We are possible to estimate shape and posture from an image in case of it is a rigid body and simple shape, but generally, in case of joint object which is a combination of many rigid body, I prepare a geometric model of ass intend for object beforehand, and technique to recognize is used by model matching with an image. A posture estimate for the whole human body is necessary in order to build the virtual world, and a posture estimates of a finger become important in order to realize interface of a computer and remote control of a machine. In addition, it is necessary to estimate a position and posture with resolving power of degree that a robot can reproduce when I regarded construction of a communication system with humans as a robot. In this study, I suggest technique to estimate a control parameter from an image by a stemma camera in order to reproduce of humans movement with a robot hand. So we estimate finger and arm shape what consist of many rigid body and becomes complicated shape, that computational complexity of model matching increases. So we attempt to cut computational complexity with resolving power of degree that a limitation condition of joint angle operation range of humans and a robot hand can reproduce.
展开▼
