Object Tracking in UAV Videos by Multi-Feature Correlation filters with Saliency Proposals

The purpose of object tracking is to locate a given target in image sequence, such as people and vehicles. In recent years, with the development UAV technology, videos has engaged many scholars. It been widely used traffic control, water quality inspection, wildlife census other fields. However, low resolution, scale change, occlusion challenges have restricting tracker. To solve aforementioned problems, we put forward multi-feature correlation filters saliency proposals. First, use histogram oriented gradient (HOG) features, gray (I) features color names (CN) heighten representation information target, so that our algorithm can accurately small targets. Then, introduce proposals reposition occluded target. Finally, dynamic update weights instead fixed mitigate adverse effects caused by template degradation. Experiments demonstrate tracker achieved satisfactory accuracy AUC scores reached 0.462, 0.417, 0.425 on UAV123@10FPS, UAV20L, UAVDT datasets respectively.