Research On Building Information Extraction Based On GF-2 Remote Sensing Satellite

In recent years,China has made remarkable progress in the field of remote sensing satellite applications.The successful launch of GF-2 remote sensing satellite has changed the long-term dependence of high-resolution observation data in China on imports and achieved the leap of remote sensing satellites to the age of sub-meter.Up to now,GF-2 satellite has been launched and running for more than four years,and the application of GF-2 satellite data has always been a research hotspot.As the basis of remote sensing satellite image data application,relevant information extraction,analysis and research work is also undergoing continuous development and modification.Building foundation pit,building monomer and building height are the basic external characteristic information of building construction.In the process of urban construction and supervision,the satellite remote sensing images could be applied to extract the information of building foundation pit,building monomer and building height.Therefore how to use high-resolution remote sensing satellite to effectively extract building information and its related processing technology become the focus and hotspot of urban remote sensing research at present.The present thesis selects satellite image data of three demonstration communities in Yangzhou city as experimental data based on GF-2 satellite image data.In the present thesis,firstly,a series of image preprocessing operations are carried out on the satellite image data in these regions,and the pre-processed images are used in later stage for the monomer recognition of object oriented building images,shadow information extraction and building height inversion,so as to verify the feasibility of information extraction from GF-2 satellite image data,evaluate the high accuracy thereof and prove that it meets the application precision requirements.The specific contents mainly include:The GF-2 satellite image data is taken as the source data and the original high-resolution satellite images on February I1,2015 of GF-2 in Yangzhou are extracted for image preprocessing research which mainly includes radiometric calibration,atmospheric correction,orthorectification,image fusion,image cropping and geometric correction.After radiometric calibration,atmospheric correction and geometric correction are performed on the images,then multi-spectral data with resolution of 4m and panchromatic data with resolution of lm are combined to perform image fusion by NNDiffuse Pan Sharpening method and obtain high-resolution remote sensing fusion images,after which,the obtained fusion images are cropped into the desired regional images and finally corrected geometrically to reduce the accuracy error caused by geometric deformation.Building foundation pit and monomer extraction based on GF-2 satellite are studied.The object oriented information extraction method based on specific rules is used to extract the monomer and shadow information of buildings according to the geometric data,texture information and spectral features of buildings and the segmentation algorithm based on edge information.And images of three demonstration communities in Yangzhou are taken as example to verify the feasibility and high efficiency of the present extraction method.Building height inversion based on GF-2 satellite is studied.The shadow information of buildings is extracted through the calculation of the shadow feature index,and the shadow length information is calculated accordingly.According to the zenith angle and azimuth angle of the sun and satellite at the same time,the inversion model of building height is established,and the inversion calculation of the actual building height is carried out to obtain the building height information.The measured height information of three demonstration communities in Yangzhou is compared with the actual height of buildings to verify the accuracy of the inversion model of building height proposed in the present paper.In the present thesis,the object oriented feature extraction method based on rule information is studied,and the geometric features,spectral features and texture features of buildings in the image are applied to identify and extract the information of building foundation pits,building monomer and building shadow;meanwhile based on the extracted shadow length of buildings,the spatial height model of buildings by satellite remote sensing is established to invert the height of buildings.In the present thesis,three communities in Yangzhou are taken as examples,and the building information of these three Yangzhou demonstration communities in remote sensing satellite images is extracted and applied,including building foundation pit recognition,monomer recognition,shadow information extraction and building height inversion,and verification thereof,in which the average accuracy of foundation pit identify is 96.7%,the average accuracy of monomer building recognition is 98.2%,and average accuracy of height inversion is better than 95.2%.The experimental results show that the present method has a quite high precision of recognition and extraction.The object-oriented feature extraction method based on rule information is feasible,which could be render technical support and service in the application of remote sensing in urban fine management,and is of high utility value.