摘要:提出一种基于稀疏 、稠密特征转换的仿射不变特征匹配算法,其中稀疏特征包括坐标,尺度,仿射模拟参数等,稠密特征指基于图像局部区域内光学属性的局部描述符.本文算法在Affine-SIFT算法基础之上,针对在特征提取阶段仅使用稀疏特征提取的缺陷做出了改进.由于稠密信息只有在稀疏参数满一定足检测条件时才能提取到特征,导致本可以匹配到的特征(包括稀疏 、稠密参数)无法提取,将通过使用稀疏特征构造新的模拟图像,通过将稀疏特征重新稠密化,并在模拟图像基础上进一步提取稀疏特征,同时可检测到原始图像中检测不到的可匹配特征,最终达到增大特征建立匹配的概率,提升正确匹配数量的目标.经实验验证,本文提出的稀密特征转换算法相比于ASIFT算法能大量增加特征匹配的数量.除针对ASIFT方法提供扩展外,该方法也可用于扩展具有充分稀疏特征参数的其它特征提取和匹配方法,并适用于目标识别 、目标分类和三维重建等问题.%In this paper ,an affine invariant feature matching algorithm based on transforming sparse fea-tures into dense features is proposed .In the proposed method ,the sparse features include coordinates , scale ,affine parameters etc .,and the dense features include information of Gaussian kernel ,area de-scriptor .Based on the Affine-SIFT algorithm ,the proposed algorithm improves the shortcomings of sparse feature extraction in the feature extraction phase .Because the dense feature can only be obtained if the sparse parameters satisfy some given conditions ,some good features for matching schemes (inclu-ding sparse and dense parameters ) can not be obtained .In this paper ,the authors construct the new sim-ulation images by using the original sparse features obtained by ASIFT ,and then extract the new sparse features from the simulated images ,and thus the authors can obtain some new features which can not be detected by the original ASIFT method .By using this scheme ,the correct match number in the matc-hing scheme is improved .The effectiveness of the proposed method as beenverified by experiments .Be-sides that it can be applied to extend ASIFT ,the proposed method can also be applied to extend other feature extraction and matching methods with sufficiently parameters of sparse feature ,and be applied to precisely target recognition ,target classification and 3D reconstruction etc .