Optical lens splicing imaging method and system

Abstract

The application relates to the technical field of optical imaging, and discloses an optical lens splicing imaging method and system, the system comprising a calibration module, a preprocessing module, a registration module, a correction module and a fusion module; the registration module, the correction module and the fusion module are arranged, a mixed strategy of global mutual information guidance and local robust matching is used, registration error problems caused by sparse feature points in low-texture areas are solved, the registration precision is effectively improved, a global model based on a Zernike polynomial is used to compensate optical aberration of each sub-lens, the image quality of an edge area of a spliced image is effectively improved, a multi-resolution pyramid optimization mode is used to reduce the calculation complexity of global registration, real-time imaging requirements are met, and a splicing seam is effectively eliminated, and the overall smoothness of an image is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of optical imaging technology, and more specifically to an optical lens stitching imaging method and system. Background Technology

[0002] With the continuous development of science and technology, optical imaging technology has been widely used in various fields, such as remote sensing, medicine, and monitoring. However, traditional single-lens imaging systems have certain limitations in terms of resolution and field of view. A single optical lens is limited by physical size (such as aperture and focal length) and manufacturing process, making it difficult to simultaneously meet the requirements of a large field of view and high resolution. For example, astronomical telescopes need to cover a wide area of ​​the sky to capture rare celestial events, while also resolving details of distant stars. Remote sensing satellites need to perform high-resolution imaging of large areas of the Earth's surface to support resource exploration and environmental monitoring. Existing optical lens stitching imaging technology forms a large-field-of-view, high-resolution overall image by arranging multiple small-field-of-view, high-resolution sub-lenses in an array and stitching them together.

[0003] However, existing optical lens stitching imaging technology still has some shortcomings. The optical aberrations of each sub-lens exist independently and are not corrected through global optimization, resulting in a significant decrease in image quality in the edge areas of the stitched image. Furthermore, traditional feature point-based registration methods are sensitive to low-texture areas, which can easily lead to the accumulation of spatial transformation parameter errors between sub-images, resulting in ghosting or misalignment in the final stitched image. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides an optical lens stitching imaging method and system to solve the problems existing in the background art.

[0005] This invention provides the following technical solution: an optical lens stitching imaging method, comprising the following steps:

[0006] Step 1: Sub-lens parameter calibration; obtain the internal parameters and distortion model of each sub-lens, and establish the mapping relationship between the sub-lens and the global coordinate system;

[0007] Step 2: Sub-image acquisition and preprocessing; Original images are acquired synchronously through each sub-lens, and the original images acquired by each sub-lens are marked as sub-images. Denoising, preliminary distortion correction, and grayscale normalization are performed on each sub-image.

[0008] Step 3: Adaptive feature registration; Based on improved mutual information metric and robust feature matching, obtain the transformation parameters of each sub-image relative to the global coordinate system;

[0009] Step 4: Pixel-based collaborative correction; Construct a global pixel model and use optimized algorithms to collaboratively compensate for optical aberrations in each sub-lens;

[0010] Step 5: Multi-strategy fusion optimization; a strategy combining weighted average fusion and multi-resolution pyramid fusion is adopted to obtain the final fused image.

[0011] Preferably, the establishment of the mapping relationship between the sub-lens and the global coordinate system specifically involves:

[0012] An improved Zhang Zhengyou calibration method was adopted, using a checkerboard calibration board as a reference. By taking N images of the calibration board at different angles and positions, the corner coordinates were extracted and subpixel-level positioning was performed.

[0013] Let the image coordinates of the a-th sub-shot be (y a ,v a The coordinates in the world coordinate system are (X... ω ,Y ω Z ω In this embodiment, the world coordinate system is the calibration plate coordinate system. Since the calibration plate is a plane in the calibration plate coordinate system, the Z-axis is taken as... ω If = 0, then the mapping relationship can be expressed as:

[0014] Where k represents the scaling factor, Let f be the intrinsic parameter matrix. x f y c is the focal length of a pixel. x c y The coordinates of the principal point of the sub-lens image are the offsets from the origin in the pixel coordinate system. The normalized extrinsic parameter matrix consists of the rotation matrix R and the translation vector T.

[0015] Preferably, the distortion model includes a radial distortion model and a tangential distortion model; the distortion model corrects the ideal image coordinates (u) of the sub-lens. ideal ,v ideal ) and actual observed coordinates (u obs ,v obs The relationship between ) can be expressed by the formula:

[0016]

[0017] Where, K = (k1r 2 +k2r 4 +k3r 6 );r 2 =(u ideal -c x ) 2 +(v ideal -c y ) 2 ;r 2denoted as k1, k2, and k3, respectively, are radial distortion coefficients. k1 is the first-order radial distortion coefficient, k2 is the second-order radial distortion coefficient, and k3 is the third-order radial distortion coefficient. p1 and p2 are tangential distortion coefficients.

[0018] Preferably, the improved mutual information metric is defined by the formula:

[0019] MI(I1,I z ) = H(I1) + H(I2) - H(I1,I2); where H(·) represents the information entropy, H(I1,I2) represents the joint information entropy, I1 represents the preprocessed sub-image reference image, I2 represents the preprocessed sub-image to be registered image; MI(I1,I2) represents the mutual information measure between I1 and I2.

[0020] Preferably, the construction of the global pixel model in step four specifically employs Zernike polynomials;

[0021] The Zernike polynomial is expressed by the following formula:

[0022] Where n represents the polynomial order, m represents the azimuth degree, |m|≤n and n-|m| is even; The normalization coefficient is... It is a radial polynomial, S m (θ) is the angular polynomial; ρ represents the radial coordinate, and θ represents the angular coordinate;

[0023] The wavefront error of each sub-lens can then be expressed as:

[0024] Where, N max Let n represent the largest number of polynomials, where n = 0, 1, 2, ..., N. max ; This represents the Zernike coefficient of the a-th sub-shot.

[0025] Preferably, the objective function of the optimized algorithm for collaboratively compensating for the optical aberrations of each sub-lens is to maximize the gradient energy of the global image, which can be expressed by the following formula:

[0026] in, This represents the correction amount for the Zernike coefficient of the a-th sub-lens. I represents the vector differential operator. a This represents the image of the a-th sub-shot. Let (x, y) represent the ideal sub-image after correction of the a-th sub-shot, (x, y) represent the image pixels, X and Y represent the total number of image pixels, M represents the total number of sub-shots, and MB represents the objective function.

[0027] Preferably, the strategy combining weighted average fusion and multi-resolution pyramid fusion includes:

[0028] Overlapping region segmentation: Based on the transformation parameters in step three, determine the overlapping region Q of each sub-image. ab ;

[0029] Weighted graph generation: in the overlapping region Q ab Within, a weight map ω is generated based on the gradient magnitude of the sub-image. ab (x,y);

[0030] Multi-resolution pyramid fusion: A Laplacian pyramid is constructed for each sub-image. In each pyramid layer, overlapping areas are fused using a weighted average, while non-overlapping areas are directly preserved. The final fused image is obtained through inverse pyramid transformation.

[0031] Preferably, the weight graph ω ab (x, y) can be expressed by the formula:

[0032] Among them, I b Let represent the image of the b-th sub-shot, and let ∈ represent a small constant to prevent the denominator from being zero, taking ∈ = 1e-5.

[0033] An optical lens stitching imaging system includes a calibration module, a preprocessing module, a registration module, a correction module, and a fusion module;

[0034] The calibration module is used to obtain the internal parameters and distortion model of each sub-lens and establish the mapping relationship between the sub-lens and the global coordinate system;

[0035] The preprocessing module is used to synchronously acquire original images through each sub-lens, mark the original images acquired by each sub-lens as sub-images, and perform noise reduction, preliminary distortion correction and grayscale normalization on each sub-image.

[0036] The registration module is used to obtain the transformation parameters of each sub-image relative to the global coordinate system based on the improved mutual information metric and robust feature matching.

[0037] The correction module is used to construct a global pixel model and to compensate for the optical aberrations of each sub-lens through optimization algorithms, so as to ensure the global image quality consistency of the stitched image.

[0038] The fusion module is used to obtain the final fused image by employing a strategy that combines weighted average fusion and multi-resolution pyramid fusion.

[0039] The technical effects and advantages of this invention are as follows:

[0040] This invention, through steps three, four, and five, facilitates the resolution of registration errors caused by sparse feature points in low-texture regions using a hybrid strategy of global mutual information guidance and local robust matching, effectively improving registration accuracy. By employing a global model based on Zernike polynomials, it compensates for the optical aberrations of each sub-lens, effectively improving the image quality of the edge regions of the stitched image. The use of a multi-resolution pyramid optimization method reduces the computational complexity of global registration, meeting the requirements of real-time imaging, while effectively eliminating stitching seams and significantly improving the overall smoothness of the image. Attached Figure Description

[0041] Figure 1 This is a flowchart of the optical lens stitching imaging method of the present invention.

[0042] Figure 2 This is a structural diagram of the optical lens stitching imaging system of the present invention. Detailed Implementation

[0043] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The optical lens stitching imaging method and system involved in the present invention are not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] like Figure 1 As shown, the present invention provides an optical lens stitching imaging method, comprising the following steps:

[0045] Step 1: Sub-lens parameter calibration; obtain the internal parameters and distortion model of each sub-lens, and establish the mapping relationship between the sub-lens and the global coordinate system; the purpose is to obtain the accurate internal parameters and distortion model of the sub-lens, so as to provide a benchmark for subsequent registration and correction.

[0046] Step 2: Sub-image acquisition and preprocessing; Original images are acquired synchronously by each sub-lens, and the original images acquired by each sub-lens are marked as sub-images. Denoising, preliminary distortion correction, and grayscale normalization are performed on each sub-image. The purpose is to reduce noise interference, eliminate the influence of initial distortion of sub-lenses on subsequent registration, and unify the grayscale range to improve registration accuracy.

[0047] Step 3: Adaptive Feature Registration; Based on improved mutual information metric and robust feature matching, obtain the transformation parameters of each sub-image relative to the global coordinate system; The transformation parameters include, but are not limited to, rotation matrix, translation vector and scaling factor; The purpose is to solve the registration error problem caused by the sparse feature points in low-texture regions and improve the estimation accuracy of spatial transformation parameters between sub-images.

[0048] Step 4: Pixel Co-correction; Construct a global pixel model and use optimization algorithms to co-compensate the optical aberrations of each sub-lens to ensure consistent global image quality in the stitched image; The purpose is to compensate for the optical aberrations of each sub-lens and ensure consistent global image quality in the stitched image, such as no blurred edges and consistent distortion.

[0049] Step 5: Multi-strategy fusion optimization; a strategy combining weighted average fusion and multi-resolution pyramid fusion is adopted to eliminate stitching seams and improve image smoothness, resulting in the final fused image; the purpose is to eliminate stitching seams in overlapping areas of sub-images and improve the overall smoothness and detail retention of the image.

[0050] In this embodiment, it should be specifically explained that the establishment of the mapping relationship between the sub-lens and the global coordinate system is as follows:

[0051] An improved Zhang Zhengyou calibration method is adopted, using a checkerboard calibration board as a reference. By taking N images of the calibration board at different angles and positions, the corner coordinates are extracted and sub-pixel level positioning is performed; where N≥20.

[0052] Let the image coordinates of the a-th sub-shot be (u a ,v a The coordinates in the world coordinate system are (X... ω ,Y ω Z ω In this embodiment, the world coordinate system is the calibration plate coordinate system. Since the calibration plate is a plane in the calibration plate coordinate system, the Z-axis is taken as... ω If = 0, then the mapping relationship can be expressed as:

[0053] Where k represents the scaling factor, since the calibration board and the image units are inconsistent, a unified scale is required; Let f be the intrinsic parameter matrix. x f y c is the focal length of a pixel. x c y The coordinates of the principal point of the sub-lens image are the offsets from the origin in the pixel coordinate system. The normalized extrinsic parameter matrix consists of the rotation matrix R and the translation vector T, where R = [r ij ]; used to describe the rotation of the sub-lens coordinate system relative to the world coordinate system; T = [t x ,ty ,t z ] T Used to describe the position of the origin of the sub-lens coordinate system in the world coordinate system;

[0054] If the mapping relationship is expressed as the mapping relationship between the a-th sub-shot and the global coordinate system, then:

[0055] u and v in the text can be represented by u. a v a express.

[0056] In this embodiment, it should be specifically noted that the distortion model includes a radial distortion model and a tangential distortion model; the distortion model corrects the ideal image coordinates (u) of the sub-lens. ideal ,v ideal ) and actual observed coordinates (u obs ,v obs The relationship between ) can be expressed by the formula:

[0057] ;

[0058] Where, K=(k1r 2 +k2r 4 +k3r 6 );r 2 =(u ideal -c x ) 2 +(v ideal -c y ) 2 ;r 2 denoted as k1, k2, and k3, is the square of the distance from the ideal image coordinate point of the sub-lens to the principal image point of the sub-lens. k1, k2, and k3 are radial distortion coefficients, with k1 being the first-order radial distortion coefficient, k2 being the second-order radial distortion coefficient, and k3 being the third-order radial distortion coefficient. k3 is typically used in scenes with large distortion, such as fisheye lenses. p1 and p2 are tangential distortion coefficients, caused by lens assembly errors. The calibration parameters of each sub-lens are obtained by optimizing the intrinsic parameter matrix, extrinsic parameter matrix, radial distortion coefficient, and tangential distortion coefficient using the least squares method.

[0059] In this embodiment, it should be specifically noted that step two, sub-image acquisition and preprocessing, includes denoising, preliminary distortion correction, and grayscale normalization. The denoising employs non-local mean filtering to denoise each sub-image, preserving edge details while suppressing Gaussian noise. The core of the non-local mean filtering is to utilize a weighted average of similar pixel blocks in the image, with the weights determined by the Euclidean distance and grayscale differences between pixel blocks. The preliminary distortion correction performs inverse distortion transformation on each sub-image based on the distortion coefficients, mapping the actual observed coordinates back to the ideal image coordinates. The formula is expressed as:

[0060] Where, k n L represents the nth order radial distortion coefficient. n (r) 2 The polynomial basis functions representing radial distortion, L n (r) 2 =r n ; I represents the identity matrix;

[0061] The grayscale normalization maps the grayscale values ​​of each sub-image from [0,255] to [0,1], and subtracts the mean and divides by the standard deviation to eliminate the influence of differences in sensor response between different sub-lenses.

[0062] In this embodiment, it should be specifically noted that the improved mutual information metric is defined by the following formula:

[0063] MI(I1,I2)=H(I1)+H(i2)-H(I1,I2); where H(·) represents the information entropy, H(I1,I2) represents the joint information entropy, I1 represents the preprocessed sub-image reference image, and I2 represents the preprocessed sub-image to be registered; MI(I1,I2) represents the mutual information metric between I1 and I2; the mutual information measures the statistical correlation of the gray-level distributions of the two images, and the larger the mutual information metric, the more accurate the registration. The transformation parameters are gradually optimized through a multi-resolution pyramid, which ranges from low resolution to high resolution.

[0064] The robust feature matching method extracts SIFT feature points based on global registration, obtains candidate point pairs through descriptor matching, eliminates mismatched points using a random sampling consensus algorithm, calculates the affine transformation matrix and fuses it with global transformation parameters to obtain the final transformation parameters.

[0065] In this embodiment, it should be specifically noted that the calculation of the affine transformation matrix and its fusion with the global transformation parameters specifically includes:

[0066] Feature extraction: Scale-invariant feature transformation (SIFT) is used to extract key points of sub-images to solve the problem of sparse feature points in low-texture regions. SIFT generates feature descriptors by calculating the gradient orientation histogram of local regions of the image. It has scale and rotation invariance and is suitable for sub-image matching under different viewpoints.

[0067] Feature matching: By comparing the Euclidean distance of the feature descriptors of two images, the nearest and second nearest neighbor feature points of each feature point in the image to be registered are found in the parameter image. If the ratio of the nearest distance to the second nearest neighbor distance is less than a threshold, the pair of feature points is considered as candidate matching points. The threshold is usually 0.7.

[0068] Mismatch Removal: Due to noise, duplicate textures, or occlusion, there may be mismatches among the candidate matching points. The random sampling consensus algorithm is used to remove mismatch points to improve matching and robustness.

[0069] Parameter optimization: Integrate with global transformation parameters.

[0070] In this embodiment, it should be specifically explained that the feature extraction is as follows:

[0071] Scale space is calculated for I2 and I1 respectively. Extreme points are detected in the scale space as key points, and pseudo feature points with low contrast or edge response are removed. For each key point, the gradient direction histogram of its neighborhood is calculated to generate feature descriptors.

[0072] The specific steps for using the random sampling consensus algorithm to eliminate mismatches are as follows:

[0073] Four pairs of candidate matching points are randomly selected. An initial affine transformation matrix is ​​calculated using these four pairs of points. All candidate matching point pairs are traversed, and the number of interior points that meet the condition is counted. The condition is that after transformation by the initial affine transformation matrix, the Euclidean distance between the point to be registered and the reference point is less than a set threshold. The set threshold is set by those skilled in the art according to the actual situation. This embodiment does not specify the specific value of the set threshold. The above steps are repeated U times. U = 1000 can be selected. The affine transformation matrix with the most interior points is selected as the optimal affine transformation matrix. The optimal affine transformation matrix is ​​optimized again using the least squares method with all interior points to obtain the final affine transformation matrix.

[0074] In this embodiment, it should be specifically noted that the construction of the global pixel model in step four specifically involves using Zernike polynomials to describe the wavefront error of each sub-lens.

[0075] The Zernike polynomial is expressed by the following formula:

[0076] Where n represents the polynomial order, m represents the azimuth degree, |m|≤n and n-|m| is even; The normalization coefficient is... It is a radial polynomial, S m (θ) is the angular polynomial; ρ represents the radial coordinate, and θ represents the angular coordinate;

[0077] The wavefront error of each sub-lens can then be expressed as:

[0078] Where, N max Let n represent the largest number of polynomials, where n = 0, 1, 2, ..., N. max ; This represents the Zernike coefficient of the a-th sub-lens; the Zernike coefficient can be obtained through point spread function (PSF) deconvolution or interferometry.

[0079] In this embodiment, it should be specifically noted that the objective function of the optimization algorithm to collaboratively compensate for the optical aberrations of each sub-lens is to maximize the gradient energy of the global image. The gradient energy reflects the degree of detail preservation and is expressed by the formula:

[0080] in, This represents the correction amount for the Zernike coefficient of the a-th sub-lens. I represents the vector differential operator. a This represents the image of the a-th sub-shot. Let (x, y) represent the ideal sub-image after correction of the a-th sub-shot, (x, y) represent the image pixels, X and Y represent the total number of image pixels, M represents the total number of sub-shots, and MB represents the objective function.

[0081] By solving for the optimal correction amount using gradient descent or L-BFGS algorithm, the impact of aberrations of each sub-lens on the global image can be minimized.

[0082] In this embodiment, it should be specifically noted that the strategy combining weighted average fusion and multi-resolution pyramid fusion includes:

[0083] Overlapping region segmentation: Based on the transformation parameters in step three, determine the overlapping region Q of each sub-image. ab ;

[0084] Weighted graph generation: in the overlapping region Q ab Within, a weight map ω is generated based on the gradient magnitude of the sub-image. ab For regions (x, y), the larger the gradient, the higher the weight, and the more details can be preserved. This can be expressed by the formula:

[0085] Among them, I b Let represent the image of the b-th sub-shot, and ∈ represent a small constant to prevent the denominator from being zero. In this embodiment, ∈ is taken as 1e-5.

[0086] Multi-resolution pyramid fusion: A Laplacian pyramid is constructed for each sub-image. In each pyramid layer, overlapping areas are fused using a weighted average, while non-overlapping areas are directly preserved. The final fused image is obtained through inverse pyramid transformation.

[0087] like Figure 2 As shown, the present invention provides an optical lens stitching imaging system, including a calibration module, a preprocessing module, a registration module, a correction module, and a fusion module;

[0088] The calibration module is used to obtain the internal parameters and distortion model of each sub-lens and establish the mapping relationship between the sub-lens and the global coordinate system;

[0089] The preprocessing module is used to synchronously acquire original images through each sub-lens, mark the original images acquired by each sub-lens as sub-images, and perform noise reduction, preliminary distortion correction and grayscale normalization on each sub-image.

[0090] The registration module is used to obtain the transformation parameters of each sub-image relative to the global coordinate system based on the improved mutual information metric and robust feature matching.

[0091] The correction module is used to construct a global pixel model and to compensate for the optical aberrations of each sub-lens through optimization algorithms, so as to ensure the global image quality consistency of the stitched image.

[0092] The fusion module employs a strategy combining weighted average fusion and multi-resolution pyramid fusion to eliminate stitching seams and improve image smoothness, resulting in a final fused image.

[0093] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0094] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An optical lens stitching imaging method, characterized in that: Includes the following steps: Step 1: Sub-lens parameter calibration; obtain the internal parameters and distortion model of each sub-lens, and establish the mapping relationship between the sub-lens and the global coordinate system; Step 2: Sub-image acquisition and preprocessing; Original images are acquired synchronously through each sub-lens, and the original images acquired by each sub-lens are marked as sub-images. Denoising, preliminary distortion correction, and grayscale normalization are performed on each sub-image. Step 3: Adaptive feature registration; Based on improved mutual information metric and robust feature matching, obtain the transformation parameters of each sub-image relative to the global coordinate system; Step 4: Pixel-to-Pixel Collaborative Correction; Construct a global pixel model and use an optimized algorithm to collaboratively compensate for the optical aberrations of each sub-lens; Step 5: Multi-strategy fusion optimization; A strategy combining weighted average fusion and multi-resolution pyramid fusion is adopted to obtain the final fused image; The robust feature matching, based on global registration, extracts SIFT feature points, obtains candidate point pairs through descriptor matching, eliminates mismatched points using a random sampling consensus algorithm, calculates the affine transformation matrix and fuses it with global transformation parameters to obtain the final transformation parameters. The calculation of the affine transformation matrix and its fusion with global transformation parameters specifically includes: Feature extraction: Key points of the sub-image are extracted using Scale Invariant Feature Transform (SIFT). SIFT generates feature descriptors by calculating the gradient orientation histogram of local regions of the image. Feature matching: By comparing the Euclidean distance between the feature descriptors of two images, the nearest and second nearest neighbor feature points of each feature point in the image to be registered are found in the parameter image. If the ratio of the nearest distance to the second nearest neighbor distance is less than a threshold, the pair of feature points is considered as candidate matching points. False match removal: False matches are removed using a random sampling consensus algorithm; Parameter optimization: Integrate with global transformation parameters.

2. The optical lens stitching imaging method according to claim 1, characterized in that: The distortion model includes a radial distortion model and a tangential distortion model; the distortion model corrects the ideal image coordinates of the sub-lens. Compared with actual observed coordinates The relationship can be expressed by the formula: ;in, ; ; Let be the square of the distance from the ideal image coordinate point of the sub-lens to the principal point of the sub-lens image. , , The radial distortion coefficient is... The first-order radial distortion coefficient, It is the second-order radial distortion coefficient. It is the third-order radial distortion coefficient; , denoted as the tangential distortion coefficient.

3. The optical lens stitching imaging method according to claim 2, characterized in that: The improved mutual information metric is defined by the following formula: ;in, Represents information entropy. Represents joint information entropy, This represents the preprocessed sub-image reference image. This represents the preprocessed sub-image to be registered; express and Mutual information metric.

4. The optical lens stitching imaging method according to claim 3, characterized in that: In step four, the global pixel model is constructed using Zernike polynomials. The Zernike polynomial is expressed by the following formula: ;in, Denotes the order of a polynomial. Indicates the order of azimuth. and It is an even number; The normalization coefficient is... It is a radial polynomial. It is an angular polynomial; Represents radial coordinates, Indicates angular coordinates; The wavefront error of each sub-lens can then be expressed as: ;in, This represents the largest number of polynomials. ; Indicates the first Zernike coefficient for individual lenses.

5. The optical lens stitching imaging method according to claim 4, characterized in that: The objective function of the optimization algorithm for collaboratively compensating for the optical aberrations of each sub-lens is to maximize the gradient energy of the global image, which can be expressed by the formula: ;in, Indicates the first The Zernike coefficient correction amount for individual lenses Represents the vector differential operator. Indicates the first Images from individual lenses, Indicates the first The ideal sub-image after lens correction Represents image pixels, , This represents the total number of pixels in the image. Indicates the total number of sub-lenses. This represents the objective function.

6. The optical lens stitching imaging method according to claim 5, characterized in that: The strategy combining weighted average fusion and multi-resolution pyramid fusion includes: Overlapping region segmentation: Based on the transformation parameters in step three, the overlapping regions of each sub-image are determined. ; Weighted graph generation: in overlapping regions Within, a weight map is generated based on the gradient magnitude of the sub-image. ; Multi-resolution pyramid fusion: A Laplacian pyramid is constructed for each sub-image. In each pyramid layer, overlapping areas are fused using a weighted average, while non-overlapping areas are directly preserved. The final fused image is obtained through inverse pyramid transformation.

7. The optical lens stitching imaging method according to claim 6, characterized in that: The weight graph Expressed as a formula: ;in, Indicates the first Images from individual lenses, To prevent the denominator from being zero, take the small constant value. .

8. An optical lens stitching imaging system, used with any one of the optical lens stitching imaging methods according to claims 1-7, characterized in that: It includes a calibration module, a preprocessing module, a registration module, a correction module, and a fusion module; The calibration module is used to obtain the internal parameters and distortion model of each sub-lens and establish the mapping relationship between the sub-lens and the global coordinate system; The preprocessing module is used to synchronously acquire original images through each sub-lens, mark the original images acquired by each sub-lens as sub-images, and perform noise reduction, preliminary distortion correction and grayscale normalization on each sub-image. The registration module is used to obtain the transformation parameters of each sub-image relative to the global coordinate system based on the improved mutual information metric and robust feature matching. The correction module is used to construct a global pixel model and to compensate for the optical aberrations of each sub-lens through optimization algorithms, so as to ensure the global image quality consistency of the stitched image. The fusion module is used to obtain the final fused image by employing a strategy that combines weighted average fusion and multi-resolution pyramid fusion.

Images