Histogram comparison crosscalibration method applied for evaluating the quality of “Canopus-v-ir” data

Regular onboard calibration of satellite data is of great practical importance for the processing of remote sensing information. For flight cross-calibration it is proposed to use a geostationary satellite as a reference to calibrate polar orbiting satellites data on the example of the “Canopus–V–IR” and Himawary satellites. Authors use data from center for collective use “IKI–Monitoring”, united data catalogue of Planeta Research Centers for Space Hydrometeorology. The work has been done on theme «Monitoring»

Modern remote sensing devices are the measuring instruments that provide data used in in various tasks of global changes monitoring in the earth's surface and determining dangerous phenomena. The accuracy and stability of the calibration characteristics of satellite instruments are important for the development and sustainable automatic application of processing algorithms. During the operation of spacecraft, it is necessary to conduct regular flight calibration of instruments, for which various techniques are used [1][2][3][4].
The method proposed for use is based on the comparison of histograms of images from the same Earth's surface scene by instruments from different spacecraft having similar spectral characteristics and made with the same view geometry and at the same time [5]. One of the devices will be considered a reference, with good calibration. We will try to calibrate other devices according to that one.
It seems natural to assume that the sensitivity of the device has no inversion and that the magnitude of the signal recorded by the device is proportional to the value of the radiation flux received from the Earth, i.e. a brighter object on earth corresponds to a stronger signal recorded by the device. In this case, all objects with the same brightness should have the same area in the images from different devices, which in terms of the image means the same histogram value in the area corresponding to the brightness of the objects.
Thus, constructing the correspondence of the partial areas of the histograms, we obtain an unambiguous correspondence of the brightnesses of one image to the brightnesses of another, or in other words, the correspondence between the recorded values of one device and the registered values of another. With a good calibration of the reference device, we get a calibration table for the recorded values of another device. One of the difficulties of such a scheme is the condition for simultaneous shooting with the same geometry. Nevertheless, when analyzing images of orbital and geostationary spacecraft, one can find approximate fulfillment of both the conditions of the geometry of the view and the conditions of simultaneity with an accuracy of less than 5 minutes, and more accurately with planning the session.
The study was conducted using data access and processing services "VEGA-Science" [6]. For the Canopus-V-IR data, scenes near to Himawari subsatillite point were provided by "VNIIEM Corporation" JC for April 12-15, 2019. Himawari data processed at the Far Eastern center SRC "Planeta" [7] were used as a reference.
The geometry of the both views of the scene from the geostationary and polar-orbiting satellite is shown in Fig. 1. On the images taken from two satellite the areas are selected to satisfy the geometry with difference of up to 5 degrees of satellites zenith angles. Images by Canopus-V-IR are averaged to the images resolution of the AHI HIMAWARI satellite (these data are used as reference data). Further, according to the histograms of images using the method described above the calibration tables are calculated for the data "Canopus-_V-IK", MSU-IK-SRM device.
The spectral characteristics of the devices in the bands of 8.6 μm and 3.7 μm are shown in Fig. 2  An example of a pair of images used for cross-calibration with a selected fragment for histogram analysis is shown in Fig. 3. The corresponding histograms and data conversion graph (calibration table) are shown in Fig. 4. A number of factors affecting the proposed procedure should be noted. The Canopus-V-IR data has obvious significant noise in the image (Fig. 5). This fact leads to errors in the construction of calibration tables. In particular, the apparent periodic strips in the warm zone (dark and irregular shades of the strip in the images Fig.56 a, b)   The error (standard deviation) is from 1 (warm region) to 2.5 degrees K (cold region). An almost linear course of the calibration curve should be noted, and only in the temperature range of 220 K and lower does the sensitivity drop, which appears in the histogram as a small upward surge in its values.
Sufficiently large deviations from the average are not associated with a possible sharp change in weather conditions in the considered regions within 5 minutes. For all data sets, the same conversion tables can be constructed for successive observations from the Himawari satellite. An example of the table graph for this case is shown in Fig.7. It can be seen that it differs from the trivial linear transformation of the image itself, but not as much as the table graphs for the Canopus-V-IR-Himawari image pairs. Thus, the error in the construction of refined calibration characteristics can be explained by the noise of the data. At the same time, the relatively close course of the calibration curves constructed using the above technique allows one to conclude that the Canopus-V- IR images after eliminating noisy data could be calibrated with using this cross-calibration technique.
The existing image set also contains data in a 3.7 μm spectral band. But they are so noisy (Fig.8) that, due to the above considerations, this technique is not applicable for performing cross-calibration. Authors use data from center for collective use "IKI-Monitoring", united data catalogue of Planeta Research Centers for Space Hydrometeorology. The work has been done on theme «Monitoring»