轮廓特征点几何约束关系的图像配准算法  

Image Registration Using Geometric Constrain Between Contour Points

作　　者：刘克俭[1] 霍宏涛[1] 韩华[1] 贾承丽 

机构地区：[1]中国人民公安大学,北京100038 [2]北京信威通信技术股份有限公司,北京100193

出　　处：《遥感信息》2017年第3期128-132,共5页Remote Sensing Information

基　　金：高分辨率XXXX重大专项(01-Y3XXXX-XX01-14/16)

摘　　要：针对遥感影像在配准过程中精度不高的问题,提出了一种高精度的遥感图像配准方法。通过对区域轮廓的等间隔采样获得特征点集,在挖掘出轮廓特征点间几何约束关系的基础上,引入概率松弛标记算法对特征点匹配概率矩阵进行多次迭代更新,以降低点对匹配过程中的模糊性与不确定性。仿真数据与实测数据的实验结果均表明,该方法可以有效地提高图像配准精度。Aiming at the problem that the accuracy of remote sensing image registration is not high,an image registration method with high accuracy is proposed in this paper.The contour of the region is considered to be a point set,and the problem of image registration is converted to finding corresponding point pairs from contour point sets.Taking advantage of the geometric constrain between contour points,the point matching probability is updated by the probability relaxation labeling technique to refine the match.The advantage of the proposed algorithm is well demonstrated by experiments on synthetic and real remote sensing data.

关 键 词：图像配准 轮廓 概率松弛标记 间隔采样 几何约束关系 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]