摘要
利用无人机的侦察图像对目标进行精确定位的过程中,针对无人机图像与基准图像不可避免地存在尺度变化、旋转变化和光照变化,提出了一种改进SIFT的图像匹配方法。在特征点匹配时,采用简单的准欧式距离作为相似度量准则,并按照DoG空间结构由粗至精进行匹配。实验结果表明,该方法可以实现图像之间的精确匹配,对噪声、光照变化和局部场景变化具有较强的鲁棒性和实时性,为无人机目标定位提供了一种可行和有效的手段。
Aiming at there are variations of scale,rotation and illumination between unmanned aerial vehicles(UAV) images and reference images in the process of target location,an image registration method based on improved SIFT algorithm is proposed.In the feature point matching,the quasi-Euclidean distance instead of Europe is as the similarity measure of feature descriptors to improve the SIFT feature matching efficiency,and it matches feature points according to a coarse-to-fine DoG structure.Experimental results show that the presented method is robust under noises,illumination variations and local scene changes.Hence,the method can be used in target location for UAV.
出处
《舰船电子工程》
2012年第12期49-51,79,共4页
Ship Electronic Engineering
基金
国家自然基金(编号:11202239)资助
关键词
无人机
SIFT
图像匹配
目标定位
准欧式距离
unmanned aerial vehicles
SIFT
image registration
target location
quasi-Euclidean