Method and system for correcting aero-optical effects based on sub-region restoration

Abstract

The application discloses a kind of based on the restoration of sub-region aerodynamic optical effect correction method and system, method includes steps: obtaining the image affected by aerodynamic optical effect;Image is preprocessed to be denoised and edge enhanced;The turbulence intensity of each pixel point after pre-processing is calculated, the image is segmented into different regions using threshold segmentation method;Frequency feature extraction is carried out on different regions after segmentation to represent its degradation characteristics, according to the proportion of high frequency energy in total energy of each region, the region is divided into high frequency region and low frequency region, and the key region and non-key region in high frequency region are identified, different methods are used for blur kernel estimation in key region, non-key region and low frequency region;The obtained blur kernel is deconvolved, and the restored image of the whole image is obtained.The application improves the restoration quality, improves the restoration speed, has good robustness, and is an effective space-variant image restoration method.

Description

Technical Field

[0001] This invention relates to the field of image processing, and in particular to a method and system for correcting aero-optical effects based on region-by-region restoration. Background Technology

[0002] Aero-optical effects refer to the complex flow field formed between the optical radome and the incoming airflow when a high-speed aircraft equipped with an optical imaging detection system flies within the atmosphere. This flow field causes thermal diffusion, thermal radiation, and transmission interference to optical imaging, resulting in phenomena such as image saturation, blurring, and jitter. This significantly impacts target recognition and guidance. Although optimizing the radome design and using adaptive optics can reduce the impact of aero-optical effects, it cannot eliminate them entirely. Therefore, research on post-imaging image correction techniques is crucial. Consequently, research on aero-optical effect correction projects plays a powerful role in promoting the high-end development of high-speed aircraft in my country.

[0003] Traditional global correction methods have significant limitations when processing aerodynamically degraded images. These methods typically assume that the degradation is uniform across the entire image and use a uniform correction model for processing. However, in real-world aerodynamically degraded images, the degree and manner in which different regions are affected by aero-optical effects vary significantly. Therefore, it is necessary to research aero-optical effect correction methods based on region-specific restoration. Summary of the Invention

[0004] The main objective of this invention is to provide a method and system for correcting aero-optical effects based on regional restoration. This is a new approach and method that can more accurately match the degradation characteristics of different regions and achieve more effective image restoration.

[0005] The technical solution adopted in this invention is:

[0006] A method for correcting aero-optical effects based on regional restoration is provided, comprising the following steps:

[0007] S1. Acquire images affected by aero-optical effects;

[0008] S2. Perform preprocessing on the image, including denoising and edge enhancement;

[0009] S3. Calculate the turbulence intensity of each pixel in the preprocessed image, and divide the image into different regions using the threshold segmentation method;

[0010] S4. Extract frequency features from different regions after segmentation to characterize their degradation characteristics. Divide the regions into high-frequency regions and low-frequency regions according to the proportion of high-frequency energy in each region to the total energy. Identify key and non-key regions in the high-frequency regions. Use different methods to perform fuzzy kernel estimation for key regions, non-key regions and low-frequency regions.

[0011] S5. Perform deconvolution operation on the obtained blur kernel to obtain the restored image of the whole image.

[0012] Following the above technical solution, the specific steps for dividing the image into different regions in step S3 are as follows: Select points in the image with significantly different turbulence intensities as seed points, calculate whether the difference between the turbulence intensity of each seed point and its neighboring pixels exceeds a preset gradient threshold. If so, merge the neighboring pixels into the region where the current seed point is located; otherwise, divide the neighboring pixels into another region. Repeat this process until there are no neighboring pixels that meet the conditions, thus completing the growth and segmentation of the region.

[0013] Following the above technical solution, L1 regularization and iterative optimization are used to perform fuzzy kernel estimation for critical regions; kernel interpolation estimation is used to perform fuzzy kernel estimation for non-critical regions; and blind deconvolution optimization algorithm is used to perform fuzzy kernel estimation for low-frequency distortion regions.

[0014] Following the above technical solution, the fuzzy kernel estimation for key regions is specifically performed using L1 regularization and iterative optimization as follows:

[0015] A local degradation model is established for the high-frequency region, and L1 regularization is introduced to constrain the sparsity properties of the fuzzy kernel.

[0016] Based on the local degradation model and the sparsity of the fuzzy kernel, the optimization problem is transformed into an augmented Lagrangian form using the alternating direction multiplier method.

[0017] Iterative updates yield the final fuzzy kernel estimate.

[0018] Following the above technical solution, the fuzzy kernel estimation for non-critical areas using kernel interpolation is specifically as follows:

[0019] Initialize the kernel seed at the CFD mesh node to complete coarse-layer interpolation;

[0020] Optical flow is fine-tuned at the pixel level, specifically by calculating interpolation weights based on streamline distances to complete fine-layer corrections.

[0021] Following the above technical solution, step S2 specifically employs bilateral filtering for noise reduction.

[0022] Following the above technical solution, the edge enhancement in step S2 specifically involves:

[0023] Calculate the gradient magnitude and direction of the denoised image;

[0024] Non-maximum suppression is applied to the gradient magnitude to remove non-edge pixels;

[0025] Edge pixels are divided into strong edge pixels and weak edge pixels using a dual threshold detection method;

[0026] If a weak edge pixel is connected to a strong edge pixel, it is retained as the final edge.

[0027] Following the above technical solution, the specific calculation process for turbulence intensity in step S3 is as follows:

[0028] The velocity parameters of the flow field are directly measured using particle image velocimetry.

[0029] The turbulence intensity of each pixel is calculated based on the velocity parameters of the flow field and the turbulence intensity formula.

[0030] The present invention also provides an aero-optical effect correction system based on regional restoration, comprising:

[0031] The image acquisition module is used to acquire images affected by aero-optical effects;

[0032] The preprocessing module is used to perform noise reduction and edge enhancement on the image.

[0033] The region segmentation module is used to calculate the turbulence intensity of each pixel in the preprocessed image and uses a threshold segmentation method to divide the image into different regions.

[0034] The partitioned fuzzy kernel estimation module is used to extract frequency features from different segmented regions to characterize their degradation characteristics. Based on the proportion of high-frequency energy in each region to the total energy, the region is divided into high-frequency and low-frequency regions. Key and non-key regions in the high-frequency region are identified, and different methods are used for fuzzy kernel estimation for key, non-key, and low-frequency regions.

[0035] The restoration module is used to perform deconvolution operations on the obtained blur kernel to obtain the restored image of the whole image.

[0036] The present invention also provides a computer storage medium storing a computer program executable by a processor, the computer program performing the aero-optical effect correction method based on regional restoration described in the above technical solution.

[0037] The beneficial effects of this invention are as follows: Based on a region-based aero-optical effect correction method, this invention addresses the limitations of traditional global correction methods by segmenting regions according to turbulence intensity. This allows for more accurate matching of degradation characteristics in different regions, resulting in more effective image restoration. Furthermore, it solves the block effect problem caused by fixed grid division, improving restoration quality and speed, and exhibits good robustness, making it an effective method for spatially varied image restoration. In addition, employing different methods for fuzzy kernel estimation in key regions, non-key regions, and low-frequency regions reduces computational load and improves restoration efficiency.

[0038] Furthermore, when segmenting an image into different regions, points with significantly different turbulence intensities are specifically selected as seed points. Then, the difference in turbulence intensities between adjacent pixels is used to determine whether pixels belong to the same region, thereby improving the accuracy of region segmentation.

[0039] Furthermore, L1 regularization combined with iterative optimization is used to estimate the fuzzy kernel in key regions of the high-frequency region, where L1 regularization forces the fuzzy kernel to be optimized. The sparsity of the kernel suppresses artifacts caused by noise and prevents the kernel estimate from becoming too diffuse. Iterative optimization can gradually refine the kernel from coarse to fine, avoiding getting trapped in local optima. Combining the two can make the results more accurate and robust. Since L1 regularization is a non-smooth term, it is difficult to optimize directly. Therefore, the Alternating Direction Multiplier Method (ADMM) is used to transform the optimization problem into an augmented Lagrangian form.

[0040] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of an embodiment of the aero-optical effect correction method based on regional restoration according to the present invention;

[0043] Figure 2 This is a flowchart of another embodiment of the aero-optical effect correction method based on regional restoration of the present invention;

[0044] Figure 3 This is a blurred image obtained by an embodiment of the present invention, which is affected by aero-optical effects;

[0045] Figure 4 yes Figure 3 A schematic diagram after regional division;

[0046] Figure 5 This is a schematic diagram of the fuzzy kernel in each region;

[0047] Figure 6 yes Figure 5 The restored image. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] It should be noted that the illustrations provided in the embodiments of the present invention are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0050] In this invention, it should also be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "first" and "second" are used only for descriptive and distinguishing purposes and should not be construed as indicating or implying relative importance.

[0051] Furthermore, it should be noted that the features of the various embodiments of the present invention can be combined or integrated in whole or in part, and as those skilled in the art will understand, they can interact and operate in different ways. Each embodiment can be implemented independently of each other or in association with one another.

[0052] The aero-optical effect correction method based on regional restoration of the present invention can be written in C / C++.

[0053] like Figure 1 As shown, the aero-optical effect correction method based on regional restoration in this embodiment of the invention mainly includes the following steps:

[0054] S1. Acquire images affected by aero-optical effects;

[0055] S2. Perform preprocessing on the image, including denoising and edge enhancement;

[0056] S3. Calculate the turbulence intensity of each pixel in the preprocessed image, and use the threshold segmentation method to segment the image into different regions;

[0057] S4. Extract frequency features from different regions after segmentation to characterize their degradation characteristics. Divide the regions into high-frequency regions and low-frequency regions according to the proportion of high-frequency energy in each region to the total energy. Identify key and non-key regions in the high-frequency regions. Use different methods to perform fuzzy kernel estimation for key regions, non-key regions and low-frequency regions.

[0058] S5. Perform deconvolution operation on the obtained blur kernel to obtain the restored image of the whole image.

[0059] As can be seen, this embodiment, based on a region-based aero-optical effect correction method, addresses the limitations of traditional global correction methods by segmenting regions according to turbulence intensity. This allows for more accurate matching of degradation characteristics in different regions, thus achieving more effective image restoration. Furthermore, employing different methods for fuzzy kernel estimation in critical, non-critical, and low-frequency regions reduces computational load and improves restoration efficiency.

[0060] Furthermore, the specific steps for dividing the image into different regions in step S3 are as follows: select points in the image with significantly different turbulence intensities as seed points, calculate whether the difference between the turbulence intensity of each seed point and its neighboring pixels exceeds a preset gradient threshold. If so, merge the neighboring pixels into the region where the current seed point is located; otherwise, divide the neighboring pixels into another region. Repeat this process until there are no neighboring pixels that meet the conditions, thus completing the growth and segmentation of the region.

[0061] Furthermore, L1 regularization and iterative optimization are used for fuzzy kernel estimation in critical regions; kernel interpolation estimation is used for fuzzy kernel estimation in non-critical regions; and blind deconvolution optimization algorithm is used for fuzzy kernel estimation in low-frequency distortion regions. L1 regularization can force the fuzzy kernel to be optimized. The sparsity of the kernel suppresses artifacts caused by noise and avoids excessive diffusion in kernel estimation. Iterative optimization can gradually refine the kernel from coarse to fine, avoiding getting trapped in local optima. Combining the two can make the results more accurate and robust.

[0062] Furthermore, since L1 regularization is a non-smooth term and is difficult to optimize directly, the Alternating Direction Multiplier Method (ADMM) can be used to transform the optimization problem into an augmented Lagrangian form. Therefore, the specific steps for estimating the fuzzy kernel in key regions using L1 regularization and iterative optimization are as follows: Establish a local degradation model for high-frequency regions and introduce L1 regularization to constrain the sparsity of the fuzzy kernel; based on the local degradation model and the sparsity of the fuzzy kernel, use the Alternating Direction Multiplier Method to transform the optimization problem into an augmented Lagrangian form; iteratively update to obtain the final fuzzy kernel estimate.

[0063] Furthermore, the fuzzy kernel estimation using kernel interpolation for non-critical regions is specifically as follows: the kernel seed is initialized at the CFD (Computational Fluid Dynamics) grid nodes to complete coarse interpolation; optical flow fine-tuning is performed at the pixel-level grid, specifically by calculating the interpolation weights based on streamline distance to complete fine-level correction.

[0064] Furthermore, step S2 specifically employs bilateral filtering for noise reduction.

[0065] Further, the edge enhancement in step S2 specifically involves: calculating the gradient magnitude and direction of the denoised image; performing non-maximum suppression on the gradient magnitude to remove non-edge pixels; dividing edge pixels into strong edge pixels and weak edge pixels using a dual threshold detection method; and retaining weak edge pixels as final edges if they are connected to strong edge pixels.

[0066] Furthermore, the specific calculation process for turbulence intensity in step S3 is as follows: the velocity parameters of the flow field are directly measured using particle image velocimetry technology.

[0067] In another embodiment of the present invention, the turbulence intensity of each pixel can be calculated based on the velocity parameters of the flow field and the turbulence intensity formula. For example... Figure 2 As shown, based on the above method embodiments, taking a specific image input as an example, this embodiment of the aero-optical effect correction method based on region restoration includes the following steps:

[0068] Step 1: Obtain the blurred image affected by aero-optical effects, such as Figure 3 As shown, its size is 256×256;

[0069] Step 2: Use bilateral filtering for noise reduction.

[0070] Pixel values ​​after bilateral filtering It can be represented as:

[0071]

[0072] in, For spatial domain weights, This represents the range weight.

[0073] Calculate the gradient magnitude and direction of the denoised image. The gradients in the horizontal and vertical directions are: and .

[0074] gradient magnitude is .

[0075] Non-maximum suppression is applied to the gradient magnitude to remove non-edge pixels.

[0076] Non-maximum suppression in the above steps can be achieved through interpolation comparison, as shown in the following equation:

[0077]

[0078] in, Along the gradient direction The gradient magnitude of adjacent pixels obtained by linear interpolation. If the current point gradient magnitude Greater than or equal to the gradient magnitude of adjacent points obtained by interpolation along both sides of its gradient direction If the gradient magnitude is positive, the magnitude of the point is retained, meaning it is likely a local maximum, i.e., a potential edge point; otherwise, the gradient magnitude of the point is set to 0, meaning it is suppressed and is not a local maximum.

[0079] Specifically, edge pixels can be divided into strong edge pixels and weak edge pixels using a dual-threshold detection method, and then non-edge pixels are removed according to a preset edge connection rule. The threshold segmentation conditions for dual-threshold detection and edge connection are shown in the following formula:

[0080]

[0081] Its edge connection rule is that if a weak edge pixel is connected to a strong edge, it is retained as an edge. Double threshold segmentation is used, and the final edge is determined by double threshold detection and edge connection.

[0082] Step 3: Obtain velocity parameters through actual measurements, and then calculate the turbulence intensity of each pixel.

[0083] Experimental measurements can directly measure the velocity parameters of the flow field using particle image velocimetry (PIV) technology;

[0084] Particle Image Velocimetry (PIV):

[0085] The cross-correlation function of the particle patterns between two frames is:

[0086]

[0087] Peak position corresponds to average displacement ( The flow field velocity is calculated as follows:

[0088]

[0089] in For pulse interval, This is the magnification factor.

[0090] The formula for calculating turbulent kinetic energy is:

[0091]

[0092] in, , and The mean square values ​​of velocity fluctuations in the three directions are used to calculate the turbulence intensity of each pixel.

[0093]

[0094] in, , , This represents the component fluctuation value of the turbulent velocity. This is the average speed over time.

[0095] Furthermore, after obtaining the turbulence intensity of each pixel, the turbulence intensity gradient of each pixel is calculated.

[0096] Turbulence intensity gradient Reflecting its spatial rate of change, the formula is:

[0097]

[0098] in:

[0099]

[0100] Similarly, the calculation is obtained.

[0101] Step 4: Based on the calculated turbulence intensity gradient, the image is divided into different regions using a threshold segmentation method. First, points with significantly different turbulence intensities are selected as seed points. Then, for each seed point, the turbulence intensities of its neighboring pixels are checked to see if they are within a certain range. If they are, the neighboring pixels are merged into the current region. A suitable gradient threshold T is set. When the turbulence intensity gradient of a pixel is greater than the threshold T, that pixel and its neighboring pixels are divided into one region. This region typically corresponds to areas with large variations in turbulence intensity, such as shock wave regions or certain parts of the boundary layer. When the turbulence intensity gradient of a pixel is less than the threshold T, it is divided into another region, which typically corresponds to areas with relatively stable turbulence intensity. This process is repeated until there are no neighboring pixels that meet the conditions, thus completing region growth and segmentation. Figure 4 As shown.

[0102] Define turbulence intensity characteristic parameters:

[0103]

[0104] in For wavefront phase, For The local window is centered. The region growth and merging condition is:

[0105]

[0106] In the above formula, , Let K be the mean and standard deviation of the turbulence intensity in the current region. This is the tolerance factor.

[0107] Step 5: Calculate the proportion of high-frequency energy in the region to the total energy, and determine whether the region is a high-frequency region or a low-frequency region.

[0108] Let the frequency spectrum of the region be Its high-frequency energy for:

[0109]

[0110] The total energy is:

[0111]

[0112] in, Here, (U, V) represents the starting frequency coordinates of the high-frequency region, and (U, V) represents the maximum coordinates in the frequency domain. The proportion of high-frequency energy to the total energy is also considered. for:

[0113]

[0114] In this example, it is believed that A value greater than 0.7 indicates a high-frequency region.

[0115] Furthermore, key and non-key regions within the high-frequency area are identified. The specific identification method is as follows:

[0116] image In position The gradient magnitude at that point is , defined by Centered on, size is Local window :

[0117] The average gradient magnitude is:

[0118]

[0119] The gradient magnitude variance is:

[0120]

[0121] The judgment criteria are:

[0122] like ,and If it is a critical area, then it is a critical area; otherwise, it is a non-critical area.

[0123] in , These are 25% of the average gradient magnitude and the variance of the gradient magnitude across the entire map, respectively.

[0124] Key regions employ L1 regularization combined with iterative optimization kernel estimation. L1 regularization forces the fuzzy kernel to be optimized. The sparsity of the kernel suppresses artifacts caused by noise and avoids excessive diffusion in kernel estimation. Iterative optimization can gradually refine the kernel from coarse to fine, avoiding getting trapped in local optima. Combining the two can make the results more accurate and robust.

[0125] For key areas, establish a local degradation model:

[0126]

[0127] in:

[0128] Images of key regions with observed high-frequency distortion;

[0129] The local fuzzy kernel to be estimated;

[0130] : Potentially clear image;

[0131] Key region index set;

[0132] Introducing L1 regularization to constrain the fuzzy kernel and ensure its sparsity:

[0133]

[0134] in: Let be the objective function. The square of the L2 norm. The L1 regularization term forces the fuzz kernel K to be sparsified (to suppress noise and focus on the main non-zero regions). As weight, It is a total variation (TV) regularization term that guarantees the smoothness and continuity of the fuzzy kernel K.

[0135] L1 regularization is a non-smooth term, making direct optimization difficult. Therefore, the Alternating Direction Multiplier Method (ADMM) is used to transform the optimization problem into an augmented Lagrangian form:

[0136]

[0137] in:

[0138] For gradient operators;

[0139] To observe the gradient;

[0140] K is the fuzzy kernel;

[0141] Z is an auxiliary variable;

[0142] U is the dual variable;

[0143] The iterative update steps, the fuzzy kernel is updated as follows:

[0144]

[0145] Auxiliary variable update:

[0146]

[0147] For non-critical regions, an interpolation fuzzy kernel is obtained by constructing a kernel propagation function based on flow field characteristics:

[0148]

[0149] in:

[0150] ,

[0151] For interpolation weights based on streamline distance, This is the streamline distance.

[0152] Here is the flow field modulation matrix;

[0153] The kernel shape is adapted to the Hadamard product.

[0154] In the low-frequency region, blind restoration is used directly to obtain the fuzzy kernel of the sub-region, reducing the amount of computation and improving the restoration efficiency.

[0155] Furthermore, by combining the fuzzy kernels of different sub-regions, a full-image fuzzy kernel matrix is ​​obtained, such as... Figure 5 As shown.

[0156] Step 5: Obtain the restored image by deconvolving the obtained full-image blur kernel, such as... Figure 6 As shown.

[0157] To implement the above-described method embodiments, the present invention also provides an aero-optical effect correction system based on regional restoration, comprising:

[0158] The image acquisition module is used to acquire images affected by aero-optical effects;

[0159] The preprocessing module is used to perform noise reduction and edge enhancement on the image.

[0160] The region segmentation module is used to calculate the turbulence intensity of each pixel in the preprocessed image and uses a threshold segmentation method to divide the image into different regions.

[0161] The partitioned fuzzy kernel estimation module is used to extract frequency features from different segmented regions to characterize their degradation characteristics. Based on the proportion of high-frequency energy in each region to the total energy, the region is divided into high-frequency and low-frequency regions. Key and non-key regions in the high-frequency region are identified, and different methods are used for fuzzy kernel estimation for key, non-key, and low-frequency regions.

[0162] The restoration module is used to perform deconvolution operations on the obtained blur kernel to obtain the restored image of the whole image.

[0163] Each module is specifically used to implement the various steps in the methods of the above embodiments, which will not be elaborated here.

[0164] This application also provides a computer-readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, server, App application store, etc., which stores a computer program, and the program implements corresponding functions when executed by a processor. The computer-readable storage medium of this embodiment implements the aero-optical effect correction method based on regional restoration in the method embodiment when executed by a processor.

[0165] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0166] The sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0167] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for correcting the effects of aero-optics based on sub-area recovery, characterized in that, The method comprises the following steps: S1, acquiring an image affected by an aero-optical effect; S2, performing preprocessing of denoising and edge enhancement on the image; S3, calculating the turbulence intensity of each pixel point of the preprocessed image, and using a threshold segmentation method to segment the image into different regions; S4, extracting frequency characteristics of the segmented different regions to represent their degradation characteristics, dividing the regions into high-frequency regions and low-frequency regions according to the proportion of high-frequency energy in total energy of each region, and identifying key regions and non-key regions in the high-frequency regions, and using different methods to estimate blur kernels for the key regions, the non-key regions and the low-frequency regions; S5, performing deconvolution operation on the obtained blur kernel to obtain a restored image of the whole image; In step S3, the specific steps of segmenting the image into different regions are as follows: selecting points with obviously different turbulence intensities in the image as seed points, calculating whether the difference between the turbulence intensity of each seed point and that of the adjacent pixel exceeds a preset gradient threshold, if yes, merging the adjacent pixel into the region where the current seed point is located, if not, dividing the adjacent pixel into another region, and repeatedly performing the process until there is no adjacent pixel meeting the condition, thereby completing the growth and segmentation of the region; In step S4, the L1 regularization and iterative optimization are used for blur kernel estimation for the key regions, the kernel interpolation estimation is used for blur kernel estimation for the non-key regions, and the blind deconvolution optimization algorithm is used for blur kernel estimation for the low-frequency distortion regions.

2. The method of claim 1, wherein, The L1 regularization and iterative optimization are used for blur kernel estimation for the key regions, and the specific steps are as follows: A local degradation model is established for the high-frequency region, and L1 regularization is introduced to constrain the sparse characteristics of the blur kernel; According to the local degradation model and the sparse characteristics of the blur kernel, the alternating direction multiplier method is used to convert the optimization problem into an augmented Lagrangian form; Iterative updating is performed to obtain the final blur kernel estimation.

3. The method of claim 1, wherein the method is a method of correcting a pneumatic optical effect based on a sub-area restoration. The kernel interpolation estimation is used for blur kernel estimation for the non-key regions, and the specific steps are as follows: The kernel seed is initialized at the CFD grid node to complete coarse layer interpolation; Pixel-level grid is used for flow adjustment, and the interpolation weight is calculated based on the stream distance to complete fine layer correction.

4. The method of claim 1, wherein the method is a method of correcting a pneumatic optical effect based on a sub-area restoration. In step S2, the bilateral filtering denoising is used.

5. The method of claim 1, wherein the method is a method of correcting a pneumatic optical effect based on a sub-area restoration. In step S2, the edge enhancement is as follows: The gradient amplitude and direction of the denoised image are calculated; Non-maximum suppression is performed on the gradient amplitude to remove non-edge pixels; The double-threshold detection method is used to divide the edge pixels into strong edge pixels and weak edge pixels; If the weak edge pixels are connected with the strong edge pixels, they are retained as the final edge.

6. The method according to any one of claims 1-5, wherein, In step S3, the specific calculation process of the turbulence intensity is as follows: The particle image velocimetry technology is used to directly measure the velocity parameters of the flow field; The turbulence intensity of each pixel point is calculated according to the velocity parameters of the flow field and the turbulence intensity formula.

7. A system for correction of aero-optical effects based on subregion recovery, characterized in that The method comprises the following steps: An image acquisition module is configured to acquire an image affected by an aero-optical effect; A preprocessing module is configured to perform preprocessing of denoising and edge enhancement on the image; The region division module is configured to calculate the turbulence intensity of each pixel point of the preprocessed image, and divide the image into different regions by using a threshold segmentation method; specifically, points with obviously different turbulence intensity in the image are selected as seed points, and whether the difference between the turbulence intensity of each seed point and that of an adjacent pixel exceeds a preset gradient threshold is calculated; if yes, the adjacent pixel is merged into the region where the current seed point is located; if no, the adjacent pixel is divided into another region; the process is repeatedly performed until there is no adjacent pixel meeting the condition, thereby completing the growth and division of the region; The partition blur kernel estimation module is configured to extract the frequency characteristics of the divided different regions to represent the degradation characteristics thereof, divide the regions into high-frequency regions and low-frequency regions according to the proportion of the high-frequency energy of each region to the total energy, and identify key regions and non-key regions in the high-frequency regions, and estimate the blur kernel of the key regions, the non-key regions and the low-frequency regions by using different methods; specifically, the L1 regularization and iterative optimization are used to estimate the blur kernel of the key regions; the kernel interpolation estimation is used to estimate the blur kernel of the non-key regions; and the blind deconvolution optimization algorithm is used to estimate the blur kernel of the low-frequency distortion regions. The restoration module is configured to perform deconvolution operation on the obtained blur kernel to obtain a restored image of the whole image.

8. The system for correction of aero-optical effects based on subarea restoration according to claim 7, characterized in that, The preprocessing module specifically adopts the bilateral filter denoising.

9. The system for correction of aero-optical effects based on subarea restoration according to claim 7, characterized in that, The specific calculation process of the turbulence intensity in the region division module is as follows: the particle image velocimetry technology is used to directly measure the velocity parameters of the flow field; and the turbulence intensity of each pixel point is calculated according to the velocity parameters of the flow field and the turbulence intensity formula.

10. A computer storage medium, characterized in that, The computer program stored therein can be executed by the processor, and the computer program performs the aerodynamic optical effect correction method based on the region restoration according to any one of claims 1-6. The computer program stored therein can be executed by the processor, and the computer program performs the aerodynamic optical effect correction method based on the region restoration according to any one of claims 1-6.

Images