遥感图像闭序列模式挖掘算法的研究与实现

发布时间：2018-05-03 17:20

本文选题：遥感图像 + 数据挖掘　；参考：《东北大学》2014年硕士论文

【摘要】：遥感图像数据挖掘是一个有着广阔应用前景的研究领域。由于遥感图像数据库的海量特征,遥感图像数据挖掘已成为空间数据挖掘的主流。近年来,随着图像获取和图像存储技术的迅速发展,使得人们能够较为方便地得到大量有用的遥感图像数据。图像数据挖掘是用来挖掘图像数据中隐含的知识、图像内或图像间的各种关系以及其他隐藏在图像数据中的各种模式的一种技术,目前仍处于实验研究阶段,是一个新兴的、但极有发展潜力的研究领域。其中一类方法是通过卫星收集数据,并通过Apriori等基本算法以及系列算法,挖掘出不同对象不同属性间的关联规则。这意味着序列模式挖掘算法可以集成到遥感图像数据挖掘算法之中。作为遥感图像数据挖掘方法的核心,序列模式挖掘算法的性能一直是影响方法性能的瓶颈。由于Apriori算法、PrefixSpan算法在挖掘大数据集上的劣势,针对遥感图像数据集,本文提出了基于BIDE的遥感图像数据挖掘方法,并对其中的闭序列模式挖掘算法进行了深入的研究与改进,使之能够更好的挖掘遥感图像数据集。本文把BIDE算法集成到遥感图像数据挖掘方法中。这种闭序列模式挖掘算法不需要维护候选闭序列,可以直接进行闭序列检查,并且可以快速完成搜索空间削减。针对遥感图像数据集,本文对算法的各个模块进行了测试,证明了方法的有效性、高效性。对于更大规模的遥感图像数据集,BIDE算法在闭序列检查和搜索空间削减的过程中需要进行大量字符匹配和支持度计算操作。这两种操作产生了大量的时间开销。针对其弱点,本文提出一种基于位置扩展的闭序列模式挖掘算法—CSBIDEP算法,通过记录每个事件的位置信息,利用位置信息得到频繁1-序列,并对其进行直接位置扩展验证,以减少对投影数据库的扫描,节省时间的开销。针对不同规模的数据集,本文将基于位置扩展的闭序列模式挖掘算法与BIDE算法进行了比较实验。从实验结果看出,前者的时间性能有了显著地提高。
[Abstract]:Remote sensing image data mining is a promising research field. Because of the massive features of remote sensing image database, remote sensing image data mining has become the mainstream of spatial data mining. In recent years, with the rapid development of image acquisition and image storage technology, people can easily get a large number of useful remote sensing image data. Image data mining is a technique used to mine hidden knowledge in image data, relationships within and between images, and other patterns hidden in image data. It is still in the stage of experimental research and is a new technology. But there is great potential for research. One kind of method is collecting data by satellite, mining association rules between different objects and attributes by using basic algorithms such as Apriori and a series of algorithms. This means that sequential pattern mining algorithm can be integrated into remote sensing image data mining algorithm. As the core of remote sensing image data mining, the performance of sequential pattern mining is always the bottleneck. Because of the disadvantage of Apriori algorithm PrefixSpan algorithm in mining big data sets, this paper proposes a method of remote sensing image data mining based on BIDE, and makes a deep research and improvement on the closed sequence pattern mining algorithm. So that it can better mining remote sensing image data sets. In this paper, the BIDE algorithm is integrated into the remote sensing image data mining method. This closed sequence pattern mining algorithm does not need to maintain candidate closed sequences, it can directly check the closed sequences, and can quickly complete the search space reduction. Based on the remote sensing image data set, the algorithm modules are tested in this paper, and the validity and efficiency of the method are proved. For the larger remote sensing image data set Bide algorithm, a large number of character matching and support calculation operations are needed in the process of closed sequence checking and searching space reduction. These two operations have a lot of time overhead. Aiming at its weakness, this paper proposes a closed sequence pattern mining algorithm based on location expansion-CSBIDEP algorithm. By recording the location information of each event, the frequent 1- sequence can be obtained by using the location information, and it is verified by direct location expansion. To reduce the scan of the projection database, saving time and overhead. In this paper, the closed sequence pattern mining algorithm based on location expansion is compared with the BIDE algorithm for different data sets. The experimental results show that the time performance of the former has been improved significantly.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;TP751

【相似文献】