This paper addresses cooperative estimation of 3D
target motion for visual sensor networks. In one of our previous
works, we already presented a cooperative estimation algorithm
called networked visual motion observers. In this paper, we
first clarify averaging accuracy attained by the networked
estimation mechanism. Then, we analyze convergence speed of
the estimates and clarify a relation of the speed to the graph
structures. Moreover, we also reveal a tracking performance of
the estimates to target objects motion and derive a connection
between the tracking performance and a visual feedback gain in
the algorithm. Finally the effectiveness of the present estimation
algorithm is demonstrated through experiments.