Study On Object-oriented Land Use Information Extraction Based On GF-2 Image Of Jin'e Town,Long Chang Country,Si Chuan Province,China

Due to the rapid progress of domestic remote sensing technology,making the domestic satellite resolution has been high.In terms of GF-2,the resolution of the panchromatic band is 0.8 meters,multi spectral resolution of up to 3.2 meters.GF-2 satellite has been launched for more than two years,and the application of GF-2 is not enough,so this paper uses GF-2 images to extract the land use information of the selected research area.The traditional method of extracting the image information mainly uses the spectral information of pixels to extract the information of the ground objects.The object-oriented information extraction technology,which is based on the "homogeneous object" is the smallest unit information extraction,this method not only can use remote sensing data rich spatial structure feature and texture feature information,can also use the spectral information of data,and can also be combined with experience and knowledge structure of users as a basis for information extraction,this method uses the combination of the three,can improve the accuracy and effectiveness of the extraction of remote sensing information,the method in information extraction gradually play its unique advantages,showing potential applications of the method.This paper chooses GF-2 image data as the test data of the study,according to the characteristics of the data,the image preprocessing includes radiometric correction,geometric correction,image enhancement etc.this research,object-oriented classification software is eCognition based,object-oriented information extraction method through the research of the multi scale image segmentation the experiment of the study area,and find the scale parameters,suitable for extracting image features information classification rules for reference,the class hierarchy created,image data in the study area according to the needs of a large number of test and verification.The main achievements of this paper are as follows:(1)The study area GF-2 image data preprocessing research,including: radiation correction,atmospheric correction,image fusion process,feature information is preprocessed image can better reflect the object itself,preparing for the image segmentation and information extraction.(2)Various factors of multi-scale image segmentation results on a large number of tests,among which is the segmentation scale,color and shape factor,obtained the best extracting features information segmentation scale,establish the classification rules for the image data of the object oriented information extraction.(3)In order to show the advantages of the object-oriented method more clearly,the maximum likelihood method is used to carry out the experiment,and the results are compared with the object oriented method.