The Quality Analysis And Assesment Of Domestic Satellite Images With Sub-meter High Resolution

With the launch of many domestic satellites,the image effect of Chinese remote sensing to earth observation technology has been significantly improved,and the application of satellite remote sensing data has also entered the era of localization.The successful launch of GF-2satellite marks that the application level of remote sensing satellites in China has entered a new stage,that is,China has successfully entered the " Sub-meter High Resolution Time".However,there is no comprehensive and systematic conclusion about the quality of domestic sub-meter-class satellite images.As the basic data of remote sensing image,the quality of remote sensing image directly affects the application feasibility of image data.Therefore,the objective evaluation of satellite image quality is an essential part before the use of satellite images.In order to verify the domestic sub-meter high resolution satellite image data quality,select GF-7,GF-2,SV-1,BJ-2 four satellites image data,Conduct qualitative and quantitative analysis and evaluation of different image data quality of satellites,this paper main work is as follows:(1)According to the satellite image pre-processing process,we produce orthophoto correction and fusion image products,and analyze and evaluate the geometric accuracy of the image and the quality of the fused image.By using different control point layout schemes to achieve block network adjustment,analyze the uncontrolled and controlled accuracy of the image,and verify the geometric accuracy of the panchromatic image,the results show that the geometric accuracy of the four satellites meets the requirements of making a 1:10 000 scale topographic map,of which the SV-1 image plane accuracy is the highest compared to other satellites,the GF-7 and SV-1 accuracy are comparable,and the error in the GF-2 plane is relatively large;Using four fusion methods,qualitative and quantitative analysis and evaluation of image fusion quality and corresponding most suitable methods are analyzed and evaluated from the aspects of image information,clarity,and spectral fidelity.According to the selected Gram-Schmidt(GS)algorithm,principal component analysis(PCA),super-resolution Bayesian(Pansharp),nearest neighbor diffusion(NND)four fusion methods to experimence,the results show that the Pansharp method is the best fusion method for GF-7,GF-2 and BJ-2.The Gram-Schmidt(GS)method is more suitable for SV-1 fusion.(2)The spectral quality of four satellite multi-spectral images was evaluated by the ability that different objects to be identified and extracted.Select 6 different typical of elements,observe the reflectivity change curve of the same object on different images,obtain the relationship between wavelength and reflectivity,establish the corresponding area of interest for the typical class,and judge whether the separability between the objects meets the quality requirements of subsequent applications.Finally,realize the classification,analyze the overall accuracy and confusion matrix analysis of different classification methods,and further evaluate the spectral quality of multi-spectral data.The results show that the neural network classification method has the best classification accuracy for GF-7,GF-2 and SV-1 satellite images,the highest overall classification accuracy of SV-1,and the strong ability to distinguish building land,forest land and paddy field.GF-7 has obvious advantages in the identification of dry land and forest land,and BJ-2 better controls the details of roads.(3)The engineering quality of the geographical elements,such as visual features,gray features and texture features,is evaluated.In the main urban areas and suburbs with different complexity,the characteristics of basic gray scale information,information,clarity and texture characteristics are analyzed through subjective and objective analysis,according to the comparison of horizontal and vertical of 8 indicators,and the engineering quality of domestic sub-meter satellite panchromatic and multi-spectral images is comprehensively evaluated.The results show that BJ-2 image can reflect complex ground information;multi-spectral image of GF-2 satellite is more suitable for single and stable suburbs;GF-7 satellite can extract more delicate and rich ground information for urban areas;SV-1 image is effective in near-infrared band,suitable for extracting farmland type.Research through three main aspects of GF-7,GF-2,SV-1,BJ-2 pan images,multi-spectrum images,fusion images,the image of geometric accuracy,fusion image quality,spectral quality,engineering quality,through comparative analysis of the satellite image preliminary evaluation conclusion,for subsequent industry application and successor development research has certain reference value.The paper has 26 figures,34 tables,66 references.