首页 | 本学科首页   官方微博 | 高级检索  
     

基于尺度配准的航拍图像与卫星影像匹配方法
引用本文:温育杜,郭士增,刘晟,马琳. 基于尺度配准的航拍图像与卫星影像匹配方法[J]. 哈尔滨工业大学学报, 2023, 55(10): 19-26
作者姓名:温育杜  郭士增  刘晟  马琳
作者单位:哈尔滨工业大学 电子与信息工程学院,哈尔滨 150001;中国航空无线电电子研究所,上海 200233
基金项目:国家自然科学基金(2,0);航空科学基金(2020Z066015002);国家重点研发计划(2022YFC3801100)
摘    要:随着无人机航拍技术的普及,航拍图像在目标检测和跟踪的应用得到普遍研究,而航拍图像与卫星影像的匹配研究相对较少。由于成像机理和成像视角的不同,无人机航拍图像与卫星影像存在较大的尺度差异,现有的图像匹配方法难以实现航拍图像与卫星影像的有效匹配。为了解决这一问题,提出了一种实现具有大尺度差异的航拍图像与卫星影像匹配的方法。该方法通过卫星影像的经纬坐标信息和无人机航拍图像成像时刻的相机位姿信息,对无人机航拍图像进行方向和尺度的配准;然后利用航拍图像成像时刻的位置信息,对卫星影像进行粗匹配,得到包含航拍图像匹配区域在内的卫星影像子图;再利用神经网络提取配准后的航拍图像与卫星影像子图的卷积神经网络(convolutional neural networks,CNN)特征,并基于CNN特征实现航拍图像与卫星影像的精匹配。仿真实验结果表明,本文所提方法能够有效的实现大尺度差异的航拍图像与卫星影像匹配。通过将本文匹配方法与现有图像匹配算法匹配精度的对比分析,证明了本文匹配算法的有效性和优越性。

关 键 词:航拍图像  卫星影像  图像匹配  图像尺度配准
收稿时间:2022-03-29

Matching method between UAV images and satellite images based on scale registration
WEN Yudu,GUO Shizeng,LIU Sheng,MA Lin. Matching method between UAV images and satellite images based on scale registration[J]. Journal of Harbin Institute of Technology, 2023, 55(10): 19-26
Authors:WEN Yudu  GUO Shizeng  LIU Sheng  MA Lin
Affiliation:School of Electronics and Information Engineering, Harbin Institute of Technology, Harbin 150001, China;Chinese Aeronautical Radio Electronics Research, Shanghai 200233, China
Abstract:While the application of UAV images in target detection and tracking has been widely studied amid the popularization of UAV aerial photography technology, there has been relatively few researches on the matching between UAV images and satellite images. Due to their different imaging mechanisms and viewpoints, there are large-scale differences between UAV images and satellite images. Effective matching between UAV images and satellite images are difficult to realize using existing image matching methods. In order to solve this problem, this paper proposes a method for matching UAV images with large-scale differences from satellite images. This method registers the direction and scale of UAV images using the longitude and latitude coordinate of satellite images and the camera pose information of UAV images. Using the position of the UAV images, the satellite images are then roughly matched to obtain the satellite sub-images including the UAV images matching area. The CNN features of the registered UAV images and satellite sub-images are then extracted using neural networks, and the precise matching between UAV images and satellite images is realized based on CNN features. Simulation results show that the proposed method can effectively match large-scale UAV images with satellite images. The comparison in terms of matching accuracy between this matching method and existing image matching algorithms show the effectiveness and superiority of this matching algorithm.
Keywords:UAV images   satellite images   image matching   image scale registration
点击此处可从《哈尔滨工业大学学报》浏览原始摘要信息
点击此处可从《哈尔滨工业大学学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号