Infrared Small Target Detection
Overview
Infrared small target detection aims to localize dim objects that may occupy only a few pixels and resemble background clutter. Methods range from low-rank and sparse decomposition to CNN and transformer-style detectors that use multiscale features, frequency cues, physical priors, and, for video, motion across frames. Current work also explores point- or box-supervised training, pseudo-labels, and lightweight models for real-time use. A central challenge is improving detection and localization without increasing false alarms, especially across unfamiliar scenes, degraded imagery, and limited or noisy annotations.
Latest papers
43 papers7 papers in the last four weeks, up 75% on the four weeks before. 0.0% of all new papers.