1: \begin{abstract}

2: An image-based control strategy along with estimation of target motion

3: is developed to track dynamic targets without motion constraints.

4: To the best of our knowledge, this is the first work that utilizes

5: a bounding box as image features for tracking control and estimation

6: of dynamic target without motion constraint. The features generated

7: from a You-Only-Look-Once (YOLO) deep neural network can relax the

8: assumption of continuous availability of the feature points in most

9: literature and minimize the gap for applications. The challenges are

10: that the motion pattern of the target is unknown and modeling its

11: dynamics is infeasible. To resolve these issues, the dynamics of the

12: target is modeled by a constant-velocity model and is employed as

13: a process model in the Unscented Kalman Filter (UKF), but process

14: noise is uncertain and sensitive to system instability. To ensure

15: convergence of the estimate error, the noise covariance matrix is

16: estimated according to history data within a moving window. The estimated

17: motion from the UKF is implemented as a feedforward term in the developed

18: controller, so that tracking performance is enhanced. Simulations

19: are demonstrated to verify the efficacy of the developed estimator

20: and controller.

21: \end{abstract}

22: