基于相位相关法改进的ORB图像匹配算法  

Improved ORB Image Matching Algorithm Based on Phase Correlation Method

作　　者：骆大森 吴英[1,2] 陈燕苹[1] 袁正[1] 刘宇 LUO Dasen;WU Ying;CHEN Yanping;YUAN Zheng;LIU Yu(Chongqing Key Laboratory of Autonomous Navigation and Microsystems,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Institute of Intelligent Technology and Science,Chongqing University of Science and Technology,Chongqing 401331,China)

机构地区：[1]重庆邮电大学自主导航与微系统重庆市重点实验室,重庆400065 [2]重庆科技学院智能技术与科技学院,重庆401331

出　　处：《电子质量》2023年第8期22-26,共5页Electronics Quality

摘　　要：为了提高图像平移后ORB特征点匹配的准确率,提出了一种基于相位相关法改进的ORB图像匹配算法。该算法首先采用相位相关法求取左右两幅图像间的像素偏移量,以确定图像的重合部分;其次,提取图像FAST特征点并计算BRIEF描述子;接着,对要匹配的点增加窗口约束,以缩小匹配范围;然后,采用快速近似最近邻(FLANN)特征点匹配算法完成图像重合部分特征点的匹配;最后,将匹配错误的特征点采用随机采样一致性(RANSAC)算法进行删除。经实验验证,该算法将ORB特征点匹配的精确度有效的地89.7%提升至93.1%。In order to improve the accuracy of ORB feature point matching after image translation,an improved ORB image matching algorithm based on phase correlation method is proposed.Firstly,the phase correlation method is used to obtain the pixel offsets between the left and right images to determine the overlap of the image.Secondly,the is the FAST feature points of the image are extracted and the BRIEF descriptors are calculated.Thirdly,window constraints are added to the points to be matched to narrow the matching range.Then,the fast library for approximate nearest neighbor(FLANN)feature point matching algorithm is used to match the feature points in the coincidence part of the image.Finally,the mismatched feature points are deleted by random sampling consistency(RANSAC)algorithm.The experimental results show that,the algorithm can effectively improve the accuracy of ORB feature point matching from 89.7%to 93.1%.

关 键 词：ORB算法 图像匹配 相位相关法 快速近似最近邻 随机采样一致性 

分 类 号：TN911.73[电子电信—通信与信息系统]