Lanke FU

Master Thesis
Supervisors: Nikolay Kobyshev

Confidence-based Multiview Object Detection and Tracking

6 degree of freedom pose tracking of rigid object is crucial in completing dynamic tasks in robotics. One such example is dynamic pick and placement of object to an from a conveyor belt. To achieve this task, an accurate pose of the object is needed to fetch the appropriate grasps. Moreover, the dynamic nature of this problem makes time constraints strict, favoring an iterative tracking approach over a detection approach. In this work, we explore different existing tracking appraoaches and ways of making object tracking more robust.