This research derives the ground target tracking algorithm of a fixed-wing UAV equipped with a gimballed camera. The gimballed camera is controlled to track a moving target independent of the UAV motion, and this is achieved by employing a UKF to estimate the position and velocity of the target. The input of the UKF is the bounding box of the target, which is generated by YOLO, a deep neural network algorithm used to detect the target. The estimated state is essential for image tracking control, and it is achieved by the IBVS algorithm and a gimbal controller to lock the target when the UAV is in flight. Simulations are conducted through to verify the efficacy of the developed controller.