一种基于单幅图像的主平面检测算法  

A Dominant Plane Detection Algorithm Based on Single Image

作　　者：赵晓莹 刘艳丽[1,2] ZHAO Xiao-ying;LIU Yin-li(College of Computer Science,Sichuan University,Chengdu 610065;National Key Laboratory of Fundamental Science on Synthetic Vision,Sichuan University,Chengdu 610065)

机构地区：[1]四川大学计算机学院,成都610065 [2]四川大学视觉合成图形图像国家重点实验室,成都610065

出　　处：《现代计算机》2018年第13期39-42,共4页Modern Computer

摘　　要：不同于传统的基于RANSAC思想的平面检测算法,提出一种新颖的单张图像主平面检测算法。通过图像中线段交点的角度直方图峰值得到图像所对应的三维场景消失点,并根据消失点将线段分类成不同组别;再使用图像分割算法分割图像中平面区域;然后由其中两组平行线段的交点像素值选择主平面颜色并确定主平面区域。最后在两组平行线中选择两对平行线优化主平面区域边界。Different from the traditional methods of plane detection based on RANSAC,proposes a novel algorithm to detect the dominant plane in the single image.Firstly,the vanishing points of the 3D scene are extracted from the peak value of the angle histogram at the intersection of the lines in the image,and we classify the lines into different classifications.Secondly,uses the image segmentation algorithm to segment the plane regions,and then uses the pixel value of the intersection of two groups of parallel lines to find the plane color and determine the dominant plane region.Finally,two pairs of parallel lines are selected to optimize the dominant plane boundary.

关 键 词：平面检测 消失点 线段分类 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]