基于改进RANSAC算法的道路直线提取方法  

Road Straight Line Extraction Method Based on Improved RANSAC Algorithm

作　　者：吴剑亮 李艳[1,2] 高扬[1] 黄小赛[1] 

机构地区：[1]南京大学国际地球系统科学研究所,江苏南京210023 [2]江苏省地理信息技术重点实验室,江苏南京210023

出　　处：《地理空间信息》2017年第5期42-44,50,共4页Geospatial Information

基　　金：国家自然科学基金资助项目(41371331)

摘　　要：直线检测是计算机视觉领域的一项重要内容,也是遥感影像信息提取的基本过程。随机抽样一致性算法(RANSAC)常用来进行包括直线在内的目标提取,是在计算机视觉领域应用较广泛的估计算法之一,但其计算效率较低。因此提出了一种基于序贯概率的改进型多目标RANSAC算法,在利用边缘检测限定随机抽样原始样本集的基础上,运用该算法提取了高分辨率航空影像数据中的道路边界线,大大提升了计算效率。Straight line detection is an important content in the field of computer vision, and is a basic process for information extraction from remote sensing imagery. Random sample consensus(RANSAC) algorithm is one of the widely used estimation algorithm in the field of computer vision, which is usually used for target extraction including lines and other features. However the computational efficiency of RANSAC algorithm is very low. In this paper, we proposed a multitarget improved RANSAC algorithm based on sequential probability. By limiting the initial sample set as image edges, we applied the algorithm to extract the road borderline in high resolution aerial imagery.

关 键 词：RANSAC算法 直线检测 边缘检测 

分 类 号：P237[天文地球—摄影测量与遥感]