一种平移旋转图像的角点匹配方法  

Corner matching for translation and rotation image

作　　者：王小鹏[1] 房超[1] 雷涛[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,兰州730070

出　　处：《计算机工程与应用》2014年第16期173-176,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61261029;No.61202314);甘肃省高等学校硕士生导师科研项目(No.1104-4)

摘　　要：在图像角点匹配过程中,目标图像往往存在平移旋转现象,直接影响匹配效果,为此提出了一种平移旋转图像的角点匹配方法。该方法首先利用角点检测生成自相关矩阵的梯度信息与仿射变换相结合,构造确定性退火算法中的自由能函数,然后对该自由能函数进行优化,获取待匹配角点间的仿射变换参数,最后利用该变换参数实现角点匹配。实验结果表明,该方法能够在目标图像发生平移旋转的情况下,有效实现角点匹配。During the course of image corner matching, the translation and rotation of object image often affect the matching results. A corner matching method is proposed based on deterministic annealing. Harris corner detection is used to extract the corner points respectively from object and matching image, and the gradient information of the corner points from the autocorrelation matrix are also calculated. The free energy function of deterministic annealing is constructed by combining the gradient information and the affine transform. The optimization affine transform parameters are obtained by optimizing the free energy function. Corner matching between object and matching image are complemented through the affine trans-form. Simulations show that this method can effectively achieve the corner matching when the object image occurring translation and rotation.

关 键 词：角点匹配 平移旋转 确定性退火 仿射变换 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]