基于正交学习差分进化算法的遥感图像配准方法研究

发布时间：2018-04-15 09:38

本文选题：图像配准 + Harris角点检测　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：图像配准技术是图像处理的关键技术之一。图像配准的过程是将不同条件下、不同时间或者不同传感器获得的两幅或者多幅图像进行匹配的过程。图像配准技术在许多领域都有广泛研究的价值,如计算机图形学、地理探测和摄影测量和材料力学等。由于遥感图像自身的多样性以及从不同传感器获得的图像之间往往存在畸变,因此对遥感图像的配准存在一定的困难。现如今,随着图像配准技术越来越深入的研究,人们对配准的精度也有了越来越高的要求。通过对现有的遥感图像配准技术的研究,以及对实验问题的分析,本文提出了以下两种有效的遥感图像配准方法:第一种方法是一种结合了基于灰度和基于特征的图像配准方法,这种方法克服了现有的基于特征和基于灰度的图像配准方法的不足。首先它提取两幅图像中的最相似图像块,然后利用Harris角点检测法提取最相似图像块中的角点特征,接着利用角点特征对两幅图像的最相似图像块进行匹配,得到最优几何变换参数,最后根据所得到的几何变换参数对整幅图像进行配准。我们称这种方法为改进的基于互信息和Harris角点检测的图像配准方法。实验表明,本文提出的这种方法是一种有效的遥感图像配准方法。第二种方法是基于灰度的方法。基于灰度的方法通常需要对参考图像和待配准图像之间的相似度矩阵进行优化。正交学习(OL)能够构造出一个引导模型,引导算法更加有效地朝着最优解的方向进行搜索。差分进化算法(DE)是一种贪婪算法,由于它在解决遥感图像配准问题中表现出了优越的性能。因此,本文提出了一种的基于正交学习差分进化算法(OLDE)的图像配准方法,这种方法充分利用正交学习的引导特性来引导差分进化算法朝着全局最优方向进行搜索。经过实验表明,本文提出的算法在遥感图像配准中表现出了比遗传算法(GA)、粒子群算法(PSO)和差分进化算法(DE)更加优越的性能,具有更强的鲁棒性。
[Abstract]:Image registration is one of the key techniques in image processing.The process of image registration is to match two or more images under different conditions, different time or different sensors.Image registration technology has been widely studied in many fields, such as computer graphics, geographical detection, photogrammetry and mechanics of materials.Because of the diversity of remote sensing images and the distortion between images obtained from different sensors, it is difficult to register remote sensing images.Nowadays, with more and more in-depth research on image registration technology, people have higher and higher requirements for registration accuracy.Through the research of the existing remote sensing image registration technology and the analysis of the experimental problems,In this paper, two effective methods for remote sensing image registration are proposed. The first method is a combination of grayscale and feature-based image registration.This method overcomes the shortcomings of existing feature-based and gray-based image registration methods.Firstly, it extracts the most similar image blocks in two images, then extracts the corner features from the most similar image blocks by using the Harris corner detection method, and then matches the most similar image blocks of the two images by corner features.The optimal geometric transformation parameters are obtained and the whole image is registered according to the obtained geometric transformation parameters.We call this method an improved image registration method based on mutual information and Harris corner detection.Experiments show that the proposed method is an effective method for remote sensing image registration.The second method is based on gray level.Gray level based methods usually need to optimize the similarity matrix between the reference image and the image to be registered.Orthogonal learning (OLL) can construct a guide model, and the guidance algorithm can search for the optimal solution more effectively.Differential evolution algorithm (DED) is a greedy algorithm because of its superior performance in solving the problem of remote sensing image registration.Therefore, an image registration method based on orthogonal learning differential evolution algorithm (OLDE) is proposed in this paper. This method makes full use of the guidance characteristic of orthogonal learning to guide the differential evolution algorithm to search for the global optimal direction.The experiments show that the proposed algorithm has better performance than genetic algorithm, particle swarm optimization (PSO) and differential evolution algorithm (DED) in remote sensing image registration, and it has better robustness than genetic algorithm (GA), particle swarm optimization (PSO) and differential evolution algorithm (DED).
【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】：TP751

【相似文献】