implement the motion reproducing the know-how into an end-effector are inferred. These are a) mimicking the shape of the palm when touching a certain surface and b) giving appropriate articulation structures that allow pressure to be applied with each fingertip to perform the touch motion from the fingertips., language of the presentation: Japanese | |||||||
| 蔀 竜太 | 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, | 藤村 友貴 | |
|
title: Unsupervised Physics-based Neural Network for Estimating the Painted Pigments Properties from Spectral Images
abstract: Numerous mural paintings are found on tumulus and have rich historical value. However, due to natural deterioration, the pigments fade. It becomes extremely challenging to understand the pigmented area and the pigment used. In this research, we propose an unsupervised learning algorithm to estimate the thickness and ratio of pigments from spectral images, using a physics model based on deep learning. We input the measured spectrum of each pixel into a neural network to estimate the pigments thickness and ratio. Once the properties of the pigment are estimated, the spectrum can be reconstructed via a physics-based model. We optimize the neural network based on the error between the measured spectrum and the reconstructed spectrum. The effectiveness of this method is evaluated using simulation data and actual measurement data. language of the presentation: Japanese | |||||||
| 長瀬 康斗 | 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, | 藤村 友貴, | 田中 賢一郎（客員准教授） |
|
title: Passive Ranging Using Multichannel Far Infrared Measurements
abstract: In this research, we propose a method for estimating distance using far-infrared light emitted by an object have heat as a passive distance measurement technology that can be applied regardless of the texture of the object. Focusing on the difference in the attenuation rate of light in the atmosphere depending on the wavelength, the distance, temperature, and emissivity of the target are estimated at the same time by observing multi-wavelength in the far-infrared band. In the experiment, the effectiveness of the proposed method was verified by capturing the actual scene and estimating the distance. language of the presentation: Japanese 発表題目: 多波長遠赤外光観測による受動的な距離計測 発表概要: 本研究では，対象のテクスチャによらず適用可能な受動的距離計測技術として，熱を持つ物体が放射する遠赤外光を利用して距離推定を行う手法を提案する．大気における光の減衰率の波長による差異に着目し，遠赤外光帯域で多波長観測を行うことで，計測対象の距離，温度，放射率を同時に推定する．実験では，実際のシーンの撮影を行い，その距離を推定することで提案手法の有効性を検証した． | |||||||
| 福井 駿 | 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, | 藤村 友貴 | |
|
title: Soil moisture estimation via transient heat sensing and analysis
abstract: In disaster response where heavy machinery is required, there is a need for safe rapid soil assessment to acquire a wide range of ground strength information. Ground strength significantly depends on soil moisture and soil moisture depth. In this study, we propose a method to estimate soil moisture by analyzing the propagation of irradiated heat. Heat propagation depends on the soil moisture, this is observed by monitoring soil surface temperature over time by a far-infrared camera. By analyzing the observed data with a heat propagation model, soil moisture was estimated. The effectiveness of the proposed method was verified in a simulated ground environment. language of the presentation: Japanese | |||||||