Detection and Pursuit of Independently Moving Agents by Means of Vision

This research investigates the usefulness of visual flow field information for the robotic navigation task of independent motion detection. The task is considered non-trivial for the case of moving observers. Multiple modules are proposed to attempt to solve the task for a mobile robotic system which is constantly in motion. First, egomotion parameters are computed on the basis of the flow field computed from image sequences. Second, independent motion is detected after constantly monitoring the flow field and noticing deviation from the expected flow. Third, the target is fixated (gaze control, tracking) by using this attentive information for an automatic initialization of an active contour which locks onto the desired agent's apparent contours for subsequent pursuit. Finally, the object is approached if necessary. The research tries to foster the discussion on cooperative processing of the modules in order to achieve more complex behaviors in a robust way. In addition, the modules a..