The Quality Assessment Of Domestic Satellite Images Based On Contrastive Analysis

Recently, in order to meet the demands of RS and GIS projects, our country isbecoming getting rid of dependence on foreign satellites by so many domesticsatellites having been launched into space. As a representative of the emergingdomestic satellites, ZY1-02C and ZY3have the characteristics of high temporalresolution, high spatial resolution and multi scale. The article illustrated theassessment on relevant Remote Sensing characteristics of ZY1-02C and ZY3images,which included systematic preprocessing, engineering quality in contrast to foreigncomparative satellite images, spectral quality and thematic application, for thepurpose of making the domestic satellite images a further development and moreextensive applications.In the preprocessing part, the systematization, standardization and industrializat-ion of preprocessing methods were being discussed. Related investigation andtechnical specification could be used as a criterion for this part. The discussion mainlytalked about the inspection of raw data, Ortho-rectification, registration, optimal bandselection, true color and image sharpening.In the engineering quality assessment part, based on ZY1-02C and ZY3, imagestatistical characteristic, image texture characteristic and image energy characteristicwere calculated by Matlab and Envi software compared with SPOT5andreconnaissance satellite. The image statistical characteristics consisted of themaximum/minimum, the average, the standard deviation, the skewness, the clarity andthe histogram. And contrast, homogeneity, angular second moment, and entropy wereincluded in the image texture characteristics. In addition, image energy characteristicsmainly reflected the wavelet detail signal energy and the wavelet edge signal energyof the image.The spectral quality assessment part was the most important part in the article,which also concentrated on ZY1-02C and ZY3in contrast to SPOT5. The part wasaimed at making full use of the spectral character of the typical features in the satelliteimages. In order to explore the spectral quality of the different satellite images, thearticle tried to compare the ability to extract and distinguish the typical features of thedifferent satellite images as a measure. The typical features comprised water, forests,corns, rice and constructions.The spectrum image space, the spectrum character curvespace and the spectrum feature space were the three perspectives of the spectral quality assessment part. Among them, the spectrum image space and the threedimensional spectrum feature space were for distinguishing multi-features. However,the spectrum character curve space and the two dimensional spectrum feature spacewere applied to extract the single typical feature. First of all, the methods of thespectrum image space was about to contrast the feature classification accuracythrough different supervised classification means referring to the above satellitesimages. The satellites image with the higher classification accuracy towards eachfeature type was regarded as the more suitable data source to distinguish the feature,as well as with the higher spectral quality. Secondly, the spectrum character curvespace was a visual method to assess the spectral quality. The main idea of the methodwas comparing the standard spectral character curve and the curve extracted throughthe different satellite images by a variety of spectral similarity measure. The moresimilar the spectrum character curve through the satellite image was, the moresuitable the satellite image was to be the data source to extract the feature, which wasalso with high spectral quality. At last, in the spectrum feature space part, we may firstchoose20samples per feature to record their reflectance of each band in the differentsatellite images, and then project the sample points to the spectrum feature space. Onone hand, in the two dimensional spectrum feature space, there represented onefeature type of different satellites images. We may come to the conclusion that such asatellite image would have the higher spectral quality to extract this feature if thesingle feature cluster or the single feature cluster’s ellipse of such satellite image wassmall, and then the satellite image was considered to be able to extract such a feature.On the other hand, the three dimensional spectrum feature space represented all thefeature types of one satellite image. We also may reach the conclusion that such asatellite image would have the higher spectral quality if one feature cluster was smallenough, meanwhile the cluster had wide enough margins with other feature clusters,and then the satellite image was considered to be able to distinguish multi-features.The article gets the following conclusions based on the above studies:1. Preprocessing: By the inspection of ZY3raw data, the pixels greater than1023were found in the high reflectance region, yet the radiation values of ZY3imagewere quantified for10bit. By the inspection of ZY1-02C raw data, the informationloss caused by atmospheric noise or imaging scans noise was modified by frequencyfiltering. In the optimal band selection part, band Ⅱ,Ⅳ,Ⅲ can be choosed as the bestbands by single band’ s statistical analysis and multi bands combination analysis. Inthe image sharpening part, through qualitative and quantitative comparison, the articlegot to the conclusion that ZY3image sharpening should apply Gram-Schimdt or pansharpening, and ZY1-02C may use IHS or Pan sharpening.2. Engineering quality: Because of the greater quantitative radiation, ZY3issuperior to ZY1-02C, SPOT5and reconnaissance satellite as a whole, next to ZY3isreconnaissance satellite, and then SPOT5, the last is ZY1-02C. By comparison of the satellite image textures, we found ZY3is the best, and then ZY1-02C and SPOT5, thelast one is reconnaissance satellite. While SPOT5satellite images was better at theenergy character, the inferior is ZY1-02C, while the third is ZY3, and the last isreconnaissance satellite. To sum up, ZY3images are almost good at many aspects,ZY1-02C also has the same quality as SPOT5and reconnaissance satellite.3. Spectral quality: First of all, ZY1-02C, ZY3and SPOT5were able to makeuse of respective spectrum to distinguish different feature types through thesupervised classification and three dimension spectrum feature space. A conclusioncould be reached when there were all kinds of feature types in the workspace thatZY1-02C might be suitable to distinguish forests and constructions, however ZY3was able to distinguish the water, rice and constructions, and yet SPOT5had a greatadvantage of distinguishing the water and forests. In the second place, when there wasonly a single feature to be extracted, the spectrum character curve space and the twodimensional spectrum feature space might be applied to assess the spectral quality ofthe three satellites. ZY1-02C preferred to extract the forests and constructions, ZY3was better for water and corn, and SPOT5had a good effect on extracting water,forests and constructions. In conclusion, the spectral quality of domestic satellitesequaled that of SPOT5. Each satellite has its advantages of embodying the typicalfeature.4. Thematic application assessment: As a data source of the land resourcesinvestigation and monitoring, traffic information investigation and monitoring, riverboundaries investigation and monitoring and surface mining environmentalinformation investigation and monitoring, ZY1-02C and ZY3could be made full useof in project application. It is conducive for the images’ spread in case of getting suchconclusions.