Search papers, labs, and topics across Lattice.
This paper introduces the Manifold-Guided Spectral Prompt Network (MSP-Net) for hyperspectral object tracking, which overcomes limitations of traditional methods that treat hyperspectral images as simple multi-channel RGB extensions. By employing graph-driven manifold routing to reconstruct band relationships and create adaptive spectral groupings, MSP-Net enhances target feature extraction while minimizing background interference. Experimental results on the HOT2020 and HOT2023 datasets show that MSP-Net achieves AUC and Precision scores exceeding 0.80 and 0.96, respectively, demonstrating its robustness across heterogeneous sensors and challenging conditions.
MSP-Net adapts in real-time to changing target appearances and scenes, achieving unprecedented tracking precision in complex environments.
Hyperspectral object tracking leverages abundant spectral information to provide unique advantages for target discrimination in complex scenes. However, existing methods typically treat hyperspectral images as multi-channel extensions of RGB images, performing feature fusion in fixed band order. This approach leads to models dependent on specific sensor configurations while neglecting manifold relationships between bands, making generalization to heterogeneous sensors difficult. Moreover, the discriminative contribution of bands dynamically changes with target attributes and scene variations, further limiting the representational capacity of static fusion strategies. To address this, we propose the Manifold-Guided Spectral Prompt Network (MSP-Net). This network first reconstructs band relationships and forms adaptive spectral grouping through graph-driven manifold routing, then jointly integrates grouped spectral statistics with template appearance to construct target-related dynamic conditional prompts, enhancing target features while suppressing background interference. Furthermore, as tracking progresses, spectral conditions continuously evolve based on intermediate target representations, enabling target prompts to adapt in real-time to appearance and scene changes. Meanwhile, reliable historical states are used to constrain target localization and scale fluctuations, significantly improving temporal stability in cross-sensor tracking. Experiments on HOT2020 and HOT2023 demonstrate that MSP-Net achieves AUC and Precision exceeding 0.80 and 0.96, respectively, exhibiting exceptional robustness under heterogeneous sensors, target deformation, and complex background conditions. The code will be released at https://github.com/GGML668897/MSP-Net.
