融合随机抽样一致性和Hough变换的实时消失点检测  

Real-Time Vanishing Point Detector Combining RANSAC and Hough Transform

作　　者：吴建平[1] 陈珂[1] 刘业 Wu Jianping;Chen Ke;Liu Ye(School of Computer Engineering,Suzhou Vocational University,Suzhou 215104)

机构地区：[1]苏州市职业大学计算机工程学院,苏州215104

出　　处：《计算机辅助设计与图形学学报》2022年第8期1238-1251,共14页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(62173248);苏州市工业应用项目(SG1819)。

摘　　要：针对消失点拥有2个自由度的特点,提出一种由随机抽样一致性(RANSAC)和Hough变换相结合的实时消失点检测方法.首先利用随机抽样一致性从图像的边缘线段中抽取多条候选消失点局内线段,基于消失点位于局内线段的延长线上的特点确定消失点的第1个自由度,然后通过在候选局内线段的延长线上实施的Hough变换获得消失点的另一个自由度.基于数据集York Urban Dataset的实验结果表明,所提方法在地平线检测精度上比7种国内外同类算法高0.69个百分点以上,在处理速度上比同类方法提高90倍以上.As a vanishing point(VP)has 2 degrees of freedom(DOF),a real-time VP detection method based on the combination of random sample consensus(RANSAC)and Hough transform is proposed.Firstly,RANSAC is used to extract a set of candidate VP inliers from the edge segments of the image,with each candidate VP inlier fixing one DOF of a VP as the VP is on the extension line of the candidate;Then the Hough transform is conducted along the extension line of each candidate VP inlier to extract the VP’s remaining DOF.The experiment on York Urban Dataset shows that the proposed algorithm is at least 0.69percent better than the state of the art in horizon line accuracy and achieves the processing speed that is at least 90 times as fast as the state of the art.

关 键 词：消失点检测 随机抽样一致性 一维Hough变换 实时消失点检测 地平线检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]