Study On Large-scale Residential Area Extraction Based On Remote Sensing Images

With the economy development and the global population growth, the society gradually get rid of theoriginal predominantly agricultural, and continuously improve the level of urbanization, more and more people went into the cities, along with the growth and expansion of the city. The urban sprawl often infringes upon viable agricultural, residential areas or productive forest land, none of which can resist or deflect the overwhelming momentum of urbanization. Land cover land use mapping data serves as a basic inventory of the land resources for all levels of government, environmental agencies, and private industry throughout the world.Remote sensing images have been widely used from local to global scale land cover classification, target identification and thematic map production, because remote sensing images own some technical advantages, such as multi-resolution, wide coverage, repeatable observation and multi/hyperspectral and so on. Remote sensing methods can be employed to classify types of land use in a practical, economical and repetitive fashion, over large areas. With multi-temporal analyses, remote sensing images provide a unique perspective of how cities evolve. The key element for mapping rural to urban land cover land use change is the ability to discriminate between rural uses (farming, pasture forests) and urban use (residential, commercial, and recreational).With the further process of reform and opening up, the rapid economic and social development, the urban and rural economy of our country has developed greatly, along with the accelerating process of the urbanization, the residential area expands obviously. Beijing, as the capital of China, the residential area in where changes particularly obviously, since China adopted the reform of the real estate market at the end of the20th century. In order to monitor the residential area changes, we must obtain objective and accurate residential area information first. The study area in this dissertation is in Beijing. Landsat satellite images were used to extract residential areas by using unsupervised classification method, spectral analysis method, supervised classification method, object-oriented classification method and NDBI index. Then, accuracy evaluation was conducted by using Kappa and confusion matrix. Based on the supervised classification results, the study area is divided into rural residential area and urban residential area, combined with the survey data. It was found that different classification methods have different classification results in both rural residential area and urban residential area. Considering multiple classifiers parallel combination thoughts, the classification results were combined by using "winner takes all" method, and in this way, the more accurate Beijing residential area information were got.Finally, based on the extracted residential area information above, Beijing three decades settlements spatial and temporal distribution characteristics of scale, strength and form were summarized. Here we divided Beijing by the fifth ring road and found that whether within the ring road or not, Beijing’s overall residential area scale is growing. The residential area and the number of patchesis are increasing while some of small patches merge into larger plaques. The tensile strength in the ring is greater than that outside the ring.In this study, we obtained that the residential area’s scale is expanded, its expansion strength is rising and the morphology of plaque area is increasing. All of the results can provide strong supports for further simulation and prediction of Beijing.