摘要
随着各种高性能SAR成像传感器的广泛使用,海量的SAR图像为各种对地表观测提供了更加丰富的研究依据。基于SAR图像,在复杂场景中提出了一种基于图论及河道轮廓模型相融合的河道提取方法。首先考虑到河道轮廓曲线的复杂性,提出一种河道轮廓分段建模的方法,将复杂的曲线河道轮廓形式化近似为规则直线线段的组合,进而分段采用最小外接矩形窗的组合准确地对河道区域进行描述。建立了一种更加适用于河道提取的图像分割规则,实现对河道区域的准确分割。在此基础上,利用河道轮廓特征的先验知识,根据区域最小外接矩形的形态和连通性对河道区域进行识别。实验结果表明:该方法与常用的基于灰度阈值判别的方法相比,不仅能够有效提取出河道区域,提取结果有效覆盖90%以上的真实区域,还能够较好地抑制背景信息,提取结果仅包含约2%的背景区域。
In recent years,as various high-quality SAR imaging devices are widely used,mass SAR image data provides us with abundant evidence for analyses.Given the SAR images,this paper proposes a novel river extraction method which fuses the graph model and the river prior model.First,taking the complexity of the river contour into consideration,a regionalized model establishment method is proposed and a series of minimum bounding rectangles are combined to simulate the complicated river contour.Further,the image is segmented by a new regulation which is adaptive to the task of river extraction.Then,the river contour prior is utilized for river region recognition,removing the background noise.Experimental results demonstrate that in contrast to the traditional gray threshold based image segmentation method,the proposed method not only has the ability to accurately extract the river region and the extracted region cover the real river region with more than 90 percent,but also be superior to the counterpart in the performance of background noise removal and only less than 2percent of the background are left in the extracted region.
出处
《遥感技术与应用》
CSCD
北大核心
2015年第2期337-344,375,共9页
Remote Sensing Technology and Application
