Research On Water Information Extraction Method Based On GF-2 Remote Sensing Image

‘Same spectrum from different materials' and 'same material with different spectra'are a common phenomenon in remote sensing images.It will become more acute as the resolution of image data increases,which seriously affects the accuracy of water information extraction.In order to improve the responsivity of the accuracy of disaster monitoring and water information management,this thesis focuses on the water information under different data sources,and proposes corresponding improved algorithms based on domestic and foreign problems,providing high-resolution optical data in water information extraction applications.The research foundation is as follows:(1)In order to realize the application research of high-resolution optical image in water body information extraction and solve the problem that water body and shadow(especially tall buildings)and dark color objects are difficult to distinguish in the process of water body information extraction,a New Comprehensive Water Index Method(NCWI)is proposed to enhance water information in this thesis,besides,this thesis introduces the Chicken Swarm Optimization algorithm(CSO)to optimize the maximum between-class variance method(OSTU)algorithm to obtain the optimal segmentation threshold automatically,The degree of program automation and operation efficiency are greatly increased.(2)In the process of water body information extraction,the traditional algorithm only considers the spectral difference between the features,and ignores the spatial difference characteristics,which makes it difficult to distinguish when the features appear 'same spectrum from different materials'.Considering the seriousness of the boundary loss and the limitations of the OSTU algorithm,this thesis uses the object-oriented concept to get the optimal hyperplane MWI obtained by the support vector machine SVM,and extracts the object-oriented water body information through fusion with the original data.This not only makes full use of the spectral characteristics of the image and the advantages of space,but also reduces the time of manual adjustment process in morphological processing.Finally,the GF-2 satellite remote sensing image data was used to test the proposed algorithm,and the results were evaluated by the combination of field investigation and evaluation indicators.The evaluation results show that the proposed algorithm is not limited by region and specific time and can achieve better accuracy standards,both visual and index evaluation effects are more prominent;in addition,compared with the former method,the water body extraction results more accurately by using the second methods,which has a strong utility model and better automation effect.