1 January 2025 in IEEE Geoscience and Remote Sensing Letters
16 December 2024 in IEEE Transactions on Geoscience and Remote Sensing
Infrared Ship Detection (IRSD) is crucial for numerous applications but faces challenges, such as small targets and complex backgrounds, resulting in misdetections and false alarms. In order to address these challenges, we propose the Scene Semantic Prior-Assisted Multi-Task Perception Network (SMPISD-MTPNet). This network employs multi-task perception: one task is to predict targets, and another focuses on scene perception to suppress false alarms caused by background interference. In order to highlight dim and small targets, we use the Scene Semantic Extractor (SSE) to guide the network by the features extracted based on expert knowledge and the Gradient-based Module to emphasize the small targets. We apply data augmentation to the networks and employ a training trick called Soft Fine-tuning to improve the network’s generalization and suppress the distortion caused by the augmentation process. Due to the unavailability of datasets with appropriate scene labels for scene perception, we have developed a new dataset called the Infrared Ship Dataset with Scene Segmentation (IRSDSS). Additionally, we have enhanced an existing dataset by adding scene masks, creating the Enhanced Infrared Ship Detection Dataset (EISDD). Our evaluations using both IRSDSS and EISDD demonstrate that SMPISD-MTPNet exceeds contemporary state-of-the-art methods in accuracy. The source code and dataset for this research can be available at https://github.com/greekinRoma/SMPISD-MTPNet.
14 January 2026 in IEEE Geoscience and Remote Sensing Letters
Moving infrared small target detection (MISTD) remains challenging due to the large computational load caused by multiframe input and background motion interference, leading to poor real-time and accuracy performance. To address these issues, we propose STLMamba, a spatial–temporal framework that integrates long-range dependency modeling with local feature sensitivity, guided by infrared video and frequency priors. Specifically, an adaptive multidirectional frequency enhancement (AMFE) module exploits spatial frequency cues to enhance targets, while an efficient local-motion Mamba (ELM) with a learnable local dynamic Gaussian motion (LDGM) mechanism captures motion patterns and suppresses background noise. Experiments demonstrate that STLMamba achieves 43.14 frames/s in NUDT-MIRSDT and 49.22 frames/s in TSIRMT, outperforming state-of-the-art methods in both accuracy and efficiency.