基于傅立叶变换的多源高分辨率遥感数据的配准技术
文献类型：学位论文
| 作者 | 余战武 |
| 学位类别 | 工学硕士 |
| 答辩日期 | 2004-07-01 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 普林特 |
| 关键词 | 多源数据配准 方向频谱函数 傅立叶变换 滤波 主干道路特征提取 Optical very high resolution remote sensing multi-source data registration Fourier transform directional spectral function road |
| 其他题名 | Multi-source Remote Sensing Very High Resolution Data Matching using Fourier Transform |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 多源数据配准是对同一场景下不同传感器或不同手段所获数据之间 的配准，它属于图像配准技术的一重要分支，在遥感数据分析、医学图像 处理、计算机视觉等方面有着重要的应用。本文在对以前多源数据配准技 术的分析与总结基础上，结合项目的应用背景，以QuickBird卫星图像 与地理信息数据为实验数据进行了自动配准技术的研究。相关的工作及主 要贡献如下： 1．主干道路特征提取：在高分辨率光学卫星图像道路特征分析的基础 上，提出了一个新的城区主干道的提取方法。该方法将道路特征的灰 度信息和几何信息有效结合，从而极大提高了算法的鲁棒性。它是实 现图像与GIS数据自动配准的关键一步。 2．道路特征图像与GIS数据的自动配准：在假定卫星图像数据与GIS 数据中包含丰富城区主干道信息且其变换模型可用相似变换近似描 述的前提下，本文提出了一套新的基于傅立叶变换的自动配准方案。 该方法首先运用带状区域在频域中的频谱主要集中于与其主方向垂 直的方向上这一特性求取图像与GIS数据之间的旋转角度，然后采 用多尺度策略求取数据之间的缩放与平移参数。与其他傅立叶变换配 准方法相比较，其最大特点在于对图像旋转、缩放、区域变化具有很 强的自适应性。 3．小区域GIS数据与大区域图像自动匹配问题的研究：以上述配准方 法为基础，本文采用分块匹配的方法解决小区域GIS数据与大区域 图像数据的配准问题。它首先使用重叠分块策略将图像数据分割成若 干小块，该策略旨在保证存在小块图像数据与GIS数据有较大重叠 区域的情况下尽可能减少分块个数；然后采用本文所提方法寻找该分 块图像数据位置并求得其与GIS数据的变换模型参数，进而确定小 块GIS数据在大块图像中的位置。 |
| 英文摘要 | Data registration is the task that enables to project an image into the reference system of an other one, such that the two can overlap perfectly, point to point. Multi-source registration consists in using data from the same scene acquired by different sensors or by different means. It can be applied to remote sensing data fusion, medical image processing or computer vision. The objective of the present research work is to study and develop semi-automatic registration technique from multi-source data, especially from high-resolution optical image and GIS (Geographical Information System) data -also called 'digital map'. This thesis first gives a general review and analysis of existing registration methods, then presents in details the newly developed technique. The major achievements and contribution can be summarized as follow: 1. Road Feature Extraction. A new method is proposed to extract main road features. Based on the observation of road networks characteristics in optical high-resolution satellite images in urban areas, we defined a model based on both spectral information and geometric constraints. The developed simple technique can achieve semi-automatic rough extraction, even in images covering large areas. It is the first step before image-GIS data registration. 2. Automatic Registration of Image and GIS road Data. With the assumption that both satellite image and GIS data hold adequate road features, and considering that their transformation can be expressed by a rigid transformation model, we propose a new FFT-based (Finite Fourier Transform) method to match the image and the GIS road map. Taking advantage of the regularity of the road pattern and road principal directions, we define the notion of directional spectrum, computed from the Fourier spectrum. The matching of the directional spectrum enables us to retrieve the rotation parameter. Then, the scale and translation are obtained by searching for the maximum of correlation in the space domain, using a multi-scale strategy. The significant advantage of the algorithm is its capability to match the rotated and scaled images robustly even when they have a large scale change or obvious road features differences between the data. 3. Study on the localization of a small area GIS data within a large image. Based on the matching method proposed above, a patch-dividing technique is proposed in order to retrieve the positioning of a small-area GIS data within a large-area satellite image. This can be considered like a recognition problem. We first divide the image into small patches ; the maximum size of these patches should be constrained. Then we use our matching method to find out the patch that has a maximum overlapping area which the GIS map, and compute the registration parameters with regard to this patch. The GIS data can then be located within the satellite ima |
| 语种 | 中文 |
| 其他标识符 | 773 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/6803]  |
| 专题 | 毕业生_硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 余战武. 基于傅立叶变换的多源高分辨率遥感数据的配准技术[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2004. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。