Vision-Based Motion Capture for Human Support Robot in Action

摘要

We have proposed a new vision-based motion capture method. The robot which is implemented this method measure subjects while the robot is moving. The goal of our work is to apply our achievements to human support robots, such as pet robots and communication robots in the future. We named the robot as "MoCap Robot". We have designed MoCap Robot to be operated in the following manner: (1) detect markers attached to the subject by image processing techniques, (2) move around to detect markers, (3) automatically estimate robot's position by using dead reckoning and image processing techniques, (4) estimate the actual position of the markers attached to the subject by comparing robot's position with markers' positions, (5) estimate the subject's motion by comparing the markers' positions and a human model, (6) output subject's motion. Our next task is to develop a prototype of MoCap Robot and evaluate the effectiveness of our method.