FingerVision for Tactile Behaviors , Manipulation , and Haptic Feedback Teleoperation
FingerVision for Tactile Behaviors , Manipulation , and Haptic Feedback Teleoperation
复制标题
用于触觉行为、操纵和触觉反馈远程操作的 FingerVision
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Akihiko Yamaguchi
中科院分区：
文献类型：
--
作者：
Akihiko Yamaguchi
本文介绍了一种基于视觉的触觉传感器FingerVision及其应用，以吸引触觉研究人员使用FingerVision。FingerVision是由山口和阿特金森在（1）中提出的。其概念是将接近视觉和触觉传感相结合。与其他基于视觉的触觉传感器，如（2） - （15）不同，FingerVision内部的摄像头可以透过皮肤看到传感器外部，这增加了传感模式。FingerVision的结构简单，由弹性透明皮肤、框架和摄像头组成。标记附着在皮肤表面用于检测皮肤变形。FingerVision的特点总结如下：
（1）多模式：它可以感知力的分布、滑动、物体姿态、纹理以及从接近视觉（对附近物体的计算机视觉）获得的其他信息。需要说明的是：
（1.1）无论物体的反作用力如何，都能检测到滑动。即使物体太轻以至于无法感知力（例如折纸鹤）时，它也能感知到滑动。
（1.2）摄像头可以在碰撞前感知到物体。利用这一特性，我们可以制造出能感知附近人类的安全交互机器人。
（2）易于制造：由于其结构简单，制造容易。
（3）成本低：最昂贵的部件是摄像头。其他部件都很便宜。
（4）坚固耐用：外力作用在皮肤和框架上，不会影响到摄像头。因此它很坚固。即使皮肤损坏，更换也很便宜。
（5）传感元件（摄像头）不必覆盖整个表面：使用广角镜头（鱼眼镜头），我们可以使摄像头的分布稀疏。
（6）处理软件可以通用，这有利于高效开发。
（7）传感器参数可调节：我们可以调节力的动态范围（皮肤的硬度和厚度）
This paper introduces a vision-based tactile sensor FingerVision and its applications in order to attract haptics researchers to use FingerVision. FingerVision was proposed in (1) by Yamaguchi and Atkeson. Its concept is combining proximity vision and tactile sensing. Unlike other visionbased tactile sensors such as (2)∼(15), the cameras inside FingerVision can see the outside of the sensor through the skin, which increases the sensing modality. The structure of FingerVision is simple, consisting of elastic and transparent skin, frame, and cameras. Markers are attached on the skin surface for detecting skin deformation. The features of FingerVision are summarized as follows: (1) Multimodal: It can sense force distribution, slip, object pose, texture, and other information obtained from proximity vision (computer vision for nearby objects). The remarks are that: (1.1) Slip can be detected regardless the reactive force from objects. It can sense slippage even when the object is too light to sense force (e.g. origami crane). (1.2) Cameras can sense objects before collision. With this feature, we can create safe interactive robots that are aware of nearby humans. (2) Easy to manufacture: Because of its simple structure, its fabrication is easy. (3) Low cost: The most expensive component is the camera. Other components are inexpensive. (4) Physically strong: External force applies to the skin and frame, and does not reach the camera. Thus it is physically strong. Even if the skin is damaged, replacing it is inexpensive. (5) The sensing elements (cameras) do not have to cover the whole surface: Using wide-angle lenses (fisheye lenses), we can make the camera allocations sparse. (6) Processing software could be common, which enables efficient development. (7) Sensor parameters are adjustable: We can adjust the dynamic range of force (hardness and thickness of the