Using Cart Algorithm Extract Residential From Landsat8 Images

Residents to be called a human settlements, it is a local or regional population centers, and in accordance with the size of the geographic scope or the local population into long live the city, town or village. Residential is a product of human society, the development of production, it is both the place of residence of human life, but also in the place of production and other activities. Residential modern society, in space, economic or political are inextricably linked, forming a complex system, and assume the economic, political, military, religious and other social functions to promote the economic development and social progress. Because residents played a functional role similar to the "nerve center" in modern society, and therefore also in the military field has an extremely important, is to select targets and objectives to protect key targets. Remote Sensing has wide coverage, is the potential of higher topographic map do not have the advantage, so the use of remote sensing images to study residents with a wide range of social, economic and military sense, has been a hot topic of concern.Based on the basic principles of remote sensing image analysis and comparison of existing information extraction of remote sensing image classification method, we present a Cart classification and regression tree algorithm to quickly create, extract information from residents to LandSat8 image method. The central idea is the use of residents with different backgrounds in spectral characteristics, texture and other aspects of computing and build a large number of eigenvalues and use SPSS Clementine data mining software every step of the way to find the most remote sensing topics from numerous indicators excellent decision tree model classification parameters sequentially extracted information will be residential.In this study, the middle reaches Zhangye, Linze vicinity expanded. After the reforms, the city of Black River region has undergone tremendous changes, a lot of arable land is turning into urban land and rural residential land, urbanization and gradually accelerated. But the development of the region and urban system is still in its infancy polar nuclei style, prominent urban centers and secondary cities is very weak. Study and master the urban development of this area of science to promote the "new town" of the process, land conservation, ecological protection has a very important practical significance.As used herein, the image data is image LandSat8 OLI, multi-spectral data a spatial resolution of 30 meters, the image acquisition time for the July 24,2014.Thesis on the following aspects were studied:the existing remote sensing image classification Indexes of Residential extraction methods; spectral information and texture characteristics of residents in the land; regression tree based on SPSS Clementine software and optimization.The main conclusions of this paper are:1. Thematic index in combination with a variety of remote sensing, compared to a single remote sensing image classification thematic index, significantly reducing the classification error and improve the accuracy of remote sensing image classification.2. When the remote sensing thematic index and other indicators are more features, the use of classification and regression trees to establish Cart algorithm can quickly select a classification index, which greatly improves the working efficiency. SPSS Clenmentine of use, making the data mining process easier.3. Construction of appropriate characteristics index helps get better results. We build BOBI index is analyzed in the study proposed on the basis of a large number of feature types in the region on the spectral characteristics, in the experiment with the existing NDBI thematic index with the use of the remote sensing, also played a better classification results.4. Sample selection is very important for this research methodology applied. Because this article is to rely on the use of samples to create a classification and regression tree research methods, the quality of the sample directly affects the quality of the final classification results.5. The texture feature has the function of extracting the information from the residents, but the size of the window is determined according to the concrete conditions.. In this paper, the texture and texture of different window sizes are different in the classification process.