Multi-ocular multi-object tracking for visually servoed disassembly operations

This paper addresses a main subtask of vision-based disassembly, namely the visual tracking of objects that are to be disassembled. The objective of this paper is to present and evaluate a general solution: A tracking algorithm that tracks the pose of either one ore more objects using either one or more calibrated cameras. The first degree of freedom, i.e. the number of tracked objects, enables one to migrate from hand-eye-configurations, where usually only a single object is to be tracked, to more complex robot-camera-configurations, where disassembly tool and workpiece are to be observed. The second degree of freedom, i.e. the number of observing cameras, can be exploited to reduce the uncertainty of an estimated pose in each pose component. The performance of the tracker is evaluated for the case of tracking one or two polyhedral objects using one or two calibrated cameras.