Sentinel-1 satellite constellation offers enough Synthetic Aperture Radar (SAR) images for long-term water monitoring, due to its relative large swath, great revisit frequency and high resolution. The middle and upper Yangtze River suffers serious flood disaster in 2016. It is significant to detect water and its changes of Poyang Lake, since it is one of the important flood storage lakes along the Yangtze River mainstream. However, the traditional segmentation algorithm has shortage in edge preservation and the accuracy of water detection, especially in the case of Poyang Lake, which is widely distributed and has more complex background, weak edges and blurred edges. A new Active Contour Model (ACM) algorithm based on local narrowband is proposed to solve these problems, and it is applied to Sentinel-1A observations related to Poyang Lake. First, a cascade two-level Otsu approach is adopted to obtain the initial contour. Second, the local narrowband is built along the initial contour to reduce the calculating time. Finally, a region-based ACM is introduced into the local narrowband to stop the contours at weak or blurred edges. Experiment results show that the new method has advantages in the edge preservation and obtains better segmentation results with respect to other methods.%Sentinel-1合成孔径雷达(SAR)卫星具有测绘带宽、重访周期短、分辨率高等优点,为长时间的水域监测提供数据基础.2016年长江中下游地区洪涝灾害严重,鄱阳湖是长江干流的重要调蓄性湖泊之一,基于SAR图像的鄱阳湖水域提取及其变化检测具有重要意义.然而受相干斑噪声的影响,尤其是在鄱阳湖分布较广、地物背景较复杂、弱边缘和模糊边缘较多的情况下,传统的水域分割方法边缘保持性较差、提取精度较低.针对上述问题,该文提出一种基于局部窄带的ACM边缘提取算法,并将其应用于Sentinel-1A获取的鄱阳湖水域时序观测图像中.该算法首先采用两级Otsu方法获取初始轮廓,随后在初始轮廓附近建立局部窄带,最后在窄带内采用基于区域的ACM方法进行轮廓线演化来解决弱边缘或模糊边缘问题.实验结果表明该方法在边缘保持和分割精度上具有明显优势,并且降低了计算时间.
展开▼
