遥感图像中基于Hough变换的直线提取算法  

Straight Line Extraction in Remote Sensing Image Based on Hough Transform

作　　者：邢坤[1] 何红艳[1] 岳春宇[1] 周楠[1] 

机构地区：[1]北京空间机电研究所,北京100094

出　　处：《航天返回与遥感》2015年第1期87-94,共8页Spacecraft Recovery & Remote Sensing

基　　金：国防973(61321001);高分专项(GFZX40136-03-02)资助课题

摘　　要：针对遥感图像中大型目标的直线特征提取问题,设计了一种基于类直线提取的改进霍夫(Hough)变换算法。类直线的提取相当于Hough变换的预处理,既约束了Hough变换所使用的直线上的特征点,改进了直接进行Hough变换时不同直线的特征点互相干扰的缺点,又降低了计算量。对图像空间的每条类直线分别进行Hough变换,改进投票过程的映射方式,寻找类直线中最多特征点所在的那条直线,只取一个投票峰值,在减小了计算复杂度的同时又去掉了虚假峰值的影响。实验结果表明,改进Hough变换直线特征提取算法性能好、效率高,可用于遥感图像处理领域。Focusing on the extraction of straight line features from large target in remote sensing image, this paper designs an improved Hough transform approach based on resembling lines extraction. Extraction of resembling lines corresponds the pre-processing of Hough transform, which can constrain feature points required for transform. This approach improves the interference of feature points on different lines and reduces the computational complexity. For each extracted resembling line, Hough transform is used in image space. Votes are cast using improved mapping scheme. The straight line determined by the most points in resembling line is extracted through seeking one peak value. New voting process aims at reducing the computational complexity as well as avoiding the effect of false peak values. The experimental results show that this algorithm is more efficient, and can be used in remote sensing image processing.

关 键 词：遥感图像 直线特征提取 霍夫变换 映射方式 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]