Application Of Buildings Extraction Based On Object-oriented Method In Baiyun District

In recent years, remote sensing data information extraction technology is developing rapidly, optimization of efficient image information extraction technology and improving the recognition accuracy has become the research hot spot. Many researchers at home and abroad have further study of image information extraction technology, and achieved some results. Remote sensing information extraction technology is a new method of interpretation of remote sensing information, different from traditional image interpretation method. Analysis object of object-oriented method is collection of pixels based on feature selection and extraction information extraction.The study area of this study is the part of GuangZhou Baiyun district. Consider QuickBird-2 data and GF- 1 data to be the foundation of building information accurately extracted in the study area. The texture, structure and shape features of remote sensing data and visual interpreting empirical knowledge as the basis, establish rules of characteristic library information extraction, the working efficiency of information extraction. On this basis, the classification results are accuracy by means of error matrix evaluation. For the same data are compared, and methods for the same method to compare different data sources, through the comparison of horizontal and vertical to demonstrate the application of the various methods and data. And put the classification results and building information extraction method based on the traditional classification result comparison, analysis the advantages and disadvantages of each method.The result of the experiment conclusion is as follows: in terms of method Multi-scale segmentation and the method of knowledge base for building land information extraction is more effective, and is closer to the actual value. This study in object-oriented classification method in the artificial ground objects access to information of the actual operation of the project, provides a new solution. In terms of data source, GF- 1 data information extraction result error distribution more uniform, higher consistency of classification results, QuickBird data- 2 is more suitable for small scale of research and application of the information extraction. In the practical application, for a wide range of remote sensing monitoring, Considering the economic cost and time cost(GF- 1 single wide data, low cost), on the basis of the data source GF- 1, the object-oriented classification method based on rules is more reasonable.