遥感影象中居民地信息的自动提取与制图

如何自动获得居民地矢量信息是遥感和地理信息系统领域研究的热点问题。以南京市江宁区为研究区域,基于光谱特征分析,建立决策树模型进行居民地信息的自动提取。重点研究如何将已提取的居民地图斑进行形态综合,得到满足地理信息系统数据建库与更新、遥感制图等的可视化图形数据,为灾害评估、城镇扩展和环境变化研究等提供必要的基础信息。最后,给出文章方法的实验结果并做出讨论。

Automatic Extraction and Mapping of Residential Features from Remote Sensing

How to automatically obtain visualized vector information of settlement is a key problem in the fields of remote sensing and GIS. A number of researchers have been conducted in this field for many years. They all believe that the key problems about automatic obtaining visualized vector information of settlement from remote sensing image are two ones, the first one is information extraction of settlement from remote sensing image, the other is the generalization of the extracted information. At present, the main method of settlement information extraction is man-machine interaction. This approach is based on the knowledge of the man about residential area in the image. Although the accuracy of the method is higher, the efficiency is lower. The research on the generalization of the extracted settlement is seldom reported. In this paper, taking Jiangning County of Nanjing as a case study area, extraction of residential information from SPOT images based on a simple model of decision tree through analysis of characteristics of residents and other land-use types on the image are discussed firstly. Then, the paper's attention is paid to how to generalize the residential features in classified imagery. Mathematical morphology is a science about morphology and structure based on set theory. The key problem to use it generalizing residential patches is to design a lot of fair structure elements. Through the analysis of the extracted settlement information, a series of structure elements are designed. They are used to fill the irregular holes inside patches, smooth the edges of the patches and keep a right angle turn. The map scale determines the size of structure element. Then according to given steps, the generalized result will obtained. The generalized residential vector patches information would meet the needs of GIS and remote sensing mapping and severed for the loss estimation of natural disaster and the study of urban extension and environmental change, etc. At last, the test results are obtained and some conclusions are derived.