合成孔径雷达(Synthetic Aperture Radar，SAR)是一种主动式微波成像雷达，可提供全天时、全天候、多参数的高分辨率图像，使其成为海上监视和导航的重要工具。基于SAR图像舰船目标的解译可用于支持环境监测和海上安全，推动我国合成孔径雷达系统走向实用化。本研究主要集中在SAR舰船图像的分割、舰船目标的几何特征提取以及舰船目标的检测三个方面，具体的研究内容如下：
1.受限于SAR成像机理，SAR图像往往伴随着斑点噪声的影响，且随着SAR分辨率的提升，高分辨SAR图像舰船旁瓣问题开始凸显，这严重影响舰船目标的主体分割。针对上述问题，本文提出了一种基于模糊局部信息C均值聚类(Fuzzy Local Information C-Means，FLICM)与Radon变换的SAR舰船图像分割算法。该方法首先利用FLICM将舰船目标与海洋背景进行初始分割；然后，对二值化的舰船切片进行Radon变换，确定主瓣和旁瓣的中心位置及其亮点区域半径，剔除旁瓣；最后，对带有微小旁瓣残留的舰船切片进行了形态学滤波处理，得到舰船目标精细分割结果。本文对所提方法进行了仿真对比实验，结果表明：所提方法可有效识别并剔除旁瓣，提升了域间差异性、域内一致性，明显降低了舰船目标的形状复杂度。
2.针对舰船目标周围存在旁瓣和背景杂波干扰的问题，考虑矩技术不但具备平移不变性和尺度不变性，还具备抗噪声干扰能力强的特点，采用了基于中心距估计方位角的方法，提高了方位角估计的稳健性；针对传统最小外接矩形拟合法提取舰船目标长宽参数，存在提取效率低的问题，提出了一种基于凸包的最小外接矩形拟合方法，该方法能剔除强散射点和不规则形状的影响，提高长宽特征提取精度和速度。本文对所提方法进行了仿真对比实验，结果表明：所提方法能够迅速且精确地从SAR图像中提取出舰船目标的几何参数。
3.针对传统的SAR图像目标检测方法存在检测速度慢，且对于复杂、多目标场景下检测结果虚警率偏高等问题，本文提出了一种基于几何特征的两级恒虚警率舰船目标检测方法。该方法首先对SAR舰船图像进行基于全局迭代的潜在目标预筛选；然后，对
潜在目标进行基于瑞利分布的双参数恒虚警率检测；接下来，对检测结果进行形态学连接和填充，克服二级检测后舰船目标出现的漏洞和断裂的问题；最后，结合几何特征鉴别舰船目标，得到检测结果。本文对所提方法进行了仿真对比实验，结果表明：本文方法相较于传统的目标检测方法能在多目标、复杂杂波背景下提高目标检测的准确率及检测速度。

Synthetic Aperture Radar (SAR) is an active microwave imaging radar that provides all-day, all-weather high-resolution imagery, making it an important tool for maritime surveillance and navigation. The ship target interpretation based on SAR image can be used to support environmental monitoring and maritime safety, and promote the practical application of China's SAR system. This research mainly focuses on three aspects of SAR ship image segmentation, ship target geometric feature extraction and ship target detection. Specific research contents are as follows:
1. Limited by the SAR imaging mechanism, SAR images are often affected by speckle noise, and with the improvement of SAR resolution, the ship sidelobe problem of high-resolution SAR images becomes prominent, which seriously affects the subject segmentation of ship targets. To solve the above problems, a SAR ship image segmentation algorithm based on Fuzzy Local Information C-Means (FLICM) clustering and Radon transform is proposed in this paper. First, the algorithm used FLICM to initially separate the ship target from the ocean background. Then, Radon transform is applied to the binary ship slice to determine the center position of the main lobe and sidelobe and the radius of the bright spot area, and the sidelobe is eliminated. Finally, the ship slices with tiny sidelobe residue are processed by morphological filtering, and the fine segmentation results are obtained. The simulation results show that the proposed method can effectively identify and eliminate sidelobe, improve inter-domain difference and intra-domain consistency, and reduce the shape complexity of ship target significantly.
2. To solve the problem of sidelobes and background clutter interference around the ship target, the method of estimating azimuth Angle based on center distance is adopted, considering that the moment technique not only has translation invariance and scale invariance, but also has strong anti-noise interference ability. Aiming at the low extraction efficiency of the traditional minimum external rectangle fitting method, a convex hull based minimum external rectangle
fitting method is proposed, which can eliminate the influence of strong scattering points and irregular shapes, and improve the extraction accuracy and speed of the length and width features. In this paper, the simulation results show that the proposed method can quickly and accurately extract the geometric parameters of ship targets from SAR images.
3. In view of the slow detection speed of traditional SAR image target detection methods and the high false alarm rate of detection results in complex and multi-target scenarios, this paper proposes a two-stage ship target detection method with constant false alarm rate based on geometric features. In this method, the potential targets of SAR ship images are pre-screened based on global iteration. Then, the two-parameter constant false alarm rate based on Rayleigh distribution is detected for potential targets. Next, the detection results are connected and filled by morphology to overcome the loopholes and fractures of the ship target after the second-level detection. Finally, the ship target is identified with geometric features and the detection results are obtained. The simulation results show that compared with the traditional target detection methods, the proposed method can improve the accuracy and speed of target detection under the background of multi-targets and complex clutter.