平行算子及其在图像平行结构检测中的应用

摘要

平行直线以及具有平行结构的多边形在图像中的体现非常丰富,以往的研究一般通过直线斜率相等判断两直线平行,对具有平行结构的多边形的检测需要预先给定信息或者适用范围狭窄,针对一般平行结构检测的研究一直较少.作者在充分研究平行结构特性的基础上,基于距离信息提出了平行算子的概念,实现了图像中平行结构的检测.首先,利用点线距离及点间距离的关系给出平行算子的定义;然后,获取图像中的有效像素,并基于平行算子得到所有平行点组;随后,根据方向信息合并平行点组,利用Hough变换实现图像中平行线的检测;最后,验证直线交点之间线段的存在性,根据存在线段的端点位置实现多边形检测.实验结果表明:本文定义的平行算子能够很好地检测到实际图像中的平行点组,在平行精度要求不高的条件下可以抵抗较大的噪声;基于平行算子的图像平行结构检测算法可以准确检测出图像中的平行直线及具有平行结构的多边形.本文提出的算法可准确检测出图像中的平行结构,具有普遍性.%Parallel lines and polygons consisting of parallel structure appear more in images.In the past,the parallelism of two straight lines was judged according to their line slope,and the detection of polygons consisting of parallel structure required prior information or specific polygons.The research on the detection of general parallel structures was little.After full study of characteristics of parallel structures,the definition of parallel operator based on distance information was proposed in this paper,which was used to realize the detection of parallel structures in images.Firstly,the definition of parallel operator was given by using the relation between the point-to-line distance and the point-topoint distance.Secondly,parallel point groups were obtained according to the defined parallel operator once the valid pixels were got in images.Thirdly,parallel point groups were merged based on the direction information,and parallel lines were then detected using Hough transform.Finally,the existence of the segments between intersections of lines was verified,and polygons were thus attained according to the endpoint location of existing segments.Experimental results showed the parallel operator defined in this paper could be used to detect the parallel point groups in the real image well,which could even more resist the large noise when the accuracy of parallelism was not high;the parallel structure detection algorithm based on parallel operator showed good detection of the parallel straight lines and polygons consisting of parallel structure in the image.The main work of this paper was:1)Proposed the definition of parallel operator according to the distance information between two points.2)Used the parallel operator to detect the parallel lines and polygons consisting of parallel structure in the image,which provided a new idea for related research.3)Gave an effective method to extract valid pixels which made the algorithm more robust to noise and reduced the computational load.The proposed algorithm could detect parallel structures accurately in the image and was universal in practical applications.