QUALITY EVALUATION AND APPLICATION POTENTIAL ANALYSIS OF TIANGONG-2 WIDE-BAND IMAGING SPECTROMETER

Tiangong-2 is the first space laboratory in China, which launched in September 15, 2016. Wide-band Imaging Spectrometer is a medium resolution multispectral imager on Tiangong-2. In this paper, the authors introduced the indexes and parameters of Wideband Imaging Spectrometer, and made an objective evaluation about the data quality of Wide-band Imaging Spectrometer in radiation quality, image sharpness and information content, and compared the data quality evaluation results with that of Landsat-8. Although the data quality of Wide-band Imager Spectrometer has a certain disparity with Landsat-8 OLI data in terms of signal to noise ratio, clarity and entropy. Compared with OLI, Wide-band Imager Spectrometer has more bands, narrower bandwidth and wider swath, which make it a useful remote sensing data source in classification and identification of large and medium scale ground objects. In the future, Wide-band Imaging Spectrometer data will be widely applied in land cover classification, ecological environment assessment, marine and coastal zone monitoring, crop identification and classification, and other related areas.


INTRODUCTION
On orbit data quality evaluation of remote sensing payload is a vital part in data processing and data quality control of satellite ground system (Chen et al., 2004).Many operational satellites carry out data quality analysis and evaluation regularly, to grasp the change of data quality, and improve and control the quality of data products through orbit calibration (Brian et al., 2015, Ron, 2015, Hong, 2010, Zhang et al., 2010, Zhang, 2009, Wang and Tian, 2007).
Tiangong-2 is the first true space laboratory of China, which developed and launched on September 15, 2016 by China Manned Space Project, to carry out space science and application experiment (Qin et al, 2017).Wide-band Imaging Spectrometer (MWI) is a mid-resolution multispectral imager on Tiangong-2 (Qin et al., 2017).It will mainly used for remote sensing monitoring and application in ocean, land and atmosphere.
At present, MWI has acquired abundant image data of earth observation.In order to verify its imaging quality, the quality of the image data was objectively evaluated by image quality evaluation indexs, and the evaluation results were compared with the OLI data of Landsat-8.In addition, the application potential and application direction of MWI data were also discussed.

DATA
MWI of Tiangong-2 uses push-broom imaging model and with an observation swath of 300km, its spectrum range cover visible near-infrared (VNI), short-wave infrared (SWI) and thermal infrared (INF).The camera parameters are shown in Table 1 (Qin et al, 2017 1) and Qinghai lake (Figure 2).The radiation of them are both relatively uniform.Dunhuang Desert is a well-known radiation calibration site in China.Qinghai Lake is the largest inland lake in China.

METHOD
In order to comprehensively evaluated the data quality of Tiangong-2 MWI, three types of evaluation elements are adopted in this paper.They are radiation quality, image sharpness and information content.The radiation quality is evaluated with image signal-to-noise ratio (SNR), image contrast and radiation uniform.Image sharpness is evaluated with image clarity index, and information content is evaluated with information entropy.

Image signal-to-noise ratio
Image SNR is the most commonly used index to evaluate the noise in images (Gao, 2008, Corner et al., 2003).In this paper, the image noise estimate method based on region segmentation and residuals statistical is used to calculate the SNR of the test images (Qin et al., 2014, Gao et al., 2007, Gao, 1993).This method is less affected by different feature types and mixed pixel, can get more accurate assessment result (Qin et al., 2014).The calculation equation of image SNR is as follows: where SNR = image signal-to-noise ratio M = image signal mean σ = image noise standard deviation

Image contrast
Image contrast is often used to reflect the gray contrast and the detail identification degree in images.In this paper, image contrast is measured based on the Gray Level Co-occurrence Matrix (GLCM) (Haralick et al., 1973, Guo andSong, 2005) of image, which can get evaluation results with consistency and comparability, especially for the images with different response range.The calculation equation of image contrast is as follows (Bu, 2012): where CON = image contrast i, j = image gray value after normalized L = the gray levels of normalized image p(i, j) = the GLCM element of the image

Radiation uniform
Radiation uniform is mainly used for evaluating the relative radiometric correction accuracy of images (Krause, 2004).
There are several methods for radiation uniform evaluation, in this paper, the average line standard deviation method is adopted (Hu and zhang, 2008).The calculation equation of image radiation uniform is as follows: % 100

Image clarity
Image clarity is a commonly used quality evaluation index to measure the fuzzy distortion degree of images (Wang et al., 2004).In this paper, a clarity evaluation method based on edge gradient is adopted, which can reflect the sharpness degree of feature edge in the image, and has a high correlation with the modulation transfer function (MTF) of images (Qin et al., 2015, Zhang and Zhang 2002, Schowengerdt et al., 1996).The calculation equation of image clarity is as follows: where E(i)= sharpness value at edge point i i = number of edge point k = the subscript of the gradient vector L = length of gradient vector where CLA= image clarity i = number of edge point N = the total numbers of edge points E(i)= sharpness value at edge point i

Information entropy
Information entropy can reflect the richness of image information and usually be used as a reference factor during image quality evaluation.In this paper, Shannon entropy is selected to evaluate the information entropy of the image (Zhou and Tian, 2008).The calculation equation of image information entropy is as follows:  As shown in table 3 and table 4, the evaluation results of the two images are different in some degree.Radiation uniform and contrast have strong correlation with the feature types in images, the other indexes, SNR, clarity and entropy, are less affected by the feature types, the evaluation results of which can reflect the overall quality of image data.
Except the index of radiation uniform, the evaluation results of other indexes have good consistency during different bands of the same image, which indicates that the image quality of each bands of MWI are relatively balanced.
The average SNR of the two images are both more than 40dB, and the average clarity of the two images is no less than 70, data products quality can meet the application requirements.

Comparison of data quality and application potential analysis
To further illustrate the data quality of the MWI, the quality evaluation results of MWI were compared with those of OLI in the similar bands, the results were as follow: The spatial resolution of OLI data is higher than that of MWI, thus it can present more details of ground features.While MWI has more bands, narrower bandwidth and wider swath, it can provide more spectral information of the ground objects.In the classification and recognition application of large or medium scale ground objects, it will be an effective method to combined use these two data sources.
At present, the MWI data product has been applied in many directions, such as land cover classification, coastal monitoring, lake monitoring, ecological environment evaluation, crop classification and so on, and have shown its great application potential (Liu, et al., 2017, Technology and Engineering Center for Space Utilization, Chinese Academy of Sciences, 2018).

CONCLUSIONS
The radiation quality, image sharpness and information content of Tiangong-2 MWI data are well and can meet the application requirements, although the data quality of MWI still has a gap compared with that of OLI in the overall.
There are several similar bands between MWI and OLI, researchers can combined use these two data sources in remote sensing applications.
Since the band number of MWI is more than that of OLI, it has more advantages in classification and identification of large and medium scale ground objects.
Tiangong-2 MWI data has broad application prospects, and will be widely used in land cover classification, ecological environment assessment, marine and coastal zone monitoring, crop identification and classification, and other related fields.

Figure 3 .
Figure 3. Landsat-8 OLI image of Dunhuang Desert (Colour synthesis with band 4, band 3 and band 2) mean of image gray in column j i, j = row number and column number in image m = total row number of image j i DN , = the gray value of pixel (i, j) radiation uniform j = column number in image n = total column number of image j DN = the mean of image gray in column j DN = the mean of image gray of image i = gray value of image pixel min, max= the minimum and maximum of gray value in an image p i = pixel probability in an image which gray value equal i

Table 1
).In this paper, two sets of remote sensing images of MWI are selected as test images.Both of them are after radiation corrected and geometrically corrected.The ground cover types of the two image are respectively Dunhuang Desert (Figure . Part of the parameters of MWI Landsat-8 is a famous land observation satellite launched by National Aeronautics and Space Administration of America on February 11, 2013 (U.S. Geological Survey, 2015).There are two main loads on the Landsat-8, they are Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS).MWI and OLI are both multi-spectral imagers, the spatial resolution of OLI is 30m, which is higher than that of MWI (100m for VNI, 200m for SWI, 400m for INF).MWI has more effective bands and narrower spectral bandwidth.The details

Table 4 .
Quality evaluation results of MWI data in Qinghai LakeIn table 3 and table 4, the Mean is the average value of all the bands, and the Std is the standard deviation of all the bands.

Table 6 .
Data quality evaluation results comparison between MWI and OLI in image of Qinghai LakeIn table 5 and table 6, the Mean and Std were the average value and the standard deviation of the similar bands.For MWI, the similar bands are band2, band3, band5, band7, band11 and band16.For OLI, the similar bands are form band1 to band6.As shown in table5 and table 6, the image contrast of MWI data is larger than that of OLI in both Qinghai Lake and Dunhuang Desert.In other indexes, such as SNR, clarity, entropy and radiation uniform, the data quality of MWI still has a gap compared with that of OLI.