基于Harris兴趣值的影像线特征高精度提取方法  

A highly accurate extraction method of image line features by Harris interest strength

作　　者：张云生[1] 刘阳[1] 邹峥嵘[1] 

机构地区：[1]中南大学地球科学与信息物理学院,长沙410083

出　　处：《测绘科学》2013年第4期72-73,79,共3页Science of Surveying and Mapping

基　　金：中南大学博士后科研启动金(74341016023)

摘　　要：针对影像高精度线特征提取问题,本文提出一种基于Harris兴趣值的直线提取方法:利用EDISON算法提取影像的初始边缘,并将其细化到单像素,对提取的每一边缘点在其法线方向进行基于Harris兴趣值的子像素定位,并将精确定位后的点拟合成直线。通过采用两幅真实影像进行试验,证实了本文所提方法的有效性。Aiming at the accurate extraction of image line features, a line extraction method based on Harris interest strength was proposed in the paper. Firstly, the edges were extracted using EDISON algorithm, and then the edges were thinned to one pixel thick edges. After that, each edge element was sub-pixel located by parabola fitting using Harris interest strength. Lastly, the accurately lo- cated edges were fitted to straight lines. Experiments were carried out with two images with typical textures, and the results revealed that the proposed method would be feasible for sub-pixel line extraction.

关 键 词：EDISON算法 Harris兴趣值 直线特征 精确定位 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]