基于面向对象分类的稀土矿开采动态监测研究

发布时间：2018-03-06 19:03

本文选题：遥感　切入点：面向对象分类　出处：《中国地质大学(北京)》2017年硕士论文　论文类型：学位论文

【摘要】：离子吸附型稀土矿不仅是全球范围内极其重要的战略资源,更是我国特有的宝贵矿产资源。近年来,我国稀土资源储量不断减少,如何实时准确的反映稀土矿山开发活动情况,成为实现我国离子吸附型稀土资源的可持续开发的前提。遥感技术的出现,弥补了传统监管手段的不足,特别是具有丰富空间细节信息的高分辨率遥感卫星数据,更是为稀土矿山监测提供了极大的便利。然而运用传统的目视解译法和基于像元分类法对高分辨率影像数据进行稀土矿山信息提取,或是耗费人力和时间,或是分类精度不高,均不能达到理想的效果。本文选用面向对象分类法,运用ENVI Feature Extraction模块,以2005~2015年间五期国内外高分辨率遥感数据作为数据源,以江西省寻乌地区为研究区,提取稀土开采区,并根据提取结果开展了稀土开采动态监测研究。主要成果如下:(1)在完成多源多时相遥感数据预处理工作的基础上,开展面向对象分类最佳影像分割尺度研究。结合稀土开采区的特点,通过对比分析,确定了各遥感影像的最佳分割参数及合并参数,实现了影像的多尺度分割,为面向对象分类提供数据基础。(2)结合实际情况,选择地形、光谱和几何特征建立符合稀土开采区特征的规则集,实现面向对象分类。其中,地形特征包括高程、坡度、地表切割深度及地形起伏度;光谱特征包括NDVI指数及光谱反射率;几何特征包括面积、延伸性和矩形相似性。最后结合规则集,选用隶属度函数法实现面向对象分类。(3)运用传统监督分类最大似然分类法提取研究区的稀土开采区,将其分类结果与面向对象分类结果分别进行精度评价和分析,结果表明面向对象分类法精度更高,效果更好。(4)运用先分类后比较法,开展稀土矿区开采动态监测,总结出在2005-2015十年间,寻乌稀土矿区的开采状况总体呈现先减少后增加然后趋于平稳的趋势。最后,结合研究结果,对该地区的环境保护治理提出了合理建议,为矿政管理部门提供数据支持与技术支撑。通过研究结果表明,运用面向对象分类技术,结合高分辨率影像提取稀土开采区及动态监测是完全可行的,并且结果较为理想。
[Abstract]:Ion adsorbed rare earth ore is not only an extremely important strategic resource in the world, but also a unique precious mineral resource in China. In recent years, the reserves of rare earth resources in China have been decreasing, so how to reflect the development activities of rare earth mines in real time and accurately? The emergence of remote sensing technology makes up for the shortcomings of traditional supervision methods, especially the high-resolution remote sensing satellite data with rich spatial details. It also provides great convenience for rare earth mine monitoring. However, using traditional visual interpretation method and pixel classification method to extract rare earth mine information from high-resolution image data may cost manpower and time, or the classification accuracy is not high. In this paper, object oriented classification method, ENVI Feature Extraction module, five periods of domestic and foreign high resolution remote sensing data from 2005 to 2015 are used as data sources, and the Xunwu area of Jiangxi Province is used as the research area to extract rare earth mining areas. Based on the extraction results, the dynamic monitoring of rare earth mining is carried out. The main results are as follows: 1) on the basis of completing the preprocessing of multi-source and multi-phase remote sensing data, According to the characteristics of rare earth mining area, the optimal segmentation parameters and merging parameters of each remote sensing image are determined by comparing and analyzing the characteristics of rare-earth mining area, and the multi-scale segmentation of the image is realized. In combination with the actual situation, the terrain, spectrum and geometry features are selected to set up the rule set that accords with the features of rare earth mining area, and the object oriented classification is realized, in which, the terrain features include elevation, slope, and so on. Surface cutting depth and topographic relief; spectral features include NDVI exponent and spectral reflectivity; geometric features include area, extensibility and rectangular similarity. The membership function method is used to realize the object-oriented classification. (3) the rare-earth mining area in the study area is extracted by the traditional supervised maximum likelihood classification method, and the accuracy of the classification results and the object-oriented classification results are evaluated and analyzed respectively. The results show that the object-oriented classification method has higher precision and better effect. The mining situation of Xunwu rare earth mining area shows a trend of first decreasing and then increasing and then tending to steady. Finally, combined with the results of the research, the reasonable suggestions for environmental protection and management in this area are put forward. The research results show that it is feasible to extract rare-earth mining areas and dynamic monitoring by using object-oriented classification technology combined with high-resolution images and the results are ideal.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD865

【相似文献】