Answers are grounded in this paper and cite their sources.
Infrared small target detection is highly sensitive to thermal radiation interference caused by platform thermal drift and sensor spatial nonuniformity. Direct detection methods may result in missed detections due to such interference, while sequentially correction-detection methods can yield satisfactory visual quality via correction yet inevitably amplify noise, which further interferes with small target detection and leads to false alarms. The main challenge lies in how to effectively leverage the correction results to enhance the detection accuracy while suppressing the thermal radiation interference. To address these issues and challenges, we propose a target-aware collaborative framework for correction and small target detection (TCF-Net). First, we design an asymmetric Thermal Radiation Correction Subnetwork, which is mainly composed of Lightweight Thermal Radiation Correction Blocks, effectively suppressing thermal interference while reducing the computational burden of the collaborative network. In order to enhance the feature interaction among correction and detection subnetwork, we propose a Cross-Task Target Aware Detail Enhancement Module, which leverages intermediate features from the correction branch to improve detail perception and the saliency of small targets. In addition, we introduce a Mamba Pixel-to-Block Fusion Module, which primarily consists of the Spiral and Direction-Aware Mamba Block (Spire-Mamba). This module retains global contextual awareness while capturing local spatial dependencies in the detection branch. Extensive experiments on multiple public infrared small target datasets demonstrate that TCF-Net significantly improves detection accuracy in complex thermal environments.
No takes yet. Share an insight, caveat, or question.
Shi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: