A registration method for mouth scan data and CBCT data fusion
By automatically capturing local surfaces for ICP registration and combining it with feature point registration, the problems of low registration accuracy and complex operation in the fusion of oral scan data and CBCT data are solved, achieving high-precision data fusion with low hardware requirements.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- LANCET ROBOTICS CO LTD
- Filing Date
- 2022-08-30
- Publication Date
- 2026-06-16
AI Technical Summary
In the existing technology, the registration method for fusing oral scan data and CBCT data has problems such as insufficient registration accuracy, inconvenient operation and high hardware requirements. In particular, traditional methods rely on initial pose, manual point selection is cumbersome and require a large amount of training data.
A method for automatically extracting local surfaces for local ICP registration is adopted. The initial transformation matrix is obtained through feature point registration, and ICP fine registration is performed by combining local surface data. Automatic extraction of local surface data replaces traditional manual point selection. The number of vertices in the surface data is used to determine the target and the starting surface, and the registration matrix is adjusted to reduce hardware requirements.
It achieves high-precision data fusion, simplifies the operation process, reduces the requirements for computer hardware performance, avoids the registration error caused by the difference in initial pose and surface point density in traditional methods, and does not require a large amount of training data.
Smart Images
Figure CN115471529B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of oral medical technology, and in particular to a registration method for fusing intraoral scan data and CBCT data. Background Technology
[0002] Fusion of intraoral scan and CBCT data: In the digital processing of teeth, it is usually necessary to acquire both intraoral scan and CBCT (cone beam CT) data. Intraoral scan data is relatively more accurate but does not include root data, while CBCT data has more noise but contains complete tooth data. Therefore, to obtain relatively complete tooth data and improve the accuracy of orthodontic treatment, it is necessary to accurately fuse intraoral scan and CBCT data. Existing registration methods include the following:
[0003] Traditional feature point registration involves users selecting at least three matching points on both the intraoral scan and CBCT data, and then using a feature point registration algorithm to achieve registration. The drawback is that intraoral scan and CBCT data belong to different modalities, making it difficult to find two accurately corresponding matching point sets, resulting in low registration accuracy and poor data fusion performance.
[0004] Traditional iterative nearest-point registration (hereinafter referred to as ICP registration) finds the set of points on the CBCT surface reconstruction that are closest to the intraoral scan data, and then moves the intraoral scan data to the point closest to that set. This process of finding the point set and moving the data is repeated until the CBCT surface reconstruction and the intraoral scan data are sufficiently close. The disadvantages are: 1. It is highly dependent on the initial pose; a poor initial pose may lead to falling into a local optimum, resulting in poor registration; 2. It is highly dependent on the geometric similarity between the CBCT surface reconstruction and the intraoral scan data; if the geometric similarity is low (CBCT contains tooth roots and other noise), it is very easy to fall into a local optimum, resulting in poor registration.
[0005] Feature point registration combined with ICP registration involves the user selecting at least three matching points on both the ground scan data and CBCT data, and then using a feature point registration algorithm to achieve coarse registration. Based on the coarse registration, the user then selects several more points in areas where the registration result is poor for ICP registration. The disadvantage is that the user needs to select point sets twice, making the operation relatively cumbersome.
[0006] AI-assisted segmentation and registration, such as CN114549604A, uses AI to perform CBCT segmentation and collect feature points for registration. Its disadvantages are: it requires a large amount of training data and has high hardware performance requirements.
[0007] In summary, existing registration methods suffer from drawbacks such as insufficient registration accuracy, inconvenient operation, and the need for large amounts of training data, resulting in high hardware requirements. To address these issues, we propose a registration method for fusing intraoral scan data and CBCT data. Summary of the Invention
[0008] Based on the technical problems existing in the background technology, this invention proposes a registration method for fusing oral scan data and CBCT data. By automatically extracting local surfaces for local ICP registration, the method has high registration accuracy, is easy to operate, does not require a large amount of training data, and has low hardware requirements. It solves the problems of insufficient registration accuracy, inconvenient operation, and high hardware requirements of existing registration methods.
[0009] This invention provides the following technical solution: a registration method for fusing intraoral scan data and CBCT data, comprising the following steps:
[0010] S1. Read the oral scan data and CBCT data, and extract the three-dimensional geometric figures from the oral scan data and the three-dimensional images from the CBCT data;
[0011] S2, CBCT data surface reconstruction: Extracting the three-dimensional geometry of the teeth from the three-dimensional images of CBCT data. The geometry is composed of closed triangular meshes.
[0012] S3. The user selects a feature point set. In the three-dimensional geometry of step S2, feature point set 1 is selected. In step S1, the feature points are sequentially selected at the closest positions to the three-dimensional geometry of the scanning data and feature point set 1 to obtain feature point set 2.
[0013] S4. Feature point registration: Using feature point set 2 as the starting point group and feature point set 1 as the target point group, the registration yields transformation matrix 1.
[0014] S5. Local surface data acquisition: Local surface data is automatically acquired from the three-dimensional geometric surface of the scanning data in S1 and stored in the surface data container for subsequent ICP fine registration.
[0015] S6. Local surface ICP registration: Use transformation matrix 1 to perform spatial transformation on the surface data container to generate two containers for storing surface geometric data. Use ICP registration to register the two containers for storing surface geometric data to obtain matrix 2. Adjust the obtained matrix 2 to obtain matrix 3.
[0016] S7. Apply the registration matrix, calculate matrix 3 by left multiplying matrix 1 to obtain matrix 4, and use matrix 4 to perform spatial transformation on the three-dimensional geometry of the oral scan data in step S1. Then the fusion of oral scan data and CBCT data is completed.
[0017] Preferably, the step S1 of reading the oral scan data and CBCT data specifically includes:
[0018] The user specifies the oral scan data and CBCT data to be fused as input, where:
[0019] Oral scan data is a geometric file based on triangular meshes, which contains shape information of the patient's crown surface and part of the gingival surface. The three-dimensional geometry extracted from it is referred to as "ios".
[0020] CBCT data is a series of continuous DICOM files containing three-dimensional information about a patient's teeth. The three-dimensional images extracted from these files are referred to as "CBCT_image".
[0021] Preferably, in step S2, the three-dimensional geometry of the tooth is extracted from the CBCT_image three-dimensional image. This geometry is composed of a closed triangular mesh, and each pixel representing a tooth in the CBCT_image must be surrounded by the mesh representing the tooth surface. The specific process is as follows:
[0022] The threshold "t" is set as the grayscale value of tooth enamel in CBCT images;
[0023] Iterate through each pixel in CBCT_image. If the grayscale value of the pixel is greater than or equal to t, it is determined to fall inside the surface grid and is assigned to the point set "p_in"; if the grayscale value of the pixel is less than t, it is determined to fall outside the surface grid and is assigned to the point set "p_out".
[0024] Extract all points in p_in that are adjacent to p_out to form a point set "p_surface", which serves as the vertices surrounding the mesh;
[0025] Connecting adjacent points in p_surface yields triangular patches. The closed mesh formed by all the triangular patches is the three-dimensional geometry containing the shape of the tooth surface, which will be referred to as "reconstructed_surface" thereafter.
[0026] Preferably, in step S3,
[0027] Users sequentially select no fewer than three feature points on the tooth surface of the reconstructed_surface. This set of feature points will be referred to as "landmark_target".
[0028] On the surface of a tooth in iOS, the user sequentially selects feature points at the closest positions corresponding to the landmark_target. This set of feature points is then referred to as "landmark_src".
[0029] Preferably, in step S4,
[0030] The feature point registration method is used to register the point sets landmark_target and landmark_src, where:
[0031] landmark_src serves as the starting point group;
[0032] landmark_target is a group of target points;
[0033] The 4x4 transformation matrix obtained through registration is referred to as "T_landmark".
[0034] Preferably, in step S5, local surface data is automatically captured on the IOS surface for subsequent ICP fine registration, as follows:
[0035] Create a container named clipped_surface to store local surface data;
[0036] Iterate through landmark_src and perform the following operation on each point p:
[0037] With p as the center, set the side length value to generate a cube "tmpCube";
[0038] Compare ios and tmpCube, extract the portion of ios that falls into tmpCube, and store it in clipped_surface.
[0039] Preferably, the local surface ICP registration in step S6 is as follows:
[0040] Use the transformation matrix T_landmark to perform a spatial transformation on clipped_surface;
[0041] Two containers named "icp_source" and "icp_target" are generated to store the surface geometry data.
[0042] Compare the number of vertices contained in clipped_surface and reconstructed_surface: pNum_clipped_surface and pNum_reconstructed_surface
[0043] If pNum_clipped_surface >= pNum_reconstructed_surface, assign pNum_clipped_surface to icp_target and assign pNum_reconstructed_surface to icp_source;
[0044] Otherwise, assign pNum_clipped_surface to icp_source and assign pNum_reconstructed_surface to icp_target;
[0045] Use the ICP registration method to register icp_source and icp_target, where:
[0046] icp_source serves as the starting surface for registration;
[0047] icp_target serves as the target surface for registration;
[0048] Name the obtained 4X4 transformation matrix as "T_tmp";
[0049] Adjust the obtained registration matrix T_tmp:
[0050] If pNum_clipped_surface >= pNum_reconstructed_surface, calculate the inverse matrix of T_tmp and denote it as "T_icp";
[0051] If pNum_clipped_surface < pNum_reconstructed_surface, directly assign T_tmp to the matrix "T_icp".
[0052] Preferably, in step S7, name the matrix obtained by multiplying matrix T_icp on the left by matrix T_landmark as T_result;
[0053] Use the transformation matrix T_result to perform a spatial transformation on ios, and至此 the fusion of the oral scan data and the CBCT data is completed.
[0054] This invention provides a registration method for fusing oral scan data and CBCT data. It automatically extracts local surfaces for local ICP registration, replacing the traditional "manual selection of surface points". Users do not need to select a local surface point set, which is convenient to operate. The target surface and the starting surface are determined by judging the number of vertices contained in the surface data, and the corresponding registration matrix is adjusted. This avoids the registration error caused by the large difference in surface point density between CBCT surface reconstruction and oral scan data. It does not require a large amount of training data, has low computer hardware performance requirements, and lowers the threshold for use. Attached Figure Description
[0055] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0057] like Figure 1 As shown, the present invention provides a technical solution: a registration method for fusing intraoral scan data and CBCT data, the steps of which are detailed below:
[0058] S1. Read the oral scan data and CBCT data, specifically including:
[0059] The user specifies the oral scan data and CBCT data to be fused as input, where:
[0060] Intraoralscan data is a geometric file based on triangular meshes, which contains shape information of the patient's crown surface and part of the gingival surface. The three-dimensional geometry extracted from it is referred to as "ios" (intraoralscan).
[0061] CBCT data is a series of continuous DICOM files containing three-dimensional information about a patient's teeth. The three-dimensional images extracted from these files are referred to as "CBCT_image".
[0062] S2 and CBCT data surface reconstruction;
[0063] The 3D geometry of the teeth is extracted from the CBCT_image 3D image. This geometry consists of a closed triangular mesh. Each pixel representing a tooth in the CBCT_image must be surrounded by the mesh representing the tooth surface. The specific process is as follows:
[0064] The threshold "t" is set as the grayscale value of tooth enamel in CBCT images;
[0065] Iterate through each pixel in CBCT_image. If the grayscale value of the pixel is greater than or equal to t, it is determined to fall inside the surface grid and is assigned to the point set "p_in"; if the grayscale value of the pixel is less than t, it is determined to fall outside the surface grid and is assigned to the point set "p_out".
[0066] Extract all points in p_in that are adjacent to p_out to form a point set "p_surface", which serves as the vertices surrounding the mesh;
[0067] Connecting adjacent points in p_surface yields triangular patches. The closed mesh formed by all the triangular patches is the three-dimensional geometry containing the shape of the tooth surface, which will be referred to as "reconstructed_surface" thereafter.
[0068] S3, User selects feature point set;
[0069] Users sequentially select no fewer than three feature points on the tooth surface of the reconstructed_surface. This set of feature points will be referred to as "landmark_target".
[0070] On the surface of a tooth in iOS, the user sequentially selects feature points at the closest positions corresponding to the landmark_target. This set of feature points is then referred to as "landmark_src".
[0071] S4. Feature point registration;
[0072] The feature point registration method is used to register the point sets landmark_target and landmark_src, where:
[0073] landmark_src serves as the starting point group;
[0074] landmark_target is a group of target points;
[0075] The 4x4 transformation matrix obtained through registration is referred to as "T_landmark".
[0076] S5, Local Surface Data Acquisition;
[0077] In step S5, local surface data is automatically captured on the IOS surface for subsequent ICP fine registration, as follows:
[0078] Create a container named clipped_surface to store local surface data;
[0079] Iterate through landmark_src and perform the following operation on each point p:
[0080] With p as the center and 6mm (this value can also be other reasonable values) as the side length, generate a cube "tmpCube";
[0081] Compare ios and tmpCube, extract the portion of ios that falls into tmpCube, and store it in clipped_surface.
[0082] S6. Local surface ICP registration;
[0083] Use the transformation matrix T_landmark to perform a spatial transformation on clipped_surface;
[0084] Two containers named "icp_source" and "icp_target" are generated to store the surface geometry data.
[0085] Compare the number of vertices contained in clipped_surface and reconstructed_surface: pNum_clipped_surface and pNum_reconstructed_surface
[0086] If pNum_clipped_surface >= pNum_reconstructed_surface, assign pNum_clipped_surface to icp_target and pNum_reconstructed_surface to icp_source;
[0087] Otherwise, assign pNum_clipped_surface to icp_source and pNum_reconstructed_surface to icp_target;
[0088] The ICP registration method is used to register icp_source and icp_target, where:
[0089] icp_source serves as the starting surface for registration;
[0090] icp_target is the target surface for registration;
[0091] The 4x4 transformation matrix obtained from the registration is named "T_tmp";
[0092] The adjusted registration matrix T_tmp:
[0093] If pNum_clipped_surface >= pNum_reconstructed_surface, calculate the inverse matrix of T_tmp and denote it as "T_icp".
[0094] If pNum_clipped_surface < pNum_reconstructed_surface, directly assign T_tmp to the matrix "T_icp".
[0095] S7. Apply the registration matrix.
[0096] Calculate the matrix obtained by multiplying matrix T_icp on the left of matrix T_landmark and name it T_result.
[0097] Perform a spatial transformation on ios using the transformation matrix T_result. Thus, the fusion of oral scan data and CBCT data is completed.
[0098] The present invention provides a registration method for the fusion of oral scan data and CBCT data. By automatically extracting local surfaces for local ICP registration, it replaces the traditional "manual selection of surface points". The user does not need to additionally select a local surface point set, which is convenient for operation. By judging the number of vertices contained in the surface data to determine the target surface and the starting surface, and adjusting the corresponding obtained registration matrix, it avoids the registration error caused by the large difference in the surface point density between the CBCT surface reconstruction and the oral scan data. It does not require a large amount of training data and has a low requirement for computer hardware performance, reducing the usage threshold.
[0099] In addition, superimposing ICP registration on the basis of feature point registration ensures that the obtained solution is the global optimal solution, avoids falling into the local optimal solution, and compensates for the problem of low registration accuracy of traditional feature point registration.
[0100] The following specifically elaborates how the registration method provided by the present invention overcomes the defects of the prior art, and then obtains the technical effects of high registration accuracy, convenient operation, no need for a large amount of training data, and low hardware requirements of the present invention, specifically as follows:
[0101] The essence of ICP registration is the registration between two point sets, and its purpose is to calculate the transformation matrix (translation plus rotation) when the source point set A moves to the target point set B and coincides with the target point set B with the highest degree.
[0102] Each iteration of the ICP algorithm searches for the point in point set B that is closest to every element in point set A, forming a temporary point set C. This temporary point set C serves as the target point for marker point registration (point set A is the source point set). The transformation matrix obtained from marker point registration is then applied to point set A. The effect of this is that point set A will become increasingly closer to point set B, and tends to be closer to the sub-parts in point set B that are more similar to itself.
[0103] However, due to the iterative search for the nearest point set within ICP, this algorithm can only guarantee a correct solution if the initial poses of the source point set A and the target point set B are close (close between truly similar sub-parts). Otherwise, the source point set A is easily attracted to a sub-part of the target point set B that is not similar to it, resulting in a large registration error.
[0104] Therefore, ICP registration generally suffers from the problem of initial ICP pose.
[0105] Another problem arises in ICP registration when applying intraoral scan data to CBCT surface reconstruction data: the selection of the "source point set" and the "target point set." In practice, only the crown surfaces of the intraoral scan surface data (consisting of a series of surface points) and the CBCT surface reconstruction data (consisting of a series of surface points) correspond. Furthermore, the CBCT surface data may differ from the intraoral scan data due to adhesive residue on some crowns during the procedure. As a result, only a very limited portion of the intraoral scan surface data and the CBCT surface reconstruction data can correspond. Directly using all intraoral scan surface data points and CBCT surface reconstruction points for ICP registration will inevitably lead to inaccuracies.
[0106] Therefore, ICP registration generally has the problem of requiring high similarity between source and target geometric data.
[0107] Solving the ICP initial pose problem:
[0108] Both traditional methods and the present invention use manual selection of corresponding feature points on the source and target surfaces to solve the problem by feature point registration.
[0109] Solving the point selection problem caused by the high similarity requirements between source and target geometric data in ICP:
[0110] Traditional methods address this problem by manually selecting a large number of points on the surface of the target geometric data that correspond to the source geometric data, and using these points as the true target data for ICP registration.
[0111] This invention employs a manual point selection method to select a small number (at least 3 sets) of corresponding points. This set of corresponding points is used both for initial feature point registration and for subsequent automatic extraction of target data surface points for ICP registration.
[0112] In step S5 of this invention, the feature points selected by the user during feature point registration are fully reused. The surface data of a cube within the selected feature points is then extracted and used as the object for ICP registration. This method can better replace the traditional manual selection method because:
[0113] (The ICP algorithm can accept surface data as input, and it will extract all points on the surface for internal calculations.)
[0114] 1. The feature points selected by the user correspond to the surface data of the oral scan and the surface reconstruction data of the CBCT (where the crown of the tooth was not bonded with adhesive during the operation). Therefore, the surface of the oral scan data within the cube around the feature point will also have a corresponding part on the surface of the CBCT surface reconstruction data.
[0115] 2. Traditional methods only require manually selecting a number of points for ICP registration, which may result in poor representativeness of the selected points and excessive local weighting. However, by extracting a portion of the surface of the scanning data for local ICP registration, all points on the extracted surface will be utilized, greatly reducing the above problems.
[0116] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A registration method for fusing intraoral scan data and CBCT data, characterized in that: Includes the following steps: S1. Read the oral scan data and CBCT data, and extract the three-dimensional geometric figures from the oral scan data and the three-dimensional images from the CBCT data; S2, CBCT data surface reconstruction: Extracting the three-dimensional geometry of the teeth from the three-dimensional images of CBCT data. The geometry is composed of closed triangular meshes. S3. The user selects a feature point set. In the three-dimensional geometry of step S2, feature point set 1 is selected. In step S1, the feature points are sequentially selected at the closest positions to the three-dimensional geometry of the scanning data and feature point set 1 to obtain feature point set 2. S4. Feature point registration: Using feature point set 2 as the starting point group and feature point set 1 as the target point group, the registration yields transformation matrix 1. S5. Local surface data extraction: On the three-dimensional geometric surface of the scanned data in S1, a cube is generated with each point in the feature point set 2 as the center and the side length value is set. The local surface data falling within the cube is automatically extracted and stored in the surface data container. S6. Local surface ICP registration: Use transformation matrix 1 to perform spatial transformation on the surface data container to generate two containers for storing surface geometric data; compare the number of vertices of the local surface in the scanning data and the surface reconstructed by CBCT, and use the surface with more vertices as the target surface for ICP registration and the surface with fewer vertices as the starting surface for ICP registration to obtain matrix 2. Adjust matrix 2 according to the comparison result of the number of vertices to obtain matrix 3. S7. Apply the registration matrix, calculate matrix 3 by left multiplying matrix 1 to obtain matrix 4, and use matrix 4 to perform spatial transformation on the three-dimensional geometry of the oral scan data in step S1. Then the fusion of oral scan data and CBCT data is completed.
2. The registration method for fusing intraoral scan data and CBCT data according to claim 1, characterized in that: The step S1, which involves reading the oral scan data and CBCT data, specifically includes: The user specifies the oral scan data and CBCT data to be fused as input, where: Oral scan data is a geometric file based on triangular meshes, which contains shape information of the patient's crown surface and part of the gingival surface. The three-dimensional geometry extracted from it is referred to as "ios". CBCT data is a series of continuous DICOM files containing three-dimensional information about a patient's teeth. The three-dimensional images extracted from these files are referred to as "CBCT_image".
3. The registration method for fusing intraoral scan data and CBCT data according to claim 2, characterized in that: In step S2, the three-dimensional geometry of the teeth is extracted from the CBCT_image 3D image. This geometry is composed of a closed triangular mesh. Each pixel representing a tooth in the CBCT_image must be surrounded by the mesh representing the tooth surface. The specific process is as follows: The threshold "t" is set as the grayscale value of tooth enamel in CBCT images; Iterate through each pixel in CBCT_image. If the grayscale value of the pixel is greater than or equal to t, it is determined to fall inside the surface grid and is assigned to the point set "p_in"; if the grayscale value of the pixel is less than t, it is determined to fall outside the surface grid and is assigned to the point set "p_out". Extract all points in p_in that are adjacent to p_out to form a point set "p_surface", which serves as the vertices surrounding the mesh; Connecting adjacent points in p_surface yields triangular patches. The closed mesh formed by all the triangular patches is the three-dimensional geometry containing the shape of the tooth surface, which will be referred to as "reconstructed_surface" thereafter.
4. The registration method for fusing intraoral scan data and CBCT data according to claim 3, characterized in that: In step S3 Users sequentially select no fewer than three feature points on the tooth surface of the reconstructed_surface. This set of feature points will be referred to as "landmark_target". On the surface of a tooth on iOS, the user sequentially selects feature points at the closest positions corresponding to the landmark_target. This set of feature points is then referred to as "landmark_src".
5. The registration method for fusing intraoral scan data and CBCT data according to claim 4, characterized in that: In step S4 The feature point registration method is used to register the point sets landmark_target and landmark_src, where: landmark_src serves as the starting point group; landmark_target is a group of target points; The 4x4 transformation matrix obtained through registration is referred to as "T_landmark".
6. The registration method for fusing intraoral scan data and CBCT data according to claim 5, characterized in that: In step S5, local surface data is automatically captured on the IOS surface for subsequent ICP fine registration, as follows: Create a container named clipped_surface to store local surface data; Iterate through landmark_src and perform the following operation on each point p: With p as the center, set the side length value to generate a cube "tmpCube"; Compare ios and tmpCube, extract the portion of ios that falls into tmpCube, and store it in clipped_surface.
7. The registration method for fusing intraoral scan data and CBCT data according to claim 6, characterized in that: The local surface ICP registration in step S6 is as follows: Use the transformation matrix T_landmark to perform a spatial transformation on clipped_surface; Two containers for storing surface geometry data are generated and named "icp_source" and "icp_target". Compare the number of vertices contained in clipped_surface and reconstructed_surface: pNum_clipped_surface and pNum_reconstructed_surface If pNum_clipped_surface >= pNum_reconstructed_surface, assign pNum_clipped_surface to icp_target and pNum_reconstructed_surface to icp_source; Otherwise, assign pNum_clipped_surface to icp_source and pNum_reconstructed_surface to icp_target; The ICP registration method is used to register icp_source and icp_target, where: The icp_source serves as the starting surface for registration; The icp_target serves as the target surface for registration; The 4X4 transformation matrix obtained from registration is named "T_tmp"; Adjust the obtained registration matrix T_tmp: If pNum_clipped_surface >= pNum_reconstructed_surface, calculate the inverse matrix of T_tmp and denote it as "T_icp"; If pNum_clipped_surface < pNum_reconstructed_surface, directly assign T_tmp to the matrix "T_icp".
8. The registration method for fusing intraoral scan data and CBCT data according to claim 7, characterized in that: In step S7 described above, the matrix obtained by multiplying the matrix T_icp by the matrix T_landmark on the left is named T_result; Perform a spatial transformation on ios using the transformation matrix T_result.至此口扫数据与CBCT数据融合完成。 It should be noted that the last part "至此口扫数据与CBCT数据融合完成。" in Chinese seems to be an incomplete or incorrect expression in the context. The English translation is based on the overall logic and the part that can be accurately translated. You may need to check and correct the original Chinese text for a more accurate and complete overall understanding.