Core Concepts
提案されたSSF-Netは、HSオブジェクトトラッキングにおいて、HSとRGBのモダリティ情報を統合し、追跡性能を効果的に向上させます。
Abstract
Hyperspectral video (HSV) offers valuable spatial, spectral, and temporal information simultaneously, making it highly suitable for handling challenges such as background clutter and visual similarity in object tracking. Existing methods primarily focus on band regrouping and rely on RGB trackers for feature extraction, resulting in limited exploration of spectral information and difficulties in achieving complementary representations of object features. The proposed SSF-Net introduces a spatial-spectral fusion network with spectral angle awareness (SAAM) to address the issue of insufficient spectral feature extraction in existing networks. It includes a spatial-spectral feature backbone (S2FB) designed to capture joint representation of texture and spectrum. Additionally, a spectral attention fusion module (SAFM) is presented to incorporate visual information into the HS spectral context for robust representation. A novel spectral angle awareness module (SAAM) investigates region-level spectral similarity between template and search images during prediction. Furthermore, a weighted prediction method combines HS and RGB predicted motions for robust tracking results.
Stats
提案されたSSF-Netは、AUCスコアが0.680であり、DP 20が0.939である。
BAE-NetのAUCスコアは0.606であり、DP 20は0.878である。
SEE-NetのAUCスコアは0.654であり、DP 20は0.907である。
SPIRITのAUCスコアは0.679であり、DP 20は0.925である。
SST-NetのAUCスコアは0.623であり、DP 20は0.917である。
Quotes
"Extensive experiments on the HOTC dataset demonstrate the effectiveness of the proposed SSF-Net, compared with state-of-the-art trackers."
"The proposed SSF-Net provides better performance compared with advanced RGB trackers."