基于FNEA与D-S证据理论的土地利用遥感分类研究
文献类型：学位论文
| 作者 | 何海清 |
| 学位类别 | 硕士 |
| 答辩日期 | 2009 |
| 授予单位 | 中国科学院水利部成都山地灾害与环境研究所 |
| 授予地点 | 成都 |
| 导师 | 李发斌 |
| 关键词 | FNEA D-S证据理论 基本可信度 土地分类 |
| 其他题名 | Study on Remote Sensing Classification of Land-Use Based on FNEA & D-S |
| 学位专业 | 地图学与地理信息系统 |
| 中文摘要 | 随着信息技术、经济及城市化快速发展，土地利用状况变化很大，各部门对土地信息的需求越来越大，建立准确、实时的土地资源数据库显得尤为重要。传统的土地调查方法已经无法准确地反映土地利用现势状况。因此，探索一种快捷实用的土地信息提取方法，准确、快速地进行土地资源调查，保证土地利用现状资料的现势性成为国土资源管理部门一项十分重要的任务。运用遥感技术摸清土地的数量及分布状况，已成为土地利用调查中普遍采用的手段。 本文在分析传统计算机自动分类和人机交互目视解译提取土地利用信息的不足之处后，提出了基于FNEA与D-S证据理论的遥感图像分类方法。该方法在区域分割算法FNEA对遥感影像分割的基础上，根据土地利用分类体系中地物光学和雷达等特征，构建D-S理论的鉴别框架、计算基本可信度，并按照一定的决策规则建立分类模型，从而划分各图斑土地利用归属类别。 该方法以图斑作为子单元进行对土地进行分类，能解决传统的基于像元分类造成多而零散的分类结果问题，与实际土地调查以图斑为单元相符合。D-S分类利用多种证据对图斑进行分类，比依靠单一信息源进行分类证据更充分、更全面，并把各个证据融合后智能化识别土地类型。以SPOT-5数据和RADARSAT-2数据进行土地利用遥感分类为例，对基于FNEA与D-S证据理论的遥感图像分类方法进行了验证，实验结果表明，该方法可以对遥感图像智能化分类，比传统计算机自动分类精度高，且与实际真实情况更接近，比人机交互的方法效率更高，因而该方法具有一定的实用性，在兼有效率和精度方面比现有方法更具优势，尤其在土地利用遥感动态监测提取新增建设用地信息中优势更加突出，能为国土资源调查、土地规划、土地执法和土地督察、资源管理与规划、环境保护提供基础信息和技术支持，并为相关工作开展提供科学依据。 |
| 英文摘要 | With the rapid development of information technology, economy and urbanization, land-use changes in a great situation, departments need more information of the land, so it is very important to set up accurate, real-time database of land resources. Traditional land survey methods can not reflect the situation of land-use accurately. Therefore, it is a very important mission for land management to find a method that can access to land information accurately and quickly. Now, it is popular means to know the land-use information by Remote Sensing (RS). In this paper a new method of RS land-use classification was proposed based on FNEA and D-S theory after analyzing the defect of current classification automatically and translating the land-use information visually. RS image was segmented to sub-objects in the method of FNEA, and a frame of discernment was established according of the Nation Land-use Classification. Evidence, the core of the frame，was extracted from the special and spectral characters of sub-objects. BPA was calculated to establish a decision-making rule, which was used to classify the sub-objects. The method classes land by use of spot as a sub-unit in line with land-use, meanwhile, the method can solve the problem that traditional pixel-based classification resulted in the classification of many scattered results. D-S classes the spots combinated a wide range of evidences, which is more fully, more comprehensive for identifying land types intelligently. An example of classifying land-use with SPOT-5 and RADARSAT-2 data was carried out. Test result shows that the method can class remote sensing intelligently, which is more high precision than existing automatic classification, and closer with the actual situation, also which is more efficient than the artificial methods, so the method has certain practial value and more advantages both the efficiency and accuracy, especially the application of dynamic monitoring land-use dynamic for new information in the construction site. So the method may provide information and scientific basis for investigating land, planning land, inspecting land and protecting environment. |
| 学科主题 | 摄影测量与遥感技术 |
| 语种 | 中文 |
| 公开日期 | 2010-10-13 |
| 分类号 | F29;F32 |
| 源URL | [http://ir.imde.ac.cn/handle/131551/2154]  |
| 专题 | 成都山地灾害与环境研究所_成都山地所知识仓储（2009年以前） |
| 推荐引用方式 GB/T 7714 | 何海清. 基于FNEA与D-S证据理论的土地利用遥感分类研究[D]. 成都. 中国科学院水利部成都山地灾害与环境研究所. 2009. |
入库方式： OAI收割
来源：成都山地灾害与环境研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。