OCSCNet-Tracker: Hyperspectral Video Tracker Based on Octave Convolution and Spatial–Spectral Capsule Network

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(4), P. 693 - 693

Published: Feb. 18, 2025

In the field of hyperspectral video tracking (HVT), occclusion poses a challenging issue without satisfactory solution. To address this challenge, current study explores application capsule networks in HVT and proposes an approach based on octave convolution spatial–spectral network (OCSCNet). Specifically, module is designed to learn features from images by integrating spatial spectral information. Hence, unlike traditional convolution, which limited learning features, proposed strategy also focuses modeling features. The integrates information distinguish among underlying categories their similarity. enhances separability establishes relationships between different components targets at various scales. Finally, confidence threshold judgment utilizes initial adjacent frames for relocating lost target. Experiments conducted HOT2023 dataset illustrate that model outperforms state-of-the-art methods, achieving success rate 65.2% precision 89.3%. addition, extensive experimental results visualizations further demonstrate effectiveness interpretability OCSCNet.

Language: Английский

SiamTU: Hyperspectral video tracking against appearance changes using improved SiameGAT with adaptive template updating

Infrared Physics & Technology, Journal Year: 2025, Volume and Issue: unknown, P. 105811 - 105811

Published: March 1, 2025

Language: Английский