CAJ | 학술논문

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。
제출일충기우SURF산자화특정편이일치성준칙적무인비행기홍외서렬영상배준적방법。해방법채용SURF산자대특정점진행제취화필배。SURF산자부고필점적국부신식,몰유고필특정점집적궤하신식。본문채용특정편이일치성준칙래척제오필배점대,제고특정점대적필배정도,소득도적특정점집채용LM비선성우화진행단응구진구해배준,여기타산법상비,본방법괄합우홍외서렬영상쾌속병접,유교호적실용효과화로봉성。
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.