Detection and Tracking Method for Maneuvering Targets Using Improved Gradient-based Optical Flow Field

In order to address the problem of larger error on detecting and tracking the maneuvering targets by using the common approach on the optical flow field, this paper presented a novel method based on the improved gradient-based optical flow field. According to the Horn-Schunck algorithm, the weight of optical flow was explored on the gradient threshold of sub-image to boost the detecting performance. With respect to the targets tracking, the unscented Kalman filtering gain was optimized. Further, the adjustment coefficient of unscented transformation was adaptively derived based on the trace of state variance matrix of the Sigma sampling points. The experiments indicated that the proposed method can effectively detect and track the maneuvering targets with the accurate estimation.