基于改进SIFT的无人机影像匹配方法  

UAV Image Matching Method Based on Improved SIFT

作　　者：王一 齐皓 王瀚铮 程佳 WANG Yi;QI Hao;WANG Hanzheng;CHENG Jia(School of Electrical Engineering,North China University of Science and Technology,Tangshan 063210,China;Tangshan City Metal Component Production Line Intelligent Technology Innovation Center,Tangshan 063210,China)

机构地区：[1]华北理工大学电气工程学院,河北唐山063210 [2]唐山市金属构件产线智能化技术创新中心,河北唐山063210

出　　处：《无线电工程》2023年第10期2337-2344,共8页Radio Engineering

基　　金：河北省高等学校科学技术研究项目(ZD2022114);唐山市科技计划项目(21130212C);华北理工大学专业学位研究生综合改革项目(YB18230221-17)。

摘　　要：针对传统尺度不变特征变换(Scale-Invariant Feature Fransform,SIFT)算法在匹配无人机影像时错误率较高、运行速度慢等问题,提出了一种基于改进SIFT的无人机影像匹配方法。利用联合双边滤波对原始影像进行预处理,在滤除噪声的同时保留边缘信息;将二进制描述算法强化描述符(Boosted Efficient Binary Local Image Descriptor,BEBLID)与改进SIFT算法相结合,通过基于机器学习的采样模型对特征点构造具有强描述性的二进制描述子,在提升匹配正确率的同时加快匹配速度;使用Vicinity-KNN算法进行特征点粗匹配,结合自适应局部仿射匹配算法对匹配的特征点进一步提纯。为了验证该方法的有效性,将ORB、SIFT、ISIFT、IKAZE算法与所提方法进行对比实验,结果表明,该方法在匹配时间和精度方面均具有很好的性能,为影像高质量匹配提供了技术基础。To solve the problems of high error rate and slow operation speed of traditional Scale-Invariant Feature Fransform(SIFT)algorithm in matching UAV images,a UAV image matching method based on improved SIFT is proposed.First,the original image is preprocessed by using joint bilateral filtering,and the edge information is retained while filtering the noise.Then,the Boosted Efficient Binary Local Image Descriptor(BEBLID)is combined with the improved SIFT algorithm for the first time,and a strong descriptive binary descriptor is constructed for the feature points through the sampling model based on machine learning,which improves the matching accuracy and speeds up the matching speed.Finally,Vicinity-KNN algorithm is used for rough matching of feature points,and adaptive local affine matching algorithm is combined to further purify the matched feature points.In order to verify the effectiveness of this method,the ORB,SIFT,ISIFT and IKAZE algorithms are compared with the proposed algorithm.The experimental results show that this method has good performance in matching time and accuracy,and provides a technical basis for high-quality image matching.

关 键 词：影像匹配 尺度不变特征变换 二进制描述算法强化描述符 自适应局部仿射匹配 

分 类 号：P237[天文地球—摄影测量与遥感]