基于RBF的视觉定位图像局部特征匹配算法  

A RBF-Based Image Local Feature Matching Algorithm in Visual Localization System

作　　者：殷锡亮 秦丹阳[3] 马琳[1] 谭学治[1] YIN Xiliang;QIN Danyang;MA Lin;TAN Xuezhi(School of Electronics and Information Engineering,Harbin Institute of Technology,Harbin 150080,China;Department of Electronics and Information Engineering,Harbin Vocational ＆ Technical College,Harbin 150081;Electronic Engineering College,Heilongjiang University,Harbin 150080,China)

机构地区：[1]哈尔滨工业大学电子与信息工程学院,黑龙江哈尔滨150080 [2]哈尔滨职业技术学院电子与信息工程学院,黑龙江哈尔滨150081 [3]黑龙江大学电子工程学院,黑龙江哈尔滨150080

出　　处：《移动通信》2018年第8期16-20,共5页Mobile Communications

基　　金：国家自然科学基金(面上)(61771186);黑龙江省自然科学基金(F2016019)

摘　　要：现有的视觉定位系统中大多使用传统的欧几里得距离排序和门限阈值的方法或基于KD-Tree的局部特征匹配算法进行局部特征的匹配。本文使用SURF特征作为视觉定位的图像局部特征,提出一种基于RBF的局部特征匹配算法。通过对比已有的局部特征匹配算法与基于RBF的局部特征匹配算法在不同数据集上的表现,综合分析各种算法在精确度、时延上的指标,得出基于RBF的局部特征匹配算法是一种有效的局部特征匹配算法的结论。In existing visual localization systems, the conventional Euclidean distance ordering with threshold or KDTreebased local feature matching algorithm is commonly used to match local features. In this paper, the Speeded Up Robust Feature（SURF） used as the image local feature of visual localization, a local feature matching algorithm based on RBF is proposed. By comparing the performance of existing local feature matching algorithms with the proposed algorithm in different data sets, the metrics including accuracy and delay of different algorithms are comprehensively analyzed with the conclusion that the proposed local feature matching algorithm based on RBF is effective.

关 键 词：视觉定位 图像局部特征 SURF KD-TREE RBF 

分 类 号：TN919.8[电子电信—通信与信息系统]