全球生态地理分区知识库的构建与应用

发布时间：2018-04-15 13:55

本文选题：生态地理分区 + 知识库　；参考：《北京建筑大学》2017年硕士论文

【摘要】：随着人工智能的不断发展,知识库和专家系统被逐渐引入到遥感领域中,尤其是针对遥感目标地物的识别和分类。当前地表覆盖变化检测多是从基于像素级别的遥感影像,当面对全球范围内复杂多样的地类,会存在大量同物异谱和异物同谱的现象,同时由于物候原因,在变化检测的过程中都会存在大量的伪变化。想要得到高精度的变化检测结果,要对检测后的变化图斑进行伪变化识别与剔除。目前变化检测后的伪变化一般是利用人工来识别,不仅浪费时间、耗费大量资源,而且由于识别人员知识的储备不足很容易造成错判。因此,急需建立一个地类知识库,用来自动识别变化检测后的伪变化。全球生态地理分区由于其全球性、分区地类稳定性、地物变化规律性和信息量大等特点,可以用来构建知识库辅助变化检测。本文针对以往遥感影像变化检测存在的问题,提出了一种基于生态地理知识库来实现伪变化识别的方法,并从知识库的结构、内容设计,知识的表达入库,再到知识库管理系统的建立,最后设计了生态去伪插件,实现了利用规则库中的规则识别伪变化,完成了整个去伪流程的结构设计和功能实现。首先,设计并构建了包括分区伪变化规则库、地表覆盖变化规则库和专题资料库在内的生态地理分区知识库,并对知识库中的伪变化规则做了部分示例,其中对生态分区伪规则库采用面向对象的方法进行设计,有效的实现了规则类之间的继承,提高了伪变化识别的推理效率。其次,利用对象关系型数据库PostgreSQL对知识库采用数据库的形式存储,并采用C#语言和ArcEngine开发了生态地理分区知识库管理系统。最后,建立生态去伪插件实现了利用知识库中的规则识别伪变化,并选取北京市和杰克逊市两个实验区对剔除伪变化后的变化图斑做了精度验证。结果表明,利用生态地理分区知识库来进行伪变化的识别不仅提高了自动化程度,而且提高了变化检测的精度。
[Abstract]:With the development of artificial intelligence, knowledge base and expert system are gradually introduced into the field of remote sensing, especially for the recognition and classification of remote sensing objects.At present, the detection of surface cover change is mostly based on pixel level remote sensing images. In the face of complex and diverse ground species in the global scope, there will be a large number of isospectral and foreign body isospectral phenomena, at the same time, due to phenological reasons,In the process of change detection, there will be a large number of pseudo changes.In order to obtain high precision change detection results, the change pattern after detection should be identified and eliminated.At present, the pseudo-change after change detection is usually identified by manual, which not only wastes time and consumes a lot of resources, but also easily leads to misjudgment because of the lack of knowledge of recognizer.Therefore, there is an urgent need to establish a knowledge base for automatic identification of pseudo changes after change detection.The global ecogeographic zoning can be used to construct a knowledge base for aided change detection because of its global, regional stability, the regularity of ground object change and the large amount of information.In this paper, aiming at the existing problems of remote sensing image change detection in the past, a method based on ecological geographical knowledge base is put forward to realize the identification of pseudo change, and the structure of knowledge base, the content design and the representation of knowledge are put into database.Then to the establishment of the knowledge base management system, finally designed the ecological de-counterfeiting plug-in, using the rules in the rule base to identify the pseudo changes, completed the structure design and functional implementation of the whole de-counterfeiting process.First of all, the paper designs and constructs the knowledge base of ecological geography partition including partition pseudo-change rule base, surface cover change rule base and thematic database, and makes some examples of pseudo-change rule in knowledge base.Among them, the object oriented method is used to design the pseudo-rule base of ecological partition, which effectively realizes the inheritance between rule classes and improves the reasoning efficiency of pseudo-change recognition.Secondly, the object relational database (PostgreSQL) is used to store the knowledge base in the form of database, and the ecological geographical partition knowledge base management system is developed by using C # language and ArcEngine.Finally, the ecological de-counterfeiting plug-in is established to recognize the pseudo-change by using the rules in the knowledge base, and two experimental areas of Beijing and Jackson are selected to verify the accuracy of the change pattern after eliminating the pseudo-change.The results show that the recognition of pseudo-change by using the knowledge base of ecogeographic partition not only improves the degree of automation but also improves the accuracy of change detection.
【学位授予单位】：北京建筑大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】