TAPM-Net: Trajectory-Aware Perturbation Modeling for Infrared Small Target Detection


Hongyang Xie (The University of Warwick), Hongyang He (The University of Warwick), Victor Sanchez (The University of Warwick)
The 35th British Machine Vision Conference
PDFPoster

Abstract

Infrared small target detection (ISTD) remains a long-standing challenge due to weak signal contrast, limited spatial extent, and cluttered backgrounds. Despite performance improvements from convolutional neural networks (CNNs) and Vision Transformers (ViTs), current models lack the mechanism to trace how small targets trigger directional, layer-wise perturbations in the feature space—an essential cue for distinguishing signal from structured noise in infrared scenes. To address this limitation, we propose a Trajectory-Aware Mamba Propagation Network (TAMP-Net) that explicitly models the spatial diffusion behavior of target-induced feature disturbances. The proposed architecture is built upon two novel components: a Perturbation-guided Path Module (PGM) and a Trajectory-Aware State Block (TASB). The TGM constructs perturbation energy fields from multi-level features and extracts gradient-following trajectories that reflect the directionality of local responses. The resulting sequences are fed into the TASB, which is a Mamba-based state-space unit that models dynamic propagation along each path while incorporating velocity-constrained diffusion and semantic-aligned feature fusion from both word- and sentence-level embeddings. Unlike existing attention-based methods, our approach enables anisotropic, context-sensitive state transitions along spatial trajectories, maintaining global coherence at low computational cost. Experiments on NUAA-SIRST and IRSTD-1K demonstrate that TAMP-Net achieves SOTA performance in infrared small target detection.

Citation

@inproceedings{Xie_2025_BMVC,
author    = {Hongyang Xie and Hongyang He and Victor Sanchez},
title     = {TAPM-Net: Trajectory-Aware Perturbation Modeling for Infrared Small Target Detection},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_709/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection