Image registration method, device, computer equipment and storage medium

Abstract

The present invention discloses an image registration method, device, computer equipment and storage medium, and belongs to the field of machine vision defect detection technology. Aiming at the problem that it is difficult to find effective feature point matching in semiconductor surface image registration in the prior art, resulting in low registration efficiency, the present invention sets a standard image to obtain standard image parameters; inputs an original image to obtain original image parameters; uses standard image parameters and original image parameters to solve the rotation matrix and translation matrix of the original image; establishes an image rigid body transformation relationship, performs affine transformation on the original image through the rotation matrix and translation matrix, and registers the original image and the standard image. Therefore, when performing image registration, there is no need to perform feature point matching, but by solving the rotation matrix and translation matrix of the image, the image scaling scale is constrained, and the rigid body transformation of the image is realized, which effectively improves the image registration efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision defect detection, and more specifically, to an image registration method, device, computer equipment and storage medium. Background Art

[0002] Image registration is the process of matching and superimposing two or more images acquired at different times, with different sensors (imaging devices) or under different conditions. It is widely used in the fields of computer vision and image processing. Image registration mainly includes three parts: feature extraction, feature matching and parameter estimation.

[0003] In digital image processing, image registration is to achieve accurate alignment of images by comparing and aligning images acquired at different time periods or different viewing angles. Image registration is of great significance in defect detection and change analysis, and it can help detect deformation, cracks and flaws of target objects. By comparing the registered image with the reference image, the changes and defects of the target object can be accurately analyzed, providing an important basis for subsequent processing and decision-making.

[0004] In the prior art, semiconductor images often present simple surface patterns, disordered textures, or a certain pattern that repeats. Classic feature point descriptors, such as SIFT and ORB, are difficult to find effective feature points or perform feature point registration on the semiconductor surface. In contrast, for some specific structures on semiconductors, such as pins, bumping, and other circular or rectangular structures, their center coordinates can be easily found.

[0005] After searching, the Chinese patent application, application number 202010828690.5, published on December 29, 2020, discloses a method for image registration based on feature points. The method includes the following parts: a. Preprocessing the reference image and the image to be registered by the wavelet transform threshold denoising method; b. Extracting feature points by SIFT algorithm; c. Describing feature points by deformation dimensionality reduction method; d. Roughly matching feature points according to cosine similarity; e. Using the improved RANSAC algorithm to eliminate some mismatches and obtain matching point pairs with high matching accuracy. Before extracting feature points, the method preprocesses the image by the wavelet transform threshold denoising method to eliminate some noise in the image, thereby improving the purity of the feature points, and when describing the feature points, a deformation dimensionality reduction method is proposed to describe the extracted feature points, reduce the SIFT descriptor dimension, shorten the algorithm running time, and optimize the matching feature point pairs by combining rough matching with fine matching to improve the matching accuracy of the algorithm. However, when this method is applied to the field of semiconductor technology, it is difficult to find effective feature points on the semiconductor surface for matching. Summary of the invention

[0006] 1. Technical issues to be solved

[0007] In view of the problem in the prior art that it is difficult to find effective feature point matches when aligning semiconductor surface images, resulting in low alignment efficiency, the present invention provides an image alignment method, apparatus, computer equipment and storage medium. Instead of performing feature point matching, the image rotation matrix and translation matrix are solved to constrain the image scaling scale, thereby achieving rigid body transformation of the image and effectively improving the image alignment efficiency.

[0008] 2. Technical solution

[0009] The purpose of the present invention is achieved through the following technical solutions.

[0010] An image registration method comprises the following steps:

[0011] Set the standard image and obtain the standard image parameters;

[0012] Input the original image and obtain the original image parameters;

[0013] Solve the rotation matrix and translation matrix of the original image using the standard image parameters and the original image parameters;

[0014] Establish the rigid body transformation relationship of the image, perform affine transformation on the original image through rotation matrix and translation matrix, and align the original image and the standard image.

[0015] Furthermore, the standard image parameters include a standard image right singular vector; the step of obtaining the standard image right singular vector includes: collecting a standard image feature point matrix; decentralizing the standard image feature point matrix to obtain a standard image decentralized feature point matrix; performing singular value decomposition on the standard image decentralized feature point matrix to obtain a standard image right singular matrix; obtaining a standard image right singular vector through the standard image right singular matrix;

[0016] The original image parameters include the right singular vectors of the original image; the steps for obtaining the right singular vectors of the original image include: collecting the feature point matrix of the original image; decentralizing the feature point matrix of the original image to obtain the decentralized feature point matrix of the original image; performing singular value decomposition on the decentralized feature point matrix of the original image to obtain the right singular matrix of the original image; and obtaining the right singular vectors of the original image through the right singular matrix of the original image.

[0017] Furthermore, the steps of solving the rotation matrix of the original image using the standard image parameters and the original image parameters include:

[0018] Calculate the rotation angle between the right singular vector of the standard image and the right singular vector of the original image to obtain a first rotation angle;

[0019] Calculating the feature point matrix of the original image using the first rotation angle to obtain the feature point matrix of the original image after rotation;

[0020] Performing a decentralized processing on the feature point matrix of the rotated original image to obtain a decentralized feature point matrix of the rotated original image;

[0021] Perform singular value decomposition on the decentralized feature point matrix of the rotated original image to obtain the right singular matrix of the rotated original image;

[0022] Obtain the right singular vector of the rotated original image through the right singular matrix of the rotated original image;

[0023] Calculate the rotation angle between the right singular vector of the standard image and the right singular vector of the rotated original image to obtain a second rotation angle;

[0024] By determining the magnitude of the first rotation angle and the second rotation angle, a rotation angle is obtained;

[0025] The rotation matrix is ​​obtained by solving the rotation angle.

[0026] Furthermore, the rotation matrix is ​​expressed as:

[0027]

[0028] Where R represents the rotation matrix and θ represents the rotation angle.

[0029] Furthermore, the translation matrix is ​​solved by registering the standard image feature point matrix and the rotated original image feature point matrix, and the steps include:

[0030] Traverse all coordinate points in the standard image feature point matrix and obtain the point set mean of all coordinate points in the standard image feature point matrix;

[0031] Traverse all coordinate points in the feature point matrix of the rotated original image, and obtain the point set mean of all coordinate points in the feature point matrix of the rotated original image;

[0032] The horizontal coordinate offset and the vertical coordinate offset are obtained by subtracting the point set mean of all coordinate points in the standard image feature point matrix and the point set mean of all coordinate points in the rotated original image feature point matrix.

[0033] Furthermore, the translation matrix is ​​expressed as:

[0034]

[0035] Among them, T represents the translation matrix, tx represents the horizontal coordinate offset, and ty represents the vertical coordinate offset.

[0036] Furthermore, the image rigid body transformation relationship is expressed as:

[0037]

[0038] Among them, H represents the rigid body transformation relationship of the image;

[0039] Through the rigid body transformation relationship of the image, the rotation matrix and translation matrix are used to perform affine transformation on the original image. The calculation formula is:

[0040]

[0041] Among them, u represents the horizontal coordinate of the original image, v represents the vertical coordinate of the original image, u' represents the horizontal coordinate of the original image after affine transformation, and v' represents the vertical coordinate of the original image after affine transformation.

[0042] An image registration device, comprising:

[0043] Setting module, setting standard image, obtaining standard image parameters;

[0044] Input module, input the original image and obtain the original image parameters;

[0045] A calculation module, using standard image parameters and original image parameters to solve the rotation matrix and translation matrix of the original image;

[0046] The registration module establishes the rigid transformation relationship of the image, performs affine transformation on the original image through the rotation matrix and translation matrix, and registers the original image and the standard image.

[0047] A computer device comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the above-mentioned method when executing the computer program.

[0048] A computer-readable storage medium stores a computer program, and the computer program executes the above method when executed by a processor.

[0049] 3. Beneficial effects

[0050] Compared with the prior art, the advantages of the present invention are:

[0051] The image registration method, device, computer equipment and storage medium of the present invention do not need to perform feature point matching when performing image registration. Instead, they solve the rotation matrix and translation matrix of the image to constrain the image scaling and achieve rigid body transformation of the image, thereby effectively improving the efficiency of image registration and having strong practicality and wide applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a flow chart of the image registration method according to an embodiment of the present invention;

[0053] Figure 2 is a standard image according to an embodiment of the present invention;

[0054] Figure 3 is the original image of the embodiment of the present invention;

[0055] Figure 4 The registration result image of the embodiment of the present invention;

[0056] Figure 5 A differential image between the registration image and the standard image according to an embodiment of the present invention;

[0057] Figure 6 It is a difference image between the unregistered image and the standard image in the embodiment of the present invention. DETAILED DESCRIPTION

[0058] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] Example

[0060] like Figure 1 As shown, an image registration method provided by this embodiment. An image registration method includes the following steps: setting a standard image and obtaining standard image parameters; inputting an original image and obtaining original image parameters; solving the rotation matrix and translation matrix of the original image using the standard image parameters and the original image parameters; establishing an image rigid body transformation relationship, performing an affine transformation on the original image through the rotation matrix and the translation matrix, and registering the original image and the standard image.

[0061] Specifically in this embodiment, Figure 2 As shown, a standard image is set and standard image parameters are obtained. The standard image parameters include a right singular value vector of the standard image.

[0062] In this embodiment, the step of acquiring the right singular vector of the standard image includes: acquiring the standard image feature point matrix P s It should be noted that, in this embodiment, the feature point is the centroid of the solder joint of a semiconductor or other product. The standard image feature point matrix P s It is expressed as:

[0063]

[0064] Among them, x represents the horizontal coordinate of the standard image feature point matrix, y represents the vertical coordinate of the standard image feature point matrix, and n represents a natural number.

[0065] For the standard image feature point matrix P s Perform decentralization processing to obtain the standard image decentralized feature point matrix P scSpecifically, calculate the standard image feature point matrix P s The horizontal and vertical coordinates of all feature points in the standard image are calculated, and then the standard image feature point matrix P is calculated. s Subtract the mean value of the horizontal coordinates of all feature points in the standard image feature point matrix P s After subtracting the mean ordinate value from the ordinates of all feature points in the image, we get the standard image decentralized feature point matrix P. sc .

[0066] In this embodiment, the standard image feature point matrix P s Perform decentralization to obtain the standard image decentralized feature point matrix P sc It is the prior art.

[0067] Decentralize the feature point matrix P of the standard image sc Perform singular value decomposition (SVD) to obtain the right singular matrix of the standard image .

[0068] In this embodiment, the standard image is decentralized feature point matrix P sc Performing singular value decomposition is a prior art.

[0069] By standard image right singular matrix Get the right singular vector E of the standard image s Specifically, save the standard image right singular matrix The first row vector of the standard image is the right singular vector E s .

[0070] It should be noted that, in this embodiment, the right singular vector E of the standard image is s Represents the decentralized feature point matrix P of the standard image sc Rotation mapping. Therefore, the standard image decentralized feature point matrix Psc is expressed as:

[0071]

[0072] Among them, P sc represents the standard image decentralized feature point matrix, U s represents the standard image left singular matrix, W s represents the diagonal matrix containing the eigenvalues ​​in the standard image, Standard image right singular matrix.

[0073] It should be noted that, in this embodiment, it is also possible to choose to decentralize the feature point matrix P of the standard image scSingular value decomposition is performed to obtain the left singular matrix of the standard image, and the left singular vector of the standard image is obtained through the left singular matrix of the standard image, which is then used in the image registration in the subsequent steps.

[0074] like Figure 3 As shown, the original image is input and the original image parameters are obtained. The original image parameters include the right singular value vector of the original image.

[0075] In this embodiment, the steps for obtaining the right singular vector of the original image are the same as the steps for obtaining the right singular value vector of the standard image, including: collecting the feature point matrix P of the original image i ; For the original image feature point matrix P i Perform decentralized processing to obtain the decentralized feature point matrix P of the original image ic ; Decentralize the feature point matrix P of the original image ic Perform singular value decomposition to obtain the right singular matrix of the original image ; Save the right singular matrix of the original image The first row vector of the original image is the right singular vector E i .

[0076] In this embodiment, the original image decentralized feature point matrix P ic It is expressed as:

[0077]

[0078] Among them, P ic Represents the original image decentralized feature point matrix, U i represents the left singular matrix of the original image, W i represents the diagonal matrix containing the eigenvalues ​​of the original image, Represents the right singular matrix of the original image.

[0079] It should be noted that, in this embodiment, it is also possible to choose to decentralize the feature point matrix P of the original image. ic The singular value decomposition is performed to obtain the left singular matrix of the original image, and the left singular vector of the original image is obtained through the left singular matrix of the original image, which is then used in the image registration in the subsequent steps.

[0080] Therefore, in this embodiment, after obtaining the right singular vector E of the standard image s and the original image right singular vector E i Then, using the standard image right singular vector E s and the original image right singular vector E i Solving the rotation matrix of the original image includes the following steps:

[0081] Calculate the right singular vector E of the standard image sand the original image right singular vector E i The first rotation angle θ1 is obtained by calculating the rotation angle between θ1 and θ2. In this embodiment, the calculation formula of the first rotation angle θ1 is:

[0082]

[0083] Where θ1 represents the first rotation angle, E s represents the right singular vector of the standard image, E i Represents the right singular vectors of the original image.

[0084] Calculate the original image feature point matrix P using the first rotation angle θ1 i , get the feature point matrix of the rotated original image , the calculation formula is:

[0085]

[0086] Among them, x i Represents the horizontal coordinate of the feature point matrix of the original image, y i Represents the ordinate of the feature point matrix of the original image, Represents the horizontal coordinate of the feature point matrix of the original image after rotation, Represents the ordinate of the feature point matrix of the original image after rotation.

[0087] The feature point matrix of the rotated original image Perform decentralization processing to obtain the decentralized feature point matrix of the rotated original image .

[0088] In this embodiment, the feature point matrix of the rotated original image Perform decentralization to obtain the decentralized feature point matrix of the rotated original image It is the prior art.

[0089] Decentralize the feature point matrix of the rotated original image Perform singular value decomposition to obtain the right singular matrix of the rotated original image .

[0090] In this embodiment, the feature point matrix of the rotated original image is decentralized. Performing singular value decomposition is a prior art.

[0091] The right singular matrix of the original image after rotation Get the right singular vectors of the rotated original image Specifically, save the right singular matrix of the rotated original image The first row vector of the rotated original image is the right singular vector of .

[0092] Calculate the right singular vector E of the standard image s and the right singular vectors of the rotated original image The second rotation angle θ2 is obtained by calculating the rotation angle between θ2 and θ3. In this embodiment, the calculation formula of the second rotation angle θ2 is:

[0093]

[0094] Where θ2 represents the second rotation angle, E s represents the right singular vector of the standard image, Represents the right singular vectors of the original image after rotation.

[0095] By determining the magnitude of the first rotation angle θ1 and the second rotation angle θ2, the rotation angle θ is obtained. In this embodiment, the determination formula of the rotation angle θ is:

[0096]

[0097] Therefore, in this embodiment, when the absolute value of the second rotation angle θ2 is less than the absolute value of the first rotation angle θ1, the rotation angle θ is the first rotation angle θ1. At this time, the first rotation angle θ1 is substituted into the rotation matrix to obtain the rotation matrix; when the absolute value of the second rotation angle θ2 is greater than or equal to the absolute value of the first rotation angle θ1, the rotation angle θ is the negative of the first rotation angle θ1. At this time, the negative first rotation angle θ1 is substituted into the rotation matrix to obtain the rotation matrix.

[0098] In this embodiment, the rotation matrix is ​​expressed as:

[0099]

[0100] Where R represents the rotation matrix and θ represents the rotation angle.

[0101] Furthermore, through the standard image feature point matrix P s And the feature point matrix of the original image after rotation Perform registration to solve the translation matrix T, the steps include:

[0102] Traverse the standard image feature point matrix P s All coordinate points in the standard image feature point matrix P are obtained s The point set mean of all coordinate points in ;

[0103] Traverse the feature point matrix of the rotated original image All coordinate points in the rotated original image are obtained The point set mean of all coordinate points in ;

[0104] The standard image feature point matrix P s The point set mean of all coordinate points in and the feature point matrix of the rotated original image The point set mean of all coordinate points in is subtracted to obtain the horizontal coordinate offset tx and the vertical coordinate offset ty.

[0105] Therefore, in this embodiment, the translation matrix is ​​solved by the horizontal coordinate offset tx and the vertical coordinate offset ty. In this embodiment, the translation matrix is ​​expressed as:

[0106]

[0107] Among them, T represents the translation matrix, tx represents the horizontal coordinate offset, and ty represents the vertical coordinate offset.

[0108] Finally, the rigid body transformation relationship of the image is established, and the original image is affine transformed through the rotation matrix R and the translation matrix T to align the original image and the standard image.

[0109] In this embodiment, the image rigid body transformation relationship is expressed as:

[0110]

[0111] Among them, H represents the rigid body transformation relationship of the image.

[0112] Then, the rotation matrix R and the translation matrix T are brought into the image rigid body transformation relationship H to perform an affine transformation on the original image. The calculation formula is:

[0113]

[0114] Among them, u represents the horizontal coordinate of the original image, v represents the vertical coordinate of the original image, u' represents the horizontal coordinate of the original image after affine transformation, and v' represents the vertical coordinate of the original image after affine transformation.

[0115] like Figure 4 As shown, this is the registration result image of the original image and the standard image.

[0116] In this embodiment, the accuracy and effect of the registration are represented by the differential image. Figure 5 As shown in , it is the difference image between the registration result image and the standard image. The registration result image and the standard image basically coincide with each other, which indicates that the registration accuracy of the original image and the standard image is high. Figure 6 As shown, it is a difference image between the unregistered result image and the standard image. The unregistered result image and the standard image do not completely overlap, which indicates that the registration accuracy of the original image and the standard image is low.

[0117] Therefore, compared with the traditional method of solving affine matrix parameters by matching feature points, the image registration method provided in this embodiment does not require prior feature point matching. Instead, it solves the rotation matrix and translation matrix of the image to constrain the image scaling and realize the rigid body transformation of the image, thereby effectively improving the efficiency of image registration and having strong practicality and wide applicability.

[0118] This embodiment also provides an image registration device, including a setting module, an input module, a calculation module and a registration module. The setting module is used to set a standard image and obtain standard image parameters; the input module is used to input the original image and obtain the original image parameters; the calculation module uses the standard image parameters and the original image parameters to solve the rotation matrix and translation matrix of the original image; the registration module establishes the image rigid body transformation relationship, performs affine transformation on the original image through the rotation matrix and translation matrix, and registers the original image and the standard image. An image registration device provided in this embodiment can implement any of the image registration methods, and the specific working process of an image registration device can refer to the corresponding process in the embodiment of the image registration method. The method and device provided in this embodiment can be implemented in other ways. For example, the device embodiment described above is only schematic; for example, the division of a module is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual connection or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.

[0119] This embodiment also provides a computer device. A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the image registration method when executing the computer program.

[0120] This embodiment also provides a computer-readable storage medium. A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, an image registration method described in this embodiment is executed. The computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, device or device; the program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0121] The above schematically describes the invention and its implementation methods, which is not restrictive. Without departing from the spirit or basic features of the invention, the invention can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention. The actual structure is not limited thereto, and any figure mark in the claims should not limit the claims involved. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and an embodiment similar to the technical solution are designed without creativity, which should all belong to the protection scope of the present invention. In addition, the word "including" does not exclude other elements or steps, and the word "one" before the element does not exclude the inclusion of "multiple" elements. The multiple elements stated in the product claim can also be implemented by one element through software or hardware. The words first, second, etc. are used to indicate the name, and do not indicate any specific order.

Claims

1. An image registration method, comprising the following steps: Setting a standard image and obtaining standard image parameters; the standard image parameters include right singular vectors of the standard image; The step of acquiring the right singular vector of the standard image includes: collecting the standard image feature point matrix; decentralizing the standard image feature point matrix to obtain the standard image decentralized feature point matrix; performing singular value decomposition on the standard image decentralized feature point matrix to obtain the standard image right singular matrix; acquiring the standard image right singular vector through the standard image right singular matrix; Input an original image and obtain original image parameters; the original image parameters include the right singular vectors of the original image; the steps of obtaining the right singular vectors of the original image include: collecting the feature point matrix of the original image; decentralizing the feature point matrix of the original image to obtain the decentralized feature point matrix of the original image; performing singular value decomposition on the decentralized feature point matrix of the original image to obtain the right singular matrix of the original image; and obtaining the right singular vectors of the original image through the right singular matrix of the original image; The rotation matrix and translation matrix of the original image are solved by using the standard image parameters and the original image parameters, and the steps include: calculating the rotation angle between the right singular vector of the standard image and the right singular vector of the original image to obtain a first rotation angle; calculating the feature point matrix of the original image by using the first rotation angle to obtain the feature point matrix of the rotated original image; performing a decentralization process on the feature point matrix of the rotated original image to obtain a decentralized feature point matrix of the rotated original image; performing a singular value decomposition on the decentralized feature point matrix of the rotated original image to obtain a right singular matrix of the rotated original image; obtaining the right singular vector of the rotated original image through the right singular matrix of the rotated original image; calculating the rotation angle between the right singular vector of the standard image and the right singular vector of the rotated original image to obtain a second rotation angle; obtaining the rotation angle by judging the size of the first rotation angle and the second rotation angle; and obtaining the rotation matrix through the rotation angle solution; Establish the rigid body transformation relationship of the image, perform affine transformation on the original image through rotation matrix and translation matrix, and align the original image and the standard image.

2. An image registration method according to claim 1, characterized in that: The rotation matrix is ​​expressed as: Where R represents the rotation matrix and θ represents the rotation angle.

3. An image registration method according to claim 2, characterized in that: The translation matrix is ​​solved by registering the standard image feature point matrix and the rotated original image feature point matrix, and the steps include: Traverse all coordinate points in the standard image feature point matrix and obtain the point set mean of all coordinate points in the standard image feature point matrix; Traverse all coordinate points in the feature point matrix of the rotated original image, and obtain the point set mean of all coordinate points in the feature point matrix of the rotated original image; The horizontal coordinate offset and the vertical coordinate offset are obtained by subtracting the point set mean of all coordinate points in the standard image feature point matrix and the point set mean of all coordinate points in the rotated original image feature point matrix.

4. The image registration method according to claim 3, characterized in that: The translation matrix is ​​expressed as: Among them, T represents the translation matrix, tx represents the horizontal coordinate offset, and ty represents the vertical coordinate offset.

5. An image registration method according to claim 4, characterized in that: The image rigid body transformation relationship is expressed as: Among them, H represents the rigid body transformation relationship of the image; Through the rigid body transformation relationship of the image, the rotation matrix and translation matrix are used to perform affine transformation on the original image. The calculation formula is: Among them, u represents the horizontal coordinate of the original image, v represents the vertical coordinate of the original image, u' represents the horizontal coordinate of the original image after affine transformation, and v' represents the vertical coordinate of the original image after affine transformation.

6. An image registration device, characterized in that: include: Setting module, setting standard image, obtaining standard image parameters; The standard image parameters include standard image right singular vectors; The step of acquiring the right singular vector of the standard image includes: collecting the standard image feature point matrix; decentralizing the standard image feature point matrix to obtain the standard image decentralized feature point matrix; performing singular value decomposition on the standard image decentralized feature point matrix to obtain the standard image right singular matrix; acquiring the standard image right singular vector through the standard image right singular matrix; An input module inputs an original image and obtains original image parameters; the original image parameters include the right singular vectors of the original image; the steps of obtaining the right singular vectors of the original image include: collecting the feature point matrix of the original image; decentralizing the feature point matrix of the original image to obtain the decentralized feature point matrix of the original image; performing singular value decomposition on the decentralized feature point matrix of the original image to obtain the right singular matrix of the original image; and obtaining the right singular vectors of the original image through the right singular matrix of the original image; The calculation module uses the standard image parameters and the original image parameters to solve the rotation matrix and the translation matrix of the original image, and the steps include: calculating the rotation angle between the right singular vector of the standard image and the right singular vector of the original image to obtain a first rotation angle; using the first rotation angle to calculate the feature point matrix of the original image to obtain the feature point matrix of the rotated original image; performing a decentralized processing on the feature point matrix of the rotated original image to obtain the decentralized feature point matrix of the rotated original image; performing a singular value decomposition on the decentralized feature point matrix of the rotated original image to obtain the right singular matrix of the rotated original image; obtaining the right singular vector of the rotated original image through the right singular matrix of the rotated original image; calculating the rotation angle between the right singular vector of the standard image and the right singular vector of the rotated original image to obtain a second rotation angle; obtaining the rotation angle by judging the size of the first rotation angle and the second rotation angle; and obtaining the rotation matrix through the rotation angle solution; The registration module establishes the rigid transformation relationship of the image, performs affine transformation on the original image through the rotation matrix and translation matrix, and registers the original image and the standard image.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is executed.

Images