摘要
为了适应网络应用对轮廓相似度计算在速度和数据传输量方面的要求,文中提出了一种非离散的快速算法,该算法采用直线和圆弧来表达轮廓曲线.在对图形进行范化处理后,将待匹配轮廓按照一定步距进行旋转.然后依次以待匹配轮廓或目标轮廓作为模板,计算模板的所有顶点与其在另一个轮廓中对应点的距离平方的平均值.在所有角度中,平均值之和的最小值即为两个轮廓的匹配度.该算法同时适用于凸多边形和凹多边形,并具有较好的区分度和匹配准确性.和以往的离散方式相比,该算法减少了需要传递的数据量,提高了运算速度.
In order to calculate contour similarity with high speed and less data transfer in network application, this paper proposes a fast non-discrete algorithm in which lines and arcs are used to represent a contour. During the investigation, the image is normalized first, and the contour to be matched is then rotated continuously with a fixed step length. The contour to be matched and the target contour are respectively set as the template in turn to calculate the average square distance between the vertexes of the template and the expected vertexes of the matched contour. Finally, the minimum sum of the average square distances in all rotation angles is defined as the matching degree of the two contours. This algorithm can be used for both the convex and the concave contours. As compared with the traditional discrete algorithms, the proposed algorithm is of less data transfer, but of higher calculation speed and good identification and matching accuracy.
出处
《华南理工大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2006年第2期77-81,共5页
Journal of South China University of Technology(Natural Science Edition)
基金
广东省工业攻关资助项目(2003C102023)
广州市科技攻关资助项目(2004Z3-D0101)