Research On Extraction And Application Of Settlement Information Based On Remote Sensing And GIS

With the development of the times,the development of various types of remote sensing satellites with different resolutions is advancing by leaps and bounds.Due to the continuous improvement of image spatial resolution,the gradual enhancement of image acquisition ability,the better image status and other reasons,remote sensing images have begun to widely serve various industries.At the same time,with the continuous progress of science and technology,it has become a hot issue to extract the ground feature information from various image data sources with automatic methods.As the center of human production,life and other social activities,the extraction of settlement distribution can help people intuitively understand the spatial distribution of local settlements,and provide some reference for the construction of local urbanization and the optimization of spatial distribution pattern of settlements.Based on GF-1 remote sensing image and Sentinel-2 remote sensing image,this thesis takes Anju District of Suining City as the research object to extract its settlement information.(1)Based on e Cognition software,this thesis discusses the methods and techniques of image segmentation,and compares several methods of image segmentation.According to the research purpose,the multi-scale segmentation method was selected for segmentation,and then the ESP scale evaluation method was used to analyze and select the most suitable segmentation parameters of the two images.Finally,different hierarchical segmentation systems were designed for the two images according to the analysis results.(2)Through the knowledge discovery of spectral features,spectral derivative features,shape features and texture features of various kinds of ground objects in the two kinds of images,the feature knowledge base of ground objects was established.Based on the feature knowledge base,the GF-1 and Sentinel-2 remote sensing image classification and extraction systems of the study area were constructed respectively.The extraction system is applied to the homologous images of the whole Anju area to verify its feasibility,and was used for the recognition and extraction of features.(3)This thesis compares the results of object-oriented classification extraction method,sample oriented classification extraction method and cart classification method,and analyzes their advantages and disadvantages and extraction accuracy respectively.It is found that the object-oriented classification method has the best effect,so it is selected to extract the settlement information.(4)The accuracy of GF-1 and sentinel-2 are 91.28% and 85.40% respectively.From the perspective of usage,GF-1 image is more suitable for the analysis of the internal structure of the settlement,sentinel-2 image is more suitable for the research of settlement expansion and spatial distribution over the years,and the cost is lower than using GF-1image.(5)Combining the extracted results with DEM data to analyze the spatial distribution characteristics of settlements in Anju District,it is found that the settlements within the administrative scope of Anju district are mainly distributed between the elevation of258m-392 m,accounting for 97.39% of the total area of settlements in Anju district and16.04% of the total area of this level.At the same time,the settlement area of Anju district is mainly distributed in the range of 2°to 15°slope,which accounts for 85.54% of the total settlement area of Anju district and 18.05% of the total slope area.In addition,there are still some settlements located in steep slope area and high elevation area in Anju District,which have great security risks,so it should be reasonably optimized.