基于多尺度FAST-9的图像快速匹配算法

摘要

FAST-9 detector can not be invariant to scale changes. Aiming at this problem, this paper proposes a fast image matching algorithm based on multi-scale FAST-9. Feature points are extracted from the constructed Gaussian scale space by the FAST-9 detector, and main orientation is assigned based on information from a circular region around the interest point. SURF descriptor is extracted by a constructed square region aligned to the selected orientation. The matching pairs are determined with the nearest neighbor distance matching method. Experimental results show that the proposed algorithm outperforms better than SURF and SIFT algorithm in speed.%FAST-9检测子不具备尺度不变性,为此,提出一种基于多尺度FAST-9的图像快速匹配算法.对图像建立高斯尺度空间,在各图层上应用FAST-9检测子分别提取特征点,在其周围建立圆形区域并分配主方向,同时建立方形区域构造SURF描述子,利用基于最近邻匹配方法进行匹配.实验结果表明,与SURF、SIFT算法相比,该算法具有较高的匹配速度.