项目名称: 高分辨率遥感图像高精度快速配准技术研究
项目编号: No.41301492
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 天文学、地球科学
项目作者: 高峰
作者单位: 中国人民解放军国防科学技术大学
项目金额: 25万元
中文摘要: 图像配准实现同场景多幅遥感图像分析和比较处理的前提和基础,相对于中低分辨率图像的配准,高分辨率图像配准的难点在于(a)很难准确的获取控制点的位置;(b)手工选择控制点效率低、耗时长;(c)数据量的庞大和细节的复杂影响了配准的速度;(d)相对于中低分辨率图像,局部变形明显,即使选择足够多的控制点,也不能很好的消除局部变形。本项目首先深入分析高分辨率图像的主要特点及形成上述配准问题的根本原因,然后分别从特征选择、配准策略等方面入手,解决高分辨率图像控制点难以选择、配准效率低等问题,并在此基础上,建立合适的数学模型,对高分辨率图像进行切割划分,并对各划分区域采用不同的变换模型参数进行配准,进而解决局部变形的问题。最后,设计一套适合于高分辨率图像的配准方案,解决其中关键科学问题,提出一些稳健性强、自动化程度高、适应性广的高分辨率图像配准方法,有效解决高分辨率图像的配准问题。
中文关键词: 高分辨率图像;图像配准;图像分块;特征编组;不确定性
英文摘要: Image registration is the base and precondition for analyzing and comparing remote sensing image. Comparing with the registration technique for low resolution images, the difficulties for the registration for high resolution image are: (a) precisely locating control points is not as simple as with moderate resolution images; (b) manually selecting the large number of control points required for precise registration is tedious and time consuming; (c) high data volume will adversely affect the processing speed in the image registration; and (d) local geometric distortion can not be removed very well using traditional image registration methods even with enough control points. We firstly analyze the main characters of high resolution images and the reasons for the above problems. Then solve the problems of hard to select control points and low efficiency by proposing new strategies for feature extraction and registration. For solving the problem of local geometric distortion, we construct mathematics models to divide images and using different parameters to register different piece of image. At last, design a scheme for high resolution image registration. At the same time, propose some robust and automatic registration methods.
英文关键词: high resolution image;image registration;image segmentation;feature grouping;uncertainty