| 摘要: |
| 针对SAR图像海陆分割的最佳阈值难以确定的问题，提出了一种差值核估计的自适应最佳阈值SAR图像海岸线提取方法。首先将图像的频数分布直方图看作一组以灰度值为横坐标和以频数为纵坐标的数据点集，然后以任一点为中心点构建左右两个邻域；定义两邻域内点横坐标到中心点横坐标距离的核函数；以核函数为权重先计算中心点纵坐标与左邻域内所有点纵坐标之差的加权平均值，再计算右邻域内所有点纵坐标与中心点纵坐标之差的加权平均值，并定义后者与前者之差为该数据点在直方图中坡度由陡变缓的跳变幅度，最后将拥有最大跳变幅度的数据点对应的横坐标作为海陆分割的最佳阈值。为去除初始分割结果中海陆交界处的不规则形状像素集，设计基于滤波操作的后处理过程。对真实大尺度SAR图像提取结果的定量评价表明提出方法可以高精度地提取出光滑、连续的海岸线，相较其他阈值分割法有更强的实用性。 |
| 关键词: SAR图像 最佳阈值 海岸线提取 核函数 |
| DOI：DOI:10.3969/j.issn.1672-2337.2019.03.012 |
| 分类号:TN957.52 |
| 基金项目:国家自然科学基金（No.41271435,41301479）；辽宁省自然科学基金（No.2015020090）；辽宁工程技术大学测绘与地理科学学院研究生优秀学位论文培育计划项目（No.YS201606） |
|
| Coastline Extraction with Optimal Threshold from SAR Images Using Kernel Estimation |
|
WANG Dailiang，YANG Yun
|
|
School of Geomatics, Liaoning Technical University, Fuxin 123000, China
|
| Abstract: |
| Aiming at the problem that the optimal threshold for sea-land segmentation of SAR images is difficult to be determined, a coastline extraction method with adaptive optimal threshold using difference kernel estimation is proposed. First, frequency distribution histogram of the image is regarded as a set of data points which take the gray value as the abscissa and the frequency as the ordinate. Second, any point in the set of data points is taken as the center point to construct left and right neighbors, and then the kernel function related to the distances from the abscissa of center point to the abscissa of all points in left and right neighbors is defined. Third, the weighted average of the difference between the ordinate of the center point and the ordinates of all points in the left neighbor is calculated, and the weighted average of the difference between the ordinates of all points in the right neighbor and the ordinate of the center point is also calculated. The difference between the latter weighted average and the former weighted average is employed as the jump amplitude of the slope from steep to slow. Eventually, the abscissa of the point with the maximum jump amplitude is taken as the optimal threshold for sea-land segmentation. To delete the set of pixels with irregular shapes, the post-processing procedure based on filtering operation is designed. The smooth and continuous coastlines can be extracted by the proposed method with high precision, which is shown by quantitative evaluation of the coastline extraction results from real large-volume SAR images. The proposed method is more practical than other threshold-based segmentation methods. |
| Key words: SAR image optimal threshold coastline extraction kernel function |
