Research on visual recognition of clothes in collaboration with actions

随着家庭机器人和康复机器人在老龄化社会中发挥积极作用，机器人自动处理衣服等日常必需品变得更加重要。然而，由于服装的变形性很大，很难从视觉上识别其形状并获得自动处理所需的信息。在本研究中，我们通过预测动作引起的服装形状，并以模型驱动的方式分析观察到的3D数据，开发了一种与动作配合的服装视觉识别方法。

As home and rehabilitation robots are expected to take an active role in an aged society, it becomes more important for robots to automatically handle daily necessaries including clothes. However, due to large deformability of the clothes, it is very difficult to visually recognize its shape and to obtain necessary information for automatic handling. In this research, we have developed a method of visual recognition of clothing in cooperation with actions by predicting clothing shape caused by the action and analyzing observed 3D data in a model-driven way.