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

Abstract

The invention relates to the field of image data processing, in particular to a multi-view point cloud automatic registration method for tooth three-dimensional modeling. The method comprises the following steps: acquiring multi-view tooth three-dimensional point cloud data as original point cloud data, and preprocessing the original point cloud data to obtain preprocessed point cloud data; carrying out adaptive embedded feature mapping on the preprocessed point cloud data, introducing a similarity measurement method based on a feature space based on the mapped multi-view point cloud feature data, and calculating the similarity between the preprocessed point cloud data of different views to obtain similarity measurement; introducing a transformation matrix, and performing optimization processing in combination with similarity measurement to obtain an optimal transformation matrix; and transforming the preprocessed point cloud data based on the optimal transformation matrix to realize automatic registration. The technical problem of low registration precision caused by inaccurate point cloud data processing and inaccurate similarity analysis among multi-view point clouds of a traditional point cloud registration method is solved.

Description

Technical Field

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

[0002] In the fields of 3D computer vision and point cloud processing, point cloud registration is a key technology. It combines point cloud data from different viewpoints or at different points in time into a unified 3D model. It is widely used in a variety of fields, such as computer-aided design (CAD), virtual reality (VR), robotic perception, and cultural heritage preservation. Using technologies such as laser scanning, stereo vision, structured light, and depth sensors, 3D point cloud data is acquired from multiple viewpoints, forming a collection of 3D points that records the geometric form and spatial distribution of objects. However, point cloud data from different viewpoints often exhibit significant differences. In particular, when the viewpoint is rotated, translated, or scaled, point cloud data can be difficult to align, lack overlap, or be missing.

[0003] Traditional point cloud registration methods rely primarily on two types of techniques: feature-based registration methods and direct registration methods. Feature-based registration methods extract key features from point clouds (such as corners, edges, surface normals, etc.) for matching. Although these methods work well in some simple cases, they often have low feature matching accuracy for complex shapes or in noisy environments, especially when point clouds are sparse or heavily occluded. Direct registration methods typically perform registration by minimizing the error between point clouds. However, these methods are computationally intensive when processing large-scale point clouds and are prone to falling into local optimal solutions, resulting in inaccurate registration results.

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

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

[0006] The present invention provides a multi-view point cloud automatic registration method for three-dimensional tooth modeling, which specifically includes the following technical solutions: A multi-view point cloud automatic registration method for three-dimensional tooth modeling includes the following steps: S1. Acquire multi-view three-dimensional tooth point cloud data as raw point cloud data, preprocess the raw point cloud data to obtain preprocessed point cloud data; perform adaptive embedding feature mapping on the preprocessed point cloud data to obtain mapped multi-view point cloud feature data; 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 preprocessed point cloud data of different viewpoints to obtain a similarity measurement; S3. Introduce the transformation matrix and optimize it in combination with the similarity metric to obtain the optimal transformation matrix; transform the preprocessed point cloud data based on the optimal transformation matrix to achieve automatic alignment.

[0007] Preferably, the S1 specifically includes: In the implementation process of adaptive embedding feature mapping, the distance measurement between a point in the preprocessed point cloud data and its neighborhood points at any perspective is calculated, and the weight factor of the perspective and the mapping weight of the neighborhood points are combined to obtain the mapped multi-view point cloud feature data.

[0008] Preferably, the S2 specifically includes: The feature space-based similarity measurement method calculates the similarity between point cloud data preprocessed from different perspectives by calculating the distance between the mapped multi-view point cloud feature data and introducing angle and gradient information.

[0009] Preferably, the S2 specifically includes: In the implementation of the similarity measurement method based on feature space, the weight coefficients of angle difference and gradient difference are introduced to construct angle difference penalty terms and gradient difference penalty terms, and the similarity measurement is calculated by combining the feature similarity scoring terms constructed based on the distance between the mapped multi-view point cloud feature data.

[0010] Preferably, the S3 specifically includes: The transformation matrix is ​​introduced to transform the preprocessed point cloud data to obtain the transformed point cloud data.

[0011] Preferably, the S3 specifically includes: Based on the transformed point cloud data, the transformation matrix is ​​optimized through a graph-based optimization algorithm.

[0012] Preferably, the S3 specifically includes: In the implementation of the graph-based optimization algorithm, the optimal transformation matrix is ​​obtained by constructing and minimizing the global optimization objective function.

[0013] Preferably, the S3 specifically includes: The global optimization objective function is constructed based on the transformed point cloud data, combined with similarity measurement, and introduced with error smoothing term and matching similarity measurement term.

[0014] The beneficial effects of the technical solution of the present invention are: 1. Through adaptive embedding feature mapping technology, the pre-processed point cloud data at each perspective is mapped, which enhances the contrast between the pre-processed point cloud data at different perspectives. The geometric information of the pre-processed point cloud data in the feature space is described more accurately, and it can handle complex geometric shapes, greatly improving the accuracy and stability of point cloud registration. In particular, it can effectively enhance matching accuracy in the face of noise, data scattering or perspective changes.

[0015] 2. In traditional point cloud registration methods, factors such as perspective differences, data noise, and lighting conditions usually affect the matching quality of point clouds. The present invention introduces a similarity measurement method based on feature space, which can measure the distance between the mapped multi-perspective point cloud feature data, and optimize the matching process by combining perspective differences and gradient information, thereby effectively solving the matching difficulties caused by perspective changes and geometric differences, making the alignment of similar point clouds more accurate.

[0016] 3. By using a graph-based optimization algorithm to handle the registration problem of pre-processed point cloud data from multiple perspectives, the registration accuracy of point clouds from all perspectives can be maximized. Compared with the traditional method of adjusting the transformation matrix of a single perspective point cloud one by one, the graph-based optimization algorithm simultaneously considers the transformation of point clouds from multiple perspectives. Through this joint optimization method, the local optimal solution problem that may be caused by the transformation of a single perspective is avoided, and the global consistency and accuracy of the registration are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the multi-view point cloud automatic registration method for three-dimensional tooth modeling described in the present invention. DETAILED DESCRIPTION

[0018] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the technical solutions 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 making creative efforts shall fall within the scope of protection of the present invention.

[0019] Unless defined otherwise, 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 invention belongs.

[0020] The following describes in detail a specific solution of a multi-view point cloud automatic registration method for three-dimensional tooth modeling provided by the present invention with reference to the accompanying drawings.

[0021] Refer to the attached Figure 1 , which shows a flow chart of a multi-view point cloud automatic registration method for three-dimensional tooth modeling provided by one embodiment of the present invention, the method comprising the following steps: S1. Acquire multi-view three-dimensional tooth point cloud data as raw point cloud data, preprocess the raw point cloud data to obtain preprocessed point cloud data; perform adaptive embedding feature mapping on the preprocessed point cloud data to obtain mapped multi-view point cloud feature data; Using existing scanning technologies (such as structured light, lidar, stereo vision, depth sensors, etc.), three-dimensional point cloud data of teeth from different perspectives (such as the front, side, top, bottom, and other angles) is obtained as raw point cloud data, and the raw point cloud data is preprocessed to obtain preprocessed point cloud data; the preprocessing process includes data cleaning, denoising, standardization and normalization, etc. The methods used are all technical means well known to those skilled in the art and are not detailed here; Adaptively embed feature mapping is performed on the pre-processed point cloud data to obtain mapped multi-view point cloud feature data to enhance the contrast of the pre-processed point cloud data under different viewpoints in the feature space, thereby achieving more accurate matching. Based on local geometric feature mapping and weighted distance measurement method, the adaptive embedding feature mapping formula is constructed. The specific mathematical expression is: , in, Indicates the A point of view The feature vector after mapping, that is, the multi-view point cloud feature data after mapping, contains the point The geometric information of the point and the viewing angle Local geometric features under ; and Respectively represent Points of view and point The position, that is, the three-dimensional coordinates, point It is A point of view Neighborhood points of It is From each perspective, the midpoint of the preprocessed point cloud data The number of neighborhood points, that is, the number of locally relevant points, is determined based on the Euclidean distance between points and specific needs, and is not limited here; It is in A point of view The mapping weight of A point of view The contribution of feature mapping is determined according to expert experience, and the reference value range is ; It is The weight factor of each perspective is used to adjust the contribution of different perspectives to the overall point cloud features, reflecting the The importance of each perspective in the final feature map is determined according to the specific application scenario, and the reference value range is ; It is the power exponent that controls the distance metric and determines the degree of influence of the local point cloud on the feature map. It is determined according to the specific application requirements. The reference value range is ; Indicates the A point of view and point The distance metric between them plays a role in measuring local geometric relationships in feature mapping.

[0022] 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 preprocessed point cloud data of different viewpoints to obtain a similarity measurement; 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 point cloud data after preprocessing from different viewpoints to obtain the similarity measurement. The key to the similarity measurement is to measure the distance between the mapped multi-view point cloud feature data and introduce angle and gradient information to optimize the matching process to ensure accurate alignment of similar point clouds. The formula for the similarity measurement is: , in, Represents the preprocessed point cloud data under any two perspectives and The similarity metric represents the matching accuracy between the pre-processed point cloud data under two perspectives. The smaller the similarity metric, the higher the matching degree of the pre-processed point cloud data in space. and Respectively expressed in and Preprocessed point cloud data from different perspectives; and Respectively represent and The number of all points in the preprocessed point cloud data under each perspective; It is Points of view location; It is A point of view The mapped multi-view point cloud feature data vector; It is a constant to prevent division by zero errors and is used to ensure numerical stability when calculating similarity metrics. It can be ; It is the weight coefficient of the angle difference, which is used to control the influence of the angle difference on the point cloud matching similarity calculation. It is determined according to the application scenario, and the reference value range is ; Represents points under any two viewing angles and point The angle difference between the two points is calculated based on the normal vector of the point, reflecting the difference caused by the rotation of the point cloud. The calculation method of the angle difference is a technical means well known to those skilled in the art and will not be described in detail here; It is the weight coefficient of the gradient difference, which is used to control the influence of the local geometric difference (such as curvature) of the point cloud. It is determined according to specific needs. The reference value range is ; and Respectively expressed in and A point of view and point The gradient of , that is, the curvature in the local neighborhood, reflects the difference in the local surface geometry of the point cloud. The local neighborhood is determined based on the Euclidean distance between points and specific needs, and is not limited here; The square of the Euclidean distance of 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 score item used to measure whether two perspectives correspond in the feature space. It is normalized so that the feature similarity score item is not affected by the feature magnitude and is one of the core parts of the similarity metric. It is an angle difference penalty term, which is used to describe the compatibility of directional features between point clouds. The larger the angle difference, the stronger the penalty, so as to force the point cloud alignment process to maintain local directional consistency. It is a gradient difference penalty term, which is used to measure the similarity of the local geometric forms of the two viewpoint point clouds, and can effectively enhance the alignment ability of surface change details; In the process of calculating similarity, by introducing angle and gradient information, the mismatch caused by perspective differences can be effectively reduced.

[0023] S3. Introduce the transformation matrix and optimize it in combination with the similarity metric to obtain the optimal transformation matrix; transform the preprocessed point cloud data based on the optimal transformation matrix to achieve automatic alignment.

[0024] After obtaining the similarity metric, global optimization is used to further improve the registration accuracy. To eliminate the problems caused by local mismatches, a graph-based optimization algorithm is used to globally adjust the transformation of the pre-processed point cloud data at each viewpoint to ensure the overall alignment of the point cloud data from multiple views. The graph-based optimization algorithm is implemented through the following global optimization objective function, which is specifically expressed as: , in, Indicates in Preprocessed point cloud data from different perspectives By optimizing the transformation matrix, the pre-processed point cloud data of different perspectives can be aligned to a unified global coordinate system. The transformation matrix includes geometric transformations such as rotation, translation or scale transformation, which is used to describe the transformation of the pre-processed point cloud data from one coordinate system to another. Indicates in The transformed point cloud data from different perspectives; Represents the point cloud mapping in the feature space, which is represented by The feature vector of the transformed point cloud data in the feature space under different viewing angles; It is in The feature vector of the pre-processed point cloud data in the feature space under different viewing angles; It is Preprocessed point cloud data from different perspectives Hedi Preprocessed point cloud data from different perspectives The matching weight between them reflects the similarity or importance between the two point clouds and is calculated based on the Euclidean distance of the point cloud data; Is the smoothing control factor, which is used to indicate the proportion of the error smoothing term in the global optimization objective function. It is determined according to the expert experience method, and the reference value range is ; It is a scale parameter used to ensure that the error in the feature space is appropriately scaled during the geometric transformation process to avoid over-reliance on large errors during the optimization process. It is determined based on expert experience and the reference value range is ; It is the weight coefficient of the similarity metric, which is used to control the influence of the similarity metric on the global optimization objective function. It is determined based on the quality of the preprocessed point cloud data and the accuracy requirements of the registration. The reference value range is ; Indicates the Transformed point cloud data at different viewing angles With the Preprocessed point cloud data from different perspectives Similarity measure of ; It is an error smoothing term, which aims to suppress large error values ​​and prevent some abnormal viewpoint errors from seriously affecting the overall optimization goal during the optimization process. It is a matching similarity measure that constrains the degree of detail alignment between 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 for all viewpoints, ensuring the joint registration of multi-view point clouds; and based on the optimal transformation matrix of each viewpoint, the preprocessed point cloud data of the viewpoint is transformed to optimize the relative position of the preprocessed point cloud data at each viewpoint on a global scale, eliminate the influence of mismatched points, and thus achieve more accurate registration.

[0025] In summary, a multi-view point cloud automatic registration method for tooth three-dimensional modeling was completed.

[0026] The order in which the embodiments of the invention are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0027] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0028] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. 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 embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A multi-view point cloud automatic registration method for three-dimensional tooth modeling, characterized by: The following steps are involved: S1. Acquire multi-view three-dimensional point cloud data of teeth as raw point cloud data, and preprocess the raw point cloud data to obtain preprocessed point cloud data; Adaptively embed feature mapping on the preprocessed point cloud data to obtain mapped multi-view point cloud feature data; 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 preprocessed point cloud data of different viewpoints to obtain a similarity measurement; S3. Introduce the transformation matrix and optimize it in combination with the similarity metric to obtain the optimal transformation matrix; transform the preprocessed point cloud data based on the optimal transformation matrix to achieve automatic alignment.

2. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 1, characterized in that: Said S1 specifically includes: In the implementation process of adaptive embedding feature mapping, the distance measurement between a point in the preprocessed point cloud data and its neighborhood points at any perspective is calculated, and the weight factor of the perspective and the mapping weight of the neighborhood points are combined to obtain the mapped multi-view point cloud feature data.

3. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 1, characterized in that: Said S2 specifically includes: The feature space-based similarity measurement method calculates the similarity between point cloud data preprocessed from different perspectives by calculating the distance between the mapped multi-view point cloud feature data and introducing angle and gradient information.

4. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 3, characterized in that: Said S2 specifically includes: In the implementation of the similarity measurement method based on feature space, the weight coefficients of angle difference and gradient difference are introduced to construct angle difference penalty terms and gradient difference penalty terms, and the similarity measurement is calculated by combining the feature similarity scoring terms constructed based on the distance between the mapped multi-view point cloud feature data.

5. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 1, characterized in that: Said S3 specifically includes: The transformation matrix is ​​introduced to transform the preprocessed point cloud data to obtain the transformed point cloud data.

6. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 5, characterized in that: Said S3 specifically includes: Based on the transformed point cloud data, the transformation matrix is ​​optimized through a graph-based optimization algorithm.

7. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 6, characterized in that: Said S3 specifically includes: In the implementation of the graph-based optimization algorithm, the optimal transformation matrix is ​​obtained by constructing and minimizing the global optimization objective function.

8. The multi-view point cloud automatic registration method for three-dimensional tooth modeling according to claim 7, characterized in that: Said S3 specifically includes: The global optimization objective function is constructed based on the transformed point cloud data, combined with similarity measurement, and introduced with error smoothing term and matching similarity measurement term.

Images