EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Infrared Small Target Detection Based on a Temporally-Aware Fully Convolutional Neural Network

Lei Zhang, Peng Han, Jiahua Xi, Zhengrong Zuo

Remote Sensing · 26.08.2023 · 15 (17) · 4198

DOI: 10.3390/rs15174198 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 26.08.2023. Verfasst von Lei Zhang und 3 weiteren Autor:innen. Die Arbeit wurde bisher 11-mal zitiert (Crossref) und nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In the field of computer vision, the detection of infrared small targets (IRSTD) is a crucial research area that plays an important role in space exploration, infrared warning systems, and other applications. However, the existing IRSTD methods are prone to generating a higher number of false alarms and an inability to accurately locate the target, especially in scenarios with low signal-to-noise ratio or high noise interference. To address this issue, we proposes a fully convolutional-based small target detection algorithm (FCST). The algorithm builds on the anchor-free detection method FCOS and adds a focus structure and a single aggregation approach to design a lightweight feature extraction network that efficiently extracts features for small targets. Furthermore, we propose a feature refinement mechanism to emphasize the target and suppress conflicting information at multiple scales, enhancing the detection of infrared small targets. Experimental results demonstrate that the proposed algorithm achieves a detection rate of 95% and a false alarm rate of 2.32% for IRSTD tasks. To tackle even more complex scenarios, we propose a temporally-aware fully convolutional infrared small target detection (TFCST) algorithm that leverages both spatial and temporal information from sequence images. Building on a single-frame detection network, the algorithm incorporates ConvLSTM units to extract spatiotemporal contextual information from the sequence images, boosting the detection of infrared small targets. The proposed algorithm shows fast detection speed and achieves a 2.73% improvement in detection rate and an 8.13% reduction in false alarm rate relative to the baseline single-frame detection networks.

Zitieren

APA

Zhang, L., Han, P., Xi, J., & Zuo, Z. (2023). Infrared Small Target Detection Based on a Temporally-Aware Fully Convolutional Neural Network. Remote Sensing, 15(17), 4198. https://doi.org/10.3390/rs15174198

Vancouver

Zhang L, Han P, Xi J, Zuo Z. Infrared Small Target Detection Based on a Temporally-Aware Fully Convolutional Neural Network. Remote Sensing. 2023;15(17):4198. doi:10.3390/rs15174198

BibTeX
@article{Zhang2023,
  author = {Zhang, Lei and Han, Peng and Xi, Jiahua and Zuo, Zhengrong},
  title = {Infrared Small Target Detection Based on a Temporally-Aware Fully Convolutional Neural Network},
  journal = {Remote Sensing},
  year = {2023},
  volume = {15},
  number = {17},
  pages = {4198},
  doi = {10.3390/rs15174198}
}
Kostenlos starten →