抄録
This research aims to develop a robot which can shoot a mobile target following a given camera work describing a space-time robot trajectory in the target's local coordinates. The robot calculates the target position using a vision and laser-based person detector and on-line modifies the trajectory to which a path planner is applied to generate a robot motion command. Since the target moves continuously moves in the image due to the errors in target detection and motion control, we post-process the captured image sequence so that the target is always at the center of image. For this post-process, we extract Harris features in the target region and calculate optical flow for the features to estimate the target motion in the image, which is then used for horizontally translate the image. We validated the proposed method by shooting experiments for several target motions.
