An elliptical Gaussian distribution-based satellite image sharpness enhancement method

Abstract

The application discloses a satellite image definition improvement method based on an elliptical Gaussian distribution, relates to the technical field of remote sensing image processing, and comprises the following steps: generating motion data according to collected physical imaging parameters, and constructing a reference motion vector based on the motion data, wherein the motion data comprises a long-axis reference value and a reference blur direction angle; constructing a constraint space based on statistical characteristics of real satellite images, and collecting random degradation parameters from the constraint space, wherein the random degradation parameters comprise a long-axis disturbance term, a direction angle jitter term and a short-axis standard deviation; and fusing the reference motion vector and the random degradation parameters to generate a composite elliptical Gaussian degradation convolution kernel; in the construction process of the training sample pair set, the long-axis reference value and the reference blur direction angle, the random degradation parameters and the alignment reference data are introduced, so that the training sample pair is more suitable for real satellite imaging in terms of blur structure and spatial matching.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing image processing, and particularly relates to a satellite image definition enhancement method based on elliptical Gaussian distribution. BACKGROUND

[0002] With the increasing demand for satellite imaging tasks in fine ground monitoring, disaster analysis, urban change analysis and resource investigation, the role of satellite image definition in spatial target recognition and texture information analysis becomes increasingly critical. Imaging equipment usually needs to obtain continuous ground information under high-speed orbital motion, complex attitude disturbance and unstable atmospheric conditions, resulting in multi-factor coupling characteristics of motion blur and optical diffusion generated during the imaging process. Under such conditions, the processing methods that simply rely on fixed blur models, linear motion assumptions or static degradation kernels often cannot accurately describe the change pattern of real imaging degradation, making it difficult to support subsequent deep learning training tasks based on pseudo-satellite image data, thereby laying the demand background for proposing an image definition enhancement scheme that can reflect both the physical imaging process and the spatial statistical structure.

[0003] In the process of simulating imaging blur patterns based on physical imaging parameters, it is usually difficult to convert the combined effects of satellite orbit attitude angle, ground projection velocity, imaging scanning direction and exposure integration time into imaging motion characteristics that can stably describe the blur amplitude and blur direction, making the generated pseudo-satellite image data difficult to maintain correspondence with the real imaging process in terms of motion blur structure. When the shift of the blur direction angle, component noise and change of the motion scale are not fully expressed, the blur structure of the pseudo-satellite image data in the training sample pair set is prone to distortion, thereby affecting the learning ability of the subsequent reconstruction model for the real degradation pattern.

[0004] Secondly, in the process of constructing statistical degradation parameters required for degradation simulation, it is usually difficult to convert the edge gradient histogram, atmospheric disturbance standard deviation statistics and point spread function half-width value into a multi-dimensional constraint space that can be used for sampling, making the long-axis disturbance term, direction angle jitter term and short-axis standard deviation lack uniformity in statistical sources. In the absence of stable correlation between gradient section data, disturbance category data and diffusion interval data, the generated random degradation parameters are difficult to reflect the gradient distribution, disturbance structure and optical diffusion characteristics under real imaging conditions, resulting in insufficient blur diversity of the pseudo-satellite image data, thereby limiting the coverage ability of the training sample pair set for real blur conditions.

[0005] In addition, in the process of establishing the spatial correspondence between the pseudo-satellite image data and the high-definition aerial image, slight pixel-level shifts often occur due to resampling operations, local texture differences or image boundary processing, affecting the spatial consistency of the training sample pair set; if the coordinate difference values of the local regions in the pseudo-satellite image data lack effective screening, or the global alignment relationship cannot be stably estimated through the gradient intensity weight, spatial misalignment is likely to occur in the pixel-by-pixel pairing process, making it difficult for the deep learning network in the parameter iterative training stage to accurately model the true blur shape, spatial shift and local change, thereby affecting the efficiency of satellite image sharpness improvement. SUMMARY

[0006] To solve the above technical problems, the application provides a satellite image sharpness improvement method based on an elliptical Gaussian distribution, which comprises the following steps:

[0007] S1, generating motion data according to the collected physical imaging parameters, and constructing a reference motion vector based on the motion data, wherein the motion data comprises a long-axis reference value and a reference blur direction angle;

[0008] S2, constructing a constraint space based on the statistical characteristics of the real satellite image, and collecting random degradation parameters from the constraint space, wherein the random degradation parameters comprise a long-axis perturbation term, a direction angle jitter term and a short-axis standard deviation;

[0009] S3, fusing the reference motion vector and the random degradation parameters to generate a composite elliptical Gaussian degradation convolution kernel;

[0010] S4, performing degradation simulation operation on the high-definition aerial image through the composite elliptical Gaussian degradation convolution kernel to generate pseudo-satellite image data, and performing pixel-level alignment between the pseudo-satellite image data and the corresponding high-definition aerial image to construct a training sample pair set.

[0011] Further, the physical imaging parameters comprise satellite orbit attitude angle, ground projection speed, imaging scanning direction and exposure integration time; and the statistical characteristics comprise edge gradient histogram of the real satellite image, atmospheric disturbance standard deviation statistics and point spread function half-width value.

[0012] Further, the step of constructing the reference motion vector based on the motion data comprises:

[0013] S11, performing parameter recording processing through the satellite orbit attitude angle, the ground projection speed, the imaging scanning direction and the exposure integration time to generate a physical imaging parameter record;

[0014] S12, performing multiplication calculation based on the ground projection speed and the exposure integration time in the physical imaging parameter record to generate a long-axis reference value;

[0015] S13, performing angle resolving processing on the imaging scan direction in the physical imaging parameter record to generate a reference blur direction angle;

[0016] S14, performing vector construction operation according to the long axis reference value and the reference blur direction angle to generate a reference motion vector.

[0017] Further, the logic of performing angle resolving processing is:

[0018] a1, performing spatial direction decomposition processing through the imaging scan direction to generate a direction decomposition amount;

[0019] a2, performing angle offset estimation calculation according to the imaging scan direction to generate an angle offset amount;

[0020] a3, performing direction noise suppression calculation through the imaging scan direction to generate a noise suppression amount;

[0021] a4, performing angle fusion processing based on the direction decomposition amount, the angle offset amount and the noise suppression amount to generate the reference blur direction angle.

[0022] Further, the step of constructing the constraint space based on the statistical characteristics of the real satellite image includes:

[0023] S21, performing section gradient distribution determination based on the edge gradient histogram to generate gradient section data;

[0024] S22, performing disturbance amplitude classification processing according to the atmospheric disturbance standard deviation statistical quantity to generate disturbance category data;

[0025] S23, performing diffusion range interval division based on the point spread function half width value to generate diffusion interval data;

[0026] S24, performing spatial domain layering processing through the gradient section data, the disturbance category data and the diffusion interval data to construct the constraint space.

[0027] Further, the step of generating gradient section data is:

[0028] S211, performing amplitude density analysis based on each gradient amplitude section of the edge gradient histogram to generate a density mutation section record;

[0029] S212, performing amplitude section boundary determination according to the density mutation section record to obtain an amplitude boundary set;

[0030] S213, performing section division processing through the amplitude boundary set and the amplitude distribution of the edge gradient histogram to generate the gradient section data.

[0031] Further, the logic of performing amplitude density analysis is:

[0032] b1, performing segment gradient change rate analysis based on the amplitude segment frequency change rate of the edge gradient histogram to obtain change rate fluctuation data;

[0033] b2, performing amplitude segment stability determination according to the number of consecutive inversions of adjacent change rate signs in the change rate fluctuation data to obtain an unstable amplitude segment set;

[0034] b3, performing mutation segment screening processing by the unstable amplitude segment set and the frequency gradient difference of the corresponding amplitude segment in the edge gradient histogram to generate a density mutation segment record.

[0035] Further, the generation logic of the composite elliptical Gaussian degradation convolution kernel is:

[0036] S31, performing amplitude superposition processing by the major axis reference value and the major axis perturbation term to generate a major axis fusion value;

[0037] S32, performing angle correction processing based on the reference blur direction angle and the direction angle jitter term to obtain a direction fusion angle value;

[0038] S33, performing axis parameter updating according to the short axis standard deviation and the correlation coefficient field to generate a short axis update value;

[0039] S34, performing convolution kernel parameter collection processing by the major axis fusion value, the direction fusion angle value, and the short axis update value to generate a composite elliptical Gaussian degradation convolution kernel.

[0040] Further, the logic for constructing a training sample pair is:

[0041] c1, performing convolution processing by the composite elliptical Gaussian degradation convolution kernel and the high-definition aerial image to generate pseudo-satellite image data;

[0042] c2, performing spatial coordinate correspondence analysis on the pseudo-satellite image data and the high-definition aerial image to obtain alignment reference data;

[0043] c3, performing pixel-by-pixel pairing processing based on the alignment reference data to generate a training sample pair set.

[0044] Further, the step of performing spatial coordinate correspondence analysis is:

[0045] c21, performing coordinate difference value calculation based on the corresponding pixel coordinates of the pseudo-satellite image data and the high-definition aerial image to generate difference value segment data;

[0046] c22, performing local consistency screening by the difference value segment data to obtain candidate alignment unit data;

[0047] c23, performing coordinate mapping confirmation processing according to the candidate alignment unit data to generate alignment reference data.

[0048] Compared with the prior art, the present application has the beneficial effects that:

[0049] The present application constructs imaging motion features based on physical imaging parameter records, long axis reference values and reference blur direction angles, and dynamically controls the blur direction and blur scale in the subsequent degradation convolution kernel, so that the pseudo-satellite image data can be close to the spatial degradation characteristics of the real satellite image in the motion blur form, thereby establishing a blur structure mapping basis consistent with the real imaging environment in the training sample pair set; in addition, the present application obtains random degradation parameters from the constraint space composed of gradient section data, disturbance category data and diffusion interval data, so that the long axis disturbance term, the direction angle jitter term and the short axis standard deviation are consistent in origin and physical consistency at the statistical level, so as to introduce controllable blur diversity in the composite elliptical Gaussian degradation convolution kernel, thereby improving the stability and rationality of the pseudo-satellite image data in the real blur state coverage.

[0050] In addition, the present application also performs spatial coordinate corresponding analysis, local consistency screening and alignment reference data generation processing on the pseudo-satellite image data and high-definition aerial image, so that the pixel-level corresponding relationship in the training sample pair set is established on the basis of consistent spatial translation correction in the global and local, thereby reducing the sample deviation caused by resampling error, window texture difference or local drift, to improve the reliability and overall coordination of the training sample pair set in the spatial matching level, and further enhance the effective modeling ability of the deep learning network in the parameter iterative training process.

[0051] Further, the present application is based on the blur structure consistency and spatial matching consistency of the training sample pair set in the parameter iterative training stage, so that the deep learning network can stably converge under the conditions of multiple types of blur forms, different spatial position offsets and multi-scale diffusion features, and obtain a satellite image sharpening processing model capable of outputting high-fidelity reconstruction results under the matching error constraint, so that the image reconstruction task has higher reconstruction accuracy and stronger structure restoration ability when facing the motion blur, optical diffusion and local structure inconsistency of the real satellite image.

[0052] In summary, the present application introduces long axis reference values and reference blur direction angles, random degradation parameters and alignment reference data in the construction process of the training sample pair set, so that the training sample pair is more consistent with the real satellite imaging in terms of blur structure and spatial matching. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0054] Figure 1 A flow chart of a satellite image definition enhancement method based on elliptical Gaussian distribution provided by the embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0056] Please refer to Figure 1 The embodiment disclosed provides a satellite image definition enhancement method based on elliptical Gaussian distribution, which comprises the following steps.

[0057] S1, generating motion data according to the collected physical imaging parameters, and constructing a reference motion vector based on the motion data, wherein the motion data comprises a long-axis reference value and a reference blur direction angle;

[0058] In one specific embodiment, for the basic physical imaging input parameters required for constructing the subsequent motion degradation model, this step performs a structured recording process on the imaging physical quantities that can be directly observed or retrieved during the satellite imaging process.

[0059] Specifically, the physical imaging parameters comprise a satellite orbit attitude angle, a ground projection speed, an imaging scanning direction and an exposure integration time.

[0060] The satellite orbit attitude angle is jointly output by a high-precision gyroscope and a star sensor of a satellite attitude control system, and is used to describe the spatial attitude change of the satellite at the imaging moment. It is a key physical quantity reflecting the amount of optical axis deviation.

[0061] The satellite orbit attitude angle comprises a roll angle, a pitch angle and a yaw angle, and all of them have a slight disturbance at the imaging moment, with a unit of °.

[0062] The ground projection speed is the speed of the satellite along the ground projection trajectory direction at the imaging moment, which is measured by an orbit mechanics model combined with a global satellite navigation system, and can be used to describe the relative motion of the target ground object during imaging, with a unit of m / s.

[0063] Imaging scan direction is provided by the payload scanning mechanism, which is used to describe the directional relationship of the sensor scanning line relative to the ground coordinate system during imaging, with the unit of °.

[0064] It should be noted that the scan direction is different in push-broom and swing-broom cameras, but it can be recorded as a single direction angle value as input.

[0065] Exposure integration time is set by the satellite payload control system, which is basically constant in fixed imaging mode, but there is dynamic change in wide swath imaging or high-speed imaging conditions. This parameter is used to describe the cumulative time of the camera during single column pixel exposure, which is a linear factor for blur length calculation.

[0066] Specifically, the step of constructing a reference motion vector based on motion data is:

[0067] S11, performing parameter recording processing through satellite orbit attitude angle, ground projection speed, imaging scan direction and exposure integration time, to generate a physical imaging parameter record;

[0068] In a preferred embodiment, to ensure the consistency of parameters in the subsequent motion blur calculation process, the above four types of physical imaging parameters are generated into a physical imaging parameter record according to a preset parameter field structure table, as shown in Table 1:

[0069] Table 1: Parameter field structure table

[0070]

[0071] Among them, the attitude angle vector adopts a three-dimensional column vector form, which is expressed as:

[0072] ;

[0073] In the formula, , , respectively represent the roll angle, pitch angle and yaw angle of the satellite at the collection moment;

[0074] The generated physical imaging parameter record is expressed as:

[0075] ;

[0076] Exemplarily, assuming that the parameters collected at a certain imaging moment are:

[0077] Satellite orbit attitude angle: ; Ground projection speed: ; Imaging scan direction: ; Exposure integration time: ; The generated physical imaging parameter record can be expressed as:

[0078] ;

[0079] It should be noted that, in order to avoid outliers caused by attitude jumps or star sensor errors from entering the degenerate kernel calculation process, this embodiment uses the following method to verify the validity of parameters: a sliding window mean comparison is performed on the set of satellite orbit attitude angles to remove instantaneous spikes. In addition, the scanning direction and exposure time are checked by interpolation of the operation log to ensure consistency of sources. The data that passes the verification is recorded as the final physical imaging parameters.

[0080] S12, perform multiplication calculation based on the ground projection velocity and exposure integration time recorded in the physical imaging parameters to generate the major axis reference value;

[0081] The exposure integral time is converted to a unit to generate the converted exposure integral time. , Exposure time is the exposure time before unit conversion;

[0082] By multiplying the ground projection velocity in the physical imaging parameter record with the converted exposure integral time, the major axis reference value, which characterizes the motion blur scale along the imaging scanning direction during the imaging process, can be obtained, expressed as: ;

[0083] In the formula, This is the reference value for the major axis. Ground projection velocity;

[0084] It should be noted that the major axis baseline value is used to construct the major axis scale of the subsequent elliptical Gaussian degenerate convolution kernel.

[0085] S13, Perform angle analysis processing on the imaging scanning direction in the physical imaging parameter record to generate the reference blur direction angle;

[0086] Specifically, the logic for performing angle analysis and processing is as follows:

[0087] a1, performs spatial orientation decomposition processing through the imaging scanning direction to generate orientation decomposition quantities;

[0088] The directional decomposition includes horizontal and vertical components;

[0089] It is important to understand that: the imaging scanning direction, as a planar angular quantity, can be decomposed to generate directional decomposition quantities that describe the structure of the scanning direction components; expressed as: ;

[0090] in, The horizontal component is the direction of the imaging scan. This represents the vertical component of the imaging scanning direction. is the imaging scan direction.

[0091] a2, performing an angle offset estimation calculation according to the imaging scan direction, to generate an angle offset amount;

[0092] is expressed as: ;

[0093] wherein, is the angle offset amount, is an offset estimation coefficient, is a vertical component of the imaging scan direction, is an arctangent function;

[0094] It should be noted that the angle offset amount reflects the degree of offset caused by the vertical component in the imaging scan direction to the blur direction, and the offset estimation coefficient is a direction stability feature obtained by statistically processing the variance of the angle change sequence of the imaging scan direction of the real satellite image, and preferably, the value range of is [0.01, 0.10].

[0095] a3, performing a direction noise suppression calculation through the imaging scan direction, to generate a noise suppression amount;

[0096] is expressed as: ;

[0097] wherein, is the noise suppression amount, is a noise suppression coefficient, is a horizontal component of the imaging scan direction;

[0098] It should be noted that the noise suppression coefficient is obtained by statistical processing of the edge gradient noise energy proportion of the real satellite image, and preferably, the noise suppression coefficient the value range of is [0.20, 0.35].

[0099] a4, performing an angle fusion processing based on the direction decomposition amount, the angle offset amount and the noise suppression amount, to generate a reference blur direction angle;

[0100] is expressed as: ;

[0101] wherein, is the reference blur direction angle, is an arctangent function;

[0102] It should be noted that the reference blur direction angle integrates the geometric component, the offset and the noise suppression factor of the imaging scan direction, and is used to describe the imaging blur direction.

[0103] S14, performing a vector construction operation according to the long axis reference value and the reference blur direction angle to generate a reference motion vector;

[0104] based on the obtained long axis reference value and the reference blur direction angle , a reference motion vector is constructed;

[0105] is expressed as: ;

[0106] In the formula, is a reference motion vector;

[0107] It should be noted that the reference motion vector is used to represent the blur direction and blur amplitude caused by imaging motion.

[0108] S2, constructing a constraint space based on the statistical characteristics of real satellite images, and collecting random degradation parameters from the constraint space, the random degradation parameters including a long axis disturbance term, a direction angle jitter term, and a short axis standard deviation;

[0109] In order to enable the degradation simulation process to reflect the blur distribution pattern in the real imaging environment, the constraint space for sampling random degradation parameters is obtained by performing segmentation, classification and spatial layering construction on the statistical characteristics of real satellite images;

[0110] The constraint space is a multi-dimensional statistical structure, including gradient section data, disturbance category data and diffusion interval data;

[0111] Based on the constraint space, random degradation parameters satisfying physical consistency are collected according to a set sampling strategy, which are used for subsequent degradation kernel generation processing.

[0112] Specifically, the statistical characteristics include an edge gradient histogram of real satellite images, an atmospheric disturbance standard deviation statistic, and a point spread function half-width value;

[0113] The edge gradient histogram is generated by extracting edge gradient values within a fixed size window and counting after the real satellite image is collected by a linear array remote sensing camera;

[0114] The atmospheric disturbance standard deviation statistic is obtained by calculating the fluctuation value of the brightness difference of adjacent pixels based on the continuous sampling data of the cross-scan direction of the spaceborne multispectral radiometer;

[0115] The point spread function half-width value is obtained based on the measurement results of the imaging load optical calibration device in the laboratory calibration process of the point light source imaging response.

[0116] Specifically, the step of constructing the constraint space based on the statistical characteristics of real satellite images includes:

[0117] S21, perform segment gradient distribution determination based on edge gradient histogram, and generate gradient segment data;

[0118] In real satellite imagery, edge gradient histograms can reflect the changes in texture intensity in different regions. In order to transform the gradient distribution into discrete segments suitable for statistical sampling, the edge gradient histogram is segmented to generate gradient segment data.

[0119] Specifically, the steps for generating gradient segment data are as follows:

[0120] S211, perform amplitude density analysis on each gradient amplitude segment based on the edge gradient histogram to generate density abrupt change segment records;

[0121] Specifically, the logic for performing amplitude density analysis is as follows:

[0122] b1, based on the amplitude segment frequency change rate of the edge gradient histogram, perform segment gradient change rate analysis to obtain change rate fluctuation data;

[0123] The frequency of each gradient amplitude segment is counted, and the rate of change of frequency between adjacent amplitude segments is calculated, expressed as: ;

[0124] In the formula, The edge gradient histogram of the th Frequency of each amplitude range The edge gradient histogram of the th Frequency of each amplitude range For the first The rate of change of frequency in the amplitude range;

[0125] Based on multiple groups Construct a set of frequency change rates This data is then output as fluctuation data of the rate of change.

[0126] It should be noted that the frequency of a certain gradient amplitude range is the number of pixels whose pixel gradient falls within that amplitude range.

[0127] b2, based on the number of consecutive reversals of adjacent rate of change signs in the rate of change fluctuation data, perform a stability determination of the amplitude segment to obtain a set of unstable amplitude segments;

[0128] In a specific embodiment, when performing gradient change stability determination based on the number of sign reversals of the rate of change, it is necessary to retrieve a preset threshold for the number of consecutive reversals. When the number of consecutive reversals is greater than or equal to the threshold, it is marked as an unstable amplitude segment, and an unstable amplitude segment set is constructed based on multiple unstable amplitude segments.

[0129] It should be noted that: the continuous inversion number threshold is set according to the statistical analysis result of the real satellite image, the sign set of the change rate of each amplitude section is counted in the edge gradient histogram of a large number of samples, the occurrence number of continuous sign inversion is recorded, and the continuous inversion number with higher occurrence frequency is extracted as the value range of the threshold; the continuous inversion number threshold obtained by this method is a small positive integer;

[0130] Preferably, the continuous inversion number threshold value range can be between 1 and 3.

[0131] b3, by executing the mutation section screening processing of the unstable amplitude section set and the frequency gradient difference value of the corresponding amplitude section in the edge gradient histogram, generating the density mutation section record;

[0132] is expressed as: ;

[0133] In the formula, is the frequency gradient difference value of the first amplitude section, , is the frequency of the adjacent amplitude section;

[0134] When a certain amplitude section belongs to the unstable amplitude section set at the same time, and its frequency gradient difference value reaches the preset frequency gradient difference value, the amplitude section is marked as a density mutation section, and is written into the density mutation section record.

[0135] Wherein, the preset frequency gradient difference value is preferably 1% of the total number of histogram pixels;

[0136] S212, performing amplitude section boundary determination according to the density mutation section record to obtain an amplitude boundary set;

[0137] In this step, in order to divide the continuous gradient amplitude space into discrete sections with statistical representation, it is necessary to perform amplitude section boundary determination according to the generated density mutation section record;

[0138] Specifically, each density mutation section contained in the density mutation section record represents the most significant boundary position of the gradient change in the gradient histogram, which can be used to determine the amplitude section boundary, and the amplitude boundary set is obtained based on a plurality of amplitude section boundaries;

[0139] Exemplarily, after performing amplitude density analysis on the edge gradient histogram, it is assumed that the obtained density mutation section record contains three amplitude sections: ;

[0140] It should be noted that: the amplitude axis of the edge gradient histogram is divided into a plurality of continuous amplitude sections, 4 can be understood as the fourth amplitude section, and the section width is set to 0.05, then:

[0141] The amplitude range of the 4th paragraph is about [0.15, 0.20], and the amplitude section boundary is taken as 0.20;

[0142] The amplitude range of the 9th paragraph is about [0.40, 0.45], and the amplitude section boundary is taken as 0.45;

[0143] The amplitude range of the 12th paragraph is about [0.55, 0.60], and the amplitude section boundary is taken as 0.60;

[0144] The obtained amplitude section boundaries are sorted in ascending order to generate an amplitude boundary set, denoted as:

[0145] ;

[0146] Since the real satellite image has strong complexity in edge structure and texture distribution, the boundary determination method based on mutation section can effectively depict the structural change of gradient distribution, so that the generated gradient section data is more consistent with the edge response characteristics of the real image.

[0147] S213, performing section division processing on the amplitude boundary set and the amplitude distribution of the edge gradient histogram to generate gradient section data;

[0148] For each divided amplitude section, based on the frequency of each amplitude section in the corresponding edge gradient histogram, the proportion of the frequency of each amplitude section to the total frequency of the amplitude section is calculated, and these proportion values are sorted in ascending order of amplitude to form a structured set, forming an amplitude section density structure, which is used to represent the distribution form of the gradient density inside the amplitude section.

[0149] Further, the amplitude range, frequency and corresponding amplitude section density structure of each amplitude section are combined into a structured record as a component unit of the gradient section data, and finally the gradient section data is generated;

[0150] It should be noted that the amplitude range is illustrated in step S212 and is not described in detail.

[0151] S22, performing disturbance amplitude classification processing according to the atmospheric disturbance standard deviation statistic to generate disturbance category data;

[0152] The real satellite image is easily affected by atmospheric disturbance during imaging, and the disturbance will be reflected as random fluctuations of different intensities in the edge structure and brightness change of the image;

[0153] Based on the atmospheric disturbance standard deviation statistic obtained by a large number of real image statistics, the disturbance amplitude can be classified to construct different disturbance levels.

[0154] In a specific implementation, the atmospheric disturbance standard deviation statistics is subjected to disturbance amplitude classification processing according to a preset classification rule, the disturbance standard deviation value is assigned to a plurality of disturbance level intervals, and disturbance category data is obtained, which describes the disturbance amplitude range corresponding to different atmospheric disturbance intensity intervals.

[0155] S23, performing diffusion range interval division based on the point spread function half-width value to generate diffusion interval data;

[0156] The half-width value of the point spread function (PSF) can describe the diffusion range when the optical system is imaging, and is an important physical index of imaging clarity and blurring degree.

[0157] In order to make the simulated degradation convolution kernel be able to map the diffusion characteristics of the real imaging system, interval division of the diffusion range needs to be performed based on the half-width value;

[0158] In a specific implementation process, the point spread function half-width value obtained by statistics is partitioned according to the numerical value, divided into a plurality of diffusion range intervals, and diffusion interval data is formed;

[0159] The diffusion interval data describes the diffusion scale that the optical system may produce under different imaging conditions, and provides sampling basis for the short-axis standard deviation in the random degradation parameter.

[0160] The diffusion interval data established in this way can ensure that the short-axis diffusion characteristics in the degradation simulation are consistent with the performance of the real satellite image in the optical diffusion layer

[0161] S24, performing spatial domain layering processing through the gradient section data, the disturbance category data and the diffusion interval data to construct a constraint space;

[0162] In one specific embodiment, the gradient structure described by the gradient section data is the first spatial dimension, the atmospheric disturbance level described by the disturbance category data is the second spatial dimension, and the diffusion scale interval described by the diffusion interval data is the third spatial dimension;

[0163] According to the combination of the three types of data in three spatial dimensions, the combination rule of "gradient section data × disturbance category data × diffusion interval data" is used to construct a multi-dimensional space with gradient-disturbance-diffusion triple constraints, and the multi-dimensional space is output as a constraint space;

[0164] Based on the constraint space, the degradation parameters of each spatial dimension can be collected as random degradation parameters, and the collection logic is:

[0165] The preset long-axis disturbance item is extracted from the first spatial dimension of the constraint space, the preset direction angle jitter item is extracted from the second spatial dimension of the constraint space, and the preset short-axis standard deviation is extracted from the third spatial dimension of the constraint space.

[0166] It should be noted that the long axis disturbance term is obtained by statistically analyzing the edge spread width along the imaging motion direction in real satellite images. Specifically, the edge of a large number of real satellite images is scanned line by line, and the gradient spread width along the imaging direction is measured, so that the long axis disturbance term, which can be directly used for degradation simulation, is obtained, and the unit is m.

[0167] The direction angle jitter term is obtained by statistically analyzing the main direction angle fluctuation of local structures in real satellite images. Specifically, the structure direction of multiple local regions of the image is analyzed, the main direction angle of each region and the change amplitude thereof relative to the adjacent region are measured, so that the angle fluctuation representing the direction instability is obtained, and the unit is °.

[0168] The generation step of the short axis standard deviation is specifically: selecting a region with uniform texture and gentle brightness change in multiple real satellite images, using a point source or an approximately point-like spot commonly used in imaging instrument calibration, analyzing the brightness spread form of the point-like feature in the transverse direction, and recording the transverse width corresponding to the position where the brightness drops to half of the peak value. The transverse width values obtained from multiple images are sorted, and a set of transverse spread width data which can be directly used for degradation simulation is formed. After the transverse spread width data is calculated by standard deviation, the generated result is the short axis standard deviation.

[0169] Exemplarily, it is assumed that the first dimension of space, i.e., the gradient section data, contains three gradient sections, and the gradient section table is shown in Table 2.

[0170] Table 2: Gradient section table

[0171]

[0172] It is assumed that the second dimension of space, i.e., the disturbance category data, contains three disturbance levels, and the disturbance level table is shown in Table 3.

[0173] Table 3: Disturbance level table

[0174]

[0175] It is assumed that the third dimension of space, i.e., the spread interval data, contains three spread scale intervals divided based on the half-width value of the point spread function, and the spread scale interval table is shown in Table 4.

[0176] Table 4: Spread scale interval table

[0177]

[0178] Through the combination rule of “gradient section data × disturbance category data × spread interval data”, the following can be generated:

[0179] A spatial hierarchy unit

[0180] Each "unit" corresponds to a real-world possible blur state; for example, a unit can represent: gradient section G3, disturbance level T2, diffusion scale E1;

[0181] It should be noted that the specific spatial hierarchy unit is not only 27, which can be combined according to different dimension levels, and can be more than 27 or less than 27. The above content is only for example to facilitate understanding

[0182] Suppose it randomly falls on the spatial hierarchy unit (G3, T2, E1) of the constraint space;

[0183] Collect random degradation parameters, that is, extract the long-axis disturbance term from the first dimension G3 of the constraint space, extract the direction angle jitter term from the second dimension T2 of the constraint space, and extract the short-axis standard deviation from the third dimension E1 of the constraint space.

[0184] S3, fuse the reference motion vector and the random degradation parameter to generate a composite elliptical Gaussian degradation convolution kernel;

[0185] Specifically, the generation logic of the composite elliptical Gaussian degradation convolution kernel is:

[0186] S31, perform amplitude superposition processing on the long-axis reference value and the long-axis disturbance term to generate a long-axis fusion value;

[0187] The generation formula of the long-axis fusion value is: ;

[0188] In the formula, is the unconstrained long-axis fusion value, is the long-axis disturbance term;

[0189] It should be noted that the amplitude superposition processing reflects the joint action of the physical motion blur scale and the statistical blur scale in actual imaging

[0190] The lower limit constraint processing is performed on the unconstrained long-axis fusion value to obtain the long-axis fusion value;

[0191] It is expressed as: ;

[0192] is the long-axis fusion value, is the minimum acceptable long-axis scale set according to the real image blur width statistics, which prevents the degradation kernel from degrading to a very small scale, is selected and as the long-axis fusion value for output.

[0193] S32, performing angle correction processing based on the reference blur direction angle and the direction angle jitter term to obtain a direction fusion angle value;

[0194] ;

[0195] wherein, is the un-normalized direction fusion angle value, is the direction angle jitter term;

[0196] It should be noted that the superposition of this step is used to simulate the random disturbance of the scanning direction angle in the real imaging process;

[0197] The un-normalized direction fusion angle value is normalized to generate the direction fusion angle value;

[0198] is expressed as: ;

[0199] wherein, is the direction fusion angle value, and the value range is [0, ), The un-normalized direction fusion angle value is angle normalized to make the output in the interval [0, ).

[0200] S33, performing axial parameter updating according to the short axis standard deviation and the correlation coefficient field to generate a short axis updated value;

[0201] is expressed as: ;

[0202] wherein, is the un-constrained short axis updated value, is the short axis standard deviation, is the correlation coefficient field, which is used to describe the correlation degree between the major axis direction and the minor axis direction of the elliptical Gaussian kernel;

[0203] It should be noted that the axial parameter updating of this step reflects the change trend of the short axis scale under the joint action of the atmospheric diffusion structure and the optical correlation;

[0204] The correlation coefficient field is obtained by performing covariance statistical processing on the two-dimensional local blur shape of the typical edge region in the real satellite image;

[0205] Specifically, based on the edge region with uniform texture and stable contour structure in high-resolution satellite image, the local gray gradient field of the region is extracted, and the horizontal component and the vertical component of the gradient field in the horizontal direction and the vertical direction are calculated to generate a local gradient covariance matrix. Based on the local gradient covariance in the matrix and the variance of each direction corresponding to the local gradient covariance, a direction coupling degree is calculated to serve as a correlation coefficient field.

[0206] is expressed as: ;

[0207] In the formula, is the horizontal component of the local gradient, is the vertical component of the local gradient, is the local gradient covariance of the horizontal component and the vertical component, , is the variance of the horizontal component and the vertical component;

[0208] The unconstrained short axis update value The lower limit constraint is performed to generate the short axis update value;

[0209] is expressed as: ;

[0210] In the formula, is the short axis update value, is the minimum acceptable short axis scale set according to the real image diffusion width statistics, which is used to avoid the short axis scale being reduced to a physically unreasonable minimum value.

[0211] S34, a convolution kernel parameter collection process is performed on the long axis fusion value, the direction fusion angle value and the short axis update value to generate a composite elliptical Gaussian degradation convolution kernel;

[0212] is expressed as: ;

[0213] is the composite elliptical Gaussian degradation convolution, indicates that the collection process of constructing an elliptical Gaussian function is performed on the long axis fusion value, the direction fusion angle value and the short axis update value as inputs.

[0214] It should be noted that before the long axis fusion value, the direction fusion angle value and the short axis update value enter the composite elliptical Gaussian degradation convolution kernel for updating, they all need to be converted based on the imaging ground resolution to be expressed in pixel scale in the convolution kernel. The conversion logic is:

[0215] The ground resolution GSD of the satellite image is obtained, and all parameters in meters are divided by GSD to convert to pixel units.

[0216] ​S4, performing a degradation simulation operation on the high-definition aerial image by using the composite elliptical Gaussian degradation convolution kernel to generate pseudo-satellite image data, and performing pixel-level alignment on the pseudo-satellite image data and the corresponding high-definition aerial image to construct a training sample pair set;

[0217] Based on the composite elliptical Gaussian degradation convolution kernel obtained in step S3, performing a degradation simulation operation on the high-definition aerial image as a true value to generate pseudo-satellite image data with a true blur pattern, and constructing a training sample pair that can be used for deep learning network training by performing spatial coordinate corresponding analysis and pixel-by-pixel pairing processing on the pseudo-satellite image data and the high-definition aerial image.

[0218] Specifically, the logic for constructing the training sample pair is as follows:

[0219] c1, generating pseudo-satellite image data by performing convolution processing on the high-definition aerial image through the composite elliptical Gaussian degradation convolution kernel;

[0220] Obtaining the high-definition aerial image after resolution unification, denoted as: ;

[0221] wherein, and are discrete pixel coordinate indexes, and the unit is pixel. To ensure that the degradation simulation process is computable on the pixel scale, if there is a difference between the spatial resolution of the original high-definition aerial image and the ground resolution of the target satellite image, then before entering this step, perform resampling processing on the high-definition aerial image to make the pixel size consistent with the pixel size of the target satellite image. The resampling processing can use bilinear interpolation or cubic interpolation method, which will not be described here.

[0222] Denote the composite elliptical Gaussian degradation convolution kernel obtained in step S3 as: ;

[0223] wherein, is a discrete pixel displacement index inside the convolution kernel, and the size of the convolution kernel is , and are positive integers;

[0224] Performing two-dimensional discrete convolution operation on the high-definition aerial image and the composite elliptical Gaussian degradation convolution kernel to generate pseudo-satellite image data, denoted as: ;

[0225] In the formula, is the gray value of the pseudo-satellite image data at the pixel coordinate ;

[0226] It should be noted that: The pixel coordinate To maintain spatial consistency between pseudosatellite imagery data and high-resolution aerial imagery, this embodiment employs a mirror-fill strategy for image edges during convolution, i.e., when... When the image exceeds the original image range, the pixels at the boundary are mirrored about the boundary, and the expanded pixel values ​​are used as the input to the convolution, thereby avoiding the edge brightness decay caused by zero padding.

[0227] After the convolution operation is completed, pseudosatellite image data with the same size as the high-resolution aerial image is obtained. The pseudo-satellite image data maintains the same spatial resolution as the satellite image, while its blurring morphology is controlled by a composite elliptical Gaussian degenerate convolution kernel to simulate motion blur and optical diffusion in the real satellite imaging process.

[0228] c2 performs spatial coordinate correspondence analysis on pseudosatellite image data and high-resolution aerial image data to obtain alignment reference data;

[0229] In one specific embodiment, in order to build reliable training sample pairs at the pixel level, it is necessary to establish a stable spatial coordinate correspondence between pseudosatellite image data and high-resolution aerial imagery.

[0230] Since pseudosatellite imagery data is obtained by convolutional degradation of high-resolution aerial imagery, the two should ideally maintain a one-to-one correspondence on the pixel grid.

[0231] However, in engineering implementation, there may be slight pixel-level offsets due to resampling, cropping, or external preprocessing. Therefore, this step generates alignment reference data through spatial coordinate correspondence analysis to describe the overall or local pixel displacement between the two images;

[0232] Specifically, the steps for performing spatial coordinate correspondence analysis are as follows:

[0233] c21, calculates the coordinate difference based on the corresponding pixel coordinates of pseudosatellite imagery data and high-resolution aerial imagery, and generates difference segment data;

[0234] In pseudosatellite imagery and high-resolution aerial imagery, a set of sampling regions with stable texture structures is selected. These sampling regions can be divided according to a fixed-interval grid, for example, extracting a small block of region every few pixels. The center pixel coordinates of each small block of region are denoted as follows in the high-resolution aerial imagery: ;

[0235] Mark the center pixel coordinates of the corresponding small region in the pseudosatellite image data: ;

[0236] in, Index for the sampling region;

[0237] Ideally, if the two images have no additional geometric offset, then: ;

[0238] To detect and quantify potential offsets, this embodiment calculates the coordinate difference for each sampling region, expressed as: ;

[0239] In the formula, For the first The lateral difference between each sampling region For the first The longitudinal difference between each sampling region;

[0240] The coordinate differences of all sampling areas are recorded according to the area index to form difference segment data, represented as follows: ;

[0241] In the formula, For difference segment data, The total number of sampling areas represents the difference segment data used to describe the pixel displacement of pseudosatellite imagery data relative to high-resolution aerial imagery at different spatial locations.

[0242] c22, performs local consistency screening on the difference fragment data to obtain candidate alignment unit data;

[0243] Because some sampling areas may be located in low-texture areas of the image or be subject to noise interference, directly using all difference segment data for displacement estimation will lead to larger errors.

[0244] Therefore, this step performs local consistency screening on the difference fragment data to eliminate unstable regions, retain regions with consistent displacement characteristics, and generate candidate alignment unit data.

[0245] Specifically, a local analysis window is constructed centered on the spatial location of the sampling area. Several spatially adjacent sampling areas are grouped into the same local window, and the difference segment data within each local window are analyzed. Calculate the statistical measure;

[0246] Represented as: ;

[0247] ;

[0248] ;

[0249] ;

[0250] In the formula, , For window The average displacement in the horizontal and vertical directions. , This represents the degree of dispersion in the horizontal and vertical directions. This represents the number of sampling regions within the window.

[0251] When a certain local window and At the same time, when the difference is less than the dispersion threshold preset based on historical experimental data, the coordinate difference within the local window is considered to have good consistency. The local window is recorded as a stable window, and the displacement statistics corresponding to the local window are recorded as candidate alignment units.

[0252] Conversely, if the local window is not stable, it will be considered an unstable region and no candidate alignment unit will be recorded.

[0253] The average displacement information corresponding to all local windows that meet the consistency condition is recorded and denoted as: ;

[0254] In the formula, For candidate alignment unit data, For a stable set of windows that pass the local consistency screening

[0255] c23, perform coordinate mapping confirmation processing based on candidate alignment unit data to generate alignment reference data;

[0256] In one specific embodiment, in order to establish a unified spatial alignment relationship between the entire pseudosatellite image data and the high-definition aerial image, a representative global alignment displacement is calculated based on the candidate alignment unit data, and the displacement is output as alignment reference data.

[0257] Specifically, for the candidate alignment unit data set The average displacements in Weighted statistics are performed to obtain the global alignment displacement. The calculation formula is: ;

[0258] ;

[0259] In the formula, For local windows The weights, based on the local window The gradient intensity within the region is set to increase the contribution of texture-rich regions to displacement estimation;

[0260] Specifically, for local windows The horizontal and vertical gradient intensities are extracted within the local window to calculate the local window. The average gradient intensity is expressed as: ;

[0261] In the formula, is the average gradient intensity, , is the horizontal gradient intensity and the vertical gradient intensity of the pixel point , is the number of sampling regions;

[0262] The average gradient intensity is normalized, and the normalized result is taken as the weight of the local window , and is output, and the calculation formula is: ;

[0263] The global alignment displacement is taken as the alignment reference data, and the alignment reference data is used to describe the overall translation relationship of the pseudolite image data on the pixel grid relative to the high-definition aerial image.

[0264] c3, based on the alignment reference data, performs pixel-by-pixel pairing processing to generate a set of training sample pairs;

[0265] In one specific embodiment, according to the alignment reference data given by the alignment reference data , the pixel coordinates of the pseudolite image data are corrected by translation, and a one-to-one correspondence relationship between the corrected pseudolite image data and the high-definition aerial image is established at the pixel coordinate level;

[0266] Specifically, for each pixel coordinate in the pseudolite image data, the corresponding pixel coordinate in the high-definition aerial image is calculated according to the alignment reference data, and is expressed as: ;

[0267] It should be noted that when the calculated is not an integer, the pixel value of the corresponding position in the high-definition aerial image is obtained by a bilinear interpolation method, and is recorded as: ;

[0268] The pixel value of the pseudolite image data at is combined into a pair of training sample records, and all pixel coordinates in the full image range that meet the boundary conditions are paired and processed as described above, and a set of training sample pairs can be constructed, and is expressed as:

[0269] ;

[0270] In the formula, is the set of training sample pairs, is the set of pixel coordinates that are still in the image effective area after the alignment displacement transformation.

[0271] S5, performing parameter iterative training of the deep learning network based on the set of training sample pairs to generate a satellite image sharpening processing model;

[0272] In one specific embodiment, the training sample pairs generated by the above-mentioned manner are used to train the model by taking the pseudo-satellite image data as input images and taking the high-definition aerial images as target images, so as to generate the satellite image sharpening processing model.

[0273] Specifically, the step of generating the satellite image sharpening processing model includes:

[0274] S51, dividing the training sample pairs into a sharpening processing training set and a sharpening processing verification set, wherein the training sample pairs include pseudo-satellite image data and corresponding high-definition aerial images;

[0275] S52, constructing a deep learning network, taking the pseudo-satellite image data in the sharpening processing training set as input of the deep learning network, taking the corresponding high-definition aerial images as output of the deep learning network, performing parameter iterative training on the deep learning network to obtain an initial satellite image sharpening processing network;

[0276] It should be noted that the parameter iterative training includes but is not limited to forward calculation, difference measurement calculation and gradient update processing, and the deep learning network includes but is not limited to an attention-enhanced network or a Transformer image reconstruction network.

[0277] S53, verifying the initial satellite image sharpening processing network by using the sharpening processing verification set, and outputting the initial satellite image sharpening processing network with a verification error less than or equal to a preset matching error threshold as a pre-constructed satellite image sharpening processing model.

[0278] The above embodiments are only used to illustrate the technical method of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical method of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical method of the present application.

Claims

1. A method for improving the sharpness of satellite images based on an elliptical Gaussian distribution, characterized in that, The method includes: S1, calculate and generate motion data based on the collected physical imaging parameters, and construct a reference motion vector based on the motion data. The motion data includes the major axis reference value and the reference blur direction angle. S2, construct a constraint space based on the statistical characteristics of real satellite images, and collect random degradation parameters from the constraint space. The random degradation parameters include major axis perturbation term, orientation angle jitter term, and minor axis standard deviation. S3 fuses the baseline motion vector with random degradation parameters to generate a composite elliptic Gaussian degenerate convolution kernel; S4 performs degradation simulation operations on high-resolution aerial images using a compound elliptical Gaussian degradation convolution kernel to generate pseudo-satellite image data. The pseudo-satellite image data is then pixel-level aligned with the corresponding high-resolution aerial images to construct a training sample pair set.

2. The method for improving satellite image sharpness based on an elliptical Gaussian distribution according to claim 1, characterized in that, The physical imaging parameters include satellite orbital attitude angle, ground projection velocity, imaging scanning direction, and exposure integration time; the statistical features include the edge gradient histogram of the real satellite image, atmospheric disturbance standard deviation statistics, and point spread function half-width value.

3. The method for improving satellite image sharpness based on an elliptic Gaussian distribution according to claim 2, characterized in that, The steps for constructing a baseline motion vector based on motion data are as follows: S11 generates physical imaging parameter records by performing parameter recording processing based on satellite orbital attitude angle, ground projection velocity, imaging scanning direction, and exposure integration time. S12, perform multiplication calculation based on the ground projection velocity and exposure integration time recorded in the physical imaging parameters to generate the major axis reference value; S13, Perform angle analysis processing on the imaging scanning direction in the physical imaging parameter record to generate the reference blur direction angle; S14, perform vector construction operation based on the major axis reference value and the reference fuzzy direction angle to generate the reference motion vector.

4. The method for improving satellite image sharpness based on an elliptical Gaussian distribution according to claim 3, characterized in that, The logic for performing angle parsing and processing is as follows: a1, performs spatial orientation decomposition processing through the imaging scanning direction to generate orientation decomposition quantities; a2, performs angular offset estimation calculation based on the imaging scanning direction, and generates the angular offset amount; a3 performs directional noise suppression calculations based on the imaging scanning direction, generating the noise suppression amount; a4 performs angle fusion processing based on the direction decomposition amount, angle offset amount and noise suppression amount to generate the reference fuzzy direction angle.

5. The method for improving satellite image sharpness based on an elliptic Gaussian distribution according to claim 2, characterized in that, The steps for constructing a constrained space based on the statistical features of real satellite imagery include: S21, perform segment gradient distribution determination based on edge gradient histogram, and generate gradient segment data; S22, perform disturbance amplitude classification processing based on atmospheric disturbance standard deviation statistics to generate disturbance category data; S23, perform diffusion range interval division based on the half-width and height values ​​of the point diffusion function to generate diffusion interval data; S24 performs spatial domain hierarchical processing using gradient segment data, disturbance category data, and diffusion interval data to construct a constraint space.

6. The method for improving satellite image sharpness based on an elliptical Gaussian distribution according to claim 5, characterized in that, The steps for generating gradient segment data are as follows: S211, perform amplitude density analysis on each gradient amplitude segment based on the edge gradient histogram to generate density abrupt change segment records; S212, perform amplitude segment boundary determination based on density abrupt change segment records to obtain amplitude boundary set; S213 performs segmentation processing by using the amplitude distribution of the amplitude boundary set and the edge gradient histogram to generate gradient segment data.

7. The method for improving satellite image sharpness based on an elliptical Gaussian distribution according to claim 6, characterized in that, The logic for performing amplitude density analysis is as follows: b1, based on the amplitude segment frequency change rate of the edge gradient histogram, perform segment gradient change rate analysis to obtain change rate fluctuation data; b2, based on the number of consecutive reversals of adjacent rate of change signs in the rate of change fluctuation data, perform a stability determination of the amplitude segment to obtain a set of unstable amplitude segments; b3 performs abrupt segment filtering by using the frequency gradient difference between the unstable amplitude segment set and the corresponding amplitude segment in the edge gradient histogram to generate density abrupt segment records.

8. The method for improving satellite image sharpness based on an elliptic Gaussian distribution according to claim 1, characterized in that, The generation logic of the compound elliptic Gaussian degenerate convolution kernel is as follows: S31, perform amplitude superposition processing on the long axis reference value and the long axis disturbance term to generate the long axis fused value; S32, perform angle correction processing based on the reference fuzzy direction angle and direction angle jitter term to obtain the direction fusion angle value; S33, perform axial parameter update based on the minor axis standard deviation and correlation coefficient fields, and generate minor axis update value; S34 performs convolution kernel parameter aggregation processing by combining the major axis fusion value, the direction fusion angle value, and the minor axis update value to generate a composite elliptical Gaussian degenerate convolution kernel.

9. The method for improving satellite image sharpness based on an elliptical Gaussian distribution according to claim 1, characterized in that, The logic for constructing training sample pairs is as follows: c1 generates pseudo-satellite image data by performing convolution processing with high-resolution aerial imagery using a compound elliptical Gaussian degenerate convolution kernel; c2 performs spatial coordinate correspondence analysis on pseudosatellite image data and high-resolution aerial image data to obtain alignment reference data; c3 performs pixel-by-pixel pairing based on alignment reference data to generate a set of training sample pairs.

10. The method for improving satellite image sharpness based on an elliptic Gaussian distribution according to claim 9, characterized in that, The steps for performing spatial coordinate correspondence analysis are as follows: c21, calculates the coordinate difference based on the corresponding pixel coordinates of pseudosatellite imagery data and high-resolution aerial imagery, and generates difference segment data; c22, performs local consistency screening on the difference fragment data to obtain candidate alignment unit data; c23 performs coordinate mapping confirmation processing based on candidate alignment unit data to generate alignment reference data.

Images