This paper investigates vision-based cooperative estimation of a 3D targetobject pose for visual sensor networks. In our previous works, we presented anestimation mechanism called networked visual motion observer achievingaveraging of local pose estimates in real time. This paper extends themechanism so that it works even in the presence of cameras not viewing thetarget due to the limited view angles and obstructions in order to fully takeadvantage of the networked vision system. Then, we analyze the averagingperformance attained by the proposed mechanism and clarify a relation betweenthe feedback gains in the algorithm and the performance. Finally, wedemonstrate the effectiveness of the algorithm through simulation.
展开▼