Linear Rotate Subspaee based Visual Tracking Methods with Application to UAV Stand-off Target Tracking

摘要

Using airborne cameras and guiding UAVs to track moving ground targets is an important aspect of the applications of UAVs. In this paper, a visual target tracking algorithm for airborne cameras is designed based on such application scenarios. During the dynamic changes in the relative positions of the target and UAV, the algorithm continuously uses the current target image to simulate and generate the possible pseudo-image of the target after rotation, which is a linear rotate subspace, and generate pseudo-filters. In the process of tracking it determines whether to update the target image and the corresponding filter by tracking performance, and thus improves the stability of visual tracking. The proposed algorithm is tested on the data set from the UAV tracking perspective, and results show that the tracking accuracy and robustness of the proposed algorithm are higher than those of the current mainstream tracking algorithms.