In the case of poor Global Positioning System (GPS) signals in recent years, optical flow is used for UAV positioning and navigation, whose accuracy is not sufficient.To solve the problem, an improved optical flow tracking algorithm is proposed.Oriented FAST and Rotated BRIEF (ORB) algorithm is used to extract feature points and send them into Lucas-Kanade (LK) pyramid.The feature points of the next frame are estimated by using bidirectional tracking, and the feature points of the current frame are verified and supplemented at the same time.After the elimination of error points, the coordinates of the point set composed of the preliminary credible points and the supplementary credible points are obtained to estimate the UAV pose.After extracting feature points from a variety of environments, real machine experiments are conducted.The experimental results show that the improved optical flow tracking and predictive control method has better positioning and tracking effects and better accuracy in comprehensive environments, which can be better applied in more complicated environments such as military use.