Study On Extraction And Expansion Of Residential Arae Based On Remote Sensing Images And GIS Data

Earthquake disaster risk indicates the degree of earthquake hazard to the hazard-affected bodies. Residential area(ResA) is the carrier of population and buildings, which provides a basic structure of the spatialization of socioeconomic data, such as population and economy data. The rapid acquisition of ResA data by remote sensing technology, as well as the allocation of urban and rural population and buildings according to the distribution of ResA, could improve the accuracy of spatial distribution statistics of population and buildings. Therefore, the uncertainty of earthquake disaster risk assessment could be significantly reduced. Besides, with the development of social economy, the scale of urban and rural ResA changes in accordance. The study on the area expansion of ResA will show practical significance to update the basic data of earthquake risk assessment, as well as the city and rural construction planning.The thesis carried out the studies on extraction and expansion of ResA through multisource remote sensing images and GIS statistics data. The main studies as follows:First of all, this thesis summarized the research status home and abroad on extraction and expansion of ResA by remote sensing data. Based on those existing problems included that few studies had been applied on Landsat-8 multispectral remote sensing images of ResA extraction, high subjectivity on segmentation threshold choice of decision-tree classification, low accuracy of ResA identification by medium-resolution images, and lacking in multi-period landuse spatial data on ResA expansion studies, we discussed the possible solutions.Secondly, we set the Dong Chuan District, Yunnan Province as the study sample, using the Landsat-8 multispectral remote sensing images, and put forward the decision tree classification model with expert knowledge. We automatically found the binaryzation segmentation threshold value of classification decision tree based on OTSU method. Then using the classification decision tree classified the landuse, and identified the ResA objectives. The overall accuracy of classification reached 90.6%, and the Kappa coefficient was 0.87. Among these, the mapping accuracy of cities, towns, and rural ResA improved by 13%, 32%, and 35% respectively compared with 1:100000 landuse data. It indicated the effectiveness of the methods in this thesis.Thirdly, we integrated the high-resolution ZY3 images to carried out the reclassified research on urban ResA. We set Qin Zhou District, Tianshui City, Gansu Province as an example, the classification decision tree model based on OTSU method had been firstly applied on Landsat-8 images, which identified the urban ResA(the landuse overall accuracy of classification was 90.6%, and ResA mapping accuracy was 0.89). Therefore, the ZY3 images had been further applied to inside urban ResA, and re-classied the buildings to through the object-oriented method on(with the overall accuracy of classification 81%). Finally the ResA data with different levels of detail were obtained. It indicated that based on multi-source remote sensing images could make up the defects of low accuracy of medium-resolution ratio images identification.Last but not least, we put Yunnan Province as an example, through the analysis on the driving forces of GIS statistics data to the ResA expansion, and built a statistic model without multi-period landuse spatial data. Based on the assumption of future economy and population increase, this thesis estimated the scale of future ResA expansion that by the year of 2025, the area of ResA of Yunnan Province will reach 920000 hectare. In the end, based on the ResA feature polygon, we attempted to design a simple spatial expansion plan of ResA.