IMHE OpenIR  > 数字山地与遥感应用中心
山地土地覆被遥感自动制图与监测方法研究
Alternative TitleStudy of Remote Sensing Methodologies for Land Cover Mapping and Change Detection in Mountainous Areas
Language中文
雷光斌
Thesis Advisor李爱农
2016
Degree Grantor中国科学院研究生院
Place of Conferral北京
Degree Name博士
Degree Discipline自然地理学
Keyword遥感 土地覆被制图 土地覆被变化检测 面向对象 决策树 多源多时相 知识 生态系统制图
Other Abstract

土地覆被是陆地表层系统最突出的景观标志,也是承载地表各类物质、能量循环过程的重要载体,与全球变化、生物多样性、可持续发展等相关研究密切相关。土地覆被遥感制图与监测研究是土地覆被研究中的基础工作,它为人类了解土地覆被现状及动态变化过程,进而预测未来情境下土地覆被变化态势提供基础数据支持。山地是陆地表层系统重要的组成部分,拥有丰富的生物多样性资源和水资源,也是全球变化等研究关注的热点区域。然而,山地区域高度异质化的景观格局、多样化的土地覆被类型、显著的地形效应以及贫乏的高质量数据,导致山地土地覆被遥感制图与监测研究无论在理论上还是在实践中都相对滞后于该领域总体的发展趋势,成为该领域研究的薄弱环节。本文选择典型山地为研究对象,系统地开展山地土地覆被遥感监测方法研究,其研究内容包括以下三个方面:一是基于多源多时相遥感数据与知识的山地土地覆被制图;二是基于双时相多指标的山地土地覆被变化检测;三是西南地区三省一市生态系统类型制图。通过研究本文得到如下的结论:(1)多源多时相遥感数据的综合决策和地学知识的深入应用是当前土地覆被遥感监测的发展趋势之一,同时,面向对象的制图/变化检测方法已成为该领域当前研究的热点。(2)在土地覆被遥感制图方法研究中,决策树算法能够从海量的多时相遥感信息中挖掘出最能有效区分各土地覆被类型的特征,最大化多源多时相遥感影像的价值。采用分层次逐级分类的方式可以有效地减少单次分类过程中土地覆被类型数量,从而间接提高土地覆被制图精度。通过多时相遥感分类对比实验发现:生长季和非生长季相结合的多时相遥感影像较单时相或单一类型(生长季或非生长季)多时相遥感影像,更能显著地提高山地森林类型制图精度,且能降低分类决策树的复杂程度,更有利于山地森林类型的自动提取。(3)在土地覆被变化检测研究中,采用多个变化检测指标(CVA、VS、ΔNDVI、ΔMNDWI和ΔNBR)分别从生长季与非生长季遥感影像对中检测出潜在的变化区域,再通过对各检测结果的融合获得潜在变化区域,从而避免了由于时相、大气条件等因素变化造成的误判和漏判。利用土地覆被变化知识的检验进一步消除了变化检测结果中的伪变化信息,从而整体上提高了土地覆被变化检测的精度。(4)在土地覆被制图向生态系统类型扩展方法研究中,首次以30m尺度的土地覆被图和植被图为基础数据,采用硬匹配、缓冲区匹配和合并类别匹配三种匹配策略实现了两种数据的融合,生产了我国西南地区首个30m尺度的生态系统类型图,类别数量由38类提高到144类。生成的生态系统类型图既具有土地覆被图精确的边界信息,又包含了植被图中丰富的植被类型信息。通过本文的研究,解决的关键科学问题是:多源、多时相、多尺度、异构数据的融合。论文的主要创新点有:(1)创新性地将山地地学知识规则化并成功应用于山地土地覆被遥感自动制图和变化检测研究中;(2)首次将双时相多指标应用于面向对象土地覆被遥感变化检测研究中,改变了传统基于像元变化检测方法更多依赖单一时相单个指标的不足;(3)首次尝试采用信息匹配策略实现土地覆被图与植被图的融合,进而形成了生态系统类型图,初步实现了土地覆被制图向生态系统类型的扩展。

;

Land cover is not only regarded as the most outstanding landscape mark and the important carrier for all kinds of material and energy cycle processes in Earth surface, but also closely related to the various researches, like global change, biodiversity, sustainable development and so on. Land cover mapping and change detection are the basic researches in the study of land cover, which can provide basic data support for understanding the current status and dynamic change processes of land cover and for predicting its future scenarios. Mountain, an essential part of Earth surface, has rich in biodiversity resources and water resources, and is the hot region of the global change research. However, severe challenges arise for land cover mapping and change detection in mountainous areas, such as the rugged terrain, the high spatial heterogeneity of land surfaces and the frequent cloud contamination of satellite imagery, which makes it lagged behind both theoretically and practically and become the bottleneck of this research field.This paper mainly research on the remote sensing methodologies for land cover mapping and change detection in typical mountainous areas. The main contents of this research were (1) land cover mapping method based on multi-source and multi-temporal satellite images and geo-knowledge in mountainous areas; (2) land cover change detection method based on the two image pairs and multiple change indices in mountainous areas; and (3) ecosystem types mapping in Southwest China (including Sichuan, Yunnan and Guizhou provinces and Chongqing municipality).This dissertation drew such main conclusions as follows:(1) Comprehensive decision-making based on multi-source and multi-temporal satellite images and related geo-knowledge is the main trend for the land cover mapping and change detection. At the same time, the object-oriented classification and change detection methods has become the core methods of this research field.(2) Decision tree algorithm can dig out the most effective features to distinguish the various land cover classes from a mass of multi-temporal satellite images, and to maximize the value of the multi-source and multi-temporal satellite images in land cover mapping. Hierarchical classification is an effective land cover mapping method for those areas with complex and heterogeneous landscapes by increasing classification time and decreasing the number of land cover classes in a single classification process. The multi-temporal information combined with growing season and non-growing season can significantly improve the mapping accuracy of forest types in mountainous area compared with single-temporal or multi-temporal images of single season, and can simplify the classification rule sets.(3) The multiple change indices (CVA, VS, ΔNDVI, ΔMNDWI and ΔNBR) and two image pairs (one leaf-on pair and one leaf-off pair) can maximally obtain all potential land cover change areas, and avoid the false change areas by using land cover change knowledge, caused by the differences of acquired time and atmospheric conditions between image pairs.(4) The 30m-resolution ecosystem map over Southwestern China with 144 ecosystem categories was generated by fusing multi-source data and the related knowledge. The multi-source data include the 30m-resolution land cover map and the vegetation map of China at a 1:1 000 000 scale. Three fusion strategies were contained: hard matching, buffer matching and merged categories matching. It not only have plentiful ecosystem categories contained in the original vegetation map, but also have accurate boundary for each ecosystem patches belonged to the 30m-resolution land cover map.The key scientific problem contained in this dissertation is how to integrate the multi-source, multiple-temporal, multi-scale and heterogeneous data to improve the quality of land cover mapping and change detection.The main innovation points are as follows:(1) It innovatively regularized the geo-knowledge in mountainous area and successfully applied them in land cover mapping and change detection.(2) The two image pairs and multiple change indices are firstly used in object-oriented land cover change detection, which changed the traditional change detection methods mainly depended on a single image pairs and change index.(3) A new strategy for ecosystem types mapping was proposed in this dissertation, which also help to extend the field of land cover mapping into the ecosystem types. 

Document Type学位论文
Identifierhttp://ir.imde.ac.cn/handle/131551/18896
Collection数字山地与遥感应用中心
Affiliation中国科学院成都山地灾害与环境研究所
Recommended Citation
GB/T 7714
雷光斌. 山地土地覆被遥感自动制图与监测方法研究[D]. 北京. 中国科学院研究生院,2016.
Files in This Item:
File Name/Size DocType Version Access License
山地土地覆被遥感自动制图与监测方法研究.(52040KB)学位论文 开放获取CC BY-NC-SAView Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[雷光斌]'s Articles
Baidu academic
Similar articles in Baidu academic
[雷光斌]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[雷光斌]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 山地土地覆被遥感自动制图与监测方法研究.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.