基于直线空间结构特征的图像匹配方法

直线匹配在图像配准、目标识别等领域有重要应用。提出利用直线空间结构特征进行匹配的方法,算法采用粗精两级匹配的思路,粗匹配阶段用四参数仿射变换模型对待匹配图像进行粗步校正,精匹配阶段利用反映直线空间结构分布特征的直线交点方位角直方图,把直线匹配映射成两个点集的匹配,结合随机抽样共识(RANSAC)算法和参数迭代一致性方法剔除错配点对,实现图像的自动匹配。实验结果表明,本文方法能够取得较高的匹配精度。

An image registration algorithm based on space structure features of line

Line matching plays an important role in image registration,object recognition,and other applications.This paper presents an image registration algorithm based on space structure features of line.The four-parameter affine transformation model is used in coarse registration.The line matching problem is converted into a point-matching problem through construction of azimuth histogram which reflects the space structure features of line distribution in fine registration.The RANdom SAmple Consensus(RANSAC) method and parameter iteration consistency method are used to delete the error matching points.Experimental results show that the proposed algorithm is provided with high registration accuracy.