单视未标定图像的正交灭点检测算法  

Un-Calibrated Single Visual Image of Orthogonal Vanishing Point Detection Algorithm

作　　者：储珺[1] 肖旭[1] 梁辰[1] Chu Jun;Xiao Xu;Liang Chen(School of Software, Nanchang Hangkong University, Nanchang Jiangxi 330063, China)

机构地区：[1]南昌航空大学软件学院,江西南昌330063

出　　处：《图学学报》2016年第6期783-790,共8页Journal of Graphics

基　　金：国家自然科学基金项目(61263046)

摘　　要：传统方法只能计算标定图像的正交灭点,同时没有考虑图像直线检测结果的误差、直线的长度以及候选灭点与约束直线之间的位置关系对灭点检测精度的影响。针对此类问题,提出了一种针对单视未标定图像的正交灭点检测方法。首先利用J-Linkage完成灭点的初始化估计,得到假设灭点集合;然后根据假设灭点与图像直线之间的一致性约束、图像直线的长度,基于投票机制先得到精确的垂直方向灭点;后利用灭点、灭线的定义和性质,计算得到图像相机参数;根据正交灭点的特性,得到准确的水平方向和纵深方向的灭点。因引入了一种新的假设灭点和图像直线之间的一致性度量方法,正交灭点检测精度不受直线检测结果的误差、直线的长度以及候选灭点与约束直线之间的位置关系的影响,在未知图像相机参数的情况下能精准的得到三个正交灭点信息。正交灭点检测方法在室内场景下可以得到更加精确的检测结果。The traditional method can only calculate the orthogonal vanishing point of the calibrationimage, dose not take into account the error of straight line detection, the length of the straight line andthe location relationship between the candidate vanishing point and the constrained line has theimpact on the accuracy of the vanishing point detection. For such problems, this paper proposes anun-calibrated single visual image of orthogonal vanishing point detection algorithm. First, useJ-Linkage algorithm to complete the initialization of the vanishing point estimates, and get the set ofassumed vanishing points. Second, according to the consistency between the assumed vanishingpoints and lines, length of line segment to get the exact vertical vanishing point based on a votingmechanism. Third, calculate the parameters of the image camera by the nature and properties of thevanishing point and vanishing line. Then based of the orthogonal characteristics of the vanishing pointto get the exact horizontal and depth direction of the vanishing point. By the introduction of a newconsistency measurement method between the assumed vanishing point and the line of image, thismethod not only gets rid of the dependence on the camera parameters, but also reduces the impacts ofthe impacts of errors of line segment extraction, length of line segment as well as the relative position ofvanishing point and the constrained line on the precision of vanishing points detection, in the case of unknown image camera parameters this method can get accurate information about the three orthogonalvanishing point. This method this thesis applied can achieve better results in indoor scenes.

关 键 词：单视图像 未标定 投票 正交灭点 N矢量 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]