基于SURF改进算法的红外序列影像快速配准的方法研究  

Research on the Infrared Image Sequence Fast Registration Based on Improved SURF Algorithm

作　　者：吴翰书[1,2] 余旭初[1] 张鹏强[1] 杨明[1] 

机构地区：[1]信息工程大学测绘学院 [2]96215部队

出　　处：《测绘与空间地理信息》2012年第1期161-163,共3页Geomatics & Spatial Information Technology

摘　　要：提出一种基于SURF算子和特征偏移一致性准则的无人飞行器红外序列影像配准的方法。该方法采用SURF算子对特征点进行提取和匹配。SURF算子仅考虑点的局部信息,没有考虑特征点集的几何信息。本文采用特征偏移一致性准则来剔除误匹配点对,提高特征点对的匹配精度,所得到的特征点集采用LM非线性优化进行单应矩阵求解配准,与其他算法相比,本方法适合于红外序列影像快速拼接,有较好的实用效果和鲁棒性。This paper proposes a fast registration method for the UAV infrared image sequence based on SURF operator and feature shift coherence rule.This method uses SURF operator to extract and match feature points with local information.While the geometric information of feature point set is not considered,the feature shift coherence rule is adopted to remove error and thus to improve the matching accuracy of the feature point set.Levenberg-Marquardt optimization is used to solve the 8 parameters transforming mode for registration.Compared with other algorithms,this method is more suitable for fast infrared image sequence stitching and has good practical performance and robustness.

关 键 词：红外序列影像 SURF算子 特征偏移一致性准则 影像配准 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]