基于GF-1遥感影像矿区的信息提取与建模

发布时间：2018-05-06 07:29

本文选题：高分一号 + 图像增强　；参考：《华北理工大学》2016年硕士论文

【摘要】：研究以GF-1遥感数据为依托,建立成熟合理的矿山信息提取的方法和流程,尝试更高效、更快捷地对矿区信息进行提取,深入分析信息提取过程中的关键技术问题。对于高分一号遥感影像的研究主要是通过融合方法和融合效果的对比,得出用GS融合方法得到的遥感影像饱和度较高,成像清晰,地物边界明了;通过计算各波段之间的相关性和波段内部信息量的统计,结合波段指数OIF,进一步确定高分一号遥感影像的最佳成像波段为band4、band3、band2;以IDL为平台实现了Mean Shift算法对高分一号的增强,经增强之后的遥感影像具有更优的判读效果,有利于后续分类工作的开展,前期研究旨在为高分一号或类似分辨率遥感影像的图像增强提供技术支持。通过大量实验确定了研究区尾矿这一目标地类的最优分割尺度以及分割参数,并对纹理特征进行了研究,选择最佳的纹理波段参与多尺度分割,分析分割效果,实验表明同时加入对比度Contrast和相关性Correlation两个纹理特征时矿区信息的分割效果最好;通过对研究区内矿区及周边地物的光谱特征进行统计分析,建立了高分一号尾矿库信息提取的IOT模型,辅以纹理和几何特征,用面向对象的方法对矿区及周围地物进行信息提取,提取精度为83.38%,对比基于像元的最大似然分类法的提取精度68.02%,研究表明,面向对象的分类方法对高分一号矿区信息的提取具有绝对的优势。
[Abstract]:Based on the GF-1 remote sensing data, this paper establishes a mature and reasonable method and process of mining information extraction, tries to extract mining information more efficiently and quickly, and deeply analyzes the key technical problems in the process of information extraction. The research of Gaofen 1 remote sensing image is mainly based on the comparison of fusion method and fusion effect. It is concluded that the RS image saturation obtained by GS fusion method is higher, the image is clear, and the boundary of ground object is clear. By calculating the correlation between each band and the statistics of internal information in the band, and combining with the band index, we further determine that the best imaging band of the high score 1 remote sensing image is band4 band 3 band 2. Using IDL as the platform, we realize the enhancement of high score 1 by Mean Shift algorithm. The enhanced remote sensing image has better interpretation effect and is helpful to the subsequent classification work. The previous study aims to provide technical support for image enhancement of high-resolution remote sensing image with high resolution or similar resolution. Through a large number of experiments, the optimal segmentation scale and segmentation parameters of the target ground type of tailings in the study area are determined, and the texture features are studied. The best texture band is selected to participate in the multi-scale segmentation, and the segmentation effect is analyzed. The experiment shows that the segmentation effect of mining area information is the best when two texture features, contrast Contrast and correlated Correlation, are added at the same time, and the spectral characteristics of mining area and surrounding ground objects in the study area are statistically analyzed. The IOT model of information extraction of Gaofian No. 1 tailings reservoir is established, which is supplemented by texture and geometric features, and the information of mining area and surrounding ground objects is extracted by object-oriented method. The extraction accuracy is 83.38, compared with the maximum likelihood classification method based on pixel, the extraction accuracy is 68.02. The research shows that the object-oriented classification method has an absolute advantage in extracting information from Gaofen No. 1 mining area.
【学位授予单位】：华北理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：P237;TD178

【相似文献】