Intraoral scanning data correction method based on dynamic feature weight adjustment

Abstract

The present invention discloses a method for correcting intraoral scan data based on dynamic feature weight adjustment. The method comprises the following steps: S1. Initial intraoral scan data collected by an intraoral scanning device is obtained; S2. Full-mouth 3D model data generated by a cone-beam computed tomography (CBCT) device is obtained; S3. 3D model information of individual teeth is extracted; S4. Based on a dynamic feature weight adjustment algorithm, the 3D model information of individual teeth is 3D-aligned with the corresponding tooth model in the CBCT data; S5. A corrected full-mouth model is generated; S6. The physical model generated by the scan is used as a verification standard; and S7. The corrected full-mouth 3D model is output. The present invention solves the problems of accuracy and consistency in oral 3D scan data, significantly improving the accuracy of the corrected full-mouth model, and possesses high practicality and clinical application value.

Description

Technical Field

[0001] The present invention relates to the field of oral technology, and in particular to a method for correcting intraoral scan data based on dynamic feature weight adjustment. Background Art

[0002] With the development of digital medical technology, digital scanning technology in the oral diagnosis and treatment process has become an important tool in oral restoration and treatment. In particular, intraoral scanning technology can quickly obtain three-dimensional point cloud data of patients' teeth and oral soft and hard tissues, providing an efficient and non-invasive diagnosis and treatment method.

[0003] In the existing technology, intraoral scanning equipment can generate high-precision local three-dimensional point cloud data by scanning the patient's teeth and oral soft and hard tissues. However, due to the limitations of oral scanning and the complexity of the scanning environment, the problem of high local accuracy but poor global consistency often occurs. Specifically, intraoral scanning technology can accurately capture the detailed information of a single tooth, but due to the limitation of the scanning range, it cannot maintain sufficient global consistency within the entire oral cavity, especially the correction of the global characteristics of the dental arch curvature is often difficult to be consistent with the actual oral structure.

[0004] To address this issue, cone-beam computed tomography (CBCT) has been introduced into the field of oral 3D imaging in recent years. CBCT can provide oral 3D models with high global accuracy, especially in displaying complex anatomical structures. However, CBCT's resolution and detail capture capabilities are relatively weak. Relying solely on CBCT data for the details of a single tooth is difficult to meet the high requirements for both detail and global accuracy of the oral model.

[0005] At present, there have been some studies on the fusion technology of intraoral scan data and CBCT data, but most of these methods use fixed weight strategies for data fusion, which fails to effectively solve the balance problem between global and local information. When dealing with complex tooth morphology and oral structure, existing fusion methods usually fail to dynamically adjust the accuracy of scan data in different areas, resulting in the fused model may not be able to achieve an ideal balance between details and global accuracy. In addition, the existing technology for global correction and accuracy calibration of intraoral scan data usually relies on manual inspection and post-correction, lacks automated correction and optimization methods, and cannot quickly and efficiently provide correction results that meet clinical applications.

[0006] In summary, the existing technology has significant deficiencies in the fusion of intraoral scanning data and CBCT data, and the calibration of full-oral correction accuracy, making it difficult to meet the needs of high precision, global consistency and efficient correction in the oral medical process. Summary of the Invention

[0007] One purpose of the present invention is to propose a method for correcting intraoral scan data based on dynamic feature weight adjustment. The present invention solves the problems of accuracy and consistency in oral three-dimensional scan data, significantly improves the accuracy of the corrected full oral model, and has high practicality and clinical application value.

[0008] According to an embodiment of the present invention, a method for correcting intraoral scan data based on dynamic feature weight adjustment includes the following steps:

[0009] S1. Acquire initial intraoral scanning data collected by an intraoral scanning device, wherein the initial intraoral scanning data includes three-dimensional point cloud information of teeth and oral soft and hard tissues;

[0010] S2. Obtain full-oral 3D model data generated by a cone-beam computed tomography device. The full-oral 3D model has globally accurate information about the entire curvature.

[0011] S3. Segmenting the intraoral scan data to extract three-dimensional model information of a single tooth;

[0012] S4. Perform 3D registration of the 3D model of a single tooth with the corresponding tooth model in the CBCT data based on a dynamic feature weight adjustment algorithm, and determine the matching relationship of each tooth by minimizing the difference between the two.

[0013] S5. Construct a fusion model to fuse the global curvature information from the CBCT data with the detailed information of individual teeth from the intraoral scan data. Perform global correction on the intraoral scan data by dynamically adjusting the feature weight distribution strategy to generate a full-oral correction model.

[0014] S6. Verify and calibrate the accuracy of the modified full-mouth model using the physical model generated by the scan as a verification standard;

[0015] S7. Output a modified full-mouth 3D model that has the details of individual teeth while maintaining the integrity and consistency of the global dental arch curvature.

[0016] Optionally, the S4 includes the following steps:

[0017] S41. Obtaining the three-dimensional model information M of a single tooth in the intraoral scan data tooth , the three-dimensional model information contains the spatial coordinate information of each point:

[0018]

[0019] in, is the coordinate of a single tooth in three-dimensional space, n tooth is the number of points in the point cloud data of a single tooth model;

[0020] S42. Obtain the corresponding tooth model information M in the CBCT data CBCT , the tooth model information contains the coordinates of each point:

[0021]

[0022] in, is the three-dimensional spatial coordinate of the corresponding tooth model in CBCT, n CBCT is the number of points in the point cloud model in the CBCT data;

[0023] S43. 3D model information M tooth and tooth model information M CBCT Perform preliminary registration, set the initial matching relationship, and use the rigid transformation method to align the 3D model information to the corresponding tooth model in the CBCT data. The preliminary registration is performed using the rotation matrix R and the displacement vector t:

[0024]

[0025] Where R is a 3x3 rotation matrix, t=(t x ,t y ,t z ) is the translation vector, M aligned is the preliminary alignment model after registration, is the coordinate of the tooth model point after rotation and translation;

[0026] S44. Preliminary alignment model M after registration based on dynamic feature weight adjustment algorithm aligned and tooth model information M CBCT Perform registration, optimize the matching relationship between the two, and minimize the difference between the two using the minimization error function:

[0027]

[0028] Among them, ω ij The dynamic feature weight reflects the contribution of each matching point to the error. The dynamic adjustment of the weight coefficient is based on the local feature differences between point clouds. The larger the weight, the greater the impact of the point cloud matching on the overall registration result.

[0029] S45. By minimizing the error function E, an iterative optimization method is used to adjust the rotation matrix R and the translation vector t, thereby optimizing the difference between the initial alignment model and the tooth model information after registration, so that the three-dimensional spatial matching relationship of each tooth is accurate:

[0030]

[0031] Where η is the learning rate, and are the gradients of the rotation matrix and displacement vector, respectively, indicating the rate of change of the error function relative to the rotation matrix and displacement vector;

[0032] S46. Determine the matching relationship of each tooth and generate a final registered single tooth 3D model. The registered single tooth 3D model is aligned with the corresponding tooth portion in the CBCT model to generate a corrected single tooth 3D model M. final :

[0033]

[0034] in, is the point cloud data of the single tooth model after correction and registration, n final is the number of points after registration.

[0035] Optionally, the S5 includes the following steps:

[0036] S53. The tooth model information and the modified three-dimensional model M of the single tooth final Fusion is performed by dynamically adjusting the fusion method of the detail information of a single tooth and the curvature information of the entire mouth through a dynamic feature weight allocation strategy:

[0037]

[0038] Among them, ω k,i for point and The feature weights of and are the distances between the CBCT data points and the corrected 3D model points of a single tooth;

[0039] S54. Based on dynamic feature weight ω k,i The whole-mouth correction model is fused. The curvature information of the whole mouth and the detailed information of the single tooth after correction are fused by the weighted average method to generate the whole-mouth correction model:

[0040]

[0041] in, The final full-mouth modified model includes detailed information from the modified 3D model of a single tooth and global curvature information from the CBCT data.

[0042] S55. Correction of the generated full oral cavity model Optimization is performed and the accuracy of the correction model is further adjusted by minimizing the error function, so that the full-mouth correction model achieves the best balance between details and global consistency:

[0043]

[0044] in, To correct the model error, is the corrected three-dimensional coordinate from a single tooth, To correspond to the coordinates in the CBCT data, minimize Then get the whole mouth correction model

[0045] Optionally, the S6 includes the following steps:

[0046] S61. Obtain the physical model M generated by warehouse scanning physical , the physical model includes the real physical structure of the entire oral cavity and provides a standard for verifying the revised full oral cavity revised model;

[0047] S62. The modified full oral cavity correction model The physical model M generated by warehouse scanning physical Compare and calculate the difference between the two;

[0048] S63. Verify and calibrate the corrected full-mouth correction model by minimizing the error function, and adjust the corrected full-mouth correction model to improve the fit between the full-mouth correction model and the physical model scanned by the warehouse;

[0049] S64. When the error function meets the preset convergence threshold or after a set number of iterations, the final full-oral correction model is determined The final modified three-dimensional model of the entire mouth.

[0050] Optionally, the S61 includes the following steps:

[0051] S611. Obtain the physical model M generated by warehouse scanning physical :

[0052]

[0053] in, The three-dimensional space coordinates of each point in the physical model generated by the bin scan, n physical is the total number of point clouds of the physical model;

[0054] S612. Scan the physical model M generated by the warehouse physical Perform preprocessing to remove noise points caused by scanning equipment errors or environmental factors, and calculate the standard deviation of the point cloud data based on the preset threshold to calculate the noise tolerance of each point:

[0055]

[0056] in, For the p-th point in the warehouse scanning physical model to the reference coordinate (x ref ,y ref ,z ref )’s Euclidean distance;

[0057] S613. Preprocessed physical model M physical Perform coordinate system conversion and compare its coordinate system with the full oral correction model The coordinate system is aligned, and the coordinate system transformation process is done by the rotation matrix R physical and displacement vector t physical conduct:

[0058]

[0059] in, is the physical model after transformation, R physical is the rotation matrix, t physical is the translation vector.

[0060] Optionally, the S62 includes the following steps:

[0061] S621. The corrected full oral model is combined with the physical model generated by the scan. Compare and select corresponding point pairs to ensure alignment of the two in three-dimensional space. Compared with the points in the revised full-mouth model One-to-one correspondence, the matching relationship of each pair of corresponding points is expressed as:

[0062]

[0063] S622. Calculate the distance difference between each pair of corresponding points of the physical model and the modified model, and evaluate the matching accuracy of the two based on the difference;

[0064] S623. Distance differences between all corresponding point pairs Perform weighted summation to obtain the overall difference between the full-mouth modified model and the physical model:

[0065]

[0066] Among them, E compare is the total difference between the modified full-mouth model and the physical model, ω p,i is the weight of each pair of points, indicating the contribution of the point pair to the overall difference, is the distance difference between each pair of points;

[0067] S624. Based on the calculated overall difference E compare , evaluate the degree of fit between the revised full-mouth correction model and the physical model.

[0068] The beneficial effects of the present invention are:

[0069] (1) The present invention adopts a fusion strategy based on dynamic feature weight adjustment, which can adaptively adjust the feature weights of different areas according to the features of the scanned area, thereby achieving more accurate three-dimensional model correction. In the traditional fixed weight fusion method, due to the failure to dynamically adjust the weights according to the scanning quality of different areas, it is easy to cause an imbalance between local details and global consistency. In the matching between tooth details and the curvature of the entire mouth, the correction results are often not accurate enough. The dynamic weight adjustment algorithm can efficiently balance between details and global structure, significantly improving the accuracy and consistency of the corrected three-dimensional model.

[0070] (2) The present invention introduces the physical model generated by the warehouse scan as a verification standard to calibrate the accuracy of the revised full-mouth correction model, so that the revised model maintains a higher degree of consistency with the actual oral structure. The present invention automatically improves the accuracy of the correction model by comparing the revised full-mouth correction model with the physical model generated by the warehouse scan, and by optimizing and adjusting the error function, ensuring that the revised full-mouth correction model is more realistic and reliable and suitable for actual clinical applications.

[0071] (3) The present invention adopts an automated dynamic optimization process when performing three-dimensional registration and full-mouth correction, ensuring that the registered three-dimensional model can fully utilize the advantages of intraoral scanning data and CBCT data. Through the dynamic optimization method, the detailed information of a single tooth can be accurately integrated with the global curvature information of the entire mouth. At the same time, the corrected full-mouth correction model is automatically calibrated, which improves the accuracy and consistency of the model, reduces the need for manual operation and debugging, and improves the overall processing efficiency and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0073] Figure 1 This is a flow chart of a method for correcting intraoral scan data based on dynamic feature weight adjustment proposed by the present invention. DETAILED DESCRIPTION

[0074] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0075] refer to Figure 1, a method for correcting intraoral scan data based on dynamic feature weight adjustment, comprising the following steps:

[0076] S1. Obtaining initial intraoral scan data collected by an intraoral scanning device, the initial intraoral scan data including three-dimensional point cloud information of teeth and oral soft and hard tissues;

[0077] S2. Obtain full-oral 3D model data generated by a cone-beam computed tomography device. The full-oral 3D model has globally accurate information about the entire curvature.

[0078] S3. Segment the intraoral scan data and extract the 3D model information of a single tooth;

[0079] S4. Perform 3D registration of the 3D model of a single tooth with the corresponding tooth model in the CBCT data based on a dynamic feature weight adjustment algorithm, and determine the matching relationship of each tooth by minimizing the difference between the two.

[0080] S5. Construct a fusion model to fuse the global curvature information from the CBCT data with the detailed information of individual teeth from the intraoral scan data. Perform global correction on the intraoral scan data by dynamically adjusting the feature weight distribution strategy to generate a full-oral correction model.

[0081] S6. Use the physical model generated by the scan as a verification standard to verify and calibrate the accuracy of the revised full-oral model;

[0082] S7. Output a modified full-mouth 3D model that has the details of individual teeth while maintaining the integrity and consistency of the global dental arch curvature.

[0083] In this embodiment, S4 includes the following steps:

[0084] S41. Obtaining the three-dimensional model information M of a single tooth in the intraoral scan data tooth , the three-dimensional model information contains the spatial coordinate information of each point:

[0085]

[0086] in, is the coordinate of a single tooth in three-dimensional space, n tooth is the number of points in the point cloud data of a single tooth model;

[0087] S42. Obtain the corresponding tooth model information M in the CBCT data CBCT , the tooth model information contains the coordinates of each point:

[0088]

[0089] in, is the three-dimensional spatial coordinate of the corresponding tooth model in CBCT, n CBCT is the number of points in the point cloud model in the CBCT data;

[0090] S43. 3D model information M tooth and tooth model information M CBCT Perform preliminary registration, set the initial matching relationship, and use the rigid transformation method to align the 3D model information to the corresponding tooth model in the CBCT data. The preliminary registration is performed using the rotation matrix R and the displacement vector t:

[0091]

[0092] Where R is a 3x3 rotation matrix, t=(t x ,t y ,t z ) is the translation vector, M aligned is the preliminary alignment model after registration, is the coordinate of the tooth model point after rotation and translation;

[0093] S44. Preliminary alignment model M after registration based on dynamic feature weight adjustment algorithm aligned and tooth model information M CBCT Perform registration, optimize the matching relationship between the two, and minimize the difference between the two using the minimization error function:

[0094]

[0095] Among them, ω ij The dynamic feature weight reflects the contribution of each matching point to the error. The dynamic adjustment of the weight coefficient is based on the local feature differences between point clouds. The larger the weight, the greater the impact of the point cloud matching on the overall registration result.

[0096] S45. By minimizing the error function E, an iterative optimization method is used to adjust the rotation matrix R and the translation vector t, thereby optimizing the difference between the initial alignment model and the tooth model information after registration, so that the three-dimensional spatial matching relationship of each tooth is accurate:

[0097]

[0098] Where η is the learning rate, and are the gradients of the rotation matrix and displacement vector, respectively, indicating the rate of change of the error function relative to the rotation matrix and displacement vector;

[0099] S46. Determine the matching relationship of each tooth and generate a final registered single tooth 3D model. The registered single tooth 3D model is aligned with the corresponding tooth portion in the CBCT model to generate a corrected single tooth 3D model M. final :

[0100]

[0101] in, is the point cloud data of the single tooth model after correction and registration, n final is the number of points after registration.

[0102] In this embodiment, S5 includes the following steps:

[0103] S53. The tooth model information and the modified three-dimensional model M of the single tooth final Fusion is performed by dynamically adjusting the fusion method of the detail information of a single tooth and the curvature information of the entire mouth through a dynamic feature weight allocation strategy:

[0104]

[0105] Among them, ω k,i for point and The feature weights of and are the distances between the CBCT data points and the corrected 3D model points of a single tooth;

[0106] S54. Based on dynamic feature weight ω k,i The whole-mouth correction model is fused. The curvature information of the whole mouth is fused with the corrected detail information of a single tooth through the weighted average method to generate the whole-mouth correction model:

[0107]

[0108] in, The final full-mouth modified model includes detailed information from the modified 3D model of a single tooth and global curvature information from the CBCT data.

[0109] S55. Correction of the generated full oral cavity model Optimization is performed and the accuracy of the correction model is further adjusted by minimizing the error function, so that the full-mouth correction model achieves the best balance between details and global consistency:

[0110]

[0111] in, To correct the model error, is the corrected three-dimensional coordinate from a single tooth, To correspond to the coordinates in the CBCT data, minimize Then get the whole mouth correction model

[0112] In this embodiment, S6 includes the following steps:

[0113] S61. Obtain the physical model M generated by warehouse scanning physical ,The physical model includes the real physical structure of the whole mouth and provides a standard for validating the revised full mouth correction model;

[0114] S62. The modified full oral model The physical model M generated by warehouse scanning physical Compare and calculate the difference between the two;

[0115] S63. Verify and calibrate the corrected full-mouth correction model by minimizing the error function, and adjust the corrected full-mouth correction model to improve the fit between the full-mouth correction model and the physical model scanned by the warehouse;

[0116] S64. When the error function meets the preset convergence threshold or after a set number of iterations, the final full-oral correction model is determined The final modified three-dimensional model of the entire mouth.

[0117] In this embodiment, S61 includes the following steps:

[0118] S611. Obtain the physical model M generated by warehouse scanning physical :

[0119]

[0120] in, The three-dimensional space coordinates of each point in the physical model generated by the bin scan, n physical is the total number of point clouds of the physical model;

[0121] S612. Scan the physical model M generated by the warehouse physical Perform preprocessing to remove noise points caused by scanning equipment errors or environmental factors, and calculate the standard deviation of the point cloud data based on the preset threshold to calculate the noise tolerance of each point:

[0122]

[0123] in, For the p-th point in the warehouse scanning physical model to the reference coordinate (x ref ,y ref ,z ref )’s Euclidean distance;

[0124] S613. Preprocessed physical model M physical Perform coordinate system conversion and compare its coordinate system with the full oral correction model The coordinate system is aligned, and the coordinate system transformation process is done by the rotation matrix R physical and displacement vector t physical conduct:

[0125]

[0126] in, is the physical model after transformation, R physical is the rotation matrix, t physical is the translation vector.

[0127] In this embodiment, S62 includes the following steps:

[0128] S621. The corrected full oral model is combined with the physical model generated by the scan. Compare and select corresponding point pairs to ensure alignment of the two in three-dimensional space. Compared with the points in the revised full-mouth model One-to-one correspondence, the matching relationship of each pair of corresponding points is expressed as:

[0129]

[0130] S622. Calculate the distance difference between each pair of corresponding points of the physical model and the modified model, and evaluate the matching accuracy of the two based on the difference;

[0131] S623. Distance differences between all corresponding point pairs Perform weighted summation to obtain the overall difference between the full-mouth modified model and the physical model:

[0132]

[0133] Among them, E compare is the total difference between the modified full-mouth model and the physical model, ω p,i is the weight of each pair of points, indicating the contribution of the point pair to the overall difference, is the distance difference between each pair of points;

[0134] S624. Based on the calculated overall difference E compare , evaluate the degree of fit between the revised full-mouth correction model and the physical model.

[0135] Example 1:

[0136] The embodiment demonstrates an intraoral scan data correction method based on dynamic feature weight adjustment applied to the correction process of the oral three-dimensional model. This embodiment demonstrates how the present invention solves the problems of traditional methods in terms of accuracy and consistency by efficiently correcting intraoral scan data and CBCT data.

[0137] The application scenario of this embodiment is the oral restoration department of a dental hospital. The hospital uses advanced intraoral scanners and cone-beam computed tomography equipment to obtain three-dimensional oral data of the patient. The patient is a 45-year-old male who complains of tooth loss and is preparing for full-mouth dental restoration treatment. The patient's oral structure is complex, including multiple missing teeth and changes in the alveolar bone. The restoration treatment requires high precision in both details and global structure.

[0138] The patient first underwent an intraoral scan, which produced 3D point cloud data of individual teeth and gums. The scan contained approximately 350,000 data points. However, due to equipment limitations, the scan only covered the teeth and part of the gums, resulting in significant deviations in the global consistency of the intraoral scan data. The patient then underwent a CBCT scan, which provides a complete 3D model of the entire oral cavity with approximately 1.2 million data points. This model covers the global structure of the alveolar bone, teeth, and oral soft and hard tissues. However, due to resolution limitations, it was difficult to accurately capture the details of the tooth surfaces.

[0139] After acquiring the 3D point cloud data from the intraoral scan, the data was first denoised. The denoising algorithm, based on the standard deviation method, removes noise points whose distance from the mean exceeds a threshold. By calculating the standard deviation between each data point and the center point and setting a noise threshold of 3 mm, approximately 20,000 noise points were removed, resulting in a final number of 330,000 valid data points.

[0140] The preprocessing of CBCT data mainly involves removing low-quality point clouds caused by scanning angles and equipment resolution. By comparing and removing edge distortion areas, approximately 50,000 data points were removed, leaving 1.15 million full-oral three-dimensional data points.

[0141] A three-dimensional registration algorithm based on dynamic feature weight adjustment is used to perform preliminary registration of the intraoral scan data and the CBCT data. The preliminary registration is performed using the rotation matrix R and the translation vector t, and the least squares method is used to minimize the preliminary registration error. The preliminary rotation matrix R and the translation vector t are set as:

[0142]

[0143] This rotation matrix and displacement vector aligned the intraoral scan data with the CBCT data, with a preliminary registration error of 3.5 mm.

[0144] In traditional methods, fixed weights are usually used for data fusion, resulting in inconsistency between details and global structures. However, this invention uses dynamic feature weight adjustment to calculate the distance difference between each point pair and dynamically adjust the weight according to the characteristics of each point and the importance of the region. In this example, the weight ω of the tooth surface area is calculated. tooth is 0.8, and the weight of the global structure region ω CBCT is 0.2, and the error after final registration is reduced to 1.5 mm.

[0145] Based on the registration results after dynamic feature weight adjustment, the experimenters generated a full-mouth correction model. The detailed information in the intraoral scan data was fused with the global information in the CBCT data through the weighted averaging method. The final full-mouth correction model had an accuracy of 0.9 mm. Compared with traditional methods, the details and global consistency of the model have been significantly improved.

[0146] To verify the accuracy of the revised model, the researchers used a physical model generated by the scanned image as a standard for comparison. By calculating the Euclidean distance between the revised full-mouth model and the scanned image, they found an overall difference of 1.0 mm. After further optimization, the error was reduced to 0.8 mm, meeting clinical restoration requirements.

[0147]

[0148]

[0149] By comparing the data, it can be seen that the method of the present invention is superior to the traditional method in correction accuracy, time efficiency and model consistency, especially in the details and global consistency of the full-mouth correction model, and has achieved significant improvements.

[0150] Through the application of this example, researchers demonstrated the feasibility and effectiveness of a dynamic feature weight adjustment-based intraoral scan data correction method for oral 3D model correction. This method effectively addresses the inaccuracy and poor consistency of traditional methods through automated 3D registration and accurate generation of a full-oral correction model. The resulting corrected model achieves significant improvements in accuracy and time efficiency, meeting the high-precision requirements of clinical treatment.

[0151] The present invention adopts a fusion strategy based on dynamic feature weight adjustment, which can adaptively and automatically adjust the feature weights of different areas according to the characteristics of the scanning area, thereby achieving more accurate three-dimensional model correction. In the traditional fixed weight fusion method, due to the failure to dynamically adjust the weights according to the scanning quality of different areas, it is easy to cause an imbalance between local details and global consistency. The correction results in the matching between tooth details and the curvature of the entire mouth are often not accurate enough. The dynamic weight adjustment algorithm can efficiently balance between details and global structure, significantly improving the accuracy and consistency of the corrected three-dimensional model.

[0152] The present invention introduces the physical model generated by the warehouse scanning as a verification standard to calibrate the accuracy of the revised full-mouth correction model, so that the revised model maintains a higher degree of consistency with the actual oral structure. The present invention automatically improves the accuracy of the correction model by comparing the revised full-mouth correction model with the physical model generated by the warehouse scanning, and by optimizing and adjusting the error function, ensuring that the revised full-mouth correction model is more realistic and reliable and suitable for actual clinical applications.

[0153] The present invention adopts an automated dynamic optimization process when performing three-dimensional alignment and full-mouth correction, ensuring that the aligned three-dimensional model can fully utilize the advantages of intraoral scanning data and CBCT data. Through the dynamic optimization method, the detailed information of a single tooth can be accurately integrated with the global curvature information of the entire mouth. At the same time, the corrected full-mouth correction model is automatically calibrated, which improves the accuracy and consistency of the model, reduces the need for manual operation and debugging, and improves the overall processing efficiency and adaptability.

[0154] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for correcting intraoral scan data based on dynamic feature weight adjustment, characterized in that: The steps include: S1. Acquire initial intraoral scanning data collected by an intraoral scanning device, wherein the initial intraoral scanning data includes three-dimensional point cloud information of teeth and oral soft and hard tissues; S2. Obtain full-oral 3D model data generated by a cone-beam computed tomography device. The full-oral 3D model has globally accurate information about the entire curvature. S3. Segmenting the intraoral scan data to extract three-dimensional model information of a single tooth; S4. Perform 3D registration of the 3D model of a single tooth with the corresponding tooth model in the CBCT data based on a dynamic feature weight adjustment algorithm, and determine the matching relationship of each tooth by minimizing the difference between the two. S5. Construct a fusion model to fuse the global curvature information from the CBCT data with the detailed information of individual teeth from the intraoral scan data. Perform global correction on the intraoral scan data by dynamically adjusting the feature weight distribution strategy to generate a full-oral correction model. S6. Verify and calibrate the accuracy of the modified full-mouth model using the physical model generated by the scan as a verification standard; S7. Output a modified full-mouth 3D model that retains the details of individual teeth while maintaining the integrity and consistency of the global dental arch curvature; The S4 comprises the following steps: S41. Obtaining the three-dimensional model information M of a single tooth in the intraoral scan data tooth , the three-dimensional model information contains the spatial coordinate information of each point: in, is the coordinate of a single tooth in three-dimensional space, n tooth is the number of points in the point cloud data of a single tooth model; S42. Obtain the corresponding tooth model information M in the CBCT data CBCT , the tooth model information contains the coordinates of each point: in, is the three-dimensional spatial coordinate of the corresponding tooth model in CBCT, n CBCT is the number of points in the point cloud model in the CBCT data; S43. 3D model information M tooth and tooth model information M CBCT Perform preliminary registration, set the initial matching relationship, and use the rigid transformation method to align the 3D model information to the corresponding tooth model in the CBCT data. The preliminary registration is performed using the rotation matrix R and the displacement vector t: Where R is a 3x3 rotation matrix, t=(t x ,t y ,t z ) is the translation vector, M aligned is the preliminary alignment model after registration, is the coordinate of the tooth model point after rotation and translation; S44. Preliminary alignment model M after registration based on dynamic feature weight adjustment algorithm aligned and tooth model information M CBCT Perform registration, optimize the matching relationship between the two, and minimize the difference between the two using the minimization error function: Among them, ω ij The dynamic feature weight reflects the contribution of each matching point to the error. The dynamic adjustment of the weight coefficient is based on the local feature differences between point clouds. The larger the weight, the greater the impact of the point cloud matching on the overall registration result. S45. By minimizing the error function E, an iterative optimization method is used to adjust the rotation matrix R and the translation vector t, thereby optimizing the difference between the initial alignment model and the tooth model information after registration, so that the three-dimensional spatial matching relationship of each tooth is accurate: Where η is the learning rate, and are the gradients of the rotation matrix and displacement vector, respectively, indicating the rate of change of the error function relative to the rotation matrix and displacement vector; S46. Determine the matching relationship of each tooth and generate a final registered single tooth 3D model. The registered single tooth 3D model is aligned with the corresponding tooth portion in the CBCT model to generate a corrected single tooth 3D model M. final : in, is the point cloud data of the single tooth model after correction and registration, n final is the number of points after registration.

2. The intraoral scan data correction method based on dynamic feature weight adjustment according to claim 1, characterized in that: The S5 comprises the following steps: S53. The tooth model information and the modified three-dimensional model M of the single tooth final Fusion is performed by dynamically adjusting the fusion method of the detail information of a single tooth and the curvature information of the entire mouth through a dynamic feature weight allocation strategy: Among them, ω k,i for point and The feature weights of and are the distances between the CBCT data points and the corrected 3D model points of a single tooth; S54. Based on dynamic feature weight ω k,i The whole-mouth correction model is fused. The curvature information of the whole mouth and the detailed information of the single tooth after correction are fused by the weighted average method to generate the whole-mouth correction model: in, The final full-mouth modified model includes detailed information from the modified 3D model of a single tooth and global curvature information from the CBCT data. S55. Correction of the generated full oral cavity model Optimization is performed and the accuracy of the correction model is further adjusted by minimizing the error function, so that the full-mouth correction model achieves the best balance between details and global consistency: in, To correct the model error, is the corrected three-dimensional coordinate from a single tooth, To correspond to the coordinates in the CBCT data, minimize Then get the whole mouth correction model 3. The intraoral scan data correction method based on dynamic feature weight adjustment according to claim 1, characterized in that: The S6 comprises the following steps: S61. Obtain the physical model M generated by warehouse scanning physical , the physical model includes the real physical structure of the entire oral cavity and provides a standard for verifying the revised full oral cavity revised model; S62. The modified full oral cavity correction model The physical model M generated by warehouse scanning physical Compare and calculate the difference between the two; S63. Verify and calibrate the corrected full-mouth correction model by minimizing the error function, and adjust the corrected full-mouth correction model to improve the fit between the full-mouth correction model and the physical model scanned by the warehouse; S64. When the error function meets the preset convergence threshold or after a set number of iterations, the final full-oral correction model is determined The final modified three-dimensional model of the entire mouth.

4. The intraoral scan data correction method based on dynamic feature weight adjustment according to claim 3, characterized in that: The S61 includes the following steps: S611. Obtain the physical model M generated by warehouse scanning physical : in, The three-dimensional space coordinates of each point in the physical model generated by the bin scan, n physical is the total number of point clouds of the physical model; S612. Scan the physical model M generated by the warehouse physical Perform preprocessing to remove noise points caused by scanning equipment errors or environmental factors, and calculate the standard deviation of the point cloud data based on the preset threshold to calculate the noise tolerance of each point: in, For the p-th point in the warehouse scanning physical model to the reference coordinate (x ref ,y ref ,z ref )’s Euclidean distance; S613. Preprocessed physical model M physical Perform coordinate system conversion and compare its coordinate system with the full oral correction model The coordinate system is aligned, and the coordinate system transformation process is done by the rotation matrix R physical and displacement vector t physical conduct: in, is the physical model after transformation, R physical is the rotation matrix, t physical is the translation vector.

5. The intraoral scan data correction method based on dynamic feature weight adjustment according to claim 4, characterized in that: The S62 includes the following steps: S621. The corrected full oral model is combined with the physical model generated by the scan. Compare and select corresponding point pairs to ensure alignment of the two in three-dimensional space. Compared with the points in the revised full-mouth model One-to-one correspondence, the matching relationship of each pair of corresponding points is expressed as: S622. Calculate the distance difference between each pair of corresponding points of the physical model and the modified model, and evaluate the matching accuracy of the two based on the difference; S623. Distance differences between all corresponding point pairs Perform weighted summation to obtain the overall difference between the full-mouth modified model and the physical model: Among them, E compare is the total difference between the modified full-mouth model and the physical model, ω p,i is the weight of each pair of points, indicating the contribution of the point pair to the overall difference, is the distance difference between each pair of points; S624. Based on the calculated overall difference E compare , evaluate the degree of fit between the revised full-mouth correction model and the physical model.

Images