中文摘要：

线状特征是数字图像处理与模式识别中一种重要的中层描述符号,具有丰富的语义信息。针对目前提取算法的不足,提出一种稳健的高精度直线提取算法。在对分裂线段感知编组时提出一个基于假设检验的方法,从几何拓扑关系和物理光谱信息两个方面以及全局与局部两个尺度进行模糊融合和检验。并利用最小二乘模板匹配将提取直线的精度提高到亚像素。本算法可在一定程度上克服线段提取中的两类错误和定位精度低的问题,通过对航空和地面影像的试验,验证了该方法的有效性。

英文摘要：

Line feature,which has abundant semantic information,is a very important intermediate level symbol in digital image processing and pattern recognition.In order to resolve many issues in current line extraction algorithms,a robust and sub-pixel approach for extracting straight lines is presented.When fractured short line segments need to be combined into groups,which is called as perceptual organization,an algorithm based on hypothesis testing and fuzzy fusion is put forward.This algorithm combines fractured line segments into a whole one according to geometric topology,physics spectral information,global and local scale information.Then,the least square template matching algorithm（LSTM） is implemented to get the higher precise line segments.The experimental results show that the proposed algorithm is more efficient,which can get richer and sub-pixel straight lines from aerial and ground images.