Small-Target Ship Detection with Joint Spatio-Temporal Features Across Multiple Frames
Sensors · 04.06.2026 · 26 (11) · 3588
DOI: 10.3390/s26113588 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 04.06.2026. Verfasst von Ye Qian und 4 weiteren Autor:innen. Beteiligte Einrichtung: Taihu Laboratory of Deepsea Technological Science, Wuxi 214000, China. Die Arbeit nennt 60 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Detecting small ship targets in sea–sky background environments is challenging due to interference from clouds, islands, sea clutter, and the limited spatial information in long-range infrared imagery. To address these issues, this paper proposes a robust detection framework that integrates multi-scale spatial feature enhancement with temporal trajectory analysis. First, a candidate target extraction method based on a multi-scale differential histogram of oriented gradients is introduced. By exploiting gradient distribution differences between targets and surrounding backgrounds, our method effectively enhances target responses while suppressing structured background edges. This response is further fused with a log-spectrum-based saliency map to improve target contrast and reduce clutter. Next, a candidate trajectory extraction algorithm based on inverse optical flow matching is developed to utilize temporal consistency. Optical flow-based grayscale compensation predicts target intensity changes between frames, while Kalman filtering estimates motion states and performs trajectory association. Finally, a multi-feature trajectory filtering strategy is designed, combining motion entropy stability, peak signal-to-noise ratio, and trajectory lifecycle to distinguish true targets from false alarms. Experimental results on eight infrared maritime sequences demonstrate superior performance. The proposed method achieves an average Background Suppression Factor (BSF) of 45.2 and an average Signal-to-Clutter Ratio Gain (SCRG) of 22.3 × 103, representing a substantial improvement over all baseline algorithms. Receiver Operating Characteristic analysis further confirms a mean detection rate exceeding 90% at a false-alarm rate of 10−3 across all sequences, confirming improved detection performance and robustness in complex maritime environments.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Taihu Laboratory of Deepsea Technological Science, Wuxi 214000, China.
- China Ship Scientific Research Center, Wuxi 214000, China.
- State Key Laboratory of Deep-Sea Manned Vehicles, Wuxi 214000, China.
- Nanjing Research Institute of Electronic Engineering, Nanjing 210000, China.
- School of Electronic and Optical Engineering, Nanjing University of Science and Technology, Nanjing 210094, China.
- School of Information and Communication Engineering, North University of China, Taiyuan 030051, China.
ORCID-Profile: Ye Qian
Zitieren
APA
Qian, Y., Hu, Z., Zhang, B., Yang, W., & Chen, Q. (2026). Small-Target Ship Detection with Joint Spatio-Temporal Features Across Multiple Frames. Sensors, 26(11), 3588. https://doi.org/10.3390/s26113588
Vancouver
Qian Y, Hu Z, Zhang B, Yang W, Chen Q. Small-Target Ship Detection with Joint Spatio-Temporal Features Across Multiple Frames. Sensors. 2026;26(11):3588. doi:10.3390/s26113588
BibTeX
@article{Qian2026,
author = {Qian, Ye and Hu, Zhen and Zhang, Bo and Yang, Wenguang and Chen, Qian},
title = {Small-Target Ship Detection with Joint Spatio-Temporal Features Across Multiple Frames},
journal = {Sensors},
year = {2026},
volume = {26},
number = {11},
pages = {3588},
doi = {10.3390/s26113588}
}
EUVIMED