中文摘要：

高分辨率影像中的伪直线是人造目标提取过程中增加计算复杂性和降低提取结果可靠性的一个主要因素。针对现有基于边缘检测的直线提取算法不能有效剔除伪直线的缺点，提出了一种可行的方法用来剔除伪卣线。该方法体现在直线提取的两个不同阶段：一是在边缘点编组前，通过引入12个过滤模板来剔除部分编组过程中可能拟合成伪直线的边缘点；二是对于已提取出来的每条直线，依据它们两侧邻域图像的灰度分布特性来判定直线的真伪。大量实验结果表明，该方法能够有效剔除由图像噪声或不规则天然地物拟合得到的绝大部分伪直线，而几乎不影响提取对应人造目标的真直线。

英文摘要：

Appearance of the false lines is one of the major disturbing factors, which increases the complexity of the computation and reduces the stability of the extraction result, in the process of extracting artificial targets from highresolution images. A robust two-step method is brought forward to filter the false lines extracted by the edge-based algorithms for extracting lines. Based on analysis of characteristics of false lines, in consideration that the lines extracted from artificial targets are regularly extraordinary, this method operates in two phase of the algorithms for extracting lines. One is that prior to grouping edges with co-lineation relationships, 12 models are employed to filter a proportion of edges, from which the false lines may be fitted later. The other is that after the process of extracting lines, the strength of the similarity of the image characteristic close to the line is defined firstly, in terms of which the lines extracted are divided into the true ones or the false ones, and the false lines will be filtered. Lots of experiments show that this method can filter a majority of the false lines extracted from image ,which are corresponding with image noise or irregular natural objects, while the true lines which are corresponding with artificial targets are almost preserved.