【摘要】 视觉影像经常存在纹理情况复杂和动态模糊等情况,严重降低了连续影像间各类特征的相似性,导致传统的匹配算法难以获得准确、稳定、分布良好的影像匹配,影响后续影像处理中各类信息的获取。针对上述情况,本文提出了一种改进SURF视觉影像匹配方法。该方法包括特征提取、初始匹配和对应匹配3个步骤。首先,利用SURF特征匹配方法提取足够且分布良好的特征点;其次,进行初始匹配,得到一些正确的匹配点对及影像对之间的初始投影变换关系;最后,采用几何对应匹配策略进行匹配传播,得到更可靠的匹配结果。通过图像对之间的几何关系,几何对应匹配能够发现比初始SURF算法更合适的匹配结果。对TUM数据综合试验表明,该算法简单快速,匹配精度高。更多还原

【Abstract】 Visual images always have the problem of complex texture and dynamic blur,and seriously reduces the similarity of various features between continuous images,and makes it difficult for traditional matching algorithms to obtain accurate,stable and well distributed image matching,then affects the acquisition of all kinds of information in subsequent image processing. In order to solve these problems,this paper proposes an improved surf visual image matching method. The method includes three steps: feature extraction,initial matching and corresponding matching. Firstly,SURF feature matching method is used to extract enough and well distributed feature points; secondly,initial matching is carried out to obtain some correct matching point pairs and the initial projection transformation relationship between image pairs; finally,geometric correspondence matching strategy is used for matching propagation to obtain more reliable matching results. Through the geometric relationship between image pairs,geometric correspondence matching can find more suitable matching results than the original surf algorithm. The experimental results of the TUM data synthesis show that the algorithm is simple and fast,and the matching accuracy is high.更多还原