Moving Target Tracking Algorithm Based on Scale Invariant Optical Flow Method

Aiming at the dynamic changes of objects and movement scene when the objects are tracked, this thesis proposes the target track algorithm which combines SIFT and optical flow, and improves template updating strategy. The standard SIFT method extracts characteristic values which is scale invariant and rotation invariant. In the matching process, KD tree (k-dimensional tree) and the BBF (Best Bin First) algorithm can be used in searching. In the process of updating, in order to solve the phenomenon about frequently updated template to the template drift, the idea of updating the template when the tracking number changes to a certain extent is put forward. The optical flow field is used in locating the feature point of positive weight in the trace template only. The result show that compared with the optical flow, the algorithm in this thesis is less complicated in estimating the moving time of the feature points and can achieve better tracking. The algorithm in this thesis is easy to achieve and it has good robustness to the movement of significant.