基于稀疏结构和SIFT特征的SAR图像配准研究

发布时间：2018-06-01 19:38

本文选题：SAR图像 + SIFT　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：图像配准就是将在不同时间、不同传感器、不同视角及不同成像条件下获取的同一场景下的多幅图像进行匹配和叠加的过程。图像配准技术是图像处理的关键技术之一,已被广泛应用于遥感数据分析、计算机视觉、医学图像分析、地理探测及模式识别众多领域。合成孔径雷达(SAR)具有全天时全天候的工作状态、远距离高分辨、强大的穿透能力等优点,其应用已经深入到军事和民用的各个领域分支。但是由于SAR图像中斑点噪声的存在,对后期SAR图像的理解与解译工作带来了极大的困难,使得已经有效运用于光学遥感图像配准算法处理SAR图像时效果不佳。本文通过分析国内外配准技术发展现状和SAR图像的特点,提出了利用SAR图像中丰富的结构信息来获取结构区域,再对这些区域提取SIFT特征点实现图像配准,通过实验证明了算法的可行性。本文的完成的工作如下:1.提出了一种基于块匹配和SIFT特征的SAR图像配准方法。由于相干斑噪声的存在不仅破坏了成像的灰度分布而且削弱了目标特性,使得SAR图像配准的特征提取和特征匹配过程受到干扰。该算法利用基于PPB距离的块匹配算法来分析SAR图像的多方向结构区域,对稳定性高、结构信息丰富的多方向结构区域提取SIFT特征点,有效削弱了相干斑的干扰,降低了匹配错误率。2.提出了一种基于结构稀疏度和SIFT特征提取的快速SAR图像配准方法。该方法利用图像的细化结构和方向结构特征得到各像素点的结构稀疏度,再对图像中结构稀疏区域提取SIFT特征点完成特征匹配,并利用随机一致性估计算法滤除错误匹配对。该算法有效地降低了匹配错误率,提升了配准速率。
[Abstract]:Image registration is the process of matching and superposing multiple images in the same scene at different time, different sensor, different angle of view and different imaging conditions. Image registration is one of the key techniques in image processing. It has been widely used in remote sensing data analysis, computer vision, medical image analysis, geographical detection and pattern recognition. Synthetic Aperture Radar (SAR) has many advantages such as all-day, all-weather, long range and high resolution, strong penetration ability and so on. It has been widely used in military and civilian fields. However, due to the presence of speckle noise in SAR images, it is very difficult to understand and interpret SAR images in the later stage, which makes the image registration algorithm for optical remote sensing images less effective in processing SAR images. Based on the analysis of the development of registration technology and the characteristics of SAR images at home and abroad, this paper proposes to obtain structural regions by using rich structural information in SAR images, and then extract SIFT feature points from these regions to achieve image registration. The feasibility of the algorithm is proved by experiments. The work accomplished in this paper is as follows: 1. A SAR image registration method based on block matching and SIFT features is proposed. Because the presence of speckle noise not only destroys the gray distribution of the image but also weakens the target characteristics, the feature extraction and feature matching process of SAR image registration is disturbed. The block matching algorithm based on PPB distance is used to analyze the multi-directional structural region of SAR image. The SIFT feature points are extracted from the multi-directional structural region with high stability and rich structure information, which effectively weakens the interference of speckle. Reduced matching error rate. 2. 2. A fast SAR image registration method based on structural sparsity and SIFT feature extraction is proposed. In this method, the thinning structure and directional structure feature of the image are used to obtain the structural sparsity of each pixel, then the SIFT feature points are extracted from the sparse region of the image to complete the feature matching, and the random consistency estimation algorithm is used to filter out the error matching pairs. The algorithm can effectively reduce the matching error rate and improve the registration rate.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【共引文献】