基于改进的SIFT图像快速拼接方法  

Method of image fast stitching based on improved SIFT

作　　者：王钎沣 夏国静 牛鑫鑫[1] 孙阿猛 田波 夏萍[1,2] WANG Qianfeng;XIA Guojing;NIU Xinxin;SUN Ameng;TIAN Bo;XIA Ping(Anhui Agricultural University,Hefei 230036,China;Anhui Intelligent Agricultural Machinery Equipment Engineering Laboratory,Hefei 230036,China)

机构地区：[1]安徽农业大学,安徽合肥230036 [2]智能农机装备工程实验室,安徽合肥230036

出　　处：《现代电子技术》2022年第8期159-162,共4页Modern Electronics Technique

基　　金：安徽省国际科技合作计划项目:基于遥感和GIS的巢湖流域土地利用变化及其生态环境效应(1604b0602029);国家自然科学基金青年基金项目(11802003)。

摘　　要：针对使用传统的尺度不变特征变换(SIFT)算法进行图像拼接时出现的特征点误匹配以及拼接用时长的问题,文中提出一种改进的SIFT图像快速拼接方法。通过使用RANSAC算法对SIFT算法在特征点匹配时进行多次优化迭代,对误匹配点进行剔除和提纯改进,实现将待配准的图像精确匹配。然后,将改进后算法的匹配结果与传统SIFT算法、Harris拼接算法的拼接结果进行对比。结果表明,改进的SIFT算法在图像拼接质量、准确性上较传统的算法均有所提高,匹配使用特征点个数减少,拼接接缝和拼接重影消失,拼接速度提高了约14倍。文中改进方法的图像拼接质量好且拼接时间较短,能够较好地实现农田和建筑物的图像拼接。In allusion to the problems of feature point mismatching and long stitching time-consumption when using the traditional SIFT(scale invariant feature transformation) algorithm for image stitching,an improved SIFT image fast stitching method is proposed. The RANSAC algorithm is used to optimize and iterate the SIFT algorithm for many times in feature point matching,to eliminate,purify and improve the wrong matching points,so as to realize the accurate matching of the image to be registered. The matching results of improved algorithm are compared with the stitching results of the traditional SIFT algorithm and Harris stitching algorithm. The experimental results show that,in comparison with the traditional algorithm,the improved SIFT algorithm can improve the quality and accuracy of image stitching,reduce the number of matching feature points,disappear stitching joints and stitching ghosting,and increase the stitching speed by about 14 times. It shows that the improved algorithm has better image stitching quality and shorter stitching time,which can better realize the image stitching of farmland and buildings.

关 键 词：图像拼接 改进SIFT 无人机图像 特征匹配 图像采集 迭代优化 对比分析 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]