基于等距环局部二值模式及粗细结合的遥感图像匹配研究

发布时间：2018-05-05 06:41

本文选题：遥感图像匹配 + 局部不变特征　；参考：《西南交通大学》2017年硕士论文

【摘要】：基于局部不变特征的匹配方法是近年来图像处理研究的一个热点,被广泛的应用于各个领域,本文将局部不变特征方法应用于遥感领域,主要研究遥感图像的匹配算法。遥感图像包括同源遥感图像与异源遥感图像,它们的匹配方法都有着各自的研究难点。同源遥感图像匹配的难点主要是实现算法的实时性,而异源遥感图像主要问题是解决图像间的非线性灰度差异,基于以上问题,本文分别提出了以下的算法。为了解决遥感图像匹配的实时性问题,目前的解决方案主要采用二进制描述子方法进行匹配。描述子通常需要计算主方向来获得旋转不变性,但主方向的计算往往存在误差,会导致匹配效果不理想。针对上述问题,本文提出一种基于等距环局部二值模式的二进制描述子方法。在描述子构造与描述子匹配过程中分别使用了等距环局部二值采样模式与分组匹配策略,使得构造的描述子具有旋转不变性,解决主方向问题和加速匹配过程。实验结果表明,与传统二进制描述子相比,本文方法可以有效解决主方向问题,并且具有更快的匹配速度及更高的准确率。异源遥感图像的匹配是一个难点,主要原因是异源图像由于成像原理等因素的不同,造成获得的图像之间有较大的非线性灰度差异,使得传统的基于灰度与基于梯度的匹配方法无法较好的进行异源图像的匹配。针对上述问题,本文提出了一种粗细匹配结合的方法。在粗匹配阶段消除图像间的位移、旋转、尺度变化,然后,在精匹配阶段消除图像间的非线性灰度差异,并使用模板匹配提高匹配的精确度。实验表明,本文提出的算法可以有效地解决异源遥感图像的匹配。
[Abstract]:The matching method based on local invariant feature is a hot topic in image processing in recent years and has been widely used in various fields. In this paper, the local invariant feature method is applied to remote sensing, and the matching algorithm of remote sensing image is mainly studied. Remote sensing images include homologous remote sensing images and heterogenous remote sensing images, and their matching methods have their own difficulties. The difficulty of homologous remote sensing image matching is mainly to realize the real-time performance of the algorithm, while the main problem of heterogeneous remote sensing image is to solve the nonlinear gray difference between images. Based on the above problems, the following algorithms are proposed in this paper. In order to solve the real-time problem of remote sensing image matching, the current solution mainly uses binary description sub-method to match. The descriptor usually needs to calculate the principal direction to obtain rotation invariance, but the calculation of the principal direction often has errors, which will lead to unsatisfactory matching effect. In order to solve the above problems, this paper presents a binary descriptor submethod based on the local binary mode of equidistant ring. In the process of descriptor construction and descriptor matching, the local binary sampling mode and grouping matching strategy of equidistant ring are used respectively, which makes the constructed descriptor rotation-invariant, solves the main direction problem and accelerates the matching process. Experimental results show that compared with the traditional binary descriptor, the proposed method can effectively solve the principal direction problem, and has faster matching speed and higher accuracy. The matching of heterogenous remote sensing images is a difficult problem. The main reason is that there is a large nonlinear gray difference between the different images due to the different imaging principles. It makes the traditional grayscale and gradient-based matching methods can not better match the heterogeneous images. In order to solve the above problems, this paper presents a method of combining coarse and fine matching. In the coarse matching stage, the displacement, rotation and scale change of the images are eliminated. Then, in the fine matching stage, the nonlinear gray difference between images is eliminated, and the accuracy of matching is improved by template matching. Experiments show that the proposed algorithm can effectively solve the matching of heterogeneous remote sensing images.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP751

【参考文献】