Abstract | ||
---|---|---|
Visual tracking has been an active and complicated research area in computer vision for recent decades. In the area of unmanned aerial vehicle (UAV) application, establishing a robust tracking model is still a challenge. The kernelized correlation filter (KCF) is one of the state-of-the-art object trackers. However, it could not reasonably handle the severe special situations in UAV application during tracking process, especially when targets undergo significant appearance changes due to camera shaking or deformation. In this paper, we proposed a new compounded feature to track the object by combining saliency feature and color features for the conspicuousness of the objects in the videos captured by UAVs. Considering the speed of real-time application, we use a spectrum-based saliency detection method - quaternion type-II DCT image signatures. In addition, severe drifting can be detected and adjusted by the relocation mechanism. Extensive experiments on the UAV tracking sequences show that the proposed method significantly improves KCF, and achieves better performance than other state-of-the-art trackers. |
Year | Venue | Field |
---|---|---|
2018 | PRCV | BitTorrent tracker,Computer vision,Correlation filter,Salience (neuroscience),Computer science,Quaternion,Discrete cosine transform,Eye tracking,Video tracking,Artificial intelligence |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
16 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Jinyu Yang | 1 | 32 | 7.81 |
Wenrui Ding | 2 | 10 | 3.11 |
Chunlei Liu | 3 | 7 | 4.82 |
Zechen Ha | 4 | 0 | 0.34 |