基于改进ORB和RANSAC算法的无人机影像特征匹配方法  

UAV image feature matching method based on improved ORB and RANSAC algorithm

作　　者：丁进选 王丽欣 苗星雷 DING Jinxuan;WANG Lixin;MIAO Xinglei(Shenzhen Investigation & Research Institute Co., Ltd, Shenzhen 518003, China;Heilongjiang Provincial Society of Surveying and Mapping Geographic Information,Harbin 150081,China;Harbin North Defence Equipment Co.,Ltd,Harbin 150050,China)

机构地区：[1]深圳市长勘勘察设计有限公司,广东深圳518003 [2]黑龙江省测绘地理信息学会,黑龙江哈尔滨150080 [3]哈尔滨北方防务装备股份有限公司,黑龙江哈尔滨150050

出　　处：《测绘工程》2021年第4期66-69,75,共5页Engineering of Surveying and Mapping

摘　　要：针对ORB和RANSAC算法存在的不足,文中提出通过对Harris角点函数的改进和将空间一致性原理思想引入RANSAC算法,提出一种改进的无人机影像特征匹配方法,并选择两组代表性的像对进行实验。实验结果表明,改进算法在特征点识别、误匹配剔除和尺度不变性方面的性能均有明显的提升,为无人机影像匹配提供借鉴。Aiming at the shortcomings of ORB and RANSAC algorithms,this paper proposes an improved UAV image feature matching method by improving Harris corner function,introduces the idea of spatial consistency into RANSAC algorithm,and selects two groups of representative image pairs for experiment.The experimental result shows that the performance of the improved algorithm in feature point recognition,mismatch elimination and scale invariance are significantly improved,which provides a reference for UAV image matching.

关 键 词：ORB RANSAC 特征匹配 尺度 空间一致性 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]