应用不变特征获取均匀控制点的遥感影像配准算法  

A remote sensing image registration algorithm by obtaining uniform control points based on invariant feature

作　　者：倪希亮[1,2] 丁琳[1] 江涛[2] 胡顺石[3] 

机构地区：[1]中国科学院遥感应用研究所,北京100101 [2]山东科技大学,山东青岛266510 [3]湖南师范大学,长沙410006

出　　处：《测绘科学》2011年第2期70-72,共3页Science of Surveying and Mapping

基　　金：国家重点基础研究发展规划项(2007CB714406);国家科技支撑计划项目(2007BAH15B01);国家航天局航天遥感论证中心(O7K00110KJ)

摘　　要：本文针对大幅影像配准时计算量大、配准点分布严重不均匀以及错配率高等问题进行了研究,提出一种可保持配准点均匀分布的快速配准算法。首先对基准影像和待配准影像进行分块提取SIFT特征算子,然后局部自适应地对影像块中的匹配点进行几何约束,从局部选取最优匹配点以达到提高图像匹配准确率的目标。通过对大量不同类型的影像进行配准实验,实验结果表明,当图像存在传感器差异甚至较高分辨率差异时,本文算法也能够快速准确地达到配准目的。In this paper, a study aiming to the problems of large images＇ high computing complexity, serious non-uniform distribution of matching points and highly wrong regstration rate were carried out. Then a fast matching algorithm was proposed. The algorithm could keep matching points uniformly distributed. First, it obtained SIFT points feature descriptor from base image and warp image by cutting up them. Then, it selected the best block matching points adaptively using geometric constraint to improve the image matching accuracy. Numerous experiments were conducted for different type images. The results showed that the proposed method was faster and more accurate. Besides, when the sensors and resolutions of the images are very different, the method could also obtain better result.

关 键 词：SIFT 尺度空间 特征算子 图像匹配 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]