Multi-scale radiation cascade remote sensing satellite in-orbit relative radiation calibration method and system

Abstract

The invention discloses a multi-scale radiation cascade remote sensing satellite in-orbit relative radiation calibration method and system. The method comprises the steps that S1, sequence image alignment is completed based on an SIFT algorithm or other image alignment methods; s2, completing earth surface truth value estimation based on a mean value estimation method or other probability estimation methods, and obtaining a relative radiometric calibration reference image; s3, extracting a gray value of each probe element of the area array on the sequence image; s4, constructing a radiation calibration model of the area array sensor; s5, completing radiometric calibration model calculation through a multi-scale method to obtain radiometric calibration coefficients of all scales; and S6, combining the radiation calibration coefficients of all scales. The in-orbit relative radiometric calibration work of the area array type remote sensing satellite is completed through the steps S1 to S6, the in-orbit relative radiometric calibration coefficient is obtained, and input is provided for radiometric correction of an original image. The method is suitable for in-orbit relative radiometric calibration of area array type remote sensing satellites, and has the advantages of being high in adaptability, high in frequency, high in precision, more convenient and low in cost.

Description

TECHNICAL FIELD

[0001] The present application relates to an on-orbit relative radiometric calibration method for optical remote sensing satellites, in particular to an on-orbit relative radiometric calibration method and system for a remote sensing satellite carrying a planar array type imaging sensor. BACKGROUND

[0002] As of March 2025, more than 1100 earth observation remote sensing satellites have been launched worldwide, of which optical remote sensing satellites account for more than 80%, and have become the mainstream of earth observation remote sensing data. Linear array sensors and planar array sensors are two main types of imaging sensors carried by optical remote sensing satellites. The former is composed of a linear array of imaging elements, which realizes push-broom imaging by flying forward, efficiently obtaining continuous strip images; multiple linear arrays with different spectral ranges can realize multi / hyperspectral imaging. The latter is composed of a planar array of imaging elements, which realizes push-broom or staring imaging by flying forward or being fixedly pointed, obtaining continuous frame sequence images, and can realize real-time monitoring of regional dynamic targets. With their respective imaging advantages, linear arrays and planar arrays have become the mainstream imaging sensors carried by current military, civilian and commercial remote sensing satellites.

[0003] On-orbit relative radiometric calibration is a key technology to ensure satellite radiometric quality and is an indispensable part of the ground processing system after satellite launch. Relative radiometric calibration requires the use of a known calibration reference that covers all imaging elements to calibrate the radiometric response model of each element. Compared to linear array sensors that only need to calibrate one-dimensional elements, calibrating two-dimensional elements of planar array sensors is more difficult in terms of calibration reference acquisition. At the same time, existing statistical calibration and yaw calibration methods are designed for linear array sensors and cannot be applied to planar array remote sensing sensors. The uniform field calibration method based on large-area uniform natural terrain on the Earth's surface is suitable for planar array sensors and is currently the most commonly used method for on-orbit calibration of planar array remote sensing sensors. However, the uniform field used for calibration needs to meet the following requirements: it needs to have a large enough area to completely cover all imaging elements of the remote sensing sensor; it needs to have good uniformity to meet the accuracy requirements of calibration; and it needs to have multiple different reflectivities to cover the dynamic range of the sensor imaging. Therefore, this method still has many limitations for on-orbit application and it is difficult to achieve high-frequency and high-precision on-orbit relative radiometric calibration. In summary, it is very important to study on-orbit relative radiometric calibration technology for planar array remote sensing satellites to improve on-orbit relative radiometric calibration accuracy and frequency, and to ensure the radiometric quality of planar array satellite images and improve the application efficiency of satellites. SUMMARY

[0004] The present application is directed to a multi-scale radiation cascade on-orbit relative radiometric calibration method for area array remote sensing satellite sensors. Firstly, the sequence images are aligned to obtain the motion trajectory of the area array detector elements on the sequence images; secondly, the relative radiometric calibration reference is calculated based on the aligned sequence images; and finally, a multi-scale radiation cascade radiometric calibration model is constructed to realize the on-orbit relative radiometric calibration of the area array remote sensing satellite.

[0005] The present application aims to provide an on-orbit high-precision relative radiometric calibration method suitable for remote sensing satellite area array sensors, which specifically comprises the following steps: Step 1: Selecting calibration sequence images and aligning the sequence images; Step 2: Obtaining the relative radiometric calibration reference image through estimation method based on the aligned sequence images; Step 3: Obtaining the motion trajectory of the area array detector elements on the time sequence images to obtain the gray value of each detector element on each sequence image; Step 4: Constructing the radiometric calibration model of each detector element of the area array sensor based on the gray value of the area array detector elements and the relative radiometric calibration reference; Step 5: Solving the radiometric calibration model in step 4 through multi-scale method to obtain the relative radiometric calibration coefficients of each detector element of the area array sensor at different scales; Step 6: Merging the radiometric calibration coefficients at each scale obtained in step 5 through multi-scale radiation cascade method to complete the on-orbit relative radiometric calibration of the area array remote sensing satellite.

[0006] Further, the specific implementation of step 1 is as follows: (11) Sequence image selection: selecting large-area uniform field images or non-uniform field images; (12) Sequence image alignment method: selecting image matching method, including SIFT matching algorithm and other image alignment methods, for uniform field sequence images, at least realizing sequence image alignment accuracy better than 1 pixel, and for non-uniform field sequence images, at least realizing sequence image alignment accuracy better than 0.5 pixel.

[0007] Further, the specific implementation of sequence image alignment is as follows: 1) Selecting the image with the largest overlapping area with other sequence images in the sequence image data as the reference image, and using scale-invariant feature transformation algorithm to block match other images with the reference image to obtain the corresponding point pairs between each sequence image and the reference image , initializing the image inter-plane affine transformation model relationship based on the matched sequence image corresponding point pairs, as follows:

[0008]

[0009] In the formula to are the coefficients of the affine transformation model; the affine model relationship between the i-th scene in the sequence images and the reference image is defined as , then there is:

[0010] Using the above affine transformation model to calculate the residuals of the matching homologous point pairs between the sequence images , there is the following formula: <�

[0011]

[0012]

[0013]

[0014] where and represent the root mean square error, which measures the average geometric deviation of all homologous point pairs in the x and y directions during the image alignment process. N refers to the total number of point pairs used for calculation, and respectively represent the x coordinates of the -th homologous point on the reference image and the sequence image, and respectively represent the y coordinates of the -th homologous point on the reference image and the sequence image; and are the standard deviations in the x and y directions respectively, indicating the registration accuracy of the statistical homologous points; Taking 3 as the threshold to eliminate gross errors and obtain new homologous point pairs between the sequence images , re-initializing the affine transformation model between the sequence images, then there is:

[0015] 2) Traverse all the sequence images, and the one-to-one correspondence relationship between the pixels of the image sequence and the reference image can be obtained , that is, the precise relative position relationship between the sequence images.

[0016] Furthermore, in step 2, the mean estimation method or other probability estimation methods are selected to make the obtained relative radiometric calibration reference image have no obvious image blurring and trailing problems. The specific implementation method of the mean estimation method is as follows: 2.1, Based on the aligned sequence image data, then there is: <00�\078>

[0017]

[0018] In the formula N For the number of images, M The number of detector elements in the array; Indicates the first i Image number n The detector number for each pixel image. The first image is the estimated true radiance of ground features. n Each pixel's grayscale value In the In the image, by the first The image captured by the probe is the n grayscale value of each pixel; 2.2 Finally, the true surface values ​​are estimated, resulting in a relative radiometric calibration reference image. .

[0019] Furthermore, in step 3, the method for extracting grayscale values ​​from the area array probes in each frame sequence is as follows: 3.1, Given that the coordinates of the area array element D on the reference image are... ; 3.2, Coordinates of Detector D in a certain frame of image This can be obtained from step 1:

[0020] The term refers to the functional relationship, specifically the function representing the precise relative positional relationship of time-series images. 3.3, Based on the coordinates of probe D in a certain frame of image This allows you to extract grayscale values. .

[0021] Furthermore, in step 4, the radiation calibration model for each element of the array sensor is as follows:

[0022] In the formula , For the first array sensor i The relative radiometric calibration coefficients of each detector element j Image sequence number N is the number of images in the sequence, where the coefficient is... When the value is 0, only the gain coefficient is calculated. .

[0023] Furthermore, step 5 includes the following sub-steps: 5.1, Determine the number of scale layers as follows: s is the s-th layer, and ScaleNumber is the maximum scale; 5.2 The aligned sequence image obtained in step 1 and the relative radiometric calibration reference image obtained in step 2 are downsampled to obtain... Sequences of images at different scales; 5.3 Combine the sequence of images at the first-level scale with the relative radiometric calibration reference image, and use step 4 to obtain the radiometric calibration model of the first level; 5.4. Solve the radiometric calibration model in 5.3 using the least squares method to obtain the radiometric calibration coefficients of the first layer. ; 5.5 Similarly, calculate the relative radiometric calibration coefficients of the s-th layer of the image sequence. .

[0024] 8. The on-orbit relative radiometric calibration method for multi-scale radiometric cascade remote sensing satellites as described in claim 7, characterized in that: in step 5.5, the relative radiometric calibration coefficients of the s-th layer of the image sequence... The calculation method is as follows: 5.5.1, using the relative radiometric calibration coefficients of the (s-1)th layer, the radiometric correction of the s-th layer scale sequence image is completed based on the following formula.

[0025] In the formula For the array sensor at the s-1 level scale, the first... i The relative radiometric calibration coefficient of each detector element.

[0026] 5.5.2 Combine the sequence of images at the s-th layer scale with the relative radiometric calibration reference image, and use step 4 to establish the radiometric calibration model for the s-th layer; 5.5.3, using the least squares method to solve the radiation calibration model in 5.5.2, obtain the radiation calibration coefficients of the s-th layer. .

[0027] Furthermore, step 6 includes the following sub-steps: 6.1 For the i-th detector element of the area array sensor, the combined formula for the relative radiation calibration coefficients of each layer is as follows:

[0028]

[0029] In the formula Let be the relative radiation calibration coefficient of the s-th layer. and is the final relative radiometric calibration coefficient of the i-th detector element of the area array sensor; 6.2 Repeat step 6.1 for all imaging elements of the array sensor to obtain the final relative radiometric calibration coefficients of all elements of the array sensor, thus completing the on-orbit relative radiometric calibration of the array sensor.

[0030] This invention also provides a multi-scale radiometric cascade remote sensing satellite on-orbit relative radiometric calibration system, comprising: The system includes a processor and a memory. The memory stores program instructions, and the processor calls the stored instructions in the memory to execute the on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in the above technical solution.

[0031] Compared with the prior art, the present invention has the following characteristics and beneficial effects: (1) On-orbit calibration can select uniform field sequence images or non-uniform scene images, which reduces the constraints of calibration on image data.

[0032] (2) The satellite does not need to have on-board calibration processing capabilities.

[0033] (3) Calibration can be completed using any frame sequence, which greatly increases the frequency of on-orbit calibration of array-type remote sensing satellites.

[0034] (4) Improved the on-orbit calibration accuracy of array-type remote sensing satellites. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the specific implementation steps in an embodiment of the present invention; Figure 2 This is a schematic diagram of a partial radiation calibration reference extracted in an embodiment of the present invention; Figure 3 This is a schematic diagram of the radiation calibration model of each detector element of the multi-scale array sensor in an embodiment of the present invention; Figure 4 This is a statistical chart showing the relative radiation correction accuracy achievable by the present invention; Figure 5 The figures shown are before and after relative radiation correction in the embodiments of the present invention, where (a) and (c) are before relative radiation correction, and (b) and (d) are after relative radiation correction. Detailed Implementation

[0036] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0037] like Figure 1 As shown in the figure, the on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade provided by the present invention specifically includes the following steps: S1, Select a sequence of images for calibration, and align the sequence of images using a certain method, as follows: 1.1 Selection of image sequence: The image sequence can be a large-area uniform field image or a non-uniform field image.

[0038] 1.2 When selecting a large-area uniform field image, the following principles should be followed: 1) It has a large enough area to completely cover all the imaging elements of the remote sensing sensor. A large enough area means that the image area can cover the sensor swath width. 2) It has good uniformity to meet the calibration accuracy requirements. Good uniformity means that the uniformity is better than the calibration accuracy. 3) Multiple different reflectivity values ​​are available to broadly cover the sensor's imaging dynamic range.

[0039] 1.3 When selecting a non-uniform field image, please note the following information: 1) Distribution of image sequence: The selected image sequence needs to be evenly distributed along the track and perpendicular to the track to avoid the low-frequency radiation error in a certain direction not being effectively calibrated; 2) Requirements for land cover types in image sequence: The selection of image sequence should avoid dynamic targets such as clouds and waves, as well as areas with grayscale steps or pixel saturation such as buildings and roofs, to prevent the phenomenon of land cover trailing relative to the calibration benchmark. 3) Bidirectional reflectance characteristics of ground objects: When remote sensing satellites conduct multi-frame time-series observations, there are differences in the observation angles of the ground area. Due to factors such as the bidirectional reflectance characteristics of ground objects and the time difference in observation, the brightness of the same ground object may vary in different frame sequences. If the difference in radiation caused by the bidirectional reflectance of ground objects is within ±3° of the zenith angle of the observed ground object within a short period of time, it can be considered that the difference in radiation caused by the bidirectional reflectance of ground objects is negligible.

[0040] The reflectivity of ground features is related to the observation angle and is called the bidirectional reflectivity (BRDF). In a sequence of imagery, the satellite's observation angle of the same ground feature changes slightly over time, causing changes in the brightness of the feature not caused by the sensor. To accurately perform relative radiometric calibration, it is necessary to separate the differences in the sensor's own response from the brightness changes caused by the BRDF effect. When the change in the observed zenith angle is within a small range of ±3°, the brightness changes caused by the BRDF effect can be considered negligible. This allows the calibration algorithm to attribute the brightness differences between image sequences primarily to the inconsistency in the response of sensor elements, thereby simplifying the model and improving calibration accuracy.

[0041] 1.4 The image alignment method can be an image matching method, such as the SIFT matching algorithm, or other image alignment methods, to obtain the image coordinates of the sequence. Coordinates of relative radiometric calibration reference image One-to-one correspondence, Where n and m are the image width and height, the goal is to achieve an alignment accuracy of at least 1 pixel (uniform field image sequence) and 0.5 pixels (non-uniform field image sequence). Taking the SIFT matching algorithm as an example, the specific steps for image sequence alignment are as follows: 1) Select the image with the largest overlap with other image sequences in the image sequence data as the reference image. Use the Scale-invariant feature transform (SIFT) algorithm to perform block matching between other images and the reference image to obtain corresponding point pairs between each image sequence and the reference image. The image-side affine transformation model relationship between images is initialized based on the matched pairs of corresponding points between sequential images, as shown in the following equation:

[0042]

[0043] In the formula arrive These are the coefficients of the affine transformation model. The affine model relationship between the i-th image in the image sequence and the reference image is defined as... Then we have:

[0044] The above affine transformation model is used to calculate the residuals of matching corresponding point pairs between sequential images. We have the following formula:

[0045]

[0046]

[0047]

[0048] in and These represent the root mean square error, which measures the average geometric deviation of all corresponding point pairs in the x and y directions during image alignment. N refers to the total number of point pairs used in the calculation. and They represent the first The x-coordinates of the corresponding points on the reference image and the image sequence. and They represent the first The y-coordinates of the corresponding points on the reference image and the image sequence; and These are the standard deviations in the x and y directions, respectively, which indicate the registration accuracy of statistically corresponding points; With 3 To remove outliers using a threshold, new pairs of corresponding points between image sequences are obtained. If we reinitialize the affine transformation model between the image sequences, then we have:

[0049] 2) By traversing all image sequences, the one-to-one correspondence between each pixel of the image sequence and the reference image can be obtained. This refers to the precise relative positional relationship between images in a sequence.

[0050] S2, based on aligned sequence images, can choose mean estimation or other probability estimation methods. At least the obtained relative radiometric calibration reference image has no obvious image blurring, trailing, or other problems, thus realizing the estimation of the true surface value. The following steps illustrate the process using mean estimation as an example: 2.1 Based on aligned sequence image data, we have:

[0051]

[0052] In the formula N For the number of images, M The number of detector elements in the array; Indicates the first i Image number n The detector number for each pixel image. The first image is the estimated true radiance of ground features. n Each pixel's grayscale value In the In the image, by the first The image captured by the probe is the n The grayscale value of each pixel.

[0053] 2.2 Finally, the true surface values ​​are estimated, resulting in a relative radiometric calibration reference image. .

[0054] S3, based on the motion trajectory of the array elements in the temporal image recovered in step 1, obtain the grayscale value of each element in each frame sequence. , M represents the number of area array detectors; taking a specific area array detector D as an example, the method for extracting grayscale values ​​of the area array detectors in each frame sequence is explained as follows: 3.1, Given that the coordinates of the area array element D on the reference image are... ; 3.2, Coordinates of Detector D in a certain frame of image This can be obtained from step 1:

[0055] The term refers to the functional relationship, specifically the function representing the precise relative positional relationship of time-series images. 3.3, Based on the coordinates of probe D in a certain frame of image This allows you to extract grayscale values. .

[0056] S4. Based on the grayscale values ​​of the array elements and the relative radiometric calibration reference image obtained in steps 2 and 3, construct the radiometric calibration model for each element of the array sensor, as shown in the following equation:

[0057] In the formula , For the first array sensor i The relative radiometric calibration coefficients of each detector element j Image sequence number N is the number of images in the sequence. The coefficients... When the value is 0, only the gain coefficient is calculated. .

[0058] S5. Solve the radiometric calibration model in step 4 using a multi-scale method to obtain the relative radiometric calibration coefficients of each detector element of the array sensor at different scales. The specific steps are as follows: 5.1, Determine the number of scale layers as follows: s is the s-th layer, and ScaleNumber is the maximum scale; 5.2 The aligned sequence image obtained in step 1 and the relative radiometric calibration reference image obtained in step 2 are downsampled to obtain... Sequences of images at different scales; 5.3 Combine the sequence of images at the first-level scale with the relative radiometric calibration reference image, and use step 4 to obtain the radiometric calibration model of the first level; 5.4. Solve the radiometric calibration model in 5.3 using the least squares method to obtain the radiometric calibration coefficients of the first layer. ; 5.5, Relative radiometric calibration coefficients of the s-th layer of the image sequence The calculation method is as follows: 5.5.1, using the relative radiometric calibration coefficients of the (s-1)th layer, the radiometric correction of the s-th layer scale sequence image is completed based on the following formula.

[0059] In the formula For the array sensor at the s-1 level scale, the first... i The relative radiometric calibration coefficient of each detector element.

[0060] 5.5.2 Combine the radiometrically corrected sequence images at the s-th layer scale with the relative radiometric calibration reference image, and use step 4 to establish the radiometric calibration model for the s-th layer; 5.5.3, using the least squares method to solve the radiation calibration model in 5.5.2, obtain the radiation calibration coefficients of the s-th layer. .

[0061] S6 utilizes a multi-scale radiometric cascade method to calculate the relative radiometric calibration coefficients of each element of an optical remote sensing satellite array sensor, thus completing the on-orbit relative radiometric calibration of array-type remote sensing satellites. The specific steps are as follows: 6.1 Taking the i-th detector element of the area array sensor as an example, the combined formula for the relative radiation calibration coefficients of each layer is as follows:

[0062]

[0063] In the formula Let be the relative radiation calibration coefficient of the s-th layer. and is the final relative radiometric calibration coefficient of the i-th detector element of the area array sensor; 6.2 Repeat step 6.1 for all imaging elements of the array sensor to obtain the final relative radiometric calibration coefficients of all elements of the array sensor, thus completing the on-orbit relative radiometric calibration of the array sensor.

[0064] In practice, the above process can be automated using computer software technology. The hardware that runs this method should also be within the scope of protection of this invention.

[0065] On the other hand, embodiments of the present invention also provide a multi-scale radiometric cascade remote sensing satellite on-orbit relative radiometric calibration system, comprising: The system includes a processor and a memory. The memory stores program instructions, and the processor calls the stored instructions in the memory to execute the on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in the above technical solution.

[0066] In a specific embodiment, over 100 images in sequence were used for relative radiometric calibration experiments. Image alignment was performed using a method based on SIFT and geometric information constraints (Geo), and the accuracy of matching corresponding points between images was measured. In the verification example, the average matching error of corresponding points in the image sequence did not exceed 0.1 pixels (1). The maximum matching error is no more than 1 pixel, as shown in Table 1.

[0067] Table 1. Matching accuracy of corresponding points between image sequences

[0068] After completing the time-series image surface true value estimation, a relative radiometric calibration reference image is obtained. ,like Figure 2 As shown.

[0069] A radiometric calibration model for each detector element of a multi-scale array sensor was constructed, as follows:Figure 3 As shown, the model constructs radiometric response equations for each detector element at different resolution scales (such as raw, 1 / 2, 1 / 4, etc.) to obtain the corresponding calibration coefficients. Finally, by cascading and merging these coefficients from multiple scales, the model can provide a comprehensive set of radiometric correction parameters, thereby achieving high-precision relative radiometric calibration across the entire sensor swath and ensuring good radiometric consistency of the image data.

[0070] The calibration coefficients were applied to the area array image sequence, and the ratio of the standard deviation of the mean to the mean along the orbital and perpendicular orbital matrices was used as an objective evaluation index to evaluate the images after relative radiometric correction. The objective evaluation results are as follows: Figure 4 As shown, Our method (blue curve), which is the method we use, has the lowest RMS value, and the curve is stable with little fluctuation. L0 (black curve) represents the original image without correction, and its RMS value is the highest among all methods, with drastic fluctuations, confirming that there is significant radiometric inhomogeneity in the original data. The Uniform method (green curve) and WU method (orange curve), as the other two correction methods, are superior to the original L0 data in terms of performance, but their RMS values ​​are significantly higher than Our method, indicating that their correction effect is not as good as our method.

[0071] Visual correction effect as Figure 5 As shown, before correction, the image exhibited significant longitudinal stripe noise and bad pixels. After correction using this method, these inhomogeneities were effectively eliminated, the stripe texture disappeared, and the bad pixels were repaired. The corrected image showed significant improvement in overall brightness, color, and uniformity, resulting in higher visual quality. This demonstrates the effectiveness of the radiometric correction method in improving image uniformity.

[0072] On the other hand, embodiments of the present invention also provide a multi-scale radiometric cascade remote sensing satellite on-orbit relative radiometric calibration system, comprising: The system includes a processor and a memory. The memory stores program instructions, and the processor calls the stored instructions in the memory to execute the on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in the above technical solution.

[0073] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for on-orbit relative radiometric calibration of remote sensing satellites with multi-scale radiometric cascades, characterized in that, Includes the following steps: Step 1: Select the image sequence for calibration and align the image sequence; Step 2: Based on the aligned sequence of images, obtain the relative radiometric calibration reference image through an estimation method; Step 3: Obtain the motion trajectory of the array elements in the time-series images, and obtain the grayscale value of each element in each sequence of images; Step 4: Based on the grayscale values ​​of the array elements and the relative radiometric calibration benchmark, construct the radiometric calibration model for each element of the array sensor; Step 5: Solve the radiometric calibration model in Step 4 using a multi-scale method to obtain the relative radiometric calibration coefficients of each detector element of the array sensor at different scales. Step 6: The radiometric calibration coefficients at each scale obtained in Step 5 are merged using a multi-scale radiometric cascade method to complete the on-orbit relative radiometric calibration of the array-type remote sensing satellite.

2. The on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: The specific implementation method of step 1 is as follows: (11) Image sequence selection: Select large-area uniform field images or non-uniform field images; (12) Sequence image alignment method: The image matching method selected includes the SIFT matching algorithm and other image alignment methods. For uniform field sequence images, the sequence image alignment accuracy is at least better than 1 pixel. For non-uniform field sequence images, the sequence image alignment accuracy is at least better than 0.5 pixels.

3. The on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: The specific implementation method for image sequence alignment is as follows: 1) Select the image with the largest overlap with other image sequences in the image sequence data as the reference image. Use the scale-invariant feature transform algorithm to perform block matching between other images and the reference image to obtain corresponding point pairs between each image sequence and the reference image. The image-side affine transformation model relationship between images is initialized based on the matched pairs of corresponding points between sequential images, as shown in the following equation: In the formula arrive The coefficients are the affine transformation model coefficients; the affine model relationship between the i-th image in the image sequence and the reference image is defined as follows: Then we have: The above affine transformation model is used to calculate the residuals of matching corresponding point pairs between sequential images. We have the following formula: in and Representing the root mean square error, these values ​​measure the average geometric deviation of all corresponding point pairs in the x and y directions during image alignment. N refers to the total number of point pairs used in the calculation. and They represent the first The x-coordinates of the corresponding points on the reference image and the image sequence. and They represent the first The y-coordinates of the corresponding points on the reference image and the image sequence; and These are the standard deviations in the x and y directions, respectively, indicating the registration accuracy of statistically corresponding points; With 3 To remove outliers using a threshold, new pairs of corresponding points between image sequences are obtained. If we reinitialize the affine transformation model between the image sequences, then we have: 2) By traversing all image sequences, the one-to-one correspondence between each pixel of the image sequence and the reference image can be obtained. This refers to the precise relative positional relationship between images in a sequence.

4. The on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: Step 2: Select the mean estimation method or other probability estimation method to ensure that the obtained relative radiometric calibration reference image has no obvious image blurring or trailing problems. The specific implementation of the mean estimation method is as follows: 2.1 Based on aligned sequence image data, we have: In the formula N For the number of images, M The number of detector elements in the array; Indicates the first i Image No. n The detector number for each pixel image. The first image is the estimated true radiance of ground features. n Each pixel's grayscale value In the In the image, by the first The image captured by the probe is the n The grayscale value of each pixel; 2.2 Finally, the true surface values ​​are estimated, resulting in a relative radiometric calibration reference image. .

5. The on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: In step 3, the method for extracting grayscale values ​​from area array detectors in each frame sequence is as follows: 3.1, Given that the coordinates of the area array element D on the reference image are... ; 3.2, Coordinates of Detector D in a certain frame of image This can be obtained from step 1: The term refers to the functional relationship, specifically the function representing the precise relative positional relationship of time-series images. 3.3, Based on the coordinates of probe D in a certain frame of image This allows you to extract grayscale values. .

6. The on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: In step 4, the radiation calibration model of each detector element of the array sensor is as follows: In the formula , For the first array sensor i The relative radiometric calibration coefficients of each detector element j Image sequence number N is the number of images in the sequence, where the coefficient is... When the value is 0, only the gain coefficient is calculated. .

7. The on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: Step 5 includes the following sub-steps: 5.1, Determine the number of scale layers as follows: s is the s-th layer, and ScaleNumber is the maximum scale; 5.2 The aligned sequence image obtained in step 1 and the relative radiometric calibration reference image obtained in step 2 are downsampled to obtain... Sequences of images at different scales; 5.3 Combine the sequence of images at the first-level scale with the relative radiometric calibration reference image, and use step 4 to obtain the radiometric calibration model of the first level; 5.

4. Solve the radiometric calibration model in 5.3 using the least squares method to obtain the radiometric calibration coefficients of the first layer. ; 5.5 Similarly, calculate the relative radiometric calibration coefficients of the s-th layer of the image sequence. .

8. The method for on-orbit relative radiometric calibration of remote sensing satellites with multi-scale radiometric cascade as described in claim 7, characterized in that: Step 5.5, relative radiometric calibration coefficients of the s-th layer of the image sequence. The calculation method is as follows: 5.5.1, using the relative radiometric calibration coefficients of the (s-1)th layer, the radiometric correction of the s-th layer scale sequence image is completed based on the following formula. In the formula For the array sensor at the s-1 level scale, the first... i The relative radiometric calibration coefficients of each detector element; 5.5.2 Combine the sequence of images at the s-th layer scale with the relative radiometric calibration reference image, and use step 4 to establish the radiometric calibration model for the s-th layer; 5.5.3, using the least squares method to solve the radiation calibration model in 5.5.2, obtain the radiation calibration coefficients of the s-th layer. .

9. The method for on-orbit relative radiometric calibration of remote sensing satellites with multi-scale radiometric cascade as described in claim 1, characterized in that: Step 6 includes the following sub-steps: 6.1 For the i-th detector element of the area array sensor, the combined formula for the relative radiation calibration coefficients of each layer is as follows: In the formula Let be the relative radiation calibration coefficient of the s-th layer. and is the final relative radiometric calibration coefficient of the i-th detector element of the area array sensor; 6.2 Repeat step 6.1 for all imaging elements of the array sensor to obtain the final relative radiometric calibration coefficients of all elements of the array sensor, thus completing the on-orbit relative radiometric calibration of the array sensor.

10. A multi-scale radiometric cascaded remote sensing satellite on-orbit relative radiometric calibration system, characterized in that, include: The processor and memory, wherein the memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the on-orbit relative radiometric calibration method for remote sensing satellites with multi-scale radiometric cascade as described in any one of claims 1-9.

Images