This paper presents two visual trackers from the different paradigms oflearning and registration based tracking and evaluates their application inimage based visual servoing. They can track object motion with four degrees offreedom (DoF) which, as we will show here, is sufficient for many finemanipulation tasks. One of these trackers is a newly developed learning basedtracker that relies on learning discriminative correlation filters while theother is a refinement of a recent 8 DoF RANSAC based tracker adapted with a newappearance model for tracking 4 DoF motion. Both trackers are shown to provide superior performance to several state ofthe art trackers on an existing dataset for manipulation tasks. Further, a newdataset with challenging sequences for fine manipulation tasks captured fromrobot mounted eye-in-hand (EIH) cameras is also presented. These sequences havea variety of challenges encountered during real tasks including jittery cameramovement, motion blur, drastic scale changes and partial occlusions.Quantitative and qualitative results on these sequences are used to show thatthese two trackers are robust to failures while providing high precision thatmakes them suitable for such fine manipulation tasks.
展开▼
