【作者】 邵锐；

【导师】 巫兆聪；

【作者基本信息】 武汉大学 ， 地图学与地理信息系统， 2005， 硕士

【摘要】 遥感影像是地表三维景物的二维平面成像,由于地球系统的复杂性和开放性,以及遥感信息传输过程的信息衰减,决定了遥感影像具有复杂、不确定、不精确和信息不完全的特性。智能遥感影像信息处理和分析的关键问题之一就是如何对于这些不精确、不完全信息进行分类分析和知识获取。常规的遥感信息处理的理论与方法——光谱和空间分析方法,基本遵循数理统计分析的理论模式,其统计方法的本质决定了它们的局限性,已经很难适应这一要求。近几年发展起来的粗集理论,为解决这一难题提供了新的途径。 粗集理论是一种新的处理含糊和不确定性问题的数学工具。它作为一种新的软计算方法,特别适合于不完整、不确定知识和数据的表达、学习和归纳,是具有发展潜力的智能信息处理方法。国内外智能数据处理领域对该理论和应用有着广泛和深入的研究,并不乏成功应用实例的报道。应用不确定性理论和方法研究遥感信息的处理、分析和结果评估是当前遥感信息智能处理领域的主流研究趋势之一。本文拟应用粗集的理论和方法,将遥感影像表达的信息看成一个知识系统,研究基于粗集理论的遥感影像预处理方法;通过利用粗集理论的非确定性问题处理能力,避免或减少主观性假设和人工设置经验性参数的影响,提高系统处理非确定性问题的能力。 论文从粗集理论的不可分辨关系(等价关系)、近似集合和分类概念出发,研究基于粗集理论的图像滤波、图像增强的算法,提出保护边缘的粗集滤波增强算法;应用粗集与聚类算法相结合的方法对遥感影像的进行处理。具体的研究工作和创新之处主要体现在: (1)通过对粗集理论的理论基础和遥感信息的不确定特性的初步分析,论述利用该理论来处理遥感信息的可行性以及优势。 (2)从粗集理论的不可分辨关系、近似集合和分类概念出发,研究图像滤波和图像增强的一些粗集算法,提出一种保护边缘的粗集均值滤波算法,实验证明其有效性。 (3)分析聚类初值对于K-means聚类结果的影响,通过应用粗集的等价关系概念来选取聚类所需要的初始类的个数和均值,从而提高聚类的效率和准确性。 (4)从DBSCAN算法入手,分析该算法的优势和面临的主要问题,通过应用粗集理论的分类概念,对算法进行改进,使改进后的DBSCAN算法适合对于大数据量的遥感影像进行聚类处理,降低算法对计算机主存的要求和对输入参数Eps的依赖。 由此可见,良好的处理机制有助于提高遥感影像的处理效果和效率。在遥感影像处理过程中,粗集理论还有很多可以发挥效用的地方,如影像分类规则的提取;与神经网络相结合建立遥感影像分类模型等等。本文所做的工作还很有限的和很粗浅,更多还原

【Abstract】 Remote Sensing image has attribution of complication, uncertainty and inadequate information. The uncertainty is a key issue for Remote Sensing theory and application, especially in classification. Evaluating and processing the uncertainty of the Remote Sensing information is an important task for RS application. As the general methods of Remote Sensing procession, spectrum and space analysis have limit because of essence of statistical method. So we choose Rough Sets for cluster that has developed in recent years to process Remote Sensing Image in this dissertation.Rough set theory is a new mathematic approach to uncertain and vague data analysis. It is, no doubt, one of the most challenging areas of modern computer applications nowadays and a new very important and rapidly growing area of research and applications. Rough set theory is applied to knowledge discovery, data reduction decision support, pattern recognition and others, and it has proved to be a very effective new mathematic approach. The theory found many, interesting real-life application in medicine, banking, engineering and others.This dissertation, firstly, introduces the scheme of rough set theory and its application. Then several improved research methods of processing of remote sensing image are put forward. A method is given to deal with image filter and enhancement based on rough set theory and this method can preserve the edge detail of image very well. Finally, rough set theory based cluster algorithms apply to the processing of remote sensing image. The detailed research work can be sum up as the following:1) According to analyses of rough set theory and the uncertainty characteristic in Remote Sensing information, the preponderance and feasibility are discussed in processing remote sensing image based on this theory.2) Regarded a remote sensing image as knowledge system, this paper proposes image filter and enhancement algorithms. These algorithms, based on rough set theory, can preserve the edge detail of image very well.3) Using equivalence relations of attributes of remote sensing image, rough set theory offers the number and the centroids of the clusters, which initialize the K-means clustering. And then the image is segmented by K-means clustering algorithm. Experiment results proved its validity and the method could be applied to remote sensing image processing.4) DBSCAN, a density based clustering algorithm, can efficiently discover clusters of arbitrary shape and effectively handle noise. However, it requires large volume ofmemory support and needs a lot of I/O costs on dealing with large-scale databases. So we put forward an improved method of DBSCAN based on rough set theory. The experimental results show that the new algorithm is adapted to processing of remote sensing image.更多还原