基于改进水平截集算法的SAR图像海岸线检测

Coastline Detection Method in SAR Images Based on anImproved Level Set Algorithm

合成孔径雷达(SAR)图像海岸线检测,在自动导航、地图绘制等海洋应用方面具有重要意义。水平截集算法是一种基于人类视觉特性的边缘检测方法。由于它具备检测效果好、抗噪能力强等优点,因而在海岸线检测方法的研究和应用两方面倍受关注。然而,水平截集算法由于迭代方式复杂等原因在应用于分辨率较大的图像时,检测速度比较慢,限制了它在工程应用中的实用性。针对SAR图像,提出一种基于水平截集算法的改进算法,先在低分辨率图像中用水平截集算法进行粗略检测,得到贴近真实海岸线的轮廓线,然后将轮廓线映射到高分辨率图像上,继续用水平截集算法进行检测,最后得到精确的结果。实验中使用RadarsatScanSAR图像证明该方法可大大加快检测速度,通过与原水平截集算法的检测效果进行对比,新方法没有降低检测效率。

Coastline detection in Synthetic Aperture Radar (SAR) images plays a significant role in marine applications, such as autonomous navigation, geographic mapping, etc. Level set algorithm is an edge detection method, based on visual characteristic of human being. Level set algorithm has been attracting much attention of research and application in coastline detection due to its detection effectiveness and quality of anti-noise. However, when applied to high resolution image, the level set algorithm will cost a lot of time to detect the contour because of its complicated iterative manner. In this paper, we present an improved method based on level set algorithm used especially in SAR images. Firstly, the coarse detection procedure is processed using level set algorithm in a low resolution image to obtain the contours closing to the real coastline of the image. Then, the coarse contour in the low resolution image is projected to the high resolution image and the coastline detection continued by level set algorithm in a finely way to obtain the final fined results. Experimental result using Radarsat ScanSAR image indicates that this novel improved method based on level set algorithm can enhance the detection speed, compared with the original level set algorithm, without the loss of detection effectiveness.