几何活动轮廓模型用于高分辨率遥感影像海岸线自动提取

Automatic Waterline Extraction in VHR Imagery Using Geometric Active Contour Model

潮湿的潮滩以及近陆浑浊的海水往往使得遥感影像中的海岸线表现为模糊边界,此外潮滩表面复杂的微地形以及大量潮沟存在,使得传统基于边缘检测或分类的海岸线提取方法都难以获得连续、完整的岸线.针对上述问题,提出了一种结合区域信息改进的几何活动轮廓模型从遥感影像中快速、自动提取海岸线的方法.该方法首先利用二维最大方差阈值化分割并辅以形态学滤波对海岸线进行粗定位,以此作为初始演化轮廓,再利用改进的几何活动轮廓模型对海岸线进行精确提取.通过SPOT5全色影像的仿真试验证明该方法能够快速、准确地提取海岸线,得到的矢量岸线可直接用于地理信息系统(GIS)分析.

Due to the complicated micro-relief and tidal-creeks of the tidal flat,most methods of general-purpose edge detection and image segmentation result in fuzzy boundary in VHR images caused by the moist land and the turbid seawater rather than a continuous and complete shoreline.A new method using modified region based Geometric Active Contour（GAC） model has been proposed to automatically extract the shoreline efficiently.More precisely,a boundary of land and water in the original image is obtained by using two dimensional maximum between-cluster variance algorithm and the morphology filter.Then the boundaries are used as the initial contour of the modified region-based GAC model.To test the applicability of the new approach,it was applied to two subsets of SPOT5 imagery.The positional precision of coastline has been proved to be satisfactory.Finally a vector-based coastline results,which could be directly used for the subsequent GIS analysis.