CN116258752A

CN116258752A - Registration method, registration apparatus, electronic device, and computer-readable storage medium

Info

Publication number: CN116258752A
Application number: CN202310213125.1A
Authority: CN
Inventors: 关沛峰; 黄炜倬; 平书伟
Original assignee: Guangzhou Aimuyi Technology Co ltd
Current assignee: Guangzhou Aimuyi Technology Co ltd
Priority date: 2023-03-07
Filing date: 2023-03-07
Publication date: 2023-06-13

Abstract

The application provides a registration method, a registration device, electronic equipment and a computer readable storage medium, and relates to the field of surgical navigation, wherein the method comprises the following steps: determining a first space principal axis direction of the image space point cloud and a second space principal axis direction of the operation space point cloud, correcting the first space principal axis direction to obtain multiple candidate directions, determining respective first conversion matrixes corresponding to the multiple candidate directions according to the multiple candidate directions and the second space principal axis direction, registering the image space point cloud and the operation space point cloud according to the respective first conversion matrixes corresponding to the multiple candidate directions to obtain respective second conversion matrixes corresponding to the multiple candidate directions, determining a coarse registration conversion matrix according to the respective first conversion matrixes and the second conversion matrixes corresponding to the multiple candidate directions, and performing fine registration on the image space point cloud and the operation space point cloud according to the coarse registration conversion matrix. The surface automatic registration method and device achieve automatic registration of the surfaces, and registration time is shortened.

Description

Registration method, registration apparatus, electronic device, and computer-readable storage medium

Technical Field

The present application relates to the field of surgical navigation, and more particularly, to a registration method, registration apparatus, electronic device, and computer-readable storage medium in the field of surgical navigation.

Background

Spatial registration is the most critical technique for surgical navigation systems and can be generally classified into point registration and markerless point registration. Point registration generally requires the attachment or placement of marker points on the surface of a navigation target (e.g., patient, manikin, etc.) and subsequent surface scanning, which increases the radiation dose and surgical costs of the navigation target, specifically prior to surgical navigation. With respect to point registration, no-mark point registration completes space registration by collecting point clouds of the navigation target surface, and thus, no-mark point registration is also called surface registration.

Surface registration is generally safer and more reliable, does not cause additional trauma to the navigation target due to placement of the marker points, and does not cause interruption of surgical navigation due to falling of the marker points. Currently, surface-based surgical navigation mostly uses anatomical landmarks for coarse registration, providing a suitable initial pose for fine registration. However, manually selecting anatomical landmarks to obtain an initial pose increases registration time and does not enable automatic registration of surfaces.

Disclosure of Invention

The application provides a registration method, a registration device, electronic equipment and a computer readable storage medium, wherein the method can realize automatic registration of a navigation target surface and shorten registration time of the surface.

In a first aspect, a method of registration is provided, the registration method comprising: determining a first spatial principal axis direction of the image spatial point cloud and a second spatial principal axis direction of the surgical spatial point cloud; correcting the direction of the first space main shaft to obtain a plurality of directions to be selected; determining a first conversion matrix corresponding to each of the multiple directions to be selected according to the multiple directions to be selected and the second space main axis direction; registering the image space point cloud and the operation space point cloud according to the first conversion matrixes corresponding to the multiple candidate directions respectively to obtain second conversion matrixes corresponding to the multiple candidate directions respectively; determining a coarse registration transformation matrix according to the first transformation matrix and the second transformation matrix which correspond to the multiple directions to be selected; and carrying out fine registration on the image space point cloud and the operation space point cloud according to the coarse registration conversion matrix.

With reference to the first aspect, in some possible implementations, the determining the first spatial principal axis direction of the image spatial point cloud and the second spatial principal axis direction of the surgical spatial point cloud includes: downsampling the image space points to obtain a first simplified point cloud; downsampling the operation space point cloud to obtain a second simplified point cloud; determining a first covariance matrix of the first simplified point cloud and a second covariance matrix of the second simplified point cloud; singular value decomposition is carried out on the first covariance matrix to obtain a first principal axis orthogonal matrix corresponding to the first simplified point cloud, and the first principal axis orthogonal matrix represents the first space principal axis direction; and performing singular value decomposition on the second covariance matrix to obtain a second principal axis orthogonal matrix corresponding to the second simplified point cloud, wherein the second principal axis orthogonal matrix represents the second space principal axis direction.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the step of correcting the first spatial principal axis direction to obtain multiple candidate directions includes: determining a main shaft correction matrix by adopting a main shaft direction conversion mode; and respectively correcting the main axis direction of the first space by adopting the main axis correction matrix to obtain the multiple candidate directions.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the step of registering the image space point cloud and the surgical space point cloud according to the first conversion matrices corresponding to the multiple candidate directions to obtain the second conversion matrices corresponding to the multiple candidate directions includes: respectively converting the first simplified point clouds by adopting first conversion matrixes corresponding to the multiple directions to be selected to obtain multiple groups of conversion point clouds; dividing each group of the conversion point cloud and the second simplified point cloud into a group to obtain a plurality of point cloud groups; and registering the conversion point clouds and the second simplified point clouds in each point cloud group respectively to obtain second conversion matrixes corresponding to the multiple candidate directions.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the step of determining a coarse registration transformation matrix according to the first transformation matrix and the second transformation matrix corresponding to each of the multiple candidate directions includes: determining a third conversion matrix corresponding to each of the multiple candidate directions according to the first conversion matrix and the second conversion matrix corresponding to each of the multiple candidate directions; determining a direction to be selected meeting the main shaft judging condition from the multiple directions to be selected; and determining the third conversion matrix corresponding to the direction to be selected meeting the principal axis judgment condition as the coarse registration conversion matrix.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the step of determining, from the multiple candidate directions, a candidate direction that meets a spindle decision condition includes: acquiring registration point pairs corresponding to the image space point cloud and the operation space point cloud; determining target distances corresponding to the multiple candidate directions according to the coordinates of the two points in the registration point pair and the third conversion matrixes corresponding to the multiple candidate directions, wherein the target distances are average Euclidean distances between the two points in the registration point pair; determining the minimum target distance in the target distances corresponding to the multiple candidate directions respectively; and determining the selected direction corresponding to the minimum target distance from the multiple selected directions as the selected direction meeting the main shaft judging condition.

With reference to the first aspect and the foregoing implementation manner, in some possible implementation manners, the acquiring a first set of outer surface points of a navigation target in an operation space; acquiring a second external surface point set of the three-dimensional model of the navigation target in an image space; filtering the first outer surface point set to obtain the operation space point cloud; and carrying out downsampling treatment on the second external surface point set to obtain the image space point cloud.

In a second aspect, there is provided a registration device comprising:

the direction determining module is used for determining a first space main axis direction of the image space point cloud and a second space main axis direction of the operation space point cloud;

the direction correction module is used for correcting the direction of the first space main shaft to obtain a plurality of directions to be selected;

the first calculation module is used for determining a first conversion matrix corresponding to each of the multiple directions to be selected according to the multiple directions to be selected and the second space main axis direction;

the second calculation module is used for registering the image space point cloud and the operation space point cloud according to the first conversion matrixes corresponding to the multiple candidate directions respectively to obtain second conversion matrixes corresponding to the multiple candidate directions respectively;

The third calculation module is used for determining a coarse registration conversion matrix according to the first conversion matrix and the second conversion matrix which correspond to the multiple directions to be selected;

and the point cloud registration module is used for carrying out fine registration on the image space point cloud and the operation space point cloud according to the coarse registration conversion matrix.

With reference to the second aspect, in some possible implementations, the direction determining module includes:

the first downsampling unit is used for downsampling the image space points to obtain a first simplified point cloud;

the second downsampling unit is used for downsampling the operation space point cloud to obtain a second simplified point cloud;

a covariance matrix calculation unit configured to determine a first covariance matrix of the first simplified point cloud and a second covariance matrix of the second simplified point cloud;

a first decomposition unit, configured to perform singular value decomposition on the first covariance matrix to obtain a first principal axis orthogonal matrix corresponding to the first simplified point cloud, where the first principal axis orthogonal matrix represents the first spatial principal axis direction;

and the second decomposition unit is used for carrying out singular value decomposition on the second covariance matrix to obtain a second principal axis orthogonal matrix corresponding to the second simplified point cloud, wherein the second principal axis orthogonal matrix represents the second space principal axis direction.

With reference to the second aspect, in some possible implementations, the direction correction module includes:

the correction matrix construction unit is used for determining a main axis correction matrix by adopting a main axis direction conversion mode;

and the main shaft direction correction unit is used for respectively correcting the main shaft directions of the first space by adopting the main shaft correction matrix to obtain the multiple directions to be selected.

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the foregoing second computing module includes:

the point cloud conversion unit is used for respectively converting the first simplified point clouds by adopting a first conversion matrix corresponding to each of the multiple directions to be selected to obtain multiple groups of conversion point clouds;

a point cloud grouping unit, configured to divide each group of the conversion point cloud and the second simplified point cloud into a group, so as to obtain a plurality of point cloud groups;

and the first conversion matrix calculation unit is used for registering the conversion point clouds and the second simplified point clouds in each point cloud group respectively to obtain second conversion matrixes corresponding to the multiple candidate directions.

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the third computing module includes:

A second conversion matrix calculation unit, configured to determine a third conversion matrix corresponding to each of the multiple directions to be selected according to the first conversion matrix and the second conversion matrix corresponding to each of the multiple directions to be selected;

a direction selecting unit for determining a direction to be selected meeting the main shaft judging condition from the multiple directions to be selected;

and a third conversion matrix calculation unit, configured to determine a third conversion matrix corresponding to the candidate direction that satisfies the principal axis determination condition as the coarse registration conversion matrix.

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the direction selecting unit includes:

a point pair obtaining subunit, configured to obtain a registration point pair corresponding to the image space point cloud and the operation space point cloud;

a distance calculating subunit, configured to determine, according to the coordinates of each of the two points in the registration point pair and the third transformation matrix corresponding to each of the multiple candidate directions, a target distance corresponding to each of the multiple candidate directions, where the target distance is an average euclidean distance between the two points in the registration point pair;

a distance selecting subunit, configured to determine a minimum target distance among target distances corresponding to the multiple directions to be selected;

And the condition determining subunit is used for determining the candidate direction corresponding to the minimum target distance in the multiple candidate directions as the candidate direction meeting the main shaft judging condition.

With reference to the second aspect and the foregoing implementation manner, in some possible implementation manners, the registration device further includes:

a first point set acquisition unit for acquiring a first external surface point set of the navigation target in the operation space;

a second point set acquisition unit configured to acquire a second external surface point set of the three-dimensional model of the navigation target in an image space;

and the third downsampling unit is used for filtering the first outer surface point set to obtain the operation space point cloud, and downsampling the second outer surface point set to obtain the image space point cloud.

In a third aspect, an electronic device is provided that includes a memory and a processor. The memory is for storing executable program code and the processor is for calling and running the executable program code from the memory for causing the electronic device to perform the method of the first aspect or any one of the possible implementations of the first aspect.

In a fourth aspect, there is provided a computer program product comprising: computer program code which, when run on a computer, causes the computer to perform the method of the first aspect or any one of the possible implementations of the first aspect.

In a fifth aspect, a computer readable storage medium is provided, the computer readable storage medium storing computer program code which, when run on a computer, causes the computer to perform the method of the first aspect or any one of the possible implementations of the first aspect.

The registration method, the registration device, the electronic equipment and the computer readable storage medium provided by the embodiment of the application have the following technical effects:

according to the embodiment of the application, the first space principal axis direction of the image space point cloud and the second space principal axis direction of the operation space point cloud are determined, the first space principal axis direction is corrected to obtain multiple candidate directions, the first conversion matrixes corresponding to the multiple candidate directions are determined according to the multiple candidate directions and the second space principal axis direction, the image space point cloud and the operation space point cloud are registered according to the first conversion matrixes corresponding to the multiple candidate directions, the second conversion matrixes corresponding to the multiple candidate directions are obtained, the coarse registration conversion matrix is determined according to the first conversion matrixes corresponding to the multiple candidate directions and the second conversion matrixes, the fine registration of the image space point cloud and the operation space point cloud is performed according to the coarse registration conversion matrix, and the automatic registration of the surface can be achieved without the surface labeling of a navigation target. The image space point cloud and the operation space point are subjected to coarse registration, and then the fine registration is performed, so that the registration time is shortened, the registration accuracy of the fine registration is improved, the registration reliability is enhanced, and a complicated manual registration process is avoided.

Drawings

FIG. 1 shows a schematic flow chart of a registration method provided by an embodiment of the present application;

FIG. 2 shows a schematic view of a plurality of corrected first spatial principal axes directions;

FIG. 3-1 illustrates a surgical spatial point cloud used for coarse registration;

fig. 3-2 illustrates an image space point cloud used for coarse registration;

FIG. 4-1 illustrates a surgical spatial point cloud used for fine registration;

fig. 4-2 shows an image space point cloud used for fine registration;

fig. 5 shows a schematic flow chart of coarse registration;

fig. 6 shows a schematic flow chart of the fine registration;

fig. 7 is a schematic structural view of a registration device according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The technical solutions in the present application will be clearly and thoroughly described below with reference to the accompanying drawings. Wherein, in the description of the embodiments of the present application, "/" means or is meant unless otherwise indicated, for example, a/B may represent a or B: the text "and/or" is merely an association relation describing the associated object, and indicates that three relations may exist, for example, a and/or B may indicate: the three cases where a exists alone, a and B exist together, and B exists alone, and in addition, in the description of the embodiments of the present application, "plural" means two or more than two.

The terms "first," "second," and the like, are used below for descriptive purposes only and are not to be construed as implying or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature.

The surgical navigation system can be divided into two stages, preoperative and intraoperative, wherein the preoperative image space is determined by reconstructing medical images of navigation targets (such as patients, human models and the like) in three dimensions, and the intraoperative position information of reflective marker balls on a surgical tool is detected in real time by using a CT device to determine the surgical space. The key step of the surgical navigation system is the registration of the surgical space to the image space, which directly affects the overall accuracy of the navigation, with the aim of obtaining a conversion relationship between the image space and the surgical space. The accuracy of spatial registration determines to a large extent the positioning accuracy of the surgical navigation system.

Spatial registration is the most critical technique for surgical navigation systems and can be generally classified into point registration and markerless point registration. Point registration generally requires the application or placement of marker points on the surface of the navigation target specifically prior to surgical navigation, followed by surface scanning, a process that increases the radiation dose and surgical cost of the navigation target. With respect to point registration, no-mark point registration completes space registration by collecting point clouds of the navigation target surface, and thus, no-mark point registration is also called surface registration.

Based on the problems in the related art described above, embodiments of the present application provide a registration method, a registration apparatus, an electronic device, and a computer-readable storage medium.

The following is an embodiment of a registration method provided in the embodiments of the present application.

Fig. 1 shows a schematic flow chart of a registration method provided in an embodiment of the present application. As shown in fig. 1, the registration method provided in the embodiment of the present application is applied to a surgical navigation system, and the spatial transformation involved in the registration process implemented by the registration method provided in the embodiment of the present application may be a transformation from a surgical space to an image space, or may be a transformation from an image space to a surgical space. Taking the conversion from an image space to a surgical space as an example, the registration method comprises the following scheme:

S110: a first spatial principal axis direction of the image spatial point cloud and a second spatial principal axis direction of the surgical spatial point cloud are determined.

The registration method provided by the embodiment of the application realizes the registration process: firstly, performing coarse registration of point cloud data and then performing fine registration of the point cloud data, wherein the coarse registration of the point cloud data comprises two steps, namely, performing point cloud data registration by adopting a principal component analysis registration algorithm (Principal Component Analysis, PCA) and performing point cloud data registration by adopting a deep learning registration algorithm (Feature-metric Registration, FMR).

In an exemplary embodiment, the surgical space point cloud refers to a set of surface points of a navigation target under a surgical space, and a space established by the CT apparatus (Computed Tomography, i.e., electronic computed tomography) for real-time tracking of the set of surface points of the navigation target acquired by the surgical tool is referred to as a surgical space, which may be passed through the surgical space coordinate system S _n And (3) representing. The surgical tool is provided with a reflective marking ball, and after the surgical tool contacts with the surface of the navigation target and moves under the surgical space, the CT equipment acquires the position of the reflective marking ball in real time, so that the real-time position information of the surgical tool is obtained, and the surgical space point cloud of the navigation target is obtained. The image space point cloud refers to a surface point set of a three-dimensional model of a navigation target in an image space, a space established by the operation navigation system for three-dimensionally reconstructing the navigation target is called an image space, and the image space can pass through an image space coordinate system S _ct And (3) representing. The navigation target can be a patient, a human model or other things.

After the surgical space point cloud and the image space point cloud of the navigation target are acquired, the surgical space point cloud is denoted as Q, and the image space point cloud is denoted as P.

And (3) registering the P and the Q by adopting a PCA algorithm, wherein the registration method based on the PCA mainly utilizes the principal axis direction of the point cloud data to carry out primary registration on the P and the Q. In the process of carrying out primary registration on the P and the Q, a first space principal axis direction of the image space point cloud and a second space principal axis direction of the operation space point cloud are calculated. The first space main axis direction is the direction of the main axis of the image space point cloud, and the second space main axis direction is the direction of the main axis of the operation space point cloud. The main shaft of the image space point cloud and the main shaft of the operation space point cloud comprise a first main shaft, a second main shaft and a third main shaft, and the first main shaft corresponds toThe X axis corresponds to the second main axis, the Y axis corresponds to the third main axis, and the Z axis corresponds to the third main axis. P is at S _ct The lower spindle comprises a first spindle x _p Second main shaft y _p And a third main axis z _p Q is at S _n The lower spindle comprises a first spindle x _q Second main shaft y _q And a third main axis z _q . Wherein z is _p ＝x _p ×y _p ，z _q ＝x _q ×z _q 。

In a possible implementation manner, S110 includes the following schemes:

Downsampling the image space points to obtain a first simplified point cloud;

downsampling the operation space point cloud to obtain a second simplified point cloud;

determining a first covariance matrix of the first simplified point cloud and a second covariance matrix of the second simplified point cloud;

singular value decomposition is carried out on the first covariance matrix to obtain a first principal axis orthogonal matrix corresponding to the first simplified point cloud, and the first principal axis orthogonal matrix represents the first space principal axis direction;

and performing singular value decomposition on the second covariance matrix to obtain a second principal axis orthogonal matrix corresponding to the second simplified point cloud, wherein the second principal axis orthogonal matrix represents the second space principal axis direction.

In an exemplary embodiment, P is downsampled to obtain a first reduced point cloud, denoted P ₁ The method comprises the steps of carrying out a first treatment on the surface of the Downsampling Q to obtain a second reduced point cloud, represented as Q ₁ . As shown in fig. 3-1, fig. 3-1 shows a surgical space point cloud used for coarse registration, the point cloud in fig. 3-1 being Q ₁ The method comprises the steps of carrying out a first treatment on the surface of the As shown in fig. 3-2, fig. 3-2 shows the image space point cloud used for coarse registration, the point cloud in fig. 3-2 being P ₁ 。

Obtaining P ₁ And Q ₁ Thereafter, P ₁ And Q ₁ As an input of the PCA algorithm, the PCA algorithm is applied to P ₁ And Q ₁ Solving principal components to realize P ₁ And Q is equal to ₁ A first spatial principal axis direction and a second spatial principal axis direction are obtained. By downsampling P and Q, the number of point clouds for coarse registration is reduced, so that the time for coarse registration is shortened, and the time for subsequent fine registration is shortened.

P is subjected to PCA algorithm ₁ And Q ₁ Performing the initial registration includes:

calculation of P ₁ First covariance matrix sum Q of (2) ₁ And (3) calculating main characteristic components according to the covariance matrix, and representing the principal axis direction of the point cloud through the characteristic components. Calculation of P ₁ First covariance matrix sum Q of (2) ₁ The second covariance matrix procedure of (2) is as follows:

calculation of P ₁ And Q ₁ Then determining a first covariance matrix from the first central coordinates and determining a second covariance matrix from the second central coordinates.

M ₁ Represents Q ₁ Point cloud quantity, N ₁ Representing P ₁ Point cloud quantity, N ₁ ＜N，M ₁ < M. The first center coordinate is denoted +.>

The second center coordinate is denoted +.>

The center coordinates are calculated by a formula set (1), and the formula set (1) is as follows:

the covariance matrix is calculated by the formula set (2), the formula set (2) is as follows:

Wherein the first covariance matrix is expressed as

The second covariance matrix is denoted +.>

Calculating to obtain P ₁ First covariance matrix sum Q of (2) ₁ After the second covariance matrix of (2), performing singular value decomposition (Singular Value Decomposition, SVD) on the first covariance matrix to obtain P ₁ The corresponding first principal axis orthogonal matrix and the second covariance matrix are subjected to singular value decomposition to obtain Q ₁ A corresponding second principal axis orthogonal matrix.

The first principal axis orthogonal matrix and the second principal axis orthogonal matrix are calculated by a formula set (3), and the formula set (3) is as follows:

wherein, the liquid crystal display device comprises a liquid crystal display device,

representing a first principal axis orthogonal matrix,>

representing a second principal axis orthogonal matrix,

d represents a diagonal matrix of m x n, the diagonal elements being singular values of C, V represents an orthogonal matrix of n x n, m=3, n=3.

Obtaining P ₁ Corresponding first principal axis orthogonal matrix and Q ₁ After the corresponding second principal axis orthogonal matrix, the first principal axis orthogonal matrix is used for the first principal axis orthogonal matrixThe matrix represents the first spatial principal axis direction and the second spatial principal axis direction is represented by the second principal axis orthogonal matrix, i.e. the first spatial principal axis direction and the second spatial principal axis direction are calculated.

S120: correcting the direction of the main axis of the first space to obtain multiple directions to be selected.

In the PCA algorithm, P is the sum of the ₁ And Q ₁ In the primary registration, there may be a problem in that the main axis is reversed, that is, the direction of the main axis of P is also reversed. After obtaining the first spatial principal axis direction of P and the second spatial principal axis direction of Q, the embodiment of the present application sequentially transforms the first principal axis x of P _p Second main shaft y _p Totally 2 ² And correcting the first space main shaft directions of P based on the 4 main shaft conversion relations to obtain 4 corrected first space main shaft directions, wherein the 4 corrected first space main shaft directions are the multiple candidate directions, and the candidate directions are 4. As shown in fig. 2, fig. 2 shows a schematic diagram of various corrected first spatial principal axis directions, and 4 candidate directions are shown as (a), (b), (c) and (d), respectively, for example, the 1 st candidate direction corresponds to (a), the 2 nd candidate direction corresponds to (b), the 3 rd candidate direction corresponds to (c), and the 4 th candidate direction corresponds to (d).

In a possible implementation manner, S120 includes the following schemes:

determining a main shaft correction matrix by adopting a main shaft direction conversion mode;

and respectively correcting the main axis direction of the first space by adopting the main axis correction matrix to obtain the multiple candidate directions.

In an exemplary embodiment, the pair P is due to the PCA algorithm ₁ And Q ₁ After the transformation, there may be a problem of spindle reversal, and the point cloud may not be necessarily adjusted to a proper pose, so a spindle correction matrix may be constructed, and the first spatial spindle direction of the P is transformed by the spindle correction matrix to solve the problem of spindle reversal. The first spatial principal axis direction of the principal axis correction matrix transformation P is specifically the positive direction of the principal axis of P.

The process for constructing the principal axis correction matrix comprises the following steps: determining a principal axis correction matrix by transforming principal axis directions, i.e. transforming the first principal axis x of P in turn _p And a second main axis y _p Directions, a total of 4 principal axis transformation relationships, wherein the third principal axis z _p ＝x _p ×y _p Thereby obtaining a principal axis correction matrix.

The principal axis correction matrix is represented by the set K,

i.e. K has 4 cases, including K ₁ 、k ₂ 、k ₃ And k ₄ 。

After the first space main axis direction and the second space main axis direction are calculated, k is adopted ₁ 、k ₂ 、k ₃ And k ₄ And respectively calibrating the directions of the main axes of the first space to obtain 4 directions to be selected. Wherein, the liquid crystal display device comprises a liquid crystal display device,

the 1 st candidate direction is expressed as

The 2 nd candidate direction is expressed as

The 3 rd candidate direction is expressed as

The 4 th candidate direction is expressed as

S130: and determining a first conversion matrix corresponding to each of the multiple candidate directions according to the multiple candidate directions and the second space main axis direction.

After obtaining 4 candidate directions, according to the 4 candidate directions and the second space main axis direction, calculating the first conversion matrixes respectively corresponding to the 4 candidate directions, wherein the first conversion matrixes are respectively expressed as V11, V12, V13 and V14. For example, the 1 st candidate direction corresponds to V11, the 2 nd candidate direction corresponds to V12, the 3 rd candidate direction corresponds to V13, and the 4 th candidate direction corresponds to V14.

The first conversion matrix comprises a rotation matrix R1 and a translation vector T1, after 4 candidate directions are obtained, V11 corresponding to the 1 st candidate direction, V12 corresponding to the 2 nd candidate direction, V13 corresponding to the 3 rd candidate direction and V14 corresponding to the 4 th candidate direction are calculated through a formula group (4). The formula set (4) is as follows:

v11 includes a rotation matrix R ₁₁ And translation vector T ₁₁ V12 includes a rotation matrix R ₁₂ And translation vector T ₁₂ V13 includes a rotation matrix R ₁₃ And translation vector T ₁₃ V14 includes a rotation matrix R ₁₄ And translation vector T ₁₄ . For example in V11

The calculation manner of V12, V13 and V14 is the same as that of V11, and will not be described here again.

S140: and registering the image space point cloud and the operation space point cloud according to the first conversion matrixes corresponding to the multiple candidate directions, so as to obtain second conversion matrixes corresponding to the multiple candidate directions.

Registering P and Q according to the first conversion matrixes corresponding to the 4 candidate directionsSpecifically, based on a deep learning registration algorithm (FMR), the first conversion matrix pair P corresponding to each of the 4 candidate directions is used ₁ And Q ₁ Registering to obtain second conversion matrixes corresponding to the 4 candidate directions. The second conversion matrix corresponding to the 1 st candidate direction is denoted as V21, the second conversion matrix corresponding to the 2 nd candidate direction is denoted as V22, the second conversion matrix corresponding to the 3 rd candidate direction is denoted as V23, and the second conversion matrix corresponding to the 4 th candidate direction is denoted as V24.

In a possible implementation manner, the determining S140 includes the following schemes:

respectively converting the first simplified point clouds by adopting first conversion matrixes corresponding to the multiple directions to be selected to obtain multiple groups of conversion point clouds;

dividing each group of the conversion point cloud and the second simplified point cloud into a group to obtain a plurality of point cloud groups;

and registering the conversion point clouds and the second simplified point clouds in each point cloud group respectively to obtain second conversion matrixes corresponding to the multiple candidate directions.

In an exemplary embodiment, P is respectively mapped to each of the 4 candidate directions using a first transformation matrix ₁ Converting to obtain 4 groups of conversion point clouds, wherein the conversion point clouds are expressed as P ₂ . Then, each group P ₂ And Q is equal to ₁ Dividing into one group to obtain 4 point cloud groups, namely group 1P ₂ And Q is equal to ₁ Dividing into one group to obtain a point cloud group 1 and a group 2P ₂ And Q is equal to ₁ Dividing into a group to obtain a point cloud group 2 and a group 3P ₂ And Q is equal to ₁ Dividing into a group to obtain a point cloud group 3 and a group 4P ₂ And Q is equal to ₁ And dividing the point cloud groups into a group to obtain the point cloud group 4.

P in point cloud group 1 by adopting FMR algorithm ₂ And Q is equal to ₁ Registering to obtain a second conversion matrix V21 corresponding to the 1 st direction to be selected;

p in point cloud group 2 by adopting FMR algorithm ₂ And Q is equal to ₁ Registering to obtain a second conversion matrix V22 corresponding to the 2 nd direction to be selected;

p in point cloud group 3 by adopting FMR algorithm ₂ And Q is equal to ₁ Registering to obtain a second conversion matrix V23 corresponding to the 3 rd direction to be selected;

p in point cloud group 4 by adopting FMR algorithm ₂ And Q is equal to ₁ And registering to obtain a second conversion matrix V24 corresponding to the 4 th candidate direction.

P is subjected to FMR algorithm ₂ And Q is equal to ₁ The process of performing the coarse registration is represented by equation (5):

is P ₂ And Q is equal to ₁ Characteristic metric error between, F (P) ₂ ) Is P ₂ Corresponding feature function, ++>

K is the feature dimension, F ()'s is the global feature of the computing point cloud, +.>

The second conversion matrix calculated by the FMR algorithm is formed by R ₂ And T ₂ The fusion is obtained, i.e. the second transformation matrix comprises a rotation matrix R ₂ And translation vector T ₂ 。

S150: and determining a coarse registration transformation matrix according to the first transformation matrix and the second transformation matrix which correspond to the multiple candidate directions.

After the first conversion matrix and the second conversion matrix corresponding to the 4 candidate directions are obtained, selecting a proper candidate direction from the 4 candidate directions as a target main axis direction of P, and calculating a rough registration conversion matrix through the first conversion matrix and the second conversion matrix corresponding to the target main axis direction, wherein the rough registration conversion matrix is also called rough registration application conversion matrix. The coarse registration transformation matrix is denoted as Vs, and the first transformation matrix and the second transformation matrix corresponding to the target principal axis direction are assumed to be V11 and V21, respectively, vs=v11×v21.

In a possible implementation manner, S150 includes the following schemes:

determining a third conversion matrix corresponding to each of the multiple candidate directions according to the first conversion matrix and the second conversion matrix corresponding to each of the multiple candidate directions;

determining a direction to be selected meeting the main shaft judging condition from the multiple directions to be selected;

and determining the third conversion matrix corresponding to the direction to be selected meeting the principal axis judgment condition as the coarse registration conversion matrix.

Because Q and P are obtained through different devices, it is difficult to ensure that the obtained point cloud distribution ranges are completely consistent, and the main axis directions of the two space point clouds are not completely parallel. Therefore, after the point cloud is aligned by adopting the FMR algorithm, the principal axes which are correctly matched with the principal axes of Q in the 4 principal axes corresponding to P are screened out through the principal axis judging condition evaluation which is set in advance, so that the target principal axis direction is obtained.

The third conversion matrix is expressed as V3, and after the first conversion matrix and the second conversion matrix corresponding to the 4 kinds of directions to be selected are obtained, the third conversion matrix corresponding to the 4 kinds of directions to be selected is calculated according to the first conversion matrix and the second conversion matrix corresponding to the 4 kinds of directions to be selected. For example, the third conversion matrix corresponding to the 1 st candidate direction is denoted as V31, v31=v11×v21; the third conversion matrix corresponding to the 2 nd candidate direction is denoted as V32, v32=v12×v22; the third conversion matrix corresponding to the 3 rd candidate direction is denoted as V33, v33=v13×v23; the third conversion matrix corresponding to the 4 th candidate direction is denoted as V34, v34=v14×v24.

And taking the main shaft judging condition as a selecting condition of the target main shaft direction, acquiring the to-be-selected direction meeting the main shaft judging condition from 4 to-be-selected directions, determining the to-be-selected direction meeting the main shaft judging condition as the target main shaft direction, and determining a third conversion matrix calculated through a first conversion matrix and a second conversion matrix corresponding to the target main shaft direction as a coarse registration conversion matrix. For example, the target principal axis direction is the 2 nd candidate direction, and the third conversion matrix corresponding to the 2 nd candidate direction is V32, that is, vs=v32.

In a possible implementation manner, the determining the candidate direction that meets the spindle determination condition from the multiple candidate directions includes the following schemes:

acquiring registration point pairs corresponding to the image space point cloud and the operation space point cloud;

determining target distances corresponding to the multiple candidate directions according to the coordinates of the two points in the registration point pair and the third conversion matrixes corresponding to the multiple candidate directions, wherein the target distances are average Euclidean distances between the two points in the registration point pair;

determining the minimum target distance in the target distances corresponding to the multiple candidate directions respectively;

and determining the selected direction corresponding to the minimum target distance from the multiple selected directions as the selected direction meeting the main shaft judging condition.

Each registration point pair comprises two mutually paired points in two point clouds, and the registration point pair specifically comprises P in each point cloud group after registration by an FMR algorithm ₂ And Q is equal to ₁ And obtaining 4 sets of registration point pairs, namely, 4 sets of registration point pairs corresponding to the 4 directions to be selected respectively. After a group of registration point pairs corresponding to the 4 candidate directions are obtained, coordinates of two points in each registration point pair can be obtained, and then the target distances corresponding to the 4 candidate directions are calculated according to the coordinates of the two points in each registration point pair and a third conversion matrix corresponding to the 4 candidate directions based on a calculation formula of the average Euclidean distance. The target distance is the average Euclidean distance between two points in the registration point pair. The calculation formula of the average Euclidean distance is shown as a formula (6):

is->

At S _n Corresponding points below->

Q ₁ Is obtained by tracking a surgical tool through CT equipment, has random acquisition range, inaccurate acquisition environment and Q ₁ The shape of the formed navigation object cannot be guaranteed to be P ₁ The shape of the three-dimensional model formed is sufficiently similar to result in S _RE The alignment of the main shaft in the same direction is not necessarily correctly judged. The point cloud is further aligned based on the FMR algorithm, so that the position difference in space is further shortened, and the registration time is shortened.

In order to screen out the spindle which is correctly matched with the spindle Q in the 4 spindles corresponding to P, the target spindle direction is obtained, the minimum average Euclidean distance between two points in the registered point pair is used as a spindle judgment condition, and the spindle which is correctly matched with the spindle Q in the 4 spindles corresponding to P is screened out according to the minimum average Euclidean distance, and the target spindle direction is obtained.

After the target distances corresponding to the 4 directions to be selected are calculated, 4 target distances are obtained, the minimum target distance is determined from the 4 target distances, and then the spindle pair corresponding to the minimum target distance in the 4 target distances is determined as the spindle pair meeting the spindle judging condition, so that the target spindle direction is obtained.

For example, based on equation (6) and P in point cloud group 1 ₂ And Q is equal to ₁ The calculated target distance corresponding to the 1 st candidate direction is S1 _RE The method comprises the steps of carrying out a first treatment on the surface of the Based on formula (6) and P in point cloud group 2 ₂ And Q is equal to ₁ The calculated target distance corresponding to the 2 nd candidate direction is S2 _RE The method comprises the steps of carrying out a first treatment on the surface of the Based on formula (6) and P in point cloud group 3 ₂ And Q is equal to ₁ The calculated target distance corresponding to the 3 rd candidate direction is S3 _RE The method comprises the steps of carrying out a first treatment on the surface of the Based on formula (6) and P in point cloud group 4 ₂ And Q is equal to ₁ Meter (D)The calculated target distance corresponding to the 4 th candidate direction is S4 _RE 。

Calculation of S1 _RE 、S2 _RE 、S3 _RE And S4 _RE Thereafter, assume S4 _RE The 4 th candidate direction is the candidate direction satisfying the main axis judgment condition, namely the 4 th candidate direction is the target main axis direction,

s160: and carrying out fine registration on the image space point cloud and the operation space point cloud according to the coarse registration conversion matrix.

As shown in fig. 4-1, fig. 4-1 shows the surgical space point cloud used for fine registration, the point cloud in fig. 4-1 being Q; as shown in fig. 4-2, fig. 4-2 shows the image space point cloud used for fine registration, the point cloud in fig. 4-2 being P. After the coarse registration transformation matrix is obtained, transforming the P by adopting the coarse registration transformation matrix to obtain a transformation point cloud corresponding to the P, wherein the transformation point cloud is expressed as P ₃ 。

Obtaining P ₃ Then, an ICP (Iterative Closest Point, point cloud registration algorithm) algorithm is adopted for P ₃ And Q for fine registration. Wherein ICP algorithm is implemented by searching P ₃ And Q, and by iteratively minimizing the average difference between two spatially corresponding points, for P ₃ With Q, the registration principle of the ICP algorithm is expressed by the following formula (7):

wherein V is ₄ Representing P ₃ And a globally minimized conversion matrix of Q,

q _i ∈Q。

in a possible implementation manner, before the step S110, the registration method further includes the following scheme:

acquiring a first outer surface point set of a navigation target in an operation space;

acquiring a second external surface point set of the three-dimensional model of the navigation target in an image space;

filtering the first outer surface point set to obtain the operation space point cloud;

and carrying out downsampling treatment on the second external surface point set to obtain the image space point cloud.

The first set of exterior surface points is acquired by the CT apparatus tracking the surgical tool during movement of the navigation target surface, the second set of exterior surface points is a set of points of a three-dimensional model surface of the navigation target, and the first set of exterior surface points is denoted as Q ₀ The second set of outer surface points is denoted as P ₀ . For Q ₀ Filtering to obtain Q, and comparing P ₀ And (3) performing downsampling to obtain P, so that the number of point clouds for coarse registration is reduced, and the coarse registration time is shortened. Wherein the filtering process refers to the process of Q ₀ Denoising and outlier removal are performed.

According to the technical scheme, the first space principal axis direction of the image space point cloud and the second space principal axis direction of the operation space point cloud are determined, the first space principal axis direction is corrected to obtain multiple candidate directions, the first conversion matrixes corresponding to the multiple candidate directions are determined according to the multiple candidate directions and the second space principal axis direction, the image space point cloud and the operation space point cloud are registered according to the first conversion matrixes corresponding to the multiple candidate directions, the second conversion matrixes corresponding to the multiple candidate directions are obtained, the coarse registration conversion matrix is determined according to the first conversion matrixes corresponding to the multiple candidate directions and the second conversion matrixes, and the fine registration of the image space point cloud and the operation space point cloud is performed according to the coarse registration conversion matrix. The image space point cloud and the operation space point are subjected to coarse registration, and then the fine registration is performed, so that the registration time is shortened, the registration accuracy of the fine registration is improved, the registration reliability is enhanced, and a complicated manual registration process is avoided.

In the specific technical scheme of the embodiment, firstly, the PCA algorithm is adopted to register the image space point cloud and the operation space point cloud, a first space main axis direction of the image space point cloud and a second space main axis direction of the operation space point cloud are obtained, then the first space main axis direction is corrected to obtain multiple candidate directions, then a first conversion matrix corresponding to each of the multiple candidate directions is calculated according to the multiple candidate directions and the second space main axis direction, then the operation space point cloud and the image space point cloud are registered based on the FMR algorithm, and the utilization of the rapid registration characteristics of the PCA algorithm and the FMR algorithm is realized, and the problem of long registration time of the ICP algorithm is improved. And constructing a principal axis correction matrix in the process of registering the image space point cloud and the operation space point by the PCA algorithm, solving the problem of principal axis reversal by the principal axis correction matrix, providing a reliable initial value for the ICP algorithm, completing coarse registration without manually selecting anatomical mark points, realizing automatic registration of the surface without the mark points, avoiding trauma to a navigation target during surface registration, and having the advantages of safety and reliability.

The following is another embodiment of a registration method provided in the embodiments of the present application.

As shown in fig. 5, fig. 5 shows a schematic flow chart of coarse registration, as shown in fig. 6, and fig. 6 shows a schematic flow chart of fine registration. Referring to fig. 5 and fig. 6, in the registration method provided by the embodiment of the present application, coarse registration of point cloud data is performed first, then fine registration of point cloud data is performed, the coarse registration of point cloud data includes two steps, the first step is to perform point cloud data registration by adopting a PCA algorithm, and the second step is to perform point cloud data registration by adopting an FMR algorithm. The specific registration process is as follows:

acquiring an original CT sequence of a three-dimensional model of a navigation target in an image space, extracting a point cloud from the original CT sequence, and obtaining a second external surface point set P ₀ For P ₀ Downsampling to obtain an image space point cloud P, and downsampling to obtain a first simplified point cloud P ₁ 。

Acquiring a first set Q of external surface points of a navigation target in a surgical space ₀ For Q ₀ Filtering to obtainThe operation space point cloud Q is subjected to downsampling to obtain a second simplified point cloud Q ₁ 。P ₁ And Q ₁ Is the spatial point cloud to be registered in fig. 5.

Obtaining P ₁ And Q ₁ Then, PCA algorithm is adopted for Q ₁ And P ₁ Initial registration is performed, PCA algorithm pairs P ₁ And Q ₁ Calculation of P during registration ₁ First spatial principal axis direction and Q of (2) ₁ And correcting the first space main axis direction through the constructed main axis correction matrix to obtain 4 corrected first space main axis directions, wherein each corrected first space main axis direction is called a to-be-selected direction, namely 4 to-be-selected directions exist, and then calculating a first conversion matrix corresponding to each of the multiple to-be-selected directions according to the 4 to-be-selected directions and the second space main axis direction. In the PCA algorithm, Q is the sum of ₁ And P ₁ After the initial registration is completed, a first conversion matrix corresponding to each of the 4 candidate directions is obtained.

After obtaining the first conversion matrixes corresponding to the 4 kinds of the candidate directions, adopting the first conversion matrixes corresponding to the 4 kinds of the candidate directions to respectively correspond to P ₁ Converting to obtain P ₁ A corresponding 4 sets of conversion point clouds, where the conversion point clouds are denoted as P ₂ Then each group P ₂ And Q is equal to ₁ Dividing into a group to obtain 4 point cloud groups.

Then, adopting FMR algorithm to respectively obtain P in each point cloud group ₂ And Q is equal to ₁ Registering to obtain second conversion matrixes corresponding to each candidate direction, so as to obtain second conversion matrixes corresponding to the 4 candidate directions.

Next, a third conversion matrix, denoted as V3, corresponding to each of the 4 candidate directions is calculated from the first and second conversion matrices corresponding to each of the 4 candidate directions. And (3) carrying out principal axis correction according to the formula (5), screening out the to-be-selected directions meeting principal axis judgment conditions from the 4 to-be-selected directions to obtain a target principal axis direction, and then calculating a coarse registration conversion matrix according to a first conversion matrix and a second conversion matrix corresponding to the target principal axis direction.

After the coarse registration conversion matrix is obtained, converting P by adopting the coarse registration conversion matrix to obtain a conversion matrix corresponding to P, wherein the conversion matrix corresponding to P is expressed as P ₃ Finally, performing fine registration P by adopting ICP algorithm ₃ And Q for fine registration.

The point cloud data is roughly registered by adopting the PCA algorithm and the FMR algorithm, so that the utilization of the rapid registration characteristics of the PCA algorithm and the FMR algorithm is realized, and the problem of long time consumption of ICP algorithm registration can be improved. And the problem of main shaft reversal is solved by carrying out main shaft correction, thereby providing a reliable initial value for ICP algorithm, completing coarse registration without manually selecting anatomical mark points, realizing automatic registration of the surface without mark points, avoiding trauma to navigation targets during surface registration, and having the advantages of safety and reliability.

The following are device embodiments of the present disclosure that may be used to perform method embodiments of the present disclosure. For details not disclosed in the embodiments of the apparatus of the present disclosure, please refer to the embodiments of the method of the present disclosure.

Fig. 7 shows a schematic structural diagram of a registration device according to an embodiment of the present application. Illustratively, as shown in fig. 7, the registration apparatus 700 includes:

a direction determining module 710, configured to determine a first spatial principal axis direction of the image spatial point cloud and a second spatial principal axis direction of the surgical spatial point cloud;

the direction correction module 720 is configured to correct the direction of the first spatial main axis to obtain multiple directions to be selected;

a first calculation module 730, configured to determine a first conversion matrix corresponding to each of the multiple candidate directions according to the multiple candidate directions and the second spatial principal axis direction;

the second calculation module 740 is configured to register the image space point cloud and the operation space point cloud according to the first transformation matrices corresponding to the multiple candidate directions, so as to obtain second transformation matrices corresponding to the multiple candidate directions;

a third calculation module 750, configured to determine a coarse registration transformation matrix according to the first transformation matrix and the second transformation matrix corresponding to each of the multiple candidate directions;

The point cloud registration module 760 is configured to perform fine registration on the image space point cloud and the surgical space point cloud according to the coarse registration transformation matrix.

In a possible implementation manner, the direction determining module 710 includes:

In one possible implementation, the direction correction module 720 includes:

In a possible implementation manner, the second calculating module 740 includes:

In a possible implementation manner, the third computing module 750 includes:

In one possible implementation manner, the direction selecting unit includes:

In a possible implementation manner, the registration device 700 further includes:

It should be noted that, in the registration apparatus provided in the foregoing embodiment, only the division of the functional modules is used for illustration when the registration method is executed, and in practical application, the foregoing functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the registration device and the registration method provided in the foregoing embodiments belong to the same concept, so for details not disclosed in the embodiments of the present disclosure, please refer to the embodiments of the registration method described in the present disclosure, and the details are not repeated here.

The foregoing embodiment numbers of the present disclosure are merely for description and do not represent advantages or disadvantages of the embodiments.

Fig. 8 shows a schematic structural diagram of an electronic device according to an embodiment of the present application.

Illustratively, as shown in FIG. 8, the electronic device 800 includes: a memory 801 and a processor 802, wherein the memory 801 stores executable program code 8011, and the processor 802 is configured to invoke and execute the executable program code 8011 to perform a registration method.

In this embodiment, the electronic device may be divided into functional modules according to the above method example, for example, each functional module may be corresponding to one processing module, or two or more functions may be integrated into one processing module, where the integrated modules may be implemented in a hardware form. It should be noted that, in this embodiment, the division of the modules is schematic, only one logic function is divided, and another division manner may be implemented in actual implementation.

In the case of dividing each function module with corresponding each function, the electronic device may include: the device comprises a direction determining module, a direction correcting module, a first calculating module, a second calculating module, a third calculating module, a point cloud registering module and the like. It should be noted that, all relevant contents of each step related to the above method embodiment may be cited to the functional description of the corresponding functional module, which is not described herein.

The electronic device provided in this embodiment is configured to perform the above-described registration method, so that the same effects as those of the above-described implementation method can be achieved.

In case an integrated unit is employed, the electronic device may comprise a processing module, a memory module. The processing module can be used for controlling and managing the actions of the electronic equipment. The memory module may be used to support the electronic device in executing, inter alia, program code and data.

Wherein the processing module may be a processor or controller that may implement or execute the various illustrative logical blocks, modules, and circuits described in connection with the present disclosure. A processor may also be a combination of computing functions, e.g., including one or more microprocessors, digital signal processing (digital signal processing, DSP) and microprocessor combinations, etc., and a memory module may be a memory.

The present embodiment also provides a computer readable storage medium having stored therein computer program code which, when run on a computer, causes the computer to perform the above-described related method steps for realizing a registration method in the above-described embodiments.

The present embodiment also provides a computer program product which, when run on a computer, causes the computer to perform the above-mentioned related steps to implement a registration method in the above-mentioned embodiments.

In addition, the electronic device provided by the embodiment of the application can be a chip, a component or a module, and the electronic device can comprise a processor and a memory which are connected; the memory is used for storing instructions, and when the electronic device runs, the processor can call and execute the instructions to enable the chip to execute one of the registration methods in the above embodiment.

The electronic device, the computer readable storage medium, the computer program product or the chip provided in this embodiment are used to execute the corresponding method provided above, so that the beneficial effects thereof can be referred to the beneficial effects in the corresponding method provided above, and will not be described herein.

It will be appreciated by those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional modules is illustrated, and in practical application, the above-described functional allocation may be performed by different functional modules according to needs, i.e. the internal structure of the apparatus is divided into different functional modules to perform all or part of the functions described above.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of modules or units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another apparatus, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The foregoing is merely specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes and substitutions are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A registration method, the registration method comprising:

Determining a first spatial principal axis direction of the image spatial point cloud and a second spatial principal axis direction of the surgical spatial point cloud;

correcting the direction of the first space main shaft to obtain a plurality of directions to be selected;

determining a first conversion matrix corresponding to each of the multiple directions to be selected according to the multiple directions to be selected and the second space main axis direction;

registering the image space point cloud and the operation space point cloud according to the first conversion matrixes corresponding to the multiple directions to be selected respectively to obtain second conversion matrixes corresponding to the multiple directions to be selected respectively;

determining a coarse registration transformation matrix according to the first transformation matrix and the second transformation matrix which correspond to the multiple directions to be selected respectively;

and carrying out fine registration on the image space point cloud and the operation space point cloud according to the coarse registration conversion matrix.

2. The registration method according to claim 1, wherein the step of determining a first spatial principal axis direction of the image spatial point cloud and a second spatial principal axis direction of the surgical spatial point cloud comprises:

downsampling the image space points to obtain a first simplified point cloud;

downsampling the surgical space point cloud to obtain a second simplified point cloud;

Determining a first covariance matrix of the first reduced point cloud and a second covariance matrix of the second reduced point cloud;

singular value decomposition is carried out on the first covariance matrix to obtain a first principal axis orthogonal matrix corresponding to the first simplified point cloud, wherein the first principal axis orthogonal matrix represents the first space principal axis direction;

3. The registration method according to claim 2, wherein the step of correcting the first spatial principal axis direction to obtain a plurality of candidate directions comprises:

and respectively correcting the main axis directions of the first space by adopting the main axis correction matrix to obtain the multiple directions to be selected.

4. The registration method according to claim 2, wherein the step of registering the image space point cloud and the operation space point cloud according to the first transformation matrices corresponding to the multiple candidate directions, and obtaining the second transformation matrices corresponding to the multiple candidate directions includes:

dividing each group of the conversion point cloud and the second simplified point cloud into one group to obtain a plurality of point cloud groups;

registering the conversion point clouds and the second simplified point clouds in each point cloud group respectively to obtain second conversion matrixes corresponding to the multiple candidate directions.

5. The registration method according to any one of claims 1 to 4, wherein the step of determining a coarse registration transformation matrix from the first transformation matrix and the second transformation matrix, which correspond to each of the plurality of candidate directions, includes:

determining a third conversion matrix corresponding to each of the multiple directions to be selected according to the first conversion matrix and the second conversion matrix corresponding to each of the multiple directions to be selected;

determining a direction to be selected meeting a main shaft judging condition from the multiple directions to be selected;

and determining a third conversion matrix corresponding to the direction to be selected meeting the main shaft judging condition as the coarse registration conversion matrix.

6. The registration method according to claim 5, wherein the step of determining a candidate direction satisfying a principal axis determination condition from the plurality of candidate directions includes:

determining the minimum target distance in the target distances corresponding to the multiple directions to be selected respectively;

7. The registration method according to any one of claims 1 to 4, wherein before the step of determining the target principal axis direction from the image space point cloud and the operation space point cloud, the registration method further comprises:

8. A registration device, the registration device comprising:

the second calculation module is used for registering the image space point cloud and the operation space point cloud according to the first conversion matrixes corresponding to the multiple directions to be selected respectively to obtain second conversion matrixes corresponding to the multiple directions to be selected respectively;

the third calculation module is used for determining a coarse registration conversion matrix according to the first conversion matrix and the second conversion matrix which correspond to the multiple directions to be selected respectively;

9. An electronic device, the electronic device comprising:

A memory for storing executable program code;

a processor for calling and running the executable program code from the memory, causing the electronic device to perform the method of any one of claims 1 to 7.

10. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed, implements the method according to any of claims 1 to 7.