摘要
给出了一种基于点特征的图像配准方法,可以用于复杂背景下任意角度旋转的图像之间的配准。图像配准算法分为四个步骤:用Forstner算子分别提取参考图像与待配准图像的特征点;使用旋转不变特征向量获取特征点之间的初始匹配关系;使用投票算法剔除误匹配点对,获取正确配对的同名点对;用最小二乘方法由正确配对的同名点对获取图像之间的配准参数。给出了基于点特征的图像配准方法在目标跟踪中的应用方法,并进行了仿真实验,结果证明该图像配准算法可以对复杂背景下弱对比度静止目标进行精确跟踪,而且算法具有很好的通用性。
In this paper, an image registration method based on point features is presented, which can be used to register images rotating at arbitrary angle in complex background. The method has four steps ： The first is to extract feature points of reference image and being matched image by Forstner operator, the next is to derive the initial matching relation and matching points by using rotation invariance, the third is to eliminate false matching points by voting algorithm, the last is to acquire accurate registration parameters from all exact matching points by LMS algorithm. The application of this method in target tracking is also presented, and the simulation results show that the method has high tracking precision and good universality.
出处
《航空兵器》
2008年第6期23-27,共5页
Aero Weaponry
关键词
图像配准
FORSTNER算子
特征点匹配
仿射变换
目标跟踪
image registration
Forstner operator
feature points matching
affine transformation
targets tracking
