Method and device for automatically constructing three-dimensional jaw symmetric reference plane

Abstract

The invention discloses a method and a device for automatically constructing a three-dimensional jaw symmetric reference plane. The method comprises the following steps: acquiring a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed; the three-dimensional jaw bone body data set is preprocessed, and a preprocessed three-dimensional jaw bone body data set is obtained; the preprocessed three-dimensional jaw bone body data set comprises a plurality of pieces of preprocessed three-dimensional jaw bone body sample information; and processing the preprocessed three-dimensional jaw bone body data set and the to-be-processed three-dimensional jaw bone body data information to obtain three-dimensional jaw bone symmetric reference plane information. Visibly, according to the embodiment, efficient overlapping of the three-dimensional jaw bone body and the mirror image data is achieved, automatic, efficient and accurate construction of the three-dimensional jaw bone symmetric reference plane information is facilitated, and therefore a reference basis is provided for the fields of oral clinical orthodontics, maxillofacial surgery digital diagnosis and treatment and the like, and the construction method of the three-dimensional jaw bone symmetric reference plane is provided.

Description

Technical Field

[0001] The present invention relates to the field of mandibular symmetry reference plane data processing, and in particular to a method and device for automatically constructing a three-dimensional mandibular symmetry reference plane. Background Art

[0002] Maxillofacial symmetry and coordination are important criteria for measuring facial beauty and attractiveness, and they play a vital role in all dental disciplines. Due to the multiple influences of human maxillofacial biological factors and environmental factors, there are almost no completely symmetrical individuals, and slight maxillofacial asymmetry is difficult to identify. However, severe maxillofacial asymmetry caused by developmental or acquired diseases such as trauma, tumors, and developmental malformations urgently needs to be treated through maxillofacial surgery, orthodontics, and other means to restore the function and beauty of the affected area and improve the quality of life of patients. The mandibular symmetry reference plane (SRP) is the premise of maxillofacial surgery design and orthodontic correction design, and is also an important basis for maxillofacial asymmetry analysis.

[0003] In recent years, with the rapid development of digital and intelligent stomatology, the method of automatically constructing a three-dimensional jaw symmetry reference plane has attracted widespread attention from scholars. Therefore, the method of constructing a symmetry reference plane based on the association and overlap of the three-dimensional jaw model body and mirror model (referred to as the "body-mirror association method") has received more and more attention. The core of the body-mirror association method lies in the overlap algorithm between the body and mirror model. The research on this algorithm is mainly divided into the Iterative Closest Point (ICP) algorithm without reference to jaw landmarks and the Procrustes Analysis (PA) algorithm with reference to jaw landmarks. The two algorithms are used to construct symmetry reference planes for jaw data without significant deformity, which can basically meet clinical needs. However, for oral and maxillofacial deformity data, it is necessary to manually select jaw areas or landmarks with better symmetry to improve the overlap effect, but this reduces the degree of automation of the algorithm to a certain extent. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and device for automatically constructing a three-dimensional mandibular symmetry reference plane. By achieving efficient overlap of the three-dimensional mandibular body and mirror data, it is beneficial to automatically, efficiently and accurately construct three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0005] In order to solve the above technical problems, the first aspect of the embodiments of the present invention discloses a method for automatically constructing a three-dimensional jaw symmetry reference plane, the method comprising:

[0006] S1, obtaining a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed;

[0007] S2, preprocessing the three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set; the preprocessed three-dimensional jaw body data set includes a plurality of preprocessed three-dimensional jaw body sample information;

[0008] S3, processing the preprocessed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

[0009] As an optional implementation, in the first aspect of the embodiment of the present invention, the preprocessing of the three-dimensional jaw body dataset to obtain the preprocessed three-dimensional jaw body dataset includes:

[0010] S21, performing data format unification processing on the three-dimensional jaw body dataset to obtain a first three-dimensional jaw body dataset;

[0011] S22, performing denoising processing on the first three-dimensional jaw body dataset to obtain a second three-dimensional jaw body dataset;

[0012] S23, segmenting the second three-dimensional jaw body dataset to obtain a third three-dimensional jaw body dataset;

[0013] S24, performing smoothing processing on the third three-dimensional jaw body dataset to obtain a preprocessed three-dimensional jaw body dataset.

[0014] As an optional implementation, in the first aspect of the embodiment of the present invention, the preprocessed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed are processed to obtain three-dimensional jaw symmetry reference plane information, including:

[0015] S31, using the preprocessed three-dimensional jaw body data set, training the first jaw initial model and the second jaw initial model to obtain the first jaw optimized model and the second jaw optimized model;

[0016] S32, using the first jaw optimization model and the second jaw optimization model, processing the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

[0017] As an optional implementation, in the first aspect of the embodiment of the present invention, the preprocessed three-dimensional jaw body data set is used to train the first jaw initial model and the second jaw initial model to obtain the first jaw optimized model and the second jaw optimized model, including:

[0018] S311, performing a random selection operation on the preprocessed three-dimensional jaw body data set to obtain target body sample information; the target body sample information is any preprocessed three-dimensional jaw body sample information in the preprocessed three-dimensional jaw body data set;

[0019] S312, performing horizontal flipping processing on the target body sample information to obtain target mirror sample information;

[0020] S313, using the first jaw initial model and the second jaw initial model, the target body sample information and the target mirror sample information are processed to obtain a first joint feature information set, a second joint feature information set, a rotation matrix information set and a translation matrix information set; the first joint feature information set includes a plurality of first joint feature information; the second joint feature information set includes a plurality of second joint feature information; the rotation matrix information set includes a plurality of rotation matrix information; the translation matrix information set includes a plurality of translation matrix information;

[0021] S314, obtaining true value information of the rotation matrix and true value information of the translation matrix;

[0022] S315, processing the target body sample information, the target mirror sample information, the rotation matrix true value information, the translation matrix true value information, the first joint feature information set, the second joint feature information set, the rotation matrix information set, and the translation matrix information set to obtain a loss function value;

[0023] S316, determining whether the loss function value is less than a preset loss function threshold, and obtaining a first determination result;

[0024] When the first judgment result is yes, execute S317;

[0025] When the first judgment result is no, executing S318;

[0026] S317, determining the first jaw initial model as a first jaw optimized model, determining the second jaw initial model as a second jaw optimized model, and executing S32;

[0027] S318, updating the first jaw initial model and the second jaw initial model to obtain an updated first jaw initial model and an updated second jaw initial model;

[0028] S319, determining that the updated first jaw initial model is the first jaw initial model, determining that the updated second jaw initial model is the second jaw initial model, and executing S311.

[0029] As an optional implementation, in the first aspect of the embodiment of the present invention, the target body sample information and the target mirror sample information are processed using the first jaw initial model and the second jaw initial model to obtain a first joint feature information set, a second joint feature information set, a rotation matrix information set, and a translation matrix information set, including:

[0030] S3131, preset s=1;

[0031] S3132, using the first jaw initial model, processing the target body sample information and the target mirror sample information to obtain body local feature information and mirror local feature information;

[0032] S3133, using the second initial jaw model, processing the local feature information of the main body and the local feature information of the mirror image to obtain first joint feature information and second joint feature information; the first joint feature information includes a plurality of first joint key point feature information; the second joint feature information includes a plurality of second joint key point feature information

[0033] S3134, fusing the body local feature information, the mirror image local feature information, the first joint feature information, and the second joint feature information to obtain body feature vector information and mirror image feature vector information;

[0034] S3135, calculating and processing the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information to obtain rotation matrix information and translation matrix information;

[0035] S3136, determining whether s is equal to a preset number of iterations, and obtaining a second determination result;

[0036] When the second judgment result is no, increase the s by 1 and execute S3137;

[0037] When the second judgment result is yes, execute S314;

[0038] S3137, using the rotation matrix information and the translation matrix information, performing matrix transformation processing on the target mirror sample information to obtain transformed target mirror sample information;

[0039] S3138, determining that the transformed target image sample information is the target image sample information, and executing S3132.

[0040] As an optional implementation, in the first aspect of the embodiment of the present invention, the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information are calculated and processed to obtain rotation matrix information and translation matrix information, including:

[0041] S31351, using a jaw key point correspondence relationship calculation model, calculating and processing the target body sample information, the target mirror image sample information, the body feature vector information and the mirror image feature vector information to obtain jaw key point matching relationship information;

[0042] The calculation model of the corresponding relationship of the jaw key points is:

[0043]

[0044] 1≤i≤N;

[0045] 1≤j≤M;

[0046] In the formula, DY(x i ,Y), is the jaw key point matching relationship information, DY(x i ,Y) represents the matching relationship information between the coordinate information of the i-th body sample point in the target body sample information and the coordinate information of all mirror sample points in the target mirror sample information, JX is the mirror feature vector information, is the transpose of the i-th feature vector information in the subject feature vector information, μ is the sampling parameter, τ is the attenuation factor, N is the number of subject sample points in the target subject sample information, M is the number of mirror sample points in the target mirror sample information, X is the target subject sample information, Y is the target mirror sample information, x i is the coordinate information of the i-th ontology sample point in the target ontology sample information;

[0047] S31352, performing calculation processing on the target body sample information and the target mirror sample information to obtain first matrix information and second matrix information;

[0048] S31353, using a jaw matrix decomposition calculation model, calculating and processing the first matrix information, the second matrix information and the jaw key point correspondence information to obtain rotation matrix information and translation matrix information;

[0049] Among them, the jaw matrix decomposition calculation model is:

[0050] XZ=JZ1·JZ2 T ;

[0051]

[0052] Where XZ is the rotation matrix information, PY is the translation matrix information, JZ1 is the first matrix information, JZ2 T is the transpose of the second matrix information, and ε1 is the deviation displacement information.

[0053] As an optional implementation, in the first aspect of the embodiment of the present invention, the first jaw optimization model and the second jaw optimization model are used to process the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information, including:

[0054] S321, using the first jaw optimization model and the second jaw optimization model, processing the three-dimensional jaw body data information to be processed to obtain the three-dimensional jaw mirror data information to be processed, target rotation matrix information and target translation matrix information;

[0055] S322, calculating and processing the to-be-processed three-dimensional jaw body data information, the to-be-processed three-dimensional jaw mirror data information, the target rotation matrix information and the target translation matrix information to obtain three-dimensional jaw joint data information;

[0056] S323, processing the three-dimensional mandibular joint data information to obtain three-dimensional mandibular symmetry reference plane information.

[0057] A second aspect of an embodiment of the present invention discloses a device for automatically constructing a three-dimensional jaw symmetry reference plane, the device comprising:

[0058] An acquisition module is used to acquire a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed;

[0059] A first calculation module is used to preprocess the three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set; the preprocessed three-dimensional jaw body data set includes a plurality of preprocessed three-dimensional jaw body sample information;

[0060] The second calculation module is used to process the preprocessed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

[0061] A third aspect of an embodiment of the present invention discloses another device for automatically constructing a three-dimensional jaw symmetry reference plane, the device comprising:

[0062] processor;

[0063] a memory coupled to the processor and storing executable program code;

[0064] The processor calls the executable program code stored in the memory to execute part or all of the steps of the method for automatically constructing a three-dimensional mandibular symmetry reference plane disclosed in the first aspect of the embodiment of the present invention.

[0065] The fourth aspect of an embodiment of the present invention discloses a computer-readable storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute part or all of the steps of the method for automatically constructing a three-dimensional mandibular symmetry reference plane disclosed in the first aspect of the embodiment of the present invention.

[0066] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0067] In an embodiment of the present invention, a three-dimensional mandibular body data set and three-dimensional mandibular body data information to be processed are obtained; the three-dimensional mandibular body data set is preprocessed to obtain a preprocessed three-dimensional mandibular body data set; the preprocessed three-dimensional mandibular body data set includes a plurality of preprocessed three-dimensional mandibular body sample information; the preprocessed three-dimensional mandibular body data set and the three-dimensional mandibular body data information to be processed are processed to obtain three-dimensional mandibular symmetry reference plane information. It can be seen that this embodiment is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information by realizing efficient overlap of three-dimensional mandibular body and mirror data, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0069] Figure 1 A schematic diagram of a process of automatically constructing a three-dimensional jaw symmetry reference plane disclosed in an embodiment of the present invention;

[0070] Figure 2 A schematic diagram of the structure of a device for automatically constructing a three-dimensional jaw symmetry reference plane disclosed in an embodiment of the present invention;

[0071] Figure 3 A schematic diagram of the structure of another device for automatically constructing a three-dimensional mandibular symmetry reference plane disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0073] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or equipment.

[0074] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0075] In this paper, the coordinate system corresponding to the coordinate information involved is the Cartesian coordinate system. The three-dimensional mandibular body data information obtained in the three-dimensional mandibular body data set in this paper is point cloud data information. The facial center point of the three-dimensional mandibular body data information is used as the origin of the Cartesian coordinate system. The x-axis, y-axis and z-axis in the Cartesian coordinate system are set as follows:

[0076] x-axis: Direction from left to right, with positive direction pointing to the right and negative direction pointing to the left.

[0077] y-axis: direction from front to back, with positive direction pointing forward and negative direction pointing backward.

[0078] z-axis: The direction from bottom to top, with the positive direction pointing upward and the negative direction pointing downward.

[0079] The 3D jaw symmetry reference plane in this article refers to a plane that reflects the left-right symmetry of the jaw in a 3D jaw model. It is usually used to represent and analyze the geometric morphology of the jaw and has important application value in the medical field (such as orthodontics, maxillofacial surgery, and 3D modeling).

[0080] The present invention discloses a method and device for automatically constructing a three-dimensional jaw symmetry reference plane, which is conducive to the automatic, efficient and accurate construction of three-dimensional jaw symmetry reference plane information by realizing efficient overlap of three-dimensional jaw body and mirror data, thereby providing a reference basis for the fields of oral clinical orthodontics, maxillofacial surgery digital diagnosis and treatment, and providing a method for constructing a three-dimensional jaw symmetry reference plane. The following are detailed descriptions.

[0081] Embodiment 1

[0082] See also Figure 1 , Figure 1 1 is a flow chart of a method for automatically constructing a three-dimensional jaw symmetry reference plane disclosed in an embodiment of the present invention. Figure 1 The described method for automatically constructing a three-dimensional jaw symmetry reference plane is applied to a device for automatically constructing a three-dimensional jaw symmetry reference plane, such as a local server or cloud server for automatically constructing an optimized management of a three-dimensional jaw symmetry reference plane, and the embodiments of the present invention are not limited thereto. Figure 1 As shown, the method for automatically constructing a three-dimensional jaw symmetry reference plane may include the following operations:

[0083] S1, obtaining a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed;

[0084] It should be noted that the three-dimensional mandibular body data set and the three-dimensional mandibular body data information to be processed are all collected through clinical scanning technology, specifically including CT scanning, three-dimensional laser scanning, MRI scanning, optical three-dimensional scanning and other technologies. A number of three-dimensional mandibular data information are obtained through clinical collection for training; among them, the three-dimensional mandibular body data set is used for model training in this embodiment.

[0085] S2, preprocessing the three-dimensional jaw ontology data set to obtain a preprocessed three-dimensional jaw ontology data set; the preprocessed three-dimensional jaw ontology data set includes a plurality of preprocessed three-dimensional jaw ontology sample information;

[0086] S3, processing the preprocessed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

[0087] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0088] In an optional embodiment, preprocessing the three-dimensional jaw body dataset to obtain a preprocessed three-dimensional jaw body dataset includes:

[0089] S21, performing unified data format processing on the three-dimensional jaw body data set to obtain a first three-dimensional jaw body data set;

[0090] It should be noted that the above-mentioned unified data format processing is based on algorithms or tools such as PCL and MeshLab to unify the three-dimensional mandibular body data set into point cloud data. Specifically, the embodiments of the present invention do not limit this.

[0091] Through unified data format processing, three-dimensional jaw body data sets of different sources, formats, and postures can be uniformly processed and integrated to ensure that subsequent models can effectively use these data.

[0092] S22, performing denoising processing on the first three-dimensional jaw body dataset to obtain a second three-dimensional jaw body dataset;

[0093] It should be noted that the above denoising process is performed based on statistical outlier removal technology, normal-based denoising technology, etc., and the specific embodiments of the present invention do not limit this.

[0094] Through denoising processing, isolated noise can be removed, the main structure of the point cloud can be retained, and the details and surface features of the point cloud can be retained as much as possible. It has strong generalization ability, and the denoised data can improve the accuracy and robustness of subsequent machine learning models.

[0095] S23, segmenting the second three-dimensional jaw body data set to obtain a third three-dimensional jaw body data set; the third three-dimensional jaw body data set includes a plurality of third three-dimensional jaw body data information;

[0096] It should be noted that the above segmentation processing is performed based on tools or algorithms such as a region growing segmentation method and a K-means method, and the specific embodiments of the present invention do not limit this.

[0097] It should be noted that the segmented data can improve the performance of subsequent models, such as registration, classification, segmentation, etc., reduce the interference of irrelevant information, and through segmentation, focus on specific features, reduce the amount of calculation, and improve task efficiency.

[0098] S24, performing smoothing processing on the third three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set;

[0099] It should be noted that the above-mentioned smoothing process is based on Gaussian filtering, bilateral filtering, Laplace-Beltrami smoothing and other algorithms for smoothing, and the specific embodiments of the present invention are not limited thereto.

[0100] It should be noted that through smoothing, the smoothed point cloud data can effectively improve the accuracy and stability of subsequent tasks (such as registration, segmentation, classification, etc.). At the same time, by reducing unnecessary details and the density of the point cloud, smoothing helps to improve computing efficiency.

[0101] It should be noted that the smoothing process of the third three-dimensional jaw body data set to obtain the preprocessed three-dimensional jaw body data set includes:

[0102] The three-dimensional jaw smoothing calculation model is used to smooth the third three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set:

[0103] Among them, the three-dimensional jaw smoothing calculation model is:

[0104]

[0105] 1≤k1≤LL1;

[0106] 1≤k2≤LL2;

[0107] Where H is the preprocessed 3D jaw bone dataset, H k1,k2 is the coordinate of the k2th point in the k1th preprocessed 3D jaw body sample information in the preprocessed 3D jaw body dataset, SS is the third 3D jaw body dataset, SS k1,k2 is the coordinate of the k2th point in the k1th third 3D jaw body data set, SS k1,k3 is the coordinate of the k3th point in the k1th third three-dimensional mandibular body data information in the third three-dimensional mandibular body data set, LY(k1,k2) is the neighborhood point set of the coordinate of the k2th point in the k1th third three-dimensional mandibular body data information in the third three-dimensional mandibular body data set, QJ is the global geometric center vector, τ1 is the local smoothing weight coefficient, τ2 is the global smoothing weight coefficient, ρ is the scale parameter controlling the neighborhood weight distribution, LL1 is the number of third three-dimensional mandibular body data information in the third three-dimensional mandibular body data set, and LL2 is the number of point coordinates in the k1th third three-dimensional mandibular body data information in the third three-dimensional mandibular body data set.

[0108] It should be noted that the local smoothing weight coefficient, the global smoothing weight coefficient and the scale parameter may be set by the user or obtained based on historical data, and the embodiments of the present invention do not limit this.

[0109] It should be noted that the neighborhood point set can be calculated by a radius neighborhood or a K-nearest neighbor neighborhood algorithm, which is not limited in the embodiment of the present invention.

[0110] It should be noted that the global geometric center vector can be calculated by the mean of all point cloud coordinates or the PCA algorithm, which is not limited in the embodiment of the present invention.

[0111] It should be noted that the local smoothing weight coefficient ranges from [0.1, 0.3], which is used to control the influence of neighborhood points. The larger the value, the stronger the smoothing influence of the local neighborhood on each point, the smoother the surface, but details may be lost. If the point cloud has a lot of noise, the value can be appropriately increased. If the point cloud has rich details and needs to retain more edge information, the value should be reduced. The global smoothing weight coefficient ranges from [0.05, 0.2], which is used to control the constraint strength of the global center on the point. The larger the value, the more stable the overall structure of the point cloud, but it may limit the optimization of local details. If the overall geometry of the point cloud is complex or prone to distortion, the value should be increased. If the local features are more critical, the value should be reduced. The scale parameter ranges from [0.1, 1.0] and is used to control the neighborhood weight distribution. The larger the value, the more points in the neighborhood will have a significant impact on smoothing, and the smoothing effect will be more uniform. The smaller the value, the more weights tend to be closer to the target point, and the smoothing effect will be more detailed. If the point cloud density is large, a smaller scale parameter can be selected to finely control the neighborhood weight. If the point cloud density is sparse, the scale parameter can be appropriately increased to expand the neighborhood range. By adjusting the local smoothing weight coefficient, the global smoothing weight coefficient and the scale parameter, a good balance can be achieved between denoising effect, detail retention and global structural stability, ensuring the high quality performance of the three-dimensional mandibular point cloud data after smoothing.

[0112] It should be noted that the three-dimensional mandibular smoothing calculation model effectively reduces noise and retains key anatomical features through local smoothing weight coefficients; the global smoothing weight coefficient is used to ensure the overall structural stability of the three-dimensional mandibular point cloud and avoid deformation; by controlling the scale parameters of the neighborhood weight distribution, different density areas can obtain good smoothing effects; through the setting of (1-τ1-τ2), the geometric weights of the original point coordinates can be maintained to ensure that important information of the original points is not lost during the smoothing process. Overall, the model achieves a good balance between local feature retention and global structural consistency while avoiding the complexity and computational burden brought by iteration.

[0113] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0114] In another optional embodiment, the preprocessed three-dimensional jaw body data set and the to-be-processed three-dimensional jaw body data information are processed to obtain three-dimensional jaw symmetry reference plane information, including:

[0115] S31, using the preprocessed three-dimensional jaw body data set, training the first jaw initial model and the second jaw initial model to obtain the first jaw optimized model and the second jaw optimized model;

[0116] It should be noted that the first jaw initial model and the second jaw initial model are a DGCNN model and a Transformer model respectively, and the specific details are not limited in the embodiment of the present invention.

[0117] S32, using the first jaw optimization model and the second jaw optimization model, processing the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

[0118] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0119] In another optional embodiment, the first jaw initial model and the second jaw initial model are trained using the preprocessed three-dimensional jaw body data set to obtain the first jaw optimized model and the second jaw optimized model, including:

[0120] S311, performing a random selection operation on the preprocessed three-dimensional jaw ontology data set to obtain target ontology sample information; the target ontology sample information is any preprocessed three-dimensional jaw ontology sample information in the preprocessed three-dimensional jaw ontology data set;

[0121] It should be noted that the above-mentioned target ontology sample information can also be several random preprocessed three-dimensional mandibular ontology sample information in the preprocessed three-dimensional mandibular ontology data set, wherein each preprocessed three-dimensional mandibular ontology sample information is performed according to the steps after S311 in S31. By simultaneously training the model for several preprocessed three-dimensional mandibular ontology sample information, it is beneficial to improve the accuracy and robustness of the model training.

[0122] S312, horizontally flipping the target body sample information to obtain target mirror sample information;

[0123] It should be noted that the above-mentioned horizontal flipping process is to horizontally flip the target body sample information on the YZ plane to obtain the target mirror sample information. Specifically, the YZ plane is a coordinate plane in the three-dimensional coordinate system, which is a plane composed of the y-axis and the z-axis. In three-dimensional space, the YZ plane can be regarded as a plane "perpendicular" to the x-axis, containing all points with an x ​​value of 0. You can simply imagine this plane as a surface "facing you", and the x-axis is pointing to your left or right; horizontal flipping refers to a mirror transformation along a plane so that the data is symmetrical on the plane. For the YZ plane, horizontal flipping means flipping the x-coordinate that was originally in the positive direction to the negative direction, or flipping the x-coordinate that was originally in the negative direction to the positive direction.

[0124] S313, using the first mandibular initial model and the second mandibular initial model, processing the target body sample information and the target mirror sample information to obtain a first joint feature information set, a second joint feature information set, a rotation matrix information set and a translation matrix information set; the first joint feature information set includes a plurality of first joint feature information; the second joint feature information set includes a plurality of second joint feature information; the rotation matrix information set includes a plurality of rotation matrix information; the translation matrix information set includes a plurality of translation matrix information;

[0125] S314, obtaining true value information of the rotation matrix and true value information of the translation matrix;

[0126] It should be noted that the above-mentioned acquisition of the true value information of the rotation matrix and the true value information of the translation matrix is ​​achieved by using the ICP algorithm to calculate and process the target body sample information and the target mirror sample information to obtain the true value information of the rotation matrix and the true value information of the translation matrix. Specifically, the embodiments of the present invention do not limit this.

[0127] It should be noted that the above-mentioned true value information of the rotation matrix and the true value information of the translation matrix are used as the true value of the model training to ensure that the method in this embodiment can accurately reflect the geometric features of the actual object for the alignment, registration or symmetry analysis of the three-dimensional point cloud data. In addition, the ICP algorithm is widely used in the point cloud data registration task because it is direct, efficient, stable, and can provide accurate rotation matrix and translation matrix. When aligning the target body sample and the mirror sample, the ICP algorithm can provide effective registration results and provide a true rotation and translation matrix for model training. Therefore, the true value information of the rotation matrix and the true value information of the translation matrix obtained by the ICP algorithm in this embodiment are used as the true value of the model training. However, since the ICP algorithm reduces the degree of automation of the algorithm to a certain extent, this embodiment, based on this problem, proposes a method that can fully automatically construct a three-dimensional mandibular symmetry reference plane, while ensuring that the accuracy of the obtained rotation matrix and translation matrix is ​​consistent with the ICP algorithm.

[0128] It should be noted that, in addition to using the ICP algorithm, the above-mentioned rotation matrix true value information and translation matrix true value information can also be obtained by other means, such as based on historical data information, etc., which is not limited in the embodiments of the present invention.

[0129] S315, processing the target body sample information, the target mirror sample information, the rotation matrix true value information, the translation matrix true value information, the first joint feature information set, the second joint feature information set, the rotation matrix information set, and the translation matrix information set to obtain a loss function value;

[0130] S316, determining whether the loss function value is less than a preset loss function threshold, and obtaining a first determination result;

[0131] When the first judgment result is yes, execute S317;

[0132] When the first judgment result is no, execute S318;

[0133] It should be noted that the above-mentioned preset loss function threshold may be set by a user, or may be obtained based on historical data, and the specific embodiment of the present invention does not limit this.

[0134] For example, the preset loss function threshold value range is [0.001, 0.02].

[0135] S317, determining the first jaw initial model as the first jaw optimization model, determining the second jaw initial model as the second jaw optimization model, and executing S32;

[0136] S318, updating the first jaw initial model and the second jaw initial model to obtain an updated first jaw initial model and an updated second jaw initial model;

[0137] It should be noted that the above-mentioned updating process is implemented based on the back propagation and gradient descent algorithm, and the specific embodiments of the present invention do not limit this.

[0138] S319, determining the updated initial model of the first jaw as the initial model of the first jaw, determining the updated initial model of the second jaw as the initial model of the second jaw, and executing S311.

[0139] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0140] In an optional embodiment, the target body sample information and the target mirror sample information are processed using the first jaw initial model and the second jaw initial model to obtain a first joint feature information set, a second joint feature information set, a rotation matrix information set, and a translation matrix information set, including:

[0141] S3131, preset s=1;

[0142] S3132, using the first jaw initial model, processing the target body sample information and the target mirror sample information to obtain body local feature information and mirror local feature information;

[0143] It should be noted that the above processing is to input the target body sample information and the target mirror sample information into the first jaw initial model respectively to obtain the body local feature information corresponding to the target body sample information and the mirror local feature information corresponding to the target mirror sample information.

[0144] S3133, using the second mandibular initial model, processing the local feature information of the main body and the local feature information of the mirror image to obtain first joint feature information and second joint feature information; the first joint feature information includes a plurality of first joint key point feature information; the second joint feature information includes a plurality of second joint key point feature information

[0145] It should be noted that the above processing is to use the local feature information of the main body and the local feature information of the mirror image as the input of the second mandibular initial model, and calculate the first joint feature information and the second joint feature information through the second mandibular initial model, wherein the first joint feature information represents the joint feature information obtained after the local feature information of the main body and the local feature information of the mirror image are jointly learned through the Transformer model. This feature information more reflects the relationship from the main body to the mirror image point cloud, and focuses on how to jointly learn the features of the mirror image point cloud based on the local features of the main body. The second joint feature information represents another joint feature information obtained after the local feature information of the mirror image and the local feature information of the main body are jointly learned through the Transformer model. This feature information more reflects the relationship from the mirror image to the main body point cloud, and emphasizes how the mirror image point cloud helps to understand the features of the main body.

[0146] S3134, fusing the body local feature information, the mirror image local feature information, the first joint feature information, and the second joint feature information to obtain body feature vector information and mirror image feature vector information;

[0147] It should be noted that the above fusion processing is to add the local feature information of the main body and the first joint feature information to obtain the main body feature vector information, and add the local feature information of the mirror image and the second joint feature information to obtain the mirror image feature vector information. Specifically, the embodiment of the present invention does not limit this.

[0148] S3135, calculating and processing the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information to obtain rotation matrix information and translation matrix information;

[0149] S3136, determining whether s is equal to a preset number of iterations, and obtaining a second determination result;

[0150] When the second judgment result is no, s is increased by 1 and S3137 is executed;

[0151] When the second judgment result is yes, execute S314;

[0152] It should be noted that the preset number of iterations may be set by a user or may be obtained based on historical data, and the embodiments of the present invention do not limit this.

[0153] Exemplarily, the preset number of iterations is 1000.

[0154] S3137, using the rotation matrix information and the translation matrix information, performing matrix transformation processing on the target mirror sample information to obtain transformed target mirror sample information;

[0155] It should be noted that the above transformation processing is calculated using the following model;

[0156] Y′=θ1·R·Y+θ2·t;

[0157] Wherein, Y′ is the target image sample information after transformation, R is the rotation matrix information, Y is the target image sample information, t is the translation matrix information, θ1 and θ2 are the first weight factor and the second weight factor respectively.

[0158] It should be noted that the first weight factor and the second weight factor may be set by a user, or may be obtained based on historical data, and the embodiments of the present invention do not limit this.

[0159] By introducing a learnable first and second weight factor, the formula allows the model to automatically adjust the importance of the rotation matrix and the translation matrix during training. This provides the model with more flexibility to adapt to different task requirements. For example, in some cases, the rotation transformation may have a greater impact on the registration, while in other cases, the translation transformation may be more important. By introducing the weight factor, the model is able to automatically adjust the impact of rotation and translation on the final transformation result according to the actual task data, while avoiding overfitting to specific data patterns.

[0160] S3138, determine that the transformed target image sample information is the target image sample information, and execute S3132.

[0161] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0162] In an optional embodiment, the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information are calculated and processed to obtain the rotation matrix information and the translation matrix information, including:

[0163] S31351, using the jaw key point correspondence relationship calculation model, calculating and processing the target body sample information, the target mirror image sample information, the body feature vector information and the mirror image feature vector information to obtain the jaw key point matching relationship information;

[0164] Among them, the calculation model of the corresponding relationship of the jaw key points is:

[0165]

[0166] 1≤i≤N;

[0167] 1≤j≤M;

[0168] In the formula, DY(x i ,Y), is the jaw key point matching relationship information, DY(x i ,Y) represents the matching relationship information between the coordinate information of the i-th object sample point in the target object sample information and the coordinate information of all mirror sample points in the target mirror sample information, JX is the mirror feature vector information, is the transpose of the i-th feature vector information in the ontology feature vector information, μ is the sampling parameter, τ is the attenuation factor, N is the number of ontology sample points in the target ontology sample information, M is the number of mirror sample points in the target mirror sample information, X is the target ontology sample information, Y is the target mirror sample information, x i is the coordinate information of the i-th entity sample point in the target entity sample information;

[0169] It should be noted that the mandibular key point matching relationship information indicates that a best matching mirror sample point is selected for each body sample point in the target body sample information, and the matching relationship of the body sample point is encoded in a one-hot form, where if x i If it matches the jth point in Y, the jth element of the corresponding one-hot vector is 1, and the others are 0, where the body sample point is a point in the target body sample information (point cloud data), and the mirror sample point is a point in the target mirror sample information (point cloud data).

[0170] It should be noted that the sampling parameter and the attenuation factor may be set by the user or obtained based on historical data, which is not limited in the embodiment of the present invention.

[0171] Exemplarily, the adoption factor is a value sampled from a Gumbel distribution (Gumbel (0,1)), and the attenuation factor ranges from [0.01, 0.1].

[0172] It should be noted that the sampling parameters introduce randomness and disturbance, which are usually used to enhance the exploratory nature of the model. In this way, the model can explore different matching point configurations during the optimization process, thereby avoiding falling into the local optimal solution; a larger attenuation factor can avoid excessive bias towards a certain solution during model training and maintain diversity; a smaller attenuation factor will prompt the model to converge to a deterministic solution faster, which is suitable for local optimization. In this embodiment, the initial value of the attenuation factor is 0.1. As the model continues to iterate, the value of the attenuation factor will continue to decrease, thereby controlling the smoothness of the softmax output.

[0173] S31352, calculating and processing the target body sample information and the target mirror sample information to obtain first matrix information and second matrix information;

[0174] It should be noted that the above calculation process is performed by a singular value decomposition algorithm, which is not specifically limited in the embodiment of the present invention.

[0175] S31353, using a jaw matrix decomposition calculation model, calculating and processing the first matrix information, the second matrix information and the jaw key point correspondence information to obtain rotation matrix information and translation matrix information;

[0176] Among them, the jaw matrix decomposition calculation model is:

[0177] XZ=JZ1·JZ2 T ;

[0178]

[0179] In the formula, XZ is the rotation matrix information, PY is the translation matrix information, JZ1 is the first matrix information, JZ2 T is the transpose of the second matrix information, and ε1 is the deviation displacement information.

[0180] It should be noted that the deviation displacement information may be set by a user, or may be obtained based on historical data, which is not limited in the embodiment of the present invention.

[0181] It should be noted that the deviation displacement information is used to translate the calculation results of the model to prevent the output from being completely dependent on the linear combination of the input features, which can help the model better fit the training data. In some cases, without the deviation displacement information, the model may not be able to effectively represent the distribution of data that deviates from the origin because it can only fit the data through the linear combination of the input features. After adding the deviation displacement information, the model can learn and optimize more flexibly, and it helps to improve numerical stability and prevent the convergence of the model from being affected by too small or too large values ​​when performing certain weighting or summation.

[0182] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0183] In an optional embodiment, the target body sample information, the target mirror sample information, the rotation matrix true value information, the translation matrix true value information, the first joint feature information set, the second joint feature information set, the rotation matrix information set, and the translation matrix information set are processed to obtain a loss function value, including:

[0184] Using the three-dimensional jaw symmetry reference plane loss function model, the target body sample information, the target mirror sample information, the rotation matrix true value information, the translation matrix true value information, the first joint feature information set, the second joint feature information set, the rotation matrix information set and the translation matrix information set are processed to obtain the loss function value;

[0185] Among them, the three-dimensional jaw symmetry reference plane loss function model is:

[0186]

[0187] 1≤k≤R;

[0188] 1≤l≤T;

[0189] δ1+δ2+δ3=1;

[0190] 0≤δ1,δ2,δ3≤1;

[0191] In the formula, SSZ is the loss function value, LHX l,k is the kth first joint key point feature information in the lth first joint feature information in the first joint feature information set, LHY l,k is the kth second joint key point feature information in the lth second joint feature information in the second joint feature information set, JZ l is the lth rotation matrix information in the rotation matrix information set, PY l is the lth translation matrix information in the translation matrix information set, JZZ and PYZ are the true value information of the rotation matrix and the true value information of the translation matrix respectively, x k Represents the coordinate information of the kth entity sample point of the target entity sample information, y k represents the coordinate information of the kth mirror sample point of the target mirror sample information, δ1, δ2 and δ3 represent the first weight parameter, the second weight parameter and the third weight parameter respectively, ‖·‖ represents the norm, R is the number of body sample points in the target body sample information, and T is the number of rotation matrix information in the rotation matrix information set.

[0192] It should be noted that the first weight parameter, the second weight parameter and the third weight parameter may be set by a user or may be obtained based on historical data, which is not limited in the embodiment of the present invention.

[0193] It should be noted that SSZ1 l By controlling the growth rate of feature loss through logarithmic transformation, the model can focus more on optimizing the details of small losses and suppress the contribution of abnormal points or abnormal matches in feature alignment to the total loss. l The alignment loss of the transformation matrix uses an exponential form, emphasizing the impact of small errors and providing highly sensitive gradient changes, so that the model converges to a better solution faster. l The square sum form of the geometric alignment loss is retained to ensure its direct optimization effect, and greater optimization efforts are applied to points with large errors to ensure the final alignment effect. Combining the three to obtain SSZ, it enhances robustness at both the feature level and the geometric level, is more adaptable to abnormal situations in point cloud registration, and helps to stabilize convergence during the optimization process and avoid falling into local minima, which can achieve high robustness, high precision and stability in point cloud alignment tasks.

[0194] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0195] In an optional embodiment, the first jaw optimization model and the second jaw optimization model are used to process the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information, including:

[0196] S321, using the first jaw optimization model and the second jaw optimization model, the three-dimensional jaw body data information to be processed is processed to obtain the three-dimensional jaw mirror data information to be processed, the target rotation matrix information and the target translation matrix information;

[0197] S322, calculating and processing the to-be-processed three-dimensional jaw body data information, the to-be-processed three-dimensional jaw mirror data information, the target rotation matrix information, and the target translation matrix information to obtain three-dimensional jaw joint data information;

[0198] It should be noted that the three-dimensional mandibular body data information to be processed, the three-dimensional mandibular mirror data information to be processed, the target rotation matrix information and the target translation matrix information are calculated and processed to obtain the three-dimensional mandibular joint data information, which is achieved through the following steps.

[0199] S3221, using the transformation matrix model, calculating and processing the three-dimensional jaw mirror image data information, the target rotation matrix information and the target translation matrix information to obtain the three-dimensional jaw body mirror image registration data information;

[0200] Among them, the transformation matrix model is:

[0201] DY′=DR·DY+DT;

[0202] Where DY′ is the three-dimensional mandibular body mirror registration data information, DR is the target rotation matrix information, DY is the three-dimensional mandibular mirror data information to be processed, and DT is the target translation matrix information;

[0203] S3222, merging the three-dimensional mandibular body data information to be processed and the three-dimensional mandibular body mirror registration data information to obtain three-dimensional mandibular joint data information.

[0204] It should be noted that the above merging process is obtained by processing the following three-dimensional jaw merging model;

[0205] Among them, the three-dimensional jaw combined model is:

[0206] DSL = DX ∪ DY′;

[0207] Wherein, DSL is the three-dimensional mandibular joint data information, and DX is the three-dimensional mandibular body data information to be processed.

[0208] S323, processing the three-dimensional mandibular joint data information to obtain three-dimensional mandibular symmetry reference plane information.

[0209] It should be noted that the above processing uses the PCA algorithm to calculate the covariance matrix and eigenvectors of the three-dimensional mandibular joint data information, selects the eigenvector with the largest variance as the normal vector of the three-dimensional mandibular symmetry reference plane, and uses the geometric center point of the three-dimensional mandibular joint data information as the center point of the three-dimensional mandibular symmetry reference plane. The three-dimensional mandibular symmetry reference plane information is constructed by the normal vector of the three-dimensional mandibular symmetry reference plane and the center point of the three-dimensional mandibular symmetry reference plane, wherein the construction is carried out through NumPy, CloudCompare or PCL (Point Cloud Library) to obtain the three-dimensional mandibular symmetry reference plane information.

[0210] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0211] In an optional embodiment, the first jaw optimization model and the second jaw optimization model are used to process the three-dimensional jaw body data information to be processed, and the three-dimensional jaw mirror data information to be processed, the target rotation matrix information and the target translation matrix information are obtained, including:

[0212] S3211, horizontally flipping the three-dimensional jaw body data information to be processed to obtain the three-dimensional jaw mirror data information to be processed;

[0213] It should be noted that the above-mentioned horizontal flipping process is consistent with the horizontal flipping process in S312.

[0214] S3212, using the first jaw optimization model, processing the three-dimensional jaw body data information to be processed and the three-dimensional jaw mirror data information to be processed to obtain local feature information of the body to be processed and local feature information of the mirror to be processed;

[0215] It should be noted that the above processing is to input the three-dimensional mandibular body data information to be processed and the three-dimensional mandibular mirror data information to be processed into the first mandibular optimization model respectively, and obtain the local feature information of the body to be processed corresponding to the three-dimensional mandibular body data information and the local feature information of the mirror to be processed corresponding to the three-dimensional mandibular mirror data information to be processed.

[0216] S3213, using the second mandibular optimization model, processing the to-be-processed body local feature information and the to-be-processed mirror local feature information to obtain the to-be-processed first joint feature information and the to-be-processed second joint feature information;

[0217] It should be noted that the above processing is to use the local feature information of the body to be processed and the local feature information of the mirror image to be processed as the second mandibular optimization model, and calculate the first joint feature information to be processed and the second joint feature information to be processed through the second mandibular optimization model.

[0218] S3214, fusing the local feature information of the subject to be processed, the three-dimensional jaw mirror image data information to be processed, the first joint feature information to be processed, and the second joint feature information to be processed to obtain feature vector information of the subject to be processed and feature vector information of the mirror image to be processed;

[0219] It should be noted that the above-mentioned fusion processing is to add the local feature information of the body to be processed and the first joint feature information to be processed to obtain the feature vector information of the body to be processed, and to add the three-dimensional mandibular mirror data information to be processed and the second joint feature information to be processed to obtain the mirror feature vector information to be processed. Specifically, the embodiment of the present invention does not limit this.

[0220] S3215, calculating and processing the three-dimensional mandibular body data information to be processed, the three-dimensional mandibular mirror image data information to be processed, the body feature vector information to be processed and the mirror image feature vector information to be processed to obtain the target rotation matrix information and the target translation matrix information.

[0221] It should be noted that the above-mentioned calculation and processing is to use the three-dimensional mandibular body data information to be processed, the three-dimensional mandibular mirror data information to be processed, the body feature vector information to be processed and the mirror feature vector information to be processed as the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information respectively, execute steps S31351-S31353, obtain the rotation matrix information and the translation matrix information, and determine the rotation matrix information as the target rotation matrix information, and determine the translation matrix information as the target translation matrix information.

[0222] It can be seen that the method of automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to the automatic, efficient and accurate construction of three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, and providing a method for constructing a three-dimensional mandibular symmetry reference plane.

[0223] Embodiment 2

[0224] See also Figure 2 , Figure 2Schematic diagram of a device for automatically constructing a three-dimensional jaw symmetry reference plane disclosed in an embodiment of the present invention. Figure 2 The described device for automatically constructing a three-dimensional jaw symmetry reference plane is applied to an optimization system for automatically constructing a three-dimensional jaw symmetry reference plane, such as a local server or cloud server for automatically constructing a three-dimensional jaw symmetry reference plane, and the embodiments of the present invention are not limited thereto. Figure 2 As shown, the device for automatically constructing a three-dimensional jaw symmetry reference plane includes:

[0225] An acquisition module 201 is used to acquire a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed;

[0226] The first calculation module 202 is used to preprocess the three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set; the preprocessed three-dimensional jaw body data set includes a plurality of preprocessed three-dimensional jaw body sample information;

[0227] The second calculation module 203 is used to process the pre-processed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

[0228] It can be seen that the device for automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to automatically, efficiently and accurately constructing three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, etc.

[0229] Embodiment 3

[0230] See also Figure 3 , Figure 3 is a schematic diagram of another structure of a device for automatically constructing a three-dimensional jaw symmetry reference plane disclosed in an embodiment of the present invention. Figure 3 The described device for automatically constructing a three-dimensional jaw symmetry reference plane is applied to an optimization system for automatically constructing a three-dimensional jaw symmetry reference plane, such as a local server or cloud server for automatically constructing a three-dimensional jaw symmetry reference plane, and the embodiments of the present invention are not limited thereto. Figure 3 As shown, the device for automatically constructing a three-dimensional jaw symmetry reference plane includes:

[0231] Processor 301;

[0232] A memory 302 coupled to the processor 301 and storing executable program codes;

[0233] The processor 301 calls the executable program code stored in the memory 302 to execute part or all of the steps of the method for automatically constructing a three-dimensional jaw symmetry reference plane in the first embodiment.

[0234] It can be seen that the device for automatically constructing a three-dimensional mandibular symmetry reference plane as described in the embodiment of the present invention is conducive to automatically, efficiently and accurately constructing three-dimensional mandibular symmetry reference plane information, thereby providing a reference basis for the fields of oral clinical orthodontics, digital diagnosis and treatment of maxillofacial surgery, etc.

[0235] Embodiment 4

[0236] An embodiment of the present invention discloses a computer-readable storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute part or all of the steps of the method for automatically constructing a three-dimensional mandibular symmetry reference plane in embodiment one.

[0237] Embodiment 5

[0238] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute part or all of the steps in the method for automatically constructing a three-dimensional mandibular symmetry reference plane described in Example 1.

[0239] The system embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative labor.

[0240] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0241] Finally, it should be noted that the method and device for automatically constructing a three-dimensional mandibular symmetry reference plane disclosed in the embodiment of the present invention only discloses a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for automatically constructing a three-dimensional jaw symmetry reference plane, characterized in that: The method comprises: S1, obtaining a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed; S2, preprocessing the three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set; the preprocessed three-dimensional jaw body data set includes a plurality of preprocessed three-dimensional jaw body sample information; S3, processing the preprocessed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

2. The method for automatically constructing a three-dimensional jaw symmetry reference plane according to claim 1, characterized in that: The preprocessing of the three-dimensional jaw body dataset to obtain a preprocessed three-dimensional jaw body dataset includes: S21, performing data format unification processing on the three-dimensional jaw body dataset to obtain a first three-dimensional jaw body dataset; S22, performing denoising processing on the first three-dimensional jaw body dataset to obtain a second three-dimensional jaw body dataset; S23, segmenting the second three-dimensional jaw body dataset to obtain a third three-dimensional jaw body dataset; S24, performing smoothing processing on the third three-dimensional jaw body dataset to obtain a preprocessed three-dimensional jaw body dataset.

3. The method for automatically constructing a three-dimensional jaw symmetry reference plane according to claim 1, characterized in that: The step of processing the preprocessed three-dimensional jaw body data set and the to-be-processed three-dimensional jaw body data information to obtain three-dimensional jaw symmetry reference plane information includes: S31, using the preprocessed three-dimensional jaw body data set, training the first jaw initial model and the second jaw initial model to obtain the first jaw optimized model and the second jaw optimized model; S32, using the first jaw optimization model and the second jaw optimization model, processing the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

4. The method for automatically constructing a three-dimensional jaw symmetry reference plane according to claim 3, characterized in that: The method of using the preprocessed three-dimensional jaw body data set to train the first jaw initial model and the second jaw initial model to obtain the first jaw optimized model and the second jaw optimized model includes: S311, performing a random selection operation on the preprocessed three-dimensional jaw body data set to obtain target body sample information; the target body sample information is any preprocessed three-dimensional jaw body sample information in the preprocessed three-dimensional jaw body data set; S312, performing horizontal flipping processing on the target body sample information to obtain target mirror sample information; S313, using the first jaw initial model and the second jaw initial model, the target body sample information and the target mirror sample information are processed to obtain a first joint feature information set, a second joint feature information set, a rotation matrix information set and a translation matrix information set; the first joint feature information set includes a plurality of first joint feature information; the second joint feature information set includes a plurality of second joint feature information; the rotation matrix information set includes a plurality of rotation matrix information; the translation matrix information set includes a plurality of translation matrix information; S314, obtaining true value information of the rotation matrix and true value information of the translation matrix; S315, processing the target body sample information, the target mirror sample information, the rotation matrix true value information, the translation matrix true value information, the first joint feature information set, the second joint feature information set, the rotation matrix information set, and the translation matrix information set to obtain a loss function value; S316, determining whether the loss function value is less than a preset loss function threshold, and obtaining a first determination result; When the first judgment result is yes, execute S317; When the first judgment result is no, executing S318; S317, determining the first jaw initial model as a first jaw optimized model, determining the second jaw initial model as a second jaw optimized model, and executing S32; S318, updating the first jaw initial model and the second jaw initial model to obtain an updated first jaw initial model and an updated second jaw initial model; S319, determining that the updated first jaw initial model is the first jaw initial model, determining that the updated second jaw initial model is the second jaw initial model, and executing S311.

5. The method for automatically constructing a three-dimensional jaw symmetry reference plane according to claim 4, characterized in that: The first jaw initial model and the second jaw initial model are used to process the target body sample information and the target mirror sample information to obtain a first joint feature information set, a second joint feature information set, a rotation matrix information set, and a translation matrix information set, including: S3131, preset s=1; S3132, using the first jaw initial model, processing the target body sample information and the target mirror sample information to obtain body local feature information and mirror local feature information; S3133, using the second initial jaw model, processing the local feature information of the main body and the local feature information of the mirror image to obtain first joint feature information and second joint feature information; the first joint feature information includes a plurality of first joint key point feature information; the second joint feature information includes a plurality of second joint key point feature information S3134, fusing the body local feature information, the mirror image local feature information, the first joint feature information, and the second joint feature information to obtain body feature vector information and mirror image feature vector information; S3135, calculating and processing the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information to obtain rotation matrix information and translation matrix information; S3136, determining whether s is equal to a preset number of iterations, and obtaining a second determination result; When the second judgment result is no, increase the s by 1 and execute S3137; When the second judgment result is yes, execute S314; S3137, using the rotation matrix information and the translation matrix information, performing matrix transformation processing on the target mirror sample information to obtain transformed target mirror sample information; S3138, determining that the transformed target image sample information is the target image sample information, and executing S3132.

6. The method for automatically constructing a three-dimensional jaw symmetry reference plane according to claim 5, characterized in that: The calculating and processing the target body sample information, the target mirror sample information, the body feature vector information and the mirror feature vector information to obtain the rotation matrix information and the translation matrix information includes: S31351, using a jaw key point correspondence relationship calculation model, calculating and processing the target body sample information, the target mirror image sample information, the body feature vector information and the mirror image feature vector information to obtain jaw key point matching relationship information; The calculation model of the corresponding relationship of the jaw key points is: In the formula, is the jaw key point matching relationship information, DY(x i , Y) represents the matching relationship information between the coordinate information of the i-th body sample point in the target body sample information and the coordinate information of all mirror sample points in the target mirror sample information, JX is the mirror feature vector information, is the transpose of the i-th feature vector information in the subject feature vector information, μ is the sampling parameter, τ is the attenuation factor, N is the number of subject sample points in the target subject sample information, M is the number of mirror sample points in the target mirror sample information, X is the target subject sample information, Y is the target mirror sample information, x i is the coordinate information of the i-th ontology sample point in the target ontology sample information; S31352, calculating and processing the target body sample information and the target mirror sample information to obtain first matrix information and second matrix information; S31353, using a jaw matrix decomposition calculation model, calculating and processing the first matrix information, the second matrix information and the jaw key point correspondence information to obtain rotation matrix information and translation matrix information; Among them, the jaw matrix decomposition calculation model is: XZ=JZ1·JZ2T; Where XZ is the rotation matrix information, PY is the translation matrix information, JZ1 is the first matrix information, JZ2 T is the transpose of the second matrix information, and ε1 is the deviation displacement information.

7. The method for automatically constructing a three-dimensional jaw symmetry reference plane according to claim 3, characterized in that: The method of using the first jaw optimization model and the second jaw optimization model to process the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information includes: S321, using the first jaw optimization model and the second jaw optimization model, processing the three-dimensional jaw body data information to be processed to obtain the three-dimensional jaw mirror data information to be processed, target rotation matrix information and target translation matrix information; S322, calculating and processing the to-be-processed three-dimensional jaw body data information, the to-be-processed three-dimensional jaw mirror data information, the target rotation matrix information and the target translation matrix information to obtain three-dimensional jaw joint data information; S323, processing the three-dimensional mandibular joint data information to obtain three-dimensional mandibular symmetry reference plane information.

8. A device for automatically constructing a three-dimensional jaw symmetry reference plane, characterized in that: The device comprises: An acquisition module is used to acquire a three-dimensional jaw body data set and three-dimensional jaw body data information to be processed; A first calculation module is used to preprocess the three-dimensional jaw body data set to obtain a preprocessed three-dimensional jaw body data set; the preprocessed three-dimensional jaw body data set includes a plurality of preprocessed three-dimensional jaw body sample information; The second calculation module is used to process the pre-processed three-dimensional jaw body data set and the three-dimensional jaw body data information to be processed to obtain three-dimensional jaw symmetry reference plane information.

9. A device for automatically constructing a three-dimensional jaw symmetry reference plane, characterized in that: The device comprises: processor; a memory coupled to the processor and storing executable program code; The processor calls the executable program code stored in the memory to execute the method for automatically constructing a three-dimensional mandibular symmetry reference plane as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, which, when called, are used to execute the method for automatically constructing a three-dimensional mandibular symmetry reference plane as described in any one of claims 1-7.

Images