A multi-view point cloud automatic registration method for tooth three-dimensional modeling

Abstract

The present application relates to the field of image data processing, especially to a multi-view point cloud automatic registration method for tooth three-dimensional modeling. The content includes: obtaining multi-view tooth three-dimensional point cloud data as original point cloud data and preprocessing to obtain preprocessed point cloud data; adaptive embedded feature mapping is performed on the preprocessed point cloud data, based on the mapped multi-view point cloud feature data, a similarity measurement method based on feature space is introduced, the similarity between the preprocessed point cloud data of different views is calculated to obtain the similarity measurement; a transformation matrix is introduced and combined with the similarity measurement for optimization processing to obtain the optimal transformation matrix; the preprocessed point cloud data is transformed based on the optimal transformation matrix to realize automatic registration. The technical problem of low registration accuracy caused by inaccurate point cloud data processing and inaccurate similarity analysis between multi-view point clouds of traditional point cloud registration method is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image data processing, and in particular to a multi-view point cloud automatic registration method for tooth three-dimensional modeling. BACKGROUND

[0002] In the field of three-dimensional computer vision and point cloud processing, point cloud registration is a key technology that combines point cloud data from different perspectives or different time points into a unified three-dimensional model, and is widely used in various fields such as computer-aided design (CAD), virtual reality (VR), robot perception, cultural heritage protection, etc. Through technologies such as laser scanning, stereo vision, structured light, and depth sensors, point cloud data of a three-dimensional space is obtained from multiple perspectives, forming a collection of three-dimensional points that record the geometric shape and spatial distribution of the object. However, point cloud data from different perspectives often has significant differences, especially when the perspective is rotated, translated, or scaled, point cloud data may have problems such as alignment difficulty, insufficient overlap, or missing.

[0003] Traditional point cloud registration methods mainly rely on two types of technologies: feature-based registration methods and direct method-based registration methods. Feature-based registration methods match key features in point clouds (such as corner points, edges, surface normals, etc.) by extracting them, although these methods work well in some simple cases, but for complex shapes or in noisy environments, especially in sparse point clouds or severe occlusion, the accuracy of feature matching is often low; direct method-based registration methods usually minimize the error between point clouds for registration, but these methods have large computational complexity when dealing with large-scale point clouds, and are prone to local optimal solutions, resulting in inaccurate registration results.

[0004] In summary, the above-mentioned traditional point cloud registration methods also have the technical problems of inaccurate processing of point cloud data and inaccurate similarity analysis between multi-view point clouds, resulting in low registration accuracy. SUMMARY

[0005] The present application provides a multi-view point cloud automatic registration method for tooth three-dimensional modeling to solve the technical problem of low registration accuracy caused by inaccurate processing of point cloud data and inaccurate similarity analysis between multi-view point clouds in traditional point cloud registration methods.

[0006] A multi-view point cloud automatic registration method for tooth three-dimensional modeling according to the present application, specifically includes the following technical solutions:

[0007] A multi-view point cloud automatic registration method for tooth three-dimensional modeling, including the following steps:

[0008] S1. Obtain multi-view three-dimensional point cloud data of teeth as original point cloud data, and pre-process the original point cloud data to obtain pre-processed point cloud data; perform adaptive embedded feature mapping on the pre-processed point cloud data to obtain mapped multi-view point cloud feature data;

[0009] S2. Based on the mapped multi-view point cloud feature data, introduce a feature space-based similarity measurement method to calculate the similarity between the pre-processed point cloud data of different views to obtain a similarity measurement.

[0010] S3. Introduce a transformation matrix and perform optimization processing in combination with the similarity measurement to obtain an optimal transformation matrix; perform transformation on the pre-processed point cloud data based on the optimal transformation matrix to realize automatic registration.

[0011] Preferably, the S1 specifically comprises:

[0012] In the implementation process of the adaptive embedded feature mapping, the distance measurement between the points in the pre-processed point cloud data under any view and their neighborhood points is calculated, and the mapped multi-view point cloud feature data is obtained in combination with the weight factor of the view and the mapping weight of the neighborhood points.

[0013] Preferably, the S2 specifically comprises:

[0014] The feature space-based similarity measurement method calculates the similarity between the pre-processed point cloud data of different views by calculating the distance between the mapped multi-view point cloud feature data and introducing angle and gradient information.

[0015] Preferably, the S2 specifically comprises:

[0016] In the implementation process of the feature space-based similarity measurement method, the weight coefficients of the angle difference and the gradient difference are introduced to construct the angle difference penalty term and the gradient difference penalty term, and the feature similarity score term constructed based on the distance between the mapped multi-view point cloud feature data is combined to calculate the similarity measurement.

[0017] Preferably, the S3 specifically comprises:

[0018] Introduce a transformation matrix to transform the pre-processed point cloud data to obtain transformed point cloud data.

[0019] Preferably, the S3 specifically comprises:

[0020] Based on the transformed point cloud data, the transformation matrix is optimized by a graph-based optimization algorithm.

[0021] Preferably, the S3 specifically comprises:

[0022] In the implementation process of the graph-based optimization algorithm, the optimal transformation matrix is obtained by constructing and minimizing a global optimization objective function.

[0023] Preferably, the S3 specifically comprises:

[0024] The global optimization objective function is constructed based on the transformed point cloud data, combined with similarity measurement, and error smoothing term and matching similarity measurement term.

[0025] The technical scheme of the present application has the following advantages:

[0026] 1. By using the adaptive embedded feature mapping technology, the preprocessed point cloud data under each view angle is mapped and processed, the contrast between the preprocessed point cloud data under different view angles is enhanced, the geometric information of the preprocessed point cloud data in the feature space is more accurately described, complex geometric shapes can be processed, and the accuracy and stability of point cloud registration are greatly improved, especially when facing noise, data scattering or view angle change, the matching accuracy can be effectively enhanced.

[0027] 2. In the traditional point cloud registration method, factors such as view angle difference, data noise and lighting conditions usually affect the matching quality of point cloud, the present application introduces a similarity measurement method based on feature space, which can measure the distance between the mapped multi-view point cloud feature data, and optimize the matching process combined with angle difference and gradient information, thereby effectively solving the matching difficulty caused by view angle change and geometric difference, and making the alignment of similar point clouds more accurate.

[0028] 3. The graph-based optimization algorithm is used to process the registration problem of the preprocessed point cloud data under multiple view angles, which can maximize the registration accuracy of all view angle point clouds, compared with the traditional method of adjusting the transformation matrix of single view angle point cloud one by one, the graph-based optimization algorithm considers the transformation of multiple view angle point clouds at the same time, through this joint optimization method, the local optimal solution problem that may be caused by single transformation of a view angle is avoided, and the global consistency and accuracy of registration are improved. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 A flowchart of a multi-view point cloud automatic registration method for tooth three-dimensional modeling according to the present application. DETAILED DESCRIPTION

[0030] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0031] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0032] The specific scheme of the multi-view point cloud automatic registration method for tooth three-dimensional modeling provided by the present application will be specifically described below in conjunction with the drawings.

[0033] Referring to the drawings Figure 1 , which shows a multi-view point cloud automatic registration method flow chart for tooth three-dimensional modeling provided by an embodiment of the present application, the method comprises the following steps:

[0034] S1. Obtain multi-view tooth three-dimensional point cloud data as original point cloud data, and pre-process the original point cloud data to obtain pre-processed point cloud data; perform adaptive embedded feature mapping on the pre-processed point cloud data to obtain mapped multi-view point cloud feature data;

[0035] Through existing scanning technology (such as structured light, laser radar, stereo vision, depth sensor, etc.), multi-view tooth three-dimensional point cloud data is obtained from different views (such as front, side, top, bottom and other angles, etc.) as original point cloud data, and the original point cloud data is pre-processed to obtain pre-processed point cloud data; the pre-processing process includes data cleaning, denoising processing, standardization and normalization, etc., and the methods used are all known technical means to those skilled in the art, which will not be described here;

[0036] The pre-processed point cloud data is subjected to adaptive embedded feature mapping to obtain mapped multi-view point cloud feature data, so as to enhance the contrast of the pre-processed point cloud data under different views in the feature space, thereby realizing more accurate matching. Based on local geometric feature mapping and weighted distance measurement method, an adaptive embedded feature mapping formula is constructed, and the specific mathematical expression is:

[0037] ,

[0038] Wherein, represents the point cloud data under the i-th view mapped feature vector of the point , i.e., the mapped multi-view point cloud feature data, contains the geometric information of the point and the local geometric features of the point under the view represent the positions, i.e., three-dimensional coordinates, of the point and the point under the view , and the point is a neighborhood point of the point under the view ; is the number of neighborhood points, i.e., the number of local related points, of the point under the view in the preprocessed point cloud data, which is determined based on the Euclidean distance between points and is not limited here; is the mapping weight of the point under the view , which represents the contribution degree of the point under the view to the feature mapping, and is determined according to expert experience, with a reference value range of ; is the weight factor of the view , which is used to adjust the contribution degree of different views to the overall point cloud feature and reflects the importance of the view in the final feature mapping, and is determined according to specific application scenarios, with a reference value range of ; is a power index for distance measurement, which determines the influence degree of local point cloud on feature mapping, and is determined according to specific application requirements, with a reference value range of ; represents the distance measurement between the point and the point under the view , which plays a role in measuring local geometric relationships in feature mapping.

[0039] S2. Based on the mapped multi-view point cloud feature data, a feature space-based similarity measurement method is introduced to calculate the similarity between the preprocessed point cloud data of different views, to obtain a similarity measurement.

[0040] Based on the mapped multi-view point cloud feature data, a feature space-based similarity measurement method is introduced to calculate the similarity between the preprocessed point cloud data of different views, to obtain a similarity measurement. The key of the similarity measurement is to measure the distance between the mapped multi-view point cloud feature data, and to introduce angle and gradient information to optimize the matching process to ensure accurate alignment of similar point clouds. The formula of the similarity measurement is:​

[0041] ,

[0042] wherein, represents the similarity measure between the pre-processed point cloud data under any two views and , the smaller the similarity measure, the higher the matching degree of the pre-processed point cloud data in space; and respectively represent the pre-processed point cloud data under the first and the first view; and respectively represent the number of all points in the pre-processed point cloud data under the first and the first view; is the position of the point under the first view; is the mapped multi-view point cloud feature data vector of the point under the first view; is a constant to prevent division by zero error, used to ensure numerical stability when calculating the similarity measure, which can be taken as ; is a weight coefficient of the angle difference, used to control the influence degree of the angle difference on the point cloud matching similarity calculation, which is determined according to the application scenario, and the reference value range is ; represents the angle difference between the points and under any two views, which is calculated based on the normal vector of the point, reflecting the difference caused by the rotation of the point cloud, and the calculation method of the angle difference is a technical means familiar to those skilled in the art, which will not be repeated here; is a weight coefficient of the gradient difference, used to control the influence degree of the local geometric difference (such as curvature) of the point cloud, which is determined according to specific needs, and the reference value range is ; and respectively represent the gradient of the point and under the first and the first view, i.e. the curvature in the local neighborhood, reflecting the difference of the local surface geometry of the point cloud, and the local neighborhood is determined based on the Euclidean distance between points combined with specific needs, which is not limited here; The square of the Euclidean distance between the mapped multi-view point cloud feature data is used to measure similarity. The smaller the Euclidean distance, the more similar the two points are in terms of geometric properties and spatial structure. It is a feature similarity scoring item, used to measure whether two perspectives correspond in the feature space. Through normalization processing, the feature similarity scoring item is not affected by the feature magnitude and is one of the core parts of similarity measurement. It is an angle difference penalty term used to describe the compatibility of directional features between point clouds. The greater the angle difference, the stronger the penalty term, so as to force the point cloud to maintain local directional consistency during the alignment process. It is a gradient difference penalty term used to measure the similarity of point clouds from two viewpoints in local geometry, which can effectively enhance the alignment capability for details of surface changes;

[0043] By introducing angle and gradient information during the similarity calculation process, mismatches caused by differences in viewpoint can be effectively reduced.

[0044] S3. Introduce a transformation matrix and optimize it using a similarity metric to obtain the optimal transformation matrix; transform the preprocessed point cloud data based on the optimal transformation matrix to achieve automatic registration.

[0045] After obtaining the similarity metric, global optimization is used to further improve registration accuracy. To eliminate the problem caused by local mismatches, a graph-based optimization algorithm is adopted to adjust the transformation of preprocessed point cloud data at each viewpoint globally, ensuring overall alignment of point cloud data across multiple views. The graph-based optimization algorithm is implemented through the following global optimization objective function, the specific mathematical expression of which is:

[0046] ,

[0047] in, Indicates the first Preprocessed point cloud data from various perspectives The transformation matrix, by optimizing the transformation matrix, can align preprocessed point cloud data from different perspectives to a unified global coordinate system. The transformation matrix includes geometric transformations such as rotation, translation, or scaling, and is used to describe the transformation of preprocessed point cloud data from one coordinate system to another. Indicates the first Point cloud data after transformation from a single perspective; Represents a point cloud mapping in feature space, indicating the point cloud mapping in the feature space. The feature vectors of point cloud data after transformation from each perspective in the feature space; It is in the Feature vectors of preprocessed point cloud data in feature space from various perspectives; is the pre-processed point cloud data under the first view is the pre-processed point cloud data under the first view is the matching weight between the pre-processed point cloud data under the first view and the pre-processed point cloud data under the second view, reflecting the similarity or importance between the two point clouds, which is calculated based on the Euclidean distance of the point cloud data; is a smoothing control factor, used to represent the proportion of the error smoothing term in the global optimization objective function, which is determined according to expert experience method, and the reference value range is ; is a scale parameter, used to ensure that the error of the feature space in the geometric transformation process is properly scaled, avoiding excessive dependence on large errors in the optimization process, which is determined according to expert experience method, and the reference value range is ; is a weight coefficient of similarity measurement, used to control the influence degree of the similarity measurement term on the global optimization objective function, which is determined according to the quality of the pre-processed point cloud data and the accuracy requirement of registration, and the reference value range is ; represents the transformed point cloud data under the first view is the similarity measurement of the transformed point cloud data under the first view and the pre-processed point cloud data under the second view; is an error smoothing term, aiming to suppress larger error values to prevent some view error outliers from having a serious impact on the overall optimization objective in the optimization process;

[0048] is a matching similarity measurement term, used to constrain the degree of detail alignment of the transformed point cloud data and the reference point cloud data in the feature space; Based on the above global optimization objective function, the nonlinear least squares method is used to find the optimal transformation matrix of all views, ensuring the joint registration of multi-view point clouds; and based on the optimal transformation matrix of each view, the pre-processed point cloud data of the view is transformed to optimize the relative position of the pre-processed point cloud data under each view in the global range, eliminate the influence of mis-matching points, and thus achieve more accurate registration.

[0049]

[0050] The order of the embodiments of the application is only for description, not representing the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0051] ​​​Various embodiments are described herein with reference to the drawings. Each embodiment is described in a progressive manner, and the same or similar parts between various embodiments are cross-referenced. Each embodiment focuses on the differences from other embodiments.

[0052] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent features. The modification or replacement does not change the essence of the corresponding technical solutions, and should be included in the protection scope of the present application.

Claims

1. A multi-view point cloud automatic registration method for tooth three-dimensional modeling, characterized in that, The method comprises the following steps: S1. Obtain multi-view three-dimensional point cloud data of teeth as original point cloud data, and pre-process the original point cloud data to obtain pre-processed point cloud data; Adaptive embedded feature mapping is performed on the pre-processed point cloud data, the distance metric between a point in the pre-processed point cloud data under any view and its neighborhood points is calculated, and the mapping weight of the neighborhood points and the weight factor of the view are combined to obtain the mapped multi-view point cloud feature data, and the specific formula is: ， in, Indicates the first From a point of view The feature vector after mapping, i.e. the multi-view point cloud feature data after mapping; and They represent the first Points from different perspectives and points Location, point It is the first From a point of view , neighborhood points; It is the first From one perspective, the midpoint of the preprocessed point cloud data The number of neighboring points; It is in the From a point of view Mapping weights; It is the first Weighting factors for each perspective; It is the power exponent that controls the distance metric; Indicates the first From a point of view and points Distance metric between; S2. Based on the mapped multi-view point cloud feature data, a similarity measurement method based on feature space is introduced to calculate the similarity between the pre-processed point cloud data of different views to obtain a similarity measurement; S3. A transformation matrix is introduced to transform the pre-processed point cloud data to obtain transformed point cloud data; based on the transformed point cloud data, a graph-based optimization algorithm is applied, the similarity measurement is combined, an error smoothing term and a matching similarity measurement term are introduced, a global optimization objective function is constructed and minimized, the transformation matrix is optimized to obtain an optimal transformation matrix; the pre-processed point cloud data is transformed based on the optimal transformation matrix to realize automatic registration.

2. The multi-view point cloud automatic registration method for tooth three-dimensional modeling according to claim 1, wherein, The S2 specifically comprises: The similarity measurement method based on feature space calculates the distance between the mapped multi-view point cloud feature data, and introduces angle and gradient information to calculate the similarity between the pre-processed point cloud data of different views.

3. The multi-view point cloud automatic registration method for tooth three-dimensional modeling according to claim 2, wherein, The S2 specifically comprises: In the implementation process of the similarity measurement method based on feature space, the weight coefficients of angle difference and gradient difference are introduced, the angle difference penalty term and the gradient difference penalty term are constructed, and the feature similarity score term constructed based on the distance between the mapped multi-view point cloud feature data is combined to calculate the similarity measurement.

4. The multi-view point cloud automatic registration method for tooth three-dimensional modeling according to claim 1, wherein, The S3 specifically comprises: The specific formula of the global optimization objective function is: ， in, Indicates the first Preprocessed point cloud data from various perspectives The transformation matrix; Indicates the total number of viewpoints; Indicates the first Point cloud data after transformation from a single perspective; Represents a point cloud mapping in feature space, indicating the point cloud mapping in the feature space. The feature vectors of point cloud data after transformation from each perspective in the feature space; It is in the Feature vectors of preprocessed point cloud data in feature space from various perspectives; It is the first Preprocessed point cloud data from various perspectives and the Preprocessed point cloud data from various perspectives Matching weights between them; It is a smoothing control factor; It is a scale parameter; These are the weighting coefficients for the similarity metric; Indicates the first Point cloud data after transformation from a different perspective With the Preprocessed point cloud data from various perspectives Similarity measure; It is the error smoothing term; It is a similarity metric.

Images