CN112150555A - In-orbit relative radiation calibration method for geosynchronous orbit area-array camera - Google Patents

Abstract

The invention relates to an on-orbit relative radiation calibration method for a geosynchronous orbit area-array camera, and S1, the on-orbit relative radiation calibration of the geosynchronous orbit area-array camera takes each half year as a period; s2, dividing the image into 6 months evenly in a period, selecting images in the order from low cloud cover to high cloud cover every month, and taking the images as statistical samples to participate in the calculation of the scaling coefficient; s3, counting the gray distribution condition of each pixel, and eliminating the influence of a 0 value and a saturated gray value; s4, taking the overall gray distribution of all pixels as the expected value of the gray distribution of each pixel, and drawing an expected gray distribution curve; s5, confirming the actual ranges of the high value area, the middle value area and the low value area; and S6, fitting the gray distribution of each pixel with expected gray distribution data according to the sequence of the high value area, the medium value area and the low value area to obtain the relative radiometric calibration coefficient of the high value area, the medium value area and the low value area of each pixel, wherein the relative radiometric calibration coefficient is used for the relative radiometric correction of the on-orbit data.

Description

An on-orbit relative radiometric calibration method for geosynchronous orbital area array cameras

technical field

The invention belongs to the field of performance testing and evaluation of remote sensing satellites, and relates to an on-orbit relative radiation calibration method for a Gaofen-4 satellite area array camera.

Background technique

The relative radiometric calibration is to use the high-precision radiometric calibration datum to calibrate the error of the imaging system, and to determine the response relationship between each detector element and the detector elements. Therefore, the accuracy of the radiometric calibration datum directly affects the relative radiometric calibration accuracy. The existing remote sensing relative radiometric calibration methods include: laboratory calibration method using integrating spheres before satellite launch. Since the radiation characteristics of satellites will change after launch, this method is only used in the initial stage of orbit. Infrared remote sensing is generally equipped with an on-board calibration device. After launch, on-board calibration can be carried out based on on-board calibration lights or diffuse reflectors. If there is no calibration device on the satellite, on-orbit calibration based on the ground uniform field can be used. , 90° yaw calibration or use satellite on-orbit statistical calibration to achieve relative radiometric calibration. The above methods are applicable to linear array imaging satellites in sun-synchronous orbit.

If it involves geosynchronous orbital area array imaging satellites, the on-orbit relative radiometric calibration will be different. For example, the Gaofen-4 satellite launched at the end of 2015 was the geosynchronous orbit satellite with the highest resolution in the world at that time, with a resolution of 50m. The laboratory calibration coefficient was adopted in the early stage of orbit, but it has not been updated for more than 4 years due to the lack of an effective alternative method for relative radiation calibration in orbit. Since the on-orbit state of the remote sensor will gradually deviate from the laboratory state before launch, and the longer the time, the greater the difference may be. Therefore, it is more and more urgent to study a reasonable on-orbit relative radiation calibration coefficient to replace the laboratory calibration coefficient.

According to the experience of sun-synchronous orbit satellites, the calibration method based on on-orbit data statistics requires each pixel to have 100,000 or even more than one million sampling records, so as to obtain a more comprehensive representation of ground objects. It is relatively easy for linear array imaging of sun-synchronous orbit to meet this requirement. Taking the Panchromatic Channel of Gaofen-2 as an example, each scene image has 27,000 lines, and 4 scene images can meet the requirement of 100,000 sampling times. However, geosynchronous orbit satellites such as Gaofen-4 cannot push-broom, and all pixels are imaged at the same time each time, so the number of samplings is equal to the number of scenes in the image.

SUMMARY OF THE INVENTION

The technical problem solved by the invention is as follows: it solves the problem that geosynchronous orbit satellites such as Gaofen-4 lack the means for generating relative radiation calibration coefficients in orbit and rely on pre-launch laboratory calibration data for a long time.

The technical solution of the present invention is: a method for on-orbit relative radiation calibration of a geosynchronous orbital area array camera, which is realized in the following manner:

S1. According to the variation law of the radiation intensity reflected by the surface within a year, the on-orbit relative radiation calibration of the geosynchronous orbital area array camera is divided into two statistical periods with every half year as a cycle;

S2. A cycle is divided into 6 months on average, and images are selected in the order of cloud cover from low to high every month, and they are used as statistical samples to participate in the calculation of the calibration coefficient. In the case of multiple scenes in the staring imaging sequence, only one scene is counted; the number of selected images Determined according to the preset number of grayscale samples for each pixel;

S3. Perform grayscale statistics on the on-orbit image samples of the geosynchronous orbital area scan camera pixel by pixel, calculate the grayscale distribution of each pixel, and eliminate the influence of 0 value and saturated grayscale value;

S4. Use the overall grayscale distribution of all pixels as the expected value of the grayscale distribution of each pixel, and draw the expected grayscale distribution curve;

S5. According to the quantized bits of the image, three gray levels are designated as the initial values of the high-value area, the median area, and the center of the low-value area for clustering. After clustering, confirm the high-value area, the median area, and the low-value area. the actual range of the value area;

S6. For each pixel, fit its grayscale distribution with the expected grayscale distribution data in the order of high value area, median value area and low value area, and obtain the high value area, median value area and The relative radiometric scaling factor of the low value region, which is used for the relative radiometric correction of on-orbit data.

Preferably, the on-orbit image samples participating in the statistics in S1 are divided into two half-year statistical periods with the winter solstice day and the summer solstice day as the boundary.

Preferably, the frequency of on-orbit relative radiometric calibration should satisfy twice a year.

Preferably, the preset number of grayscale samples is not less than 60,000 times.

Preferably, distributed parallel processing is used for the image data participating in the single statistics.

Preferably, the designated high, medium and low grayscales correspond to 10%-15%, 50%, and 85-90% of the saturation value, respectively.

Preferably, the least squares model fitting is preferred for the median area calibration coefficient.

Preferably, after the S6 fitting, smoothing is performed on the grayscale data of the pixel in the transition area between the high-value area, the middle-value area, and the low-value area.

Preferably, the transition area is 5 gray levels adjacent to each other up and down with the gray level boundary between the high-value area and the median area or the gray-level area between the low-value area and the median area as the center.

The beneficial effects of the present invention compared with the prior art are:

The invention can periodically generate relative radiation calibration coefficients by using on-orbit data, and update it once every six months.

Description of drawings

Fig. 1 is the flow chart of the present invention;

Fig. 2 is a single pixel grayscale distribution probability curve of an example of the present invention;

Fig. 3 is the overall distribution probability curve of the example of the present invention;

Fig. 4 is the low value area fitting result of the embodiment of the present invention;

Fig. 5 is the fitting result of the median area of the embodiment of the present invention;

FIG. 6 is a fitting result of a high-value region according to an embodiment of the present invention.

Detailed ways

The present invention will be further elaborated below in conjunction with the examples.

A method for on-orbit relative radiation calibration of a geosynchronous orbital area array camera is shown in Figure 1. The steps are as follows:

(1) Determine the statistical period

Taking the Gaofen-4 satellite as an example, the latest calibration cycle is from the winter solstice in 2019 to the summer solstice in 2020. Since the time difference is only about one week, for the sake of convenience, the data from January to June 2020 is used here to illustrate.

(2) Determine the statistical sample

In the first half of 2020, Gaofen-4 satellite data distributed a total of 212,737 scenes to users, which means that each probe received 212,737 samples, including 47,306 samples from January to February, 31,937 samples in March, 51,740 samples in April, and 45,732 samples in May. times, distributed 36022 times in June. First, the case of temporarily adjusting the imaging gain is eliminated, and the samples using the same gain are retained. Secondly, in order to eliminate the statistical deviation caused by the scene, time phase and cloud as much as possible, only one scene is reserved for the multi-scene staring imaging sequence, and 10,000 scenes are selected in the order of cloud amount from low to high each month, as statistical samples to participate in the calibration Calculation of coefficients.

(3) Perform grayscale statistics pixel by pixel, and eliminate the influence of 0 value and saturated gray value of each pixel

Grayscale statistics are performed pixel by pixel. The full chromatographic segment of the Gaofen-4 satellite adopts 10-bit quantization, and the quantization value ranges from 0 to 1023. 0 means no signal, 1023 means the signal is saturated, and a large number of images are actually saturated at the gray value of 1022. The quantitative information at this time It is already inaccurate, and the order of magnitude of the pixels corresponding to these three grayscales jumps significantly, so the three grayscale statistical values of 0, 1022, and 1023 for each pixel are set to zero. The total number of single pixel samples after removing these three gray levels is the effective number of samples. Divide the number of gray pixels by the number of valid samples to calculate the percentage of each gray, and draw the gray distribution probability map of each pixel, taking the pixels in the first row and first column as an example, as shown in Figure 2.

(4) Statistical sample overall gray distribution

The same as above, obtain the percentage of each gray level of the whole 10240*10240 pixels on the area array, and draw the overall gray level distribution probability map as shown in Figure 3.

(5) Determine the segmentation threshold

Specify 870, 511, and 154, which are located at 85%, 50%, and 15% of the saturation gray level, respectively, as the initial values of the cluster center for cluster calculation. After clustering, three response intervals of high, medium, and low are formed. Value area response interval [904, 1023], median area response interval [213, 903], low value area response interval [0, 212].

(6) Generation of segmented radiation calibration coefficients

In each response interval of high, medium and low, according to the gray distribution data of each pixel and the whole, a piecewise linear model is used to fit the radiation characteristics of the pixel, and the relative radiation calibration coefficient of each corresponding interval is generated. The results are shown in Figures 4, 5, and 6. The fitting result in the low-value area is y=0.9971x, R ² =0.996; the fitting result in the median area is y=1.0028x, R ² =0.9857; the fitting result in the high-value area is y =1.0023x, R ² =0.9731. It should be noted that in the actual business statistics, because the brightness interval of the median area has the highest linearity of the pixel response, the least squares fitting process should be performed to obtain the radiation correction coefficient. The relative radiometric scaling factor can be used for relative radiometric correction of on-orbit data.

(7) In this example, due to the high data consistency, the grayscale of the two transition areas is still monotonically changed after the calibration coefficient is used, and no smoothing is required.

The parts not described in detail in the present invention belong to the common knowledge of those skilled in the art.

Claims (9)

1. An on-orbit relative radiation calibration method for a geosynchronous orbit area-array camera is characterized by being realized in the following way:

s1, dividing the on-orbit relative radiation calibration of the geosynchronous orbit area-array camera into two statistical time intervals by taking each half year as a cycle according to the change rule of the radiation intensity reflected by the earth surface within one year;

s2, dividing the image into 6 months evenly in a period, selecting images in the order from low cloud cover to high cloud cover every month, using the images as statistical samples to participate in the calculation of a scaling coefficient, and only counting 1 scene when a gazing imaging sequence has multiple scenes; the number of the selected images is determined according to the number of gray level samples preset by each pixel;

s3, carrying out gray level statistics on the on-orbit image samples of the geosynchronous orbit area-array camera pixel by pixel, counting the gray level distribution condition of each pixel, and eliminating the influence of a 0 value and a saturated gray level value;

s4, taking the overall gray distribution of all pixels as the expected value of the gray distribution of each pixel, and drawing an expected gray distribution curve;

s5, according to the quantization digit of the image, three gray scales are designated as initial values of the centers of a high-value area, a medium-value area and a low-value area respectively for clustering, and after clustering, the actual ranges of the high-value area, the medium-value area and the low-value area are confirmed;

and S6, fitting the gray distribution of each pixel with expected gray distribution data according to the sequence of the high value area, the medium value area and the low value area to obtain the relative radiometric calibration coefficient of the high value area, the medium value area and the low value area of each pixel, wherein the relative radiometric calibration coefficient is used for the relative radiometric correction of the on-orbit data.

2. The method of claim 1, wherein: the on-orbit image samples involved in statistics in S1 are divided into two half-year-long statistical time periods with the winter solstice and summer solstice of each year as the boundary.

3. The method of claim 1, wherein: the on-orbit relative radiation calibration frequency is satisfied twice a year.

4. The method of claim 1, wherein: the number of the preset gray level samples is not less than 6 ten thousand.

5. The method of claim 1, wherein: and the image data participating in single statistics adopts distributed parallel processing.

6. The method of claim 1, wherein: the designated high, middle and low gray levels respectively correspond to 10% -15%, 50% and 85-90% of the saturation value.

7. The method of claim 1, wherein: and the median region scaling coefficient is preferably fitted by a least square model.

8. The method of claim 1, wherein: after the fitting of S6, the pixel gray data in the transition region between the high value region, the median region, and the low value region is smoothed.

9. The method of claim 8, wherein: the transition area is 5 adjacent gray scales up and down with the gray scale boundary point of the high value area and the middle value area or the low value area and the middle value area as the center.

Images