Study On High Resolution Remote Sensing Image Fusion And Object-Oriented Information Extraction

The successful launch of the GF-2 satellite marks the Chinese remote sensing satellite into the Sub-meter "High Score Age".So far GF-2 satellite was launched only more than two years,the research on the application of GF-2 image has a certain significance.So this paper carries out the pretreatment operation of GF-2 image radiation calibration,atmospheric correction,orthorectification,image fusion and other pre-processing operations,which mainly studied the image fusion and fusion quality evaluation,and select the better fusion image for the late research on Object-oriented information extraction,it laid a certain foundation for further reserch.The main research work of this paper is as follows:First,study on the pretreatment of GF-2 Image.System analysis,summarizes the basic theory,characteristics and application of remote sensing image fusion,In-depth study analyzes the commonly used 9 kinds of fusion algorithm,which focuses on research in the selection of remote sensing image fusion algorithm in practical application.Second,in view of the production process,we mainly rely on the subjective human eye to evaluate the fusion image,and the subjective evaluation has great correlation with the observer's visual perception and experience level,and does not have the objectivity.We study the objective evaluation index and system of remote sensing image.However,in the quality evaluation,even if the same fusion algorithm is aimed at the same image,the application is different,the application direction of the image is different,the evaluation of the quality of the fusion is not the same,need to make specific evaluation for the specific image,So this paper based on the GDAL library using c#language to write the objective parameter calculation software which integrates the commonly used parameters,it can provide a stable objective quantitative standard for the quality evaluation of remote sensing image,moreover,and the objective evaluation process is more convenient,simple and fast.Finally the better fusion algorithm for the experimental data is Subtractive Resolution Merge.Third,Taking the Tongzhou District of Beijing as the experimental area,the subtractive resolution Merge fusion method was used to fuse the image,and the object-oriented classification method was used to extract the large-scale coverage information of the main urban areas in the northwest of Tongzhou.mainly studied the object-oriented multi-scale segmentation and layered extraction method,studied the characteristics of vegetation,water,roads and buildings on high resolution images,based on this,the rule set is established to classify the study area and compared with the standard nearest neighbor classification method,the results show that the accuracy of rule-based object-oriented classification results is better than that of the nearest neighbor classification in supervised classification.