多光谱遥感图像变化检测方法研究
发布时间：2019-01-10 09:04
【摘要】：遥感图像变化检测是通过分析处理不同时期同一地点获取的遥感图像而获取变化信息的技术。作为遥感图像解译的主要技术之一，遥感图像变化检测已经广泛应用于国防军情监控、资源探测、环境监测、城市规划等领域。多光谱遥感图像数据具有从可见光到红外光波段的多个接收频段，丰富的光谱信息增加了识别多种类型变化的可能性和可信度，因此多光谱遥感图像在变化检测中的应用越来越广泛。本文主要研究了多光谱遥感图像的变化检测，完成了如下两方面的工作： （1）提出了一种基于Treelet融合和图正则约束谱聚类的多光谱遥感图像变化检测方法。该方法首先对预处理后的两时相各个波段图像对应做差，得到各个波段的差异图，用非抽样离散小波分解各个波段差异图，得到各自的小波分解系数，然后将处在同一分解波段同一分解方向的所有波段的小波系数使用Treelet方法进行融合，使用逆非抽样离散小波重构融合后得到的小波系数，得到新的差异图。使用图正则约束谱聚类分割新的差异图，得到最终的变化检测结果。我们通过对3组多光谱遥感图像进行实验，验证了本方法的有效性。 （2）提出了一种基于半监督降维和显著图的多光谱遥感图像变化检测方法。该方法首先对预处理后的两时相多波段图像对应波段做差，得到差异图组，接着采用半监督降维的方式将差异图组融合为一幅新的差异图。然后对得到的新的差异图作两种处理，，一种处理是直接对新的差异图用K-means聚类，得到差异图的类别标记图，另一种处理是对新的差异图先提取显著图，对显著图用K-means聚类，得到显著图的类别标记图，将差异图的类别标记图和显著图的类别标记图借助数学形态学的理论进行融合，得到最终的变化检测结果。我们通过3组多光谱遥感图像进行实验，验证了本方法的有效性。 本论文工作得到了国家自然科学基金（60970066）、国家自然科学基金（61173092）以及中央高校基本科研业务费专项资金（K50510020025）的资助。
[Abstract]:Remote sensing image change detection is a technique to obtain change information by analyzing and processing remote sensing images obtained at the same time and the same place. As one of the main techniques of remote sensing image interpretation, remote sensing image change detection has been widely used in the field of defense monitoring, resource detection, environmental monitoring, urban planning and so on. Multispectral remote sensing image data have multiple receiving bands from visible to infrared light bands, and rich spectral information increases the possibility and reliability of identifying various types of changes. Therefore, multispectral remote sensing images are more and more widely used in change detection. In this paper, the change detection of multispectral remote sensing images is studied, and the following two aspects are accomplished: (1) A change detection method for multispectral remote sensing images based on Treelet fusion and regular constrained spectral clustering is proposed. In this method, first of all, the 02:00 phase image of each band after preprocessing is mismatched, and the difference map of each band is obtained, and the wavelet decomposition coefficient is obtained by decomposing the difference image of each band by using non-sampling discrete wavelet transform. Then all the wavelet coefficients in the same decomposition band and the same decomposition direction are fused using Treelet method. The wavelet coefficients obtained from the fusion are reconstructed by inverse non-sampling discrete wavelets, and a new difference graph is obtained. The new difference graph is segmented by regular constraint spectrum clustering, and the final change detection result is obtained. Experiments on three sets of multispectral remote sensing images show that the proposed method is effective. (2) A multispectral remote sensing image change detection method based on semi-supervised reduction and saliency map is proposed. In this method, the 02:00 multiband images after preprocessing are first divided into the corresponding bands, and the difference map group is obtained, and then the difference map group is fused into a new difference map by semi-supervised dimensionality reduction. Then two kinds of processing are made for the new difference map, one is to cluster the new difference map directly with K-means, and the other is to extract the salient map from the new difference map. In this paper, K-means clustering is used to obtain the class marker map of the significant map, and the category marker map of the difference map and the class marker map of the salient map are fused with the theory of mathematical morphology, and the final change detection results are obtained. Three sets of multispectral remote sensing images are used to verify the effectiveness of this method. This thesis is supported by the National Natural Science Foundation of China (60970066), the National Natural Science Foundation of China (61173092) and the Special Fund for basic Scientific Research operating expenses of the Central University (K50510020025).
【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
本文编号：2406131
[Abstract]:Remote sensing image change detection is a technique to obtain change information by analyzing and processing remote sensing images obtained at the same time and the same place. As one of the main techniques of remote sensing image interpretation, remote sensing image change detection has been widely used in the field of defense monitoring, resource detection, environmental monitoring, urban planning and so on. Multispectral remote sensing image data have multiple receiving bands from visible to infrared light bands, and rich spectral information increases the possibility and reliability of identifying various types of changes. Therefore, multispectral remote sensing images are more and more widely used in change detection. In this paper, the change detection of multispectral remote sensing images is studied, and the following two aspects are accomplished: (1) A change detection method for multispectral remote sensing images based on Treelet fusion and regular constrained spectral clustering is proposed. In this method, first of all, the 02:00 phase image of each band after preprocessing is mismatched, and the difference map of each band is obtained, and the wavelet decomposition coefficient is obtained by decomposing the difference image of each band by using non-sampling discrete wavelet transform. Then all the wavelet coefficients in the same decomposition band and the same decomposition direction are fused using Treelet method. The wavelet coefficients obtained from the fusion are reconstructed by inverse non-sampling discrete wavelets, and a new difference graph is obtained. The new difference graph is segmented by regular constraint spectrum clustering, and the final change detection result is obtained. Experiments on three sets of multispectral remote sensing images show that the proposed method is effective. (2) A multispectral remote sensing image change detection method based on semi-supervised reduction and saliency map is proposed. In this method, the 02:00 multiband images after preprocessing are first divided into the corresponding bands, and the difference map group is obtained, and then the difference map group is fused into a new difference map by semi-supervised dimensionality reduction. Then two kinds of processing are made for the new difference map, one is to cluster the new difference map directly with K-means, and the other is to extract the salient map from the new difference map. In this paper, K-means clustering is used to obtain the class marker map of the significant map, and the category marker map of the difference map and the class marker map of the salient map are fused with the theory of mathematical morphology, and the final change detection results are obtained. Three sets of multispectral remote sensing images are used to verify the effectiveness of this method. This thesis is supported by the National Natural Science Foundation of China (60970066), the National Natural Science Foundation of China (61173092) and the Special Fund for basic Scientific Research operating expenses of the Central University (K50510020025).
【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
【参考文献】
相关期刊论文 前5条
1 林文鹏;黄敬峰;胡小猛;赵敏;;基于MODIS温度植被角度指数的农作物估产模型研究[J];红外与毫米波学报;2010年06期
2 吴柯;牛瑞卿;王毅;杜博;;基于PCA与EM算法的多光谱遥感影像变化检测研究[J];计算机科学;2010年03期
3 程涛;朱国宾;刘玉安;;基于二维压缩感知的定向遥感和变化检测[J];红外与毫米波学报;2013年05期
4 张鹏强;余旭初;刘智;李建胜;万刚;;多时相遥感图像相对辐射校正[J];遥感学报;2006年03期
5 丁丽霞,周斌,王人潮;遥感监测中5种相对辐射校正方法研究[J];浙江大学学报(农业与生命科学版);2005年03期
本文编号：2406131
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2406131.html
