线性体提取方法及其在地质方面的研究

发布时间：2019-01-10 15:52

【摘要】：线性体是自然景观和人造景观在图像上的一种表现，通过色调反差、亮度变化、地形地貌的线状排列等特征表现出来。线状体中蕴含许多在地质应用中具有十分重要作用的地质专题信息，例如线状构造、环状构造、地层界线、地质体界线、地震带界线、矿化蚀变带界线等。在作为线性体载体的众多图像数据源中，，卫星影像是最为普及也是发展最为迅猛的一种。从卫星影像上获得线性体的传统方法是目视解译，其缺点是对从业人员要求很高、非常耗费时间和金钱，解译结果的主观性太强等，因而人们迫切要求利用计算机将目视解译方法模拟成计算机的运算模型从而达到自动解译的目的，以满足从海量的卫星遥感影像中快速获取遥感信息的需求。本文以线性体的提取及利用为目标，采用卫星遥感影像，立足于融合方法、边缘提取方法、线性体的统计分析方法几个关键点进行研究，所得的主要成果及结论如下：（1）提出WBS融合方法。基于Wrapping方法的离散曲波变换，提出了低频子带采用加权平均法、中间子带采用最大区域边缘信息度量法、高频子带采用替换法的多尺度融合策略。实验研究表明，WBS方法是一种适合噪声较为严重的遥感影像的融合方法，可以使融合影像在保证光谱保持度与其它流行方法相当的前提下，清晰度、综合质量、结构相似度等方面均远远优于其它流行方法，与此同时，对噪声的消除效果明显。（2）提出HSOD边缘提取方法。采用高斯高通滤波代替常用的低通滤波法，使边缘的识别度更高；使用最大类间方差法求阈值，实现了阈值的自动化设定，且强弱阈值结合的方法减少了边缘点的遗漏；经基于八连接通路的边缘连接方法使强弱边缘点信息得以有机结合，边缘信息更加准确、完整。实验研究表明，相较传统经典的边缘检测方法LoG、Sobel、Canny在不同图像复杂度和清晰度的影像上进行的质量评价结果显示，无论是目视还是定量评价的结果，HSOD方法均更佳。（3）提出一种卫星遥感影像的自动化解译在地质方面的解决方案。基于线性体和地质异常场特征之间存在本质的关联关系，应用前述研究的线性体提取方法，比较系统地提出了线性体提取结果在地学统计分析方面的应用模式，发展了比较完整的卫星遥感影像的自动化解译在地质方面的解决方案，并选择了新疆柯坪隆起和青海祁漫塔格虎头崖多金属矿田为研究区，对该解决方案的有效性进行了实践验证。
[Abstract]:Linear body is a kind of representation of natural landscape and artificial landscape on the image, which can be shown by contrast of tone, change of brightness, linear arrangement of landform and geomorphology and so on. The linear body contains a lot of special geological information which plays an important role in geological application, such as linear structure, ring structure, stratigraphic boundary, geological body boundary, seismic zone boundary, mineralization alteration zone boundary and so on. Among the many image data sources, satellite image is the most popular and rapidly developing. The traditional method of obtaining linear bodies from satellite images is visual interpretation. Its disadvantages are that it is very demanding to the practitioners, it is very time-consuming and expensive, and the result of interpretation is too subjective and so on. Therefore, it is urgent to use the computer to simulate the visual interpretation method into the computer operation model so as to achieve the purpose of automatic interpretation, in order to meet the needs of rapid acquisition of remote sensing information from massive satellite remote sensing images. This paper aims at the extraction and utilization of linear volume, adopts satellite remote sensing image, based on fusion method, edge extraction method and statistical analysis method of linear volume. The main results and conclusions are as follows: (1) WBS fusion method is proposed. Based on the discrete Qu Bo transform of the Wrapping method, a multi-scale fusion strategy is proposed, in which the weighted average method is used for the low frequency subband, the maximum region edge information is measured for the middle subband, and the substitution method is used for the high frequency subband. The experimental results show that the WBS method is a suitable fusion method for the noisy remote sensing images. It can make the fusion images with the same spectral retention as other popular methods, the clarity and the comprehensive quality of the fusion images can be achieved on the premise that the spectral retention degree is equal to that of other popular methods. The structural similarity is much better than other popular methods, at the same time, the effect of noise elimination is obvious. (2) HSOD edge detection method is proposed. Gao Si high-pass filter is used to replace the common low-pass filtering method to make the edge recognition degree higher, the maximum inter-class variance method is used to calculate the threshold value, the automatic setting of the threshold value is realized, and the combination of strong and weak threshold method reduces the omission of edge points. The information of strong and weak edge points can be combined organically by the method of edge connection based on eight join paths, and the edge information is more accurate and complete. The experimental results show that HSOD method is better than the traditional edge detection method LoG,Sobel,Canny in evaluating the quality of images with different image complexity and sharpness. (3) A geological solution for automatic interpretation of satellite remote sensing images is proposed. Based on the essential correlation between linear bodies and geological anomaly field characteristics, the application model of linear volume extraction results in geoscience statistical analysis is put forward systematically by using the linear volume extraction method mentioned above. A relatively complete automatic interpretation of satellite remote sensing images has been developed in the geological aspects of the solution, and the Xinjiang Keping uplift and Qimantag Tiger head Cliff Polymetallic Orefield have been selected as the study areas. The effectiveness of the solution is verified in practice.
【学位授予单位】：长安大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP751

【参考文献】