Outlier elimination method for robust visual servo control in complex environment

When track the recognized rigid target object for the Position Based Visual Servo(PBVS) system on the generic complex environment, existing optical flow based tracking approach has possibility to occur outlier due to its aperture problem which is occurred from drifting tracking points. In this paper, we propose the method how to remove the drifting tracking points using Kalman filter. We compare the measurement result with estimation result from Kalman filter and then our proposed approach detects the drifting tracking points. Then detected outliers are removed. Finally, we calculate the object pose result without detected outlier. The result of system which is removed the occurred outlier shows that the system performance was good rather than before remove the outlier. Then we can realize the improved differences between previous system with outlier and current system without outlier.