基于信息融合与匀质区域提取的SAR图像变化检测

发布时间：2018-08-06 07:53

【摘要】：变化检测问题属于图像处理领域,通常是指依据不同时间的多次观测来对一个物体的状态变化进行确定或者对某种现象的变化的过程进行确定的过程。随着遥感技术的发展,合成孔径雷达(SAR)图像也越来越成为了图像变化检测问题中的主要的数据来源。SAR图像具有全天候、全天时的工作能力,对地物有一定的穿透能力,且不受大气、气候等随机因素的影响,具有不可比拟的优点。国内外众多学者对SAR图像的变化检测问题进行了大量的研究,变化检测结果的精度也在不断地提高。无监督变化检测算法为最常使用的变化检测算法,该类算法的主要步骤为SAR图像的预处理、差异图的构造和差异图的分析。本文的研究重点为SAR图像变化检测中的差异图构造的问题,给出了两种基于信息融合的构造变化检测差异图的方法,具体如下所述:1.提出基于信息融合的SAR图像变化检测差异图的构造方法,该方法直接利用均值比算子和对数比算子得到的变化检测差异图进行差异图的构造,融合规则选取了数学融合算子,规则较为简单,但是能有效地利用两种比值算子的优点,有利于后续的差异图分析的继续进行。然而,针对复杂的SAR图像数据集,在没有其它额外处理的情况下,由于所选取的融合算子十分简单,并不能取得最优的变换检测结果。于是,考虑增加额外的处理步骤,来挖掘该融合算子所包含的丰富信息来生成最终的变化检测差异图,从而取得更加优秀的变化检测结果。因此提出了第二种基于信息融合的变化检测算法,即基于信息融合与匀质区域提取的变化检测差异图构造方法。2.提出基于信息融合与匀质区域提取的变化检测差异图构造方法。该方法是对提出的方法一的进一步改进,对方法一得到到变化检测差异图在多尺度空间下进行匀质区域的提取,从而得到最终的变化检测差异图。该方法的仿真结果也验证了我们的预期,的确能够取得更为精确地变化检测结果。除了将提出的算法应用在常见的SAR图像变化检测数据集上,也将算法应用在了科研项目中的黄河入海口数据集。组成该数据集的图像由于相互之间的噪声水平差异巨大,通常的SAR图像变化检测算法难以进行有效的检测。而提出的方法二,由于在信息融合得到的差异图上增加了匀质区域的提取,能够取得了较好的效果。
[Abstract]:The problem of change detection belongs to the field of image processing. It usually refers to the process of determining the state change of an object or the change of a certain phenomenon based on multiple observations at different times. With the development of remote sensing technology, synthetic Aperture Radar (SAR) images have become the main data source in image change detection. And not affected by the atmosphere, climate and other random factors, has incomparable advantages. Many scholars at home and abroad have done a lot of research on the change detection of SAR images, and the accuracy of the change detection results has been improved. Unsupervised change detection algorithm is the most commonly used change detection algorithm. The main steps of this algorithm are preprocessing of SAR image, construction of difference map and analysis of difference graph. This paper focuses on the construction of difference map in SAR image change detection, and gives two methods based on information fusion to construct change detection difference map, which is described as follows: 1. In this paper, a method of constructing difference map of SAR image change detection based on information fusion is proposed. The difference graph is constructed by means of the mean ratio operator and logarithmic ratio operator. The fusion rule selects the mathematical fusion operator. The rules are simple, but the advantages of the two ratio operators can be effectively utilized, which is beneficial to the continuation of the subsequent difference graph analysis. However, in the case of complex SAR image data sets and without other additional processing, the selected fusion operator is very simple and can not obtain the optimal transform detection results. Therefore, we consider adding additional processing steps to mine the abundant information contained in the fusion operator to generate the final change detection difference graph, so as to obtain better change detection results. Therefore, the second change detection algorithm based on information fusion is proposed, that is, the method of constructing difference map of change detection based on information fusion and homogeneous region extraction. Based on information fusion and homogeneous region extraction, a method for constructing difference map of change detection is proposed. This method is a further improvement of the proposed method 1. The first method obtains the difference map of change detection and extracts the homogeneous region in multi-scale space to obtain the final difference map of change detection. The simulation results of this method also verify our expectation and can obtain more accurate change detection results. In addition to applying the proposed algorithm to the common SAR image change detection data set, the algorithm is also applied to the Yellow River estuary data set in the scientific research project. Because of the huge difference of noise level between the images which make up the dataset, it is difficult to detect the change of SAR image effectively. In the second method, due to the addition of homogeneous region to the difference map obtained by information fusion, good results can be obtained.
【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

【相似文献】