基于改进的SAR-SIFT的SAR图像配准  

SAR Image Registration Based on Improved SAR-SIFT

作　　者：王福全 李晓天 朱袁杰 WANG Fuquan;LI Xiaotian;ZHU Yuanjie(Zhejiang Land Surveying and Planning Company,Hangzhou Zhejiang 310030,China;Zhejiang Academy of Surveying and Mapping Science and Technology,Hangzhou Zhejiang 311121,China;Ningbo Alato Digital Technology Company,Ningbo Zhejiang 315000,China)

机构地区：[1]浙江省国土勘测规划有限公司,浙江杭州310030 [2]浙江省测绘科学技术研究院,浙江杭州311121 [3]宁波市阿拉图数字科技有限公司,浙江宁波315000

出　　处：《北京测绘》2020年第12期1799-1803,共5页Beijing Surveying and Mapping

摘　　要：在SAR图像配准中,针对SAR-SIFT算法在SAR图像配准中通过计算整个局部区域的梯度值,然后进行统计分析确定主方向,导致时间复杂度高且易受外界干扰的问题。本文对算法中的主方向分配进行改进。首先,对特征点邻域中的像素方向梯度进行累加,构成区域Harris矩阵,再用区域Harris矩阵代替梯度直方图来计算特征点邻域的主方向。然后对主方向取绝对值,抑制强度反转现象。实验结果表明:与SAR-SIFT算法相比,改进的SAR-SIFT算法匹配性能更好,计算效率更高。In SAR image registration,the SAR-SIFT algorithm calculates the gradient value of the entire local area in SAR image registration,and then performs statistical analysis to determine the main direction,which leads to the problem of high time complexity and vulnerability to external interference.This paper improves the main direction distribution in the algorithm.First,the pixel direction gradients in the neighborhood of feature points are accumulated to form a regional Harris matrix,and then the regional Harris matrix is used to replace the gradient histogram to calculate the main direction of the neighborhood of the feature point.Then take the absolute value of the main direction to suppress the intensity reversal phenomenon.Experimental results show that,compared with the SAR-SIFT algorithm,the improved SAR-SIFT algorithm has better matching performance and higher calculation efficiency.

关 键 词：合成孔径雷达 尺度不变特征变换 主方向 Harris矩阵 图像配准 

分 类 号：P237[天文地球—摄影测量与遥感]