基于多尺度模型的数据库影像特征匹配

A feature matching method in remote-sensing image database based on multi-scale model

针对传统的遥感影像数据库中影像特征匹配效率低下的问题,该文提出一种基于多尺度模型的数据库影像特征匹配方法。为遥感影像数据库构建多尺度模型,能较高效率地在数据库中快速定位候选的匹配影像;基于搜索得到的候选影像,采用快速排序算法逐层筛选最佳匹配影像,从而实现了数据库匹配影像的精确定位。实验结果表明:所提算法不仅能高效地完成影像间特征点匹配运算,还能从数据库中精确地筛选得到最佳的匹配影像;且随着数据库中影像总数的增加,该方法的优势表现得更加明显。这一研究将为遥感影像数据库的高效、实时及动态匹配提供可能性。

Aiming at the poor efficiency of feature points matching in remote-sensing image database,a new method based on multi-scale model was proposed in this paper.Firstly,a remote-sensing image database was established in SQL server,and the K-nearest neighbor algorithm based on the theory of resampling was used to restructure the image database to construct a multi-scale model.Then,the algorithm of SIFT was used to match the feature points of images from database in different scales.Finally,the best matching image was obtained by the identification database.For investigating the efficiency of the new method,the comparison of the matching time was made between the traditional image database matching method and the new method.The outcomes showed that the efficiency of the new method outperformed the traditional method,and with the increase of the number of images in database,the advantage of this method became better.