.png){kind=icon}
.png){kind=icon}
Return
Efficient Particle Scale Space for Robust Tracking
.png){kind=icon}
PRE
AI
DOI：10.1109/LSP.2020.2973098.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Both siamese network and correlation filter (CF) based trackers have recently achieved superior performance in tracking scenarios with various challenging factors. For the challenging scale variations, most of these state-of-the-art trackers usually employ multiple patches with different bounding boxes to estimate the target size. However, these patches are fixedly generated by the hand-crafted bounding boxes in spatial domains, which may be suboptimal to cope with scale changes due to the lack of temporal scale information. In this letter, we tackle the problem of efficient scale estimation by presenting a generic scheme that allows the adaptive generation of bounding boxes in temporal domains and improves the tracking accuracy. Specifically, we introduce the novel particle scale space by refining the conventional particle filter and extend this space to many siamese and CF trackers for robust tracking. Extensive experiments are performed on the OTB2013, OTB50, OTB100 and UAVDT datasets. The proposed variants maintain at least almost identical frame-rates with baseline trackers and perform favorably against them, as well as other state-of-the-art trackers.
Keywords:
Object tracking
siamese network
scale estimation
correlation filter
particle filter
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
9.6
Papers:
1.1W
Citations:
1.7W