Research On Segmentation Scale Of Information Extraction With Towns Based On GF-2 Image

With the extensive application made "High resolution" series of satellites have launched their data, high-resolution remote sensing satellite of the curtain has opened the era of application, how to more fully tap the potential of domestic high-resolution remote sensing images, multi-angle play a significant role related to remote sensing data, the field of scientific research and related staff are urgent responsibility. On the current application of "GF-1" image in terms of application development for the "GF-2" image is a great potential for development and scientific issues to be solved.With respect to the low-resolution remote sensing images, a high spatial resolution image with geographic information is complex and detailed, clear layout, geometric,spectrum, texture and more accurate information and so on. At present, by virtue of its high efficiency and a high degree of accuracy and other characteristics, the use of object-oriented approach to information extraction has become a trend. Its processing target is not one isolated pixels, but with a variety of features and semantic features of the relationship between "homogeneous" polygons, used properly and the image is reduced fragmentation "salt and pepper noise" occurs to a certain extent, the scale is divided Department is the impact of object-oriented technology plays a key factor in good and bad. So, how to make object-oriented technology advantages into full play,becoming the hotspot areas by dividing scale optimization technology improvements.Meanwhile, with the continuous progress of new urbanization, how to timely and comprehensive grasp of the city and the surrounding surface features detailed information on the scientific planning of town, reasonable industrial layout, improve the "smart city" and "big data" system, etc. development strategy is important.Based on the previous studies on the basis of object-oriented approach, to fully tap the score II in various related image information for object-oriented approach in key technology-segmentation scale were properly improved and combined to build multi-feature set, etc. It means a set of urban information extraction method for imageII score, proven effective. In this study, the ancient city of Xinjiang Ili Turks gossip typical study area for urban feature of complex information, information extraction is difficult, the problem of low precision, commitment to segmentation scale as a breakthrough, to further enhance the effect of the object-oriented classification and precision, it is desirable to be able to effectively improve the object-oriented technology in Xinjiang Town feature information extraction work to provide a scientific reference. The main research are as follows:1. Image segmentation is the basis of object-oriented technology and key, its segmentation quality is good or bad, will directly affect the accuracy of information extraction. Based on previous research on the use of multi-scale segmentation method for image segmentation, the method is suitable for object-oriented technology, the resulting segmented object image more in line with actual boundary feature.2. Multiscale segmentation segmentation scale selection information extraction results have an important impact. In order to solve common object-oriented classification of "less divided" and "over-segmentation" and other issues, this paper summarizes the previous selecting optimal segmentation scale method based on the first use ESP to be divided about the scale range of coarse screening; again the use of the feature spectrum luminance mean standard deviation divided under different scales for statistical analysis, to further narrow the range of scales; and finally determine the optimal segmentation scale parameters of each division and finalize the layers of various types of surface features that is by controlling the experiment and experience of human-computer interaction.3. The composite picture was optimal after multiscale segmentation, the need for another important part of object-oriented technology- the characteristic parameters selection to improve the importance, for the relevant characteristics of the country was complete while maintaining as much without cause within the scope of the redundancy, we use all the relevant parameters Closest classification principle, firstselect the filter, and finally determined by testing individual group optimal feature,and then use these features for debugging threshold and membership function,determine the classification rules and extract the target information.4. The information extraction process, was due to the different parts of geometry,spectrum,and so on. the first by correlation analysis and human judgment will feature extraction type layered, each layer to select the most appropriate segmentation scale;then base was different according to different types of characteristic parameters were extracted for the misclassification phenomenon minimize leakage points, the highest classification accuracy mentioned; and finally supervised classification results under the classification results with the same experimental conditions do primary precision contrast, again control experiments utilizing segmentation and multi-scale feature set on classification results in order to achieve further comparative analysis to verify the effectiveness of the technology improvements.5. China’s urbanization development is inseparable from remote sensing technology, domestic and foreign scholars have high-resolution remote sensing applications in urban terrain classification, land use survey, a number of areas of environmental data, census, urban and rural planning. The use of domestic satellites obtain high-resolution, multi-spectral, high-quality image and other information on the city level is already difficult, but how to use these remote sensing to solve practical problems encountered in the process of urban development also requires human effort. Finally, the prospect of technology for application analysis, concludes the paper and look to further explore through optimized segmentation scale score of Town image information extraction technology,the country town feature information retrieval and other areas have a certain scientific reference value and practical significance.