Infrared Small Target Detection

This paper introduces SeRankDet, a deep learning model for infrared small target detection. SeRankDet leverages our novel Selective Rank-Aware Attention module for preserving salient responses, Large Selective Feature Fusion for dynamic feature integration, and Dilated Difference Convolution for improved target-background separation. Despite its lightweight design, SeRankDet delivers state-of-the-art performance.

Infrared small target detection poses unique challenges due to the scarcity of intrinsic target features and the abundance of similar …

A Sparse Differential Directionality prior (SDD) framework is proposed for infrared small target detection. SDD leverages directional characteristics to differentiate targets from background, applying mixed sparse constraints on differential directional images and continuity difference matrix derived from Tucker decomposition. Saliency coherence strategy further enhances target detectability during hierarchical decomposition.

We propose OSCAR, a one-stage cascade refinement network for single-frame infrared small target detection, along with a new benchmark consisting of the SIRST-V2 dataset, normalized contrast evaluation metric, and DeepInfrared toolkit.

This paper proposes a model-driven deep network that combines discriminative networks and conventional model-driven methods, utilizing a feature map cyclic shift scheme and bottom-up attentional modulation to highlight and preserve infrared small target features.

This study presents an open dataset for single-frame infrared small target detection and introduces an asymmetric contextual modulation module. This module utilizes a top-down global contextual feedback and a point-wise channel attention-based bottom-up modulation pathway to enhance the detection of infrared small targets.

This paper proposes a novel method for infrared small target detection in heterogeneous backgrounds. It employs an infrared patch-tensor (IPT) model and leverages both local and non-local priors. The target-background separation is modeled as a robust low-rank tensor recovery problem, enhanced with an entry-wise local-structure-adaptive and sparsity enhancing weight.