Sentinel-1宽模式的SAR图像拼接及大区域海岸线快速提取

SAR Image Stitching and Rapid Extraction of Large-Scale Coastline Using Wide-Swath Sentinel-1 Data

为实现对大区域海岸线的监测和管理,该文提出了一种使用Sentinel-1宽模式数据的SAR图像拼接及海岸线快速提取方法。首先从输入SAR图像序列中提取地理信息,基于图像序列重叠区域的灰度统计信息,经由Wallis变换消除相邻图像间的色彩差异,进而利用地理信息生成研究区域的镶嵌图像;然后以镶嵌图像作为海岸线快速提取的输入图像,建立关于镶嵌图像的高斯图像金字塔序列,对该序列的低分辨率图像使用边界追踪算法提取海岸线,再将所得海岸线结果重采样至与输入图像相同的分辨率,得到最终的大区域海岸线。针对黄海海域的实验结果表明,在SAR图像拼接中,不同输入图像在接缝处的灰度均值接近、灰度方差低,过渡自然且未出现明显镶嵌线,而提取的大区域海岸线在保证精度的前提下,处理时间明显低于手工标记等方法。

In order to monitor and manage large-scale coastal zone,this paper presents a new approach for the SAR image mosaic and rapid coastline extraction using wide-swath Sentinel-1 data.Firstly,the geographic information is extracted from the input SAR image sequence.Then,based on the gray statistics of the overlapping regions of the image sequence,the color differences between adjacent images are eliminated by Wallis transform,and the mosaic image of the studied region is generated by geographic information (longitude and latitude).Next,the mosaic image is used as the input image for rapid extraction of the coastline,and the Gauss image pyramid sequence of the mosaic image is established.The boundary tracing algorithm is used to extract the coastline from the low-resolution image of the Gaussian image pyramid sequence,and the coastline results are resampled to the same resolution as the input image to obtain the final large-scale coastline.The experimental results of the Yellow Sea show that the gray value of the joints of different input images is close,the gray variance is low,and the transition is natural without obvious inlaid lines in the SAR image stitching result.Moreover,under the premise of ensuring accuracy,the extracted large-scale coastline has significantly lower processing time than that of manual marking and other methods.