| 引用本文: | 李雨轩， 刘 峥， 冉 磊. 改进三维最大类间方差的SAR图像海陆分割算法[J]. 雷达科学与技术, 2024, 22(4): 416-426.[点击复制] |
| LI Yuxuan, LIU Zheng, RAN Lei. An Improved Algorithm for SAR Image Sea⁃Land Segmentation Based on 3D Maximum Between⁃Class Variance[J]. Radar Science and Technology, 2024, 22(4): 416-426.[点击复制] |
|
| 摘要: |
| 合成孔径雷达（Synthetic Aperture Radar, SAR）是一种在气象条件较差的条件下工作的高分辨率成像雷达。无论在民用还是军事领域，用合成孔径雷达进行海上舰船图像目标检测都能起到关键作用。海陆分割作为一种SAR图像预处理的手段，可以为后续的目标检测、识别与跟踪技术研究奠定良好的基础。本文对包含较复杂岛岸背景的SAR图像的海陆分割算法进行研究，在传统三维最大类间方差算法的基础上提出一种改进的SAR图像海陆分割算法，将原有算法第三维度邻域中值改为Prewitt算子的梯度运算，Prewitt算子适合处理灰度渐变复杂的图像，SAR图像能够更好地进行图像分割。考虑到维度上增加一维会使得算法效率降低，运用分解的思想，将三维最大类间方差拆分成3个一维最大类间方差算法。通过两组SAR图像进行实验，结果表明本文提出的改进三维最大类间方差算法具有可行性，并且在海陆分割效果方面优于一维、二维最大类间方差算法，在运算复杂度方面优于传统的三维最大类间方差算法。 |
| 关键词: 合成孔径雷达 海陆分割 最大类间方差 阈值选取 |
| DOI：DOI:10.3969/j.issn.1672-2337.2024.04.008 |
| 分类号:TN957.51 |
| 基金项目:国家自然科学基金（No.62001346） |
|
| An Improved Algorithm for SAR Image Sea⁃Land Segmentation Based on 3D Maximum Between⁃Class Variance |
|
LI Yuxuan, LIU Zheng, RAN Lei
|
|
National Key Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, China
|
| Abstract: |
| Synthetic aperture radar （SAR） is a type of high?resolution imaging radar that operates in poor meteorological conditions. Whether in the civilian or military field, the use of SAR for image target detection of ships at sea can play a key role. As a means of SAR image preprocessing, sea?land segmentation can lay a good foundation for the subsequent research on target detection, recognition and tracking technology. In this paper, an improved sea?land segmentation algorithm for SAR images with complicated island and shore backgrounds is proposed on the basis of the traditional three?dimensional maximum between?class variance algorithm, which changes the median value of the third?dimensional neighborhood of the original algorithm to the gradient operation of the Prewitt operator, which is suitable for processing images with complicated grayscale gradients, and can better segment SAR images. Considering that adding one dimension to the dimension can reduce the efficiency of the algorithm, the idea of decomposition is used to split the three?dimensional maximum between?class variance into three one?dimensional maximum between?class variance algorithms. Through experiments with two sets of SAR images, the results show that the improved three?dimensional maximum between?class variance algorithm proposed in this paper is feasible, and it is better than the one?dimensional and two?dimensional maximum between?class variance algorithms in terms of sea?land segmentation effect, and better than the traditional three?dimensional maximum between?class variance algorithm in terms of computational complexity. |
| Key words: synthetic aperture radar （SAR） sea⁃land segmentation maximum between⁃class variance thres⁃hold selection |
