Joint Wavelet Thresholding And F-NLM Denoising For High-resolution SAR Ship Detection
-
摘要: 摘 要：【目的】针对高分辨率合成孔径雷达（SAR）舰船目标多场景、多尺度、密集排布的显著特征和成像过程中相干噪声导致目标边缘细节模糊的问题，提出一种联合小波阈值和快速非局部均值滤波（F-NLM）去噪的高分辨SAR舰船检测方法。【方法】首先，利用小波阈值和F-NLM联合去噪模块对SAR图像进行预处理，降低海杂波噪声，增强检测目标细节特征和边缘信息，使提取特征更具判别性。然后，选用YOLOv7检测网络结合双向特征融合模块（Bi-FPN)，有效地对多尺度特征聚合，进一步提高模型准确率。【结果】结果显示，使用去噪数据集D-SSDD得到的检测平均精准度（AP）可达98.69%，虚警率（FAR）降低至2.37%【结论】研究表明：该高分辨率SAR舰船检测方法不仅能均匀背景杂波提高图像质量，还能提高多尺度特征信息的交互性，保证目标检测精度准确。
-
关键词:
- 关键词：SAR舰船检测 /
- 小波阈值 /
- 快速非局部均值滤波 /
- 双向特征融合模块 /
- YOLOv7
Abstract: Abstract：[Objectives] Aiming at the significant features of high-resolution synthetic aperture radar (SAR) ship targets with multi-scene, multi-scale, and dense arrangement and the problem of blurring of target edge details due to coherent noise in the imaging process, a high-resolution SAR ship detection method is proposed with joint wavelet thresholding and fast nonlocal mean filtering (F-NLM) denoising. [Methods] First, wavelet thresholding and F-NLM denoising module are utilized to preprocess the SAR image to reduce the sea clutter noise, enhance the detailed features and edge information of the detection target, and make the extracted features more discriminative. Then, YOLOv7 detection network combined with bi-directional feature fusion module (Bi-FPN) is selected to effectively aggregate multi-scale features to further improve the model accuracy. [Results] The results -
计量
- 文章访问数: 48
- 被引次数: 0
下载:
