Oral cavity three-dimensional scanning and data reconstruction method based on laser diffraction

Abstract

The invention relates to the technical field of three-dimensional modeling, in particular to an oral cavity three-dimensional scanning and data reconstruction method based on laser diffraction, and the method comprises the steps: carrying out the neighbor search of an initial oral cavity point cloud model, and obtaining a curvature ratio change curve of each search radius; determining a target curvature threshold value according to the fluctuation condition of each curvature proportion change curve; performing grading processing by using the target curvature threshold to obtain each curvature grade; obtaining an updated point cloud area of each first curvature level, each point cloud area corresponding to each matched curvature level and a point cloud area of each residual curvature level, analyzing the space size and level of the point cloud areas to determine the size of a thinning voxel, and carrying out thinning processing on the point cloud areas of the corresponding curvature levels. And obtaining a diluted oral cavity point cloud model. According to the method, the key oral tissue characteristics are effectively reserved while the thinning effect of the oral point cloud model is ensured, and data reconstruction is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional modeling, and in particular to a method for oral three-dimensional scanning and data reconstruction based on laser diffraction. Background Art

[0002] In the process of using laser diffraction to perform oral 3D scanning and achieve data reconstruction, in order to reduce data redundancy, improve the efficiency of 3D reconstruction and visualization, reduce the computational burden, and facilitate storage and transmission, it is necessary to thin out the oral point cloud data.

[0003] Traditional thinning methods determine the curvature threshold by the proportion of discrete points under different curvatures, and then perform hierarchical thinning. This method does not take into account the characteristic distribution of actual oral tissue, so that the point cloud model after thinning cannot well represent the characteristics of oral tissue, which may lead to the loss of key features, thereby affecting the accuracy of data reconstruction and clinical application effect. Summary of the Invention

[0004] In order to solve the technical problem of low accuracy of oral data reconstruction after thinning processing mentioned above, the purpose of the present invention is to provide a method for oral 3D scanning and data reconstruction based on laser diffraction. The technical solution adopted is as follows: One embodiment of the present invention provides a method for three-dimensional scanning and data reconstruction of an oral cavity based on laser diffraction, the method comprising the following steps: Performing laser scanning on the oral cavity to obtain an initial oral cavity point cloud model; performing nearest neighbor search on the initial oral cavity point cloud model with several search radii to obtain a curvature ratio change curve for each search radius; Determining a target curvature threshold according to fluctuations of each of the curvature ratio change curves; using the target curvature threshold to perform grading processing on all curvature values ​​corresponding to the target search radius to obtain various curvature levels; Performing consistency analysis on regional features of the point cloud regions before and after curvature expansion corresponding to the selected first curvature levels to update the point cloud regions of the first curvature levels, thereby obtaining updated point cloud regions for each first curvature level; wherein the first curvature level is a curvature level having a curvature value greater than a curvature threshold within a curvature range; Determining a first merging bias value based on the spatial values ​​and distribution range proportions of the selected point cloud areas of the second curvature level to determine whether the second curvature level is updated to a matching curvature level, and if so, obtaining each point cloud area corresponding to the matching curvature level; wherein the second curvature level is a curvature level lower than a preset level; The thinned voxel size is determined according to the spatial size and level of the updated point cloud area of ​​each first curvature level, the point cloud area corresponding to each matching curvature level, and the point cloud area of ​​the remaining curvature levels except the first curvature level and the matching curvature level, so as to perform thinning processing on the point cloud area of ​​the corresponding curvature level and obtain the thinned oral point cloud model.

[0005] Furthermore, the initial oral cavity point cloud model is subjected to a nearest neighbor search with several search radii to obtain a curvature ratio change curve of each search radius, including: For any search radius, performing a nearest neighbor search on the initial oral point cloud model using the search radius to obtain a set of nearest neighbor points for each discrete point; Based on the set of nearest neighbor points, a covariance matrix is ​​constructed using a covariance calculation method; eigenvalues ​​of the covariance matrix are obtained, and based on the eigenvalues, a curvature value of each discrete point within the search radius is obtained; A first proportion of the number of discrete points corresponding to each curvature value in the total number of all discrete points is determined, and a curvature proportion change curve of the search radius is constructed based on each curvature value and the first proportion corresponding to each curvature value.

[0006] Furthermore, determining the target curvature threshold according to the fluctuation of each curvature ratio change curve includes: Determine the minimum and maximum points in each of the curvature ratio change curves, and segment the curvature ratio change curves using the minimum values ​​as segmentation points to obtain initial curvature segments; For any initial curvature segment and any adjacent initial curvature segment, determine a second merging bias value according to the curvature difference between the maximum value points in the two adjacent initial curvature segments; Performing a merging analysis on each initial curvature segment using the second merging bias value to obtain each merged curvature segment; Determine the curvature values ​​corresponding to the two endpoints of each merged curvature segment as suspected curvature thresholds, and obtain each suspected curvature threshold corresponding to each search radius; Determine the selection compliance based on the second proportion of the number of occurrences of each suspected curvature threshold in the total number of all curvature proportion change curves and the first proportion of the number of discrete points corresponding to different search radii in the total number of all discrete points; All the suspected curvature thresholds are screened by the selection conformity to obtain a plurality of target curvature thresholds.

[0007] Furthermore, determining the second merging bias value according to the curvature difference between the maximum value points in two adjacent initial curvature segments includes: Determine the absolute value of the difference in curvature between the maximum points in two adjacent initial curvature segments to obtain a first merging bias factor; Obtaining the absolute value of the difference between the first proportion values ​​corresponding to the curvature maximum points in two adjacent initial curvature segments, and using the ratio of the absolute value of the difference between the first proportion values ​​to the absolute value of the maximum proportion difference as a second merging bias factor; wherein the absolute value of the maximum proportion difference is the maximum value among the absolute values ​​of the differences between all the first proportion values; The first merging bias factor and the second merging bias factor are combined to determine a second merging bias value.

[0008] Furthermore, the target curvature threshold is used to perform grading processing on all curvature values ​​corresponding to the target search radius to obtain various curvature levels, including: Obtain the number of identical values ​​between the suspected curvature threshold and the target curvature threshold corresponding to each search radius, and use the search radius corresponding to the largest of the identical values ​​as the target search radius; The target curvature thresholds are arranged in order, and all curvature values ​​corresponding to the target search radius are graded using the arranged target curvature thresholds to obtain each curvature level.

[0009] Furthermore, the consistency analysis is performed on the regional features of the point cloud regions before and after the curvature expansion corresponding to the selected first curvature levels to update the point cloud regions of the first curvature level to obtain the updated point cloud regions of each first curvature level, including: For any first curvature level, any point cloud area of ​​the first curvature level is obtained as the point cloud area before expansion; any curvature value in the initial oral point cloud model that is adjacent to the minimum curvature value in the curvature range of the first curvature level and is smaller than the minimum curvature value is updated to the curvature range of the first curvature level, and the point cloud area of ​​the updated curvature range is obtained as the point cloud area after expansion; Determining a corresponding point of each boundary point of the point cloud region before expansion in the point cloud region after expansion; determining a consistency index of the point cloud feature range before and after expansion based on the distance between each boundary point of the point cloud region before expansion and the corresponding point; Obtain the consistency index during several curvature expansions, select the optimal secondary curvature expansion based on the difference between the consistency index during each two adjacent expansions, and use the point cloud area corresponding to the optimal secondary curvature expansion as the updated point cloud area of ​​the first curvature level; wherein the two adjacent expansions indicate that the next expansion is based on the previous expansion.

[0010] Furthermore, determining a corresponding point of each boundary point of the point cloud region before expansion in the point cloud region after expansion includes: Determine the center point of the point cloud area before expansion, and then obtain the connection line between each boundary point of the point cloud area before and after expansion and the center point; When the connecting line of any boundary point of the point cloud region before expansion coincides with the connecting line of a certain boundary point of the point cloud region after expansion, the certain boundary point of the point cloud region after expansion is used as the corresponding point of the certain boundary point.

[0011] Furthermore, the determination of the consistency index of the point cloud feature range before and after the expansion based on the distance between each boundary point of the point cloud area before the expansion and its corresponding point includes: Constructing a distance fluctuation curve with the number of each boundary point as the horizontal coordinate and the distance of each boundary point as the vertical coordinate; performing dimensionality reduction segmentation processing on the vertical coordinate data of the distance fluctuation curve to obtain various dimensionality reduction segments; The consistency index of the point cloud feature range before and after expansion is determined based on the ratio of the maximum number of curve data points of all dimensionality reduction segments to the total number of boundary points of the expanded point cloud area, and the distance between the center points of each two adjacent dimensionality reduction segments.

[0012] Furthermore, the determining of the first merging bias value according to the spatial values ​​and distribution range proportions of the selected point cloud regions of the second curvature level includes: For any second curvature level, point cloud areas of other curvature levels that are adjacent to the point cloud area of ​​the second curvature level and have a level greater than the second curvature level are used as comparison point cloud areas; Determining a spatial value of a point cloud region of a second curvature level, and further determining a ratio of the number of discrete points in the point cloud region of the second curvature level to the number of discrete points in each comparison point cloud region; According to the spatial value and each of the ratios, a first combined deviation value of the second curvature level and each of the other curvature levels is determined; wherein the spatial value is negatively correlated with the first combined deviation value, and the ratio is positively correlated with the first combined deviation value.

[0013] Furthermore, determining the thinning voxel size according to the spatial size and level of each updated point cloud area of ​​the first curvature level, each point cloud area corresponding to each matching curvature level, and the point cloud areas of the remaining curvature levels except the first curvature level and the matching curvature level includes: For any undetermined curvature level, determining a ratio of the level of the undetermined curvature level to the maximum level; wherein the undetermined curvature level is the first curvature level, the matching curvature level, or the remaining curvature levels except the first curvature level and the matching curvature level; Calculating the product of the spatial value and the level ratio of the undetermined curvature level, performing negative correlation normalization processing on the product of the spatial value and the level ratio, and obtaining a size correction coefficient of the undetermined curvature level; The initial thinned-out voxel size is corrected using the size correction coefficient to determine the thinned-out voxel size of the undetermined curvature level.

[0014] The present invention has the following beneficial effects: In order to overcome the defects of the existing thinning processing for oral point cloud models, the present invention provides a method for oral three-dimensional scanning and data reconstruction based on laser diffraction. This method determines the curvature threshold by obtaining the curvature ratio change curve of each search radius, which can fully consider the curvature characteristics of the oral point cloud model under different search modes, and to a certain extent improves the numerical accuracy of the curvature threshold and the grading effect when performing curvature grading, so as to facilitate the subsequent thinning processing of the point cloud areas corresponding to the curvature grading results, which is the prerequisite for subsequent graded thinning processing.

[0015] Next, in order to retain the significant regional characteristics of oral tissue as much as possible, the point cloud area with higher curvature level is expanded to obtain updated point cloud area, and the lower curvature level is merged into the surrounding higher curvature level to obtain the point cloud area corresponding to each matching curvature level. This is equivalent to combining the characteristics of point cloud areas of different curvature levels and curvature levels, and adjusting the distribution of point cloud areas of each curvature level initially obtained, so as to achieve more obvious emphasis on some characteristic oral tissue parts.

[0016] Finally, the thinned voxel size is adaptively determined based on the adjusted point cloud region for each curvature level. This thinned voxel size is then used to thin the point cloud region corresponding to the curvature level, resulting in a thinned oral point cloud model. Because the curvature grading of the present invention takes into account the characteristic distribution of actual oral tissue, the thinned oral point cloud model can effectively represent the characteristics of oral tissue, effectively avoiding the loss of key features, and improving the accuracy of data reconstruction and the effectiveness of clinical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. 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 paying any creative work.

[0018] Figure 1A flowchart of a method for three-dimensional oral cavity scanning and data reconstruction based on laser diffraction provided in accordance with one embodiment of the present invention; Figure 2 This is a flowchart of step S21 in an embodiment of the present invention; Figure 3 This is a flowchart of step S3 in an embodiment of the present invention; Figure 4 This is a flow chart of the steps for determining a first merging bias value in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] To further illustrate the technical means and effects employed by the present invention to achieve its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementations, structures, features, and effects of the technical solutions proposed by the present invention. In the following description, references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0020] 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.

[0021] The application scenarios targeted by the present invention may be: Point cloud thinning of oral point cloud models can reduce data redundancy and improve the efficiency of 3D reconstruction and visualization, making it suitable for clinical applications such as orthodontic planning and implant surgery simulation. When thinning, care must be taken to maintain the accuracy of key anatomical structures, such as crown margins and occlusal surfaces, to avoid oversimplification that could affect subsequent processing based on the oral 3D model. Furthermore, it is important to ensure that the thinned oral point cloud model still meets the accuracy requirements of subsequent processing, and to avoid introducing jagged edges or distortion.

[0022] In order to overcome the above-mentioned shortcomings of the existing oral point cloud model after thinning, an embodiment of the present invention provides an oral three-dimensional scanning and data reconstruction method based on laser diffraction, such as Figure 1 As shown, the following steps are included: S1, laser scanning the oral cavity to obtain an initial oral cavity point cloud model; performing a nearest neighbor search on the initial oral cavity point cloud model with several search radii to obtain a curvature ratio change curve for each search radius.

[0023] The curvature distribution of the initial oral point cloud model neighbor search performed with different search radii is different. In order to facilitate the analysis of each curvature distribution to determine the target curvature threshold, it is necessary to obtain the curvature ratio change curve of each search radius. The curvature ratio change curve is the data support for determining the target curvature threshold; determining the target curvature threshold through the curvature ratio change curve of different search radii helps to overcome the defect of low accuracy of the traditional method of determining the curvature threshold by only the ratio of discrete points under different curvatures under a single search radius, thereby improving the accuracy of curvature grading.

[0024] The above step S1 can be implemented through steps S11 to S12 (not shown): S11, performing laser scanning on the oral cavity to obtain an initial oral cavity point cloud model.

[0025] Before laser scanning, the oral cavity is pre-treated. Specifically, the edges of the preparation are exposed using a gingival retraction cord, and the teeth are cleaned and dried to prevent the laser reflection from being disturbed by liquid.

[0026] Specifically, during the scanning process, a Z-shaped scanning path was used for the posterior teeth, sequentially covering the occlusal, labial, buccal, and lingual / palatal surfaces. Simultaneously, maxillary and mandibular displacement data were acquired. Using a point cloud stitching algorithm, partially overlapping scan data were registered. Translation vectors and rotation matrices were used to spatially align multi-angle data, resulting in a rough point cloud model of the oral cavity, recorded as the initial oral point cloud model.

[0027] It should be noted that the scanning head needs to be preheated in advance to prevent the generation of fog in the hot and humid environment of the oral cavity that affects the imaging; follow the standard process of working dentition first, then the opposing dentition, and finally the occlusal relationship.

[0028] S12, performing a nearest neighbor search on the initial oral cavity point cloud model with several search radii, and obtaining a curvature ratio change curve of each search radius.

[0029] For any search radius, the steps of obtaining the curvature ratio change curve of the search radius include: S121, performing a nearest neighbor search on the initial oral point cloud model using the search radius to obtain a set of nearest neighbor points for each discrete point.

[0030] In point cloud processing, the radius r of the K-nearest-neighbor search is used to define a spherical range to find the nearest neighbor within a certain radius from a given point. The detailed implementation of the nearest-neighbor search is prior art and falls outside the scope of this invention, so it will not be elaborated here.

[0031] S122, based on the nearest neighbor point set, construct a covariance matrix using a covariance calculation method; obtain the eigenvalues ​​of the covariance matrix, and obtain the curvature value of each discrete point within the search radius based on the eigenvalues.

[0032] Each discrete point has a corresponding covariance matrix. By extracting the eigenvalues ​​of the covariance matrix, the curvature value of each discrete point can be determined. Each discrete point has a corresponding curvature value, and different discrete points may have the same curvature value. The curvature value calculation process is prior art and is beyond the scope of protection of this invention, so it will not be elaborated here.

[0033] S123, determining a first proportion of the number of discrete points corresponding to each curvature value in the total number of all discrete points, and constructing a curvature proportion change curve of the search radius based on each curvature value and the first proportion corresponding to each curvature value.

[0034] Here, the curvature ratio variation curve shows how the number of discrete points corresponding to a given curvature value changes as the curvature value increases. When determining the curvature threshold for an oral point cloud model, using the curvature ratio variation curve of the search radius as basic data analysis can intuitively reflect the correlation between the statistical laws of curvature distribution and anatomical features, thus guiding the selection of the curvature threshold.

[0035] In this embodiment, the curvature ratio variation curve is obtained by fitting the curvature ratio variation curve using the curvature values ​​of different numerical values ​​as the horizontal axis and the corresponding discrete point ratio value of each curvature value as the vertical axis (i.e., the first ratio value) through least squares method. By analyzing the curvature ratio variation curve, the curvature distribution variation of different points in the oral point cloud model can be analyzed. The curvature distribution variation can be used to analyze the curvature distribution characteristics and determine the most appropriate curvature threshold.

[0036] It should be noted that the curvatures of the horizontal axis of the curvature ratio change curve are arranged in a certain order, such as from small to large. The neighboring search results of each search radius are analyzed and executed according to the above steps S121 to S123 to obtain the curvature ratio curve of each search radius.

[0037] So far, this embodiment has obtained the curvature ratio change curve of each search radius.

[0038] S2, determining a target curvature threshold according to the fluctuation of each curvature ratio change curve; using the target curvature threshold to perform graded processing on all curvature values ​​corresponding to the target search radius to obtain each curvature level.

[0039] The purpose of grouping curvature features with different curvature ranges can be achieved by determining a curvature threshold. When obtaining the curvature threshold based on the proportion of curvature in different search radii, different results will be obtained. In order to avoid not considering the impact of different search methods on the obtained curvature when determining the threshold for a single search radius, the fluctuation analysis is performed based on the curvature proportion change curve of different search radii to determine the appropriate curvature threshold and classify the curvature levels of all curvatures in the initial point cloud model.

[0040] The above step S2 can be implemented through steps S21 to S22 (not shown): S21: Determine a target curvature threshold according to fluctuations of each curvature ratio change curve.

[0041] Here, there can be multiple target curvature thresholds. The target curvature threshold is a curvature threshold determined by analyzing the curvature search results of multiple different search radii. It can comprehensively consider the impact of different search methods on the obtained curvature, help improve the numerical accuracy of the search threshold, and facilitate subsequent more reliable grading of the curvature.

[0042] Step S21 above is passed Figure 2 The steps S211 to S216 shown implement: S211, determining the minimum and maximum points in each curvature ratio change curve, segmenting the curvature ratio change curve with the minimum value as the segmentation point, and obtaining each initial curvature segment.

[0043] Here, the maximum point in the curvature ratio change curve corresponds to the high-frequency concentration area of ​​the curvature distribution, such as the peak of the curvature distribution, which may represent the concentrated appearance of a significant feature, such as the edge of the tooth or the pit and fissure; the minimum value in the curvature ratio change curve is located at the bottom of the valley between the two maximum points, which can reflect the transition zone or category decomposition of the curvature distribution, that is, the natural separation point between different curvature features.

[0044] In this embodiment, the process of determining the maximum and minimum points in the curvature ratio change curve is a prior art and is not within the scope of protection of the present invention, and will not be elaborated on here. Each curvature ratio change curve has a corresponding number of initial curvature segments.

[0045] It should be noted that using minimum points as segmentation points results in a low proportion of both types of curvature features. This maximizes intra-segment consistency, meaning that curvature features within the same segment are similar, helping to ensure low heterogeneity within each curvature segment. Compared to maximum points, the curvature distribution in the transition zone changes more gently, resulting in more stable curvature segments.

[0046] S212: For any initial curvature segment and any adjacent initial curvature segment, determine a second merging bias value according to the curvature difference between the maximum value points in the two adjacent initial curvature segments.

[0047] Here, the second merging bias value refers to the possibility of performing a merging operation on two adjacent curvature segments, and the second merging bias value can be determined by analyzing the similarity of curvature distribution characteristics of two adjacent initial curvature segments.

[0048] As an exemplary embodiment, the step of determining the second merging bias value includes: The first step is to determine the absolute value of the difference in curvature between the maximum points in two adjacent initial curvature segments to obtain the first merging bias factor.

[0049] In this embodiment, the smaller the difference in the numerical values ​​of the curvature values ​​corresponding to the maximum values ​​in two adjacent initial curvature segments, the more similar the curvature sizes corresponding to the significant curvature features of the two initial curvature segments are, and the greater the possibility that the two adjacent initial curvature segments will be merged.

[0050] The second step is to obtain the absolute value of the difference between the first proportion values ​​corresponding to the curvature of the maximum points in two adjacent initial curvature segments, and use the ratio of the absolute value of the difference between the first proportion values ​​to the absolute value of the maximum proportion difference as the second merging bias factor.

[0051] In this embodiment, the maximum point corresponds to the high-frequency concentration area of ​​the curvature distribution, which can represent the concentrated appearance of a certain significant feature. Therefore, if the curvature distribution ratios of the maximum points of two adjacent initial curvature segments are similar, then the curvature characteristics of the two adjacent initial curvature segments are similar. At this time, the possibility of performing a merging operation is smaller, and it can be used to determine the second merging bias value.

[0052] The third step is to determine the second merging bias value by combining the first merging bias factor and the second merging bias factor.

[0053] In this embodiment, when analyzing the possibility of merging two adjacent initial curvature segments, the differences in curvature characteristics are analyzed from two perspectives, one is the numerical difference between the curvatures, and the other is the difference between the significant curvature distribution proportions. By integrating the two merging factors, the numerical accuracy of the second merged bias value can be effectively improved, which facilitates the subsequent merging operation of the initial curvature segments.

[0054] As an example, the calculation formula for the second combined deflection value corresponding to the jth initial curvature segment and the j+1th initial curvature segment may be: ; Where, represents the second combined deflection value corresponding to the jth initial curvature segment and the j+1th initial curvature segment, represents the absolute value of the difference in curvature between the maximum points in the jth and j+1th initial curvature segments, represents the first merged deflection factor corresponding to the jth and j+1th initial curvature segments, It represents the absolute value of the difference between the first proportion values ​​of the curvature corresponding to the maximum point in the j-th and j+1-th initial curvature segments, Represents the maximum absolute value of the difference between all first proportion values, Represents the second merge bias factor corresponding to the j-th and j+1-th initial curvature segments.

[0055] In the calculation formula of the second merge bias value, The larger the value is, the worse the curvature similarity between the maximum points in the jth and j+1th initial curvature segments is, and the smaller the possibility of merging the jth and j+1th initial curvature segments is. As the first merging bias factor, under normal circumstances There is no possibility of being zero. If there is an extreme case, a non-zero constant is added to the denominator of the first combined bias factor, and the empirical value is 0.01; It represents the ratio of the difference in the curvature distribution ratio between the maximum points in the j-th and j+1-th initial curvature segments to the difference in the maximum curvature distribution ratio. The larger the value is, the larger the second merging bias value is.

[0056] S213: Perform a merging analysis on each initial curvature segment using the second merging deviation value to obtain each merged curvature segment.

[0057] Here, in order to reduce the amount of analysis data for the target curvature threshold and improve the efficiency of determining the target curvature threshold, a merge analysis is performed using the second merge bias value determined in step S212, and a merge operation is performed on the initial curvature segments to obtain merged curvature segments.

[0058] In this embodiment, in order to facilitate comparative analysis and determine whether to perform merging processing, the second merged bias value can be normalized, for example, the second merged bias value can be normalized using a sigmoid function to obtain a normalized second merged bias value.

[0059] As an exemplary embodiment, the step of obtaining the merged curvature segments includes: For each curvature ratio change curve, the normalized second merged bias values ​​corresponding to the first and second initial curvature segments in the curve can be calculated in sequence. If the normalized second merged bias value is greater than the merged bias threshold, which can be taken as an empirical value of 0.7, the first and second initial curvature segments are merged to obtain a merged curvature segment. Then calculate the normalized second merged deviation value corresponding to the first merged curve segment and the third initial curvature segment. If the normalized second merged deviation value is still greater than the merge deviation threshold, merge the first merged curve segment and the third initial curvature segment to obtain a new merged curve segment. If the normalized second merged deviation value is not greater than the merge deviation threshold, use the first merged curve segment as the final merged curve segment. Then calculate the second merged deviation values ​​corresponding to the third and fourth initial curvature segments, and then perform a merge judgment operation. In this way, by performing the merging judgment operation in sequence according to the distribution order of each initial curvature segment in the curvature, each merged curvature segment corresponding to each curvature proportion change curve can be obtained.

[0060] S214 , determining the curvature values ​​corresponding to the two endpoints of each merged curvature segment as suspected curvature thresholds, and obtaining each suspected curvature threshold corresponding to each search radius.

[0061] In this embodiment, the curvature threshold is used to implement curvature range division. Therefore, the curvature values ​​at the two endpoints of the merged curvature segment can be used as the suspected curvature threshold. This can generate several suspected curvature thresholds corresponding to the curvature ratio change curve for each search radius. Each suspected curvature threshold may become the final target curvature threshold for dividing the curvature range.

[0062] S215 , determining the selection compliance according to the second proportion of the number of occurrences of each suspected curvature threshold in the total number of all curvature proportion change curves and the first proportion of the number of discrete points corresponding to different search radii in the total number of all discrete points.

[0063] Here, each suspected curvature threshold has a different numerical value. Each suspected curvature threshold satisfies the target curvature threshold selection requirement to a different degree. In order to select the most appropriate curvature threshold from all suspected curvature thresholds, i.e., the target curvature threshold, it is necessary to determine the degree of compliance of each suspected curvature threshold.

[0064] In this embodiment, the more times a suspected curvature threshold appears in different curvature ratio change curves, the more reliable the suspected curvature threshold is, and the higher the confidence level of the selection. The smaller the proportion of discrete points corresponding to a suspected curvature threshold relative to the total number of discrete points in the corresponding search radius, the more consistent the suspected curvature threshold is with the minimum data characteristics, the greater its likelihood of serving as a segmentation point, and the higher the selection compliance. Therefore, the second ratio value is positively correlated with the selection compliance, while the first ratio value is negatively correlated with the selection compliance.

[0065] As an example, the calculation formula for the selection compliance of the mth suspected curvature threshold can be: ; Where, Indicates the selection compliance of the mth suspected curvature threshold, represents the number of occurrences of the mth suspected curvature threshold, that is, the number of curvature ratio change curves where the mth suspected curvature threshold appears, and N represents the total number of all curvature ratio change curves. Indicates the second proportion value corresponding to the mth suspected curvature threshold, It represents the first proportion value corresponding to when the m-th suspected curvature threshold is on the curvature proportion change curve of the k-th suspected curvature threshold.

[0066] According to the calculation process of the selection conformity of the m-th suspected curvature threshold, the selection conformity of each suspected curvature threshold can be obtained.

[0067] S216 , screening all suspected curvature thresholds by selecting the degree of conformity to obtain several target curvature thresholds.

[0068] In this embodiment, a selection compliance threshold is set, and an empirical value of 0.7 may be used. To compare and analyze with the selection compliance threshold, the selection compliance of each suspected curvature threshold is normalized, such as by using a sigmoid function to normalize the selection compliance, to obtain a normalized selection compliance. After obtaining the selection compliance threshold and the normalized selection compliance, the normalized selection compliance is compared with the selection compliance threshold. The suspected curvature threshold corresponding to the selection compliance greater than the selection compliance threshold is used as the final target curvature threshold, thereby obtaining a number of target curvature thresholds.

[0069] It should be noted that in extreme cases where the number of target curvature thresholds is less than 2, in order to achieve a more accurate classification of curvature values, the value of the selected matching threshold can be adjusted, such as adjusting the selected matching threshold from 0.7 to 0.6, and performing the target curvature threshold screening operation again. Of course, implementers can set or adjust the selected matching threshold based on specific circumstances, and this is not specifically limited here.

[0070] S22 , performing classification processing on all curvature values ​​corresponding to the target search radius using the target curvature threshold to obtain a curvature range of each curvature level.

[0071] Here, the target search radius is the search radius with the largest number of suspected curvature thresholds that are identical to the target curvature threshold. The greater the number of identical threshold values, the better the search effect for the corresponding search radius. Of course, the target search radius can also be any search radius.

[0072] The above step S22 can be implemented through steps S221 to S222 (not shown): S221 , obtaining the number of identical values ​​between the suspected curvature threshold and the target curvature threshold corresponding to each search radius, and taking the search radius corresponding to the largest identical number as the target search radius.

[0073] S222 , arranging the target curvature thresholds in order, and performing a grading operation on all curvature values ​​corresponding to the target search radius using the arranged target curvature thresholds to obtain each curvature grade.

[0074] For example, the target curvature thresholds arranged from small to large are a1, a2, and a3. All curvature values ​​corresponding to the target search radius that are less than or equal to a1 are classified into one level, which is the first curvature level, level 1. Then, all curvature values ​​corresponding to the target search radius that are greater than a1 and less than or equal to a2 are classified into one level, which is the second curvature level, level 2. Finally, all curvature values ​​corresponding to the target search radius that are greater than a3 are classified into one level, which is the third curvature level, level 3. The larger the curvature, the higher the corresponding level. The third curvature level is higher than the second curvature level, and the second curvature level is higher than the first curvature level.

[0075] Thus, this embodiment obtains a more accurate division of each curvature level.

[0076] S3, performing consistency analysis on regional features of the point cloud regions before and after the curvature expansion corresponding to the selected plurality of first curvature levels to update the point cloud regions of the first curvature level and obtain updated point cloud regions of each first curvature level.

[0077] Each curvature level obtained in step S2 can be used to divide the initial oral point cloud model into point cloud areas, but the divided point cloud areas at this time may not be able to well represent the characteristics of the location of the oral tissue. For example, for the boundary range between the soft palate and the hard palate, when the curvature range is small, only the boundary position of the soft palate and the hard palate is obtained, and the divided point cloud area cannot well reflect the connection relationship between the soft palate and the hard palate. Therefore, for the point cloud area corresponding to the curvature level with a larger curvature, the regional characteristics of the corresponding point cloud area can be analyzed by adding or subtracting the changes in the point cloud area represented by the curvature within the curvature range.

[0078] In this embodiment, the curvature range of the first curvature level is the curvature range with the largest curvature. In this embodiment, the curvature level within the curvature range that includes a curvature value greater than a curvature threshold is selected as the first curvature level. The curvature threshold can be set to 0.6, so the curvature level that includes a curvature value greater than the curvature threshold of 0.6 can be selected from all curvature levels as the first curvature level. In addition, in this embodiment, all curvatures are defaulted to normalized curvatures, that is, the curvature values ​​range from 0 to 1.

[0079] It should be noted that not updating the point cloud area of ​​each curvature level can improve the efficiency of thinning processing of the oral point cloud model while ensuring the thinning effect.

[0080] The above step S3 can be Figure 3 The steps S31 to S33 shown are implemented as follows: S31. For any first curvature level, obtain any point cloud area of ​​the first curvature level as the point cloud area before expansion; update any curvature value in the initial oral point cloud model that is adjacent to the minimum curvature value in the curvature range of the first curvature level and is smaller than the minimum curvature value to the curvature range of the first curvature level, and obtain the point cloud area of ​​the updated curvature range as the point cloud area after expansion.

[0081] Here, we obtain the curvature value in the initial oral point cloud model that is smaller than the minimum curvature value corresponding to the first curvature level, i.e., the smaller curvature value. This smaller curvature value is added to the curvature range of the first curvature level to expand the curvature characteristics of the current curvature range. With the expanded curvature range, the corresponding initial point cloud region can be updated to obtain a new point cloud region. Each time a new curvature value is added, the curvature range of the first curvature level is expanded.

[0082] In this embodiment, each curvature level may correspond to multiple different point cloud areas. For the sake of convenience of description, taking any point cloud area corresponding to a single curvature level as an example, any point cloud area corresponding to any first curvature level is analyzed to complete the point cloud area update.

[0083] First, obtain the point cloud area before and after a single expansion corresponding to any first curvature level. This allows for subsequent analysis of the consistency of the performance characteristics of the point cloud area before and after expansion to determine whether the point cloud area before and after expansion is within the same feature range. Specifically, the edge contours of the point cloud area before and after expansion are analyzed to verify whether the expanded point cloud area and the point cloud area before and after expansion belong to the same geometric feature. The stronger the consistency of the regional features of the point cloud area before and after expansion, the more likely it is that the regional features before and after expansion belong to the same surface feature.

[0084] It should be noted that after completing a single curvature expansion, if expansion processing is required later, the next curvature expansion needs to be performed again based on the curvature range after the previous expansion.

[0085] S32, determining the corresponding point of each boundary point of the point cloud area before expansion in the point cloud area after expansion; and determining the consistency index of the point cloud feature range before and after expansion based on the distance between each boundary point of the point cloud area before expansion and its corresponding point.

[0086] Here, the corresponding point refers to the point located on the boundary of the point cloud area after expansion in the direction of the line connecting the boundary point and the center point of the point cloud area before expansion; the consistency index of the point cloud feature range indicates the consistency of the distance change between the boundary point and the corresponding point between the point cloud area before and after expansion.

[0087] S321 , determining a corresponding point of each boundary point of the point cloud region before expansion in the point cloud region after expansion.

[0088] The first step is to determine the center point of the point cloud area before expansion, and then obtain the connection line between each boundary point and the center point of the point cloud area before and after expansion.

[0089] In this embodiment, the center point of the point cloud region before expansion is calculated using the centroid formula. The method for determining the center point is prior art and is not protected by the present invention, so it will not be elaborated here.

[0090] In the second step, when the line connecting any boundary point of the point cloud area before expansion coincides with the line connecting a boundary point of the point cloud area after expansion, the boundary point of the point cloud area after expansion is used as the corresponding point of any boundary point.

[0091] In this embodiment, a directed line connecting the boundary points of the point cloud region before expansion and its center point is determined as a first line, and a directed line connecting the boundary points of the point cloud region after expansion and the center point of the point cloud region before expansion is determined as a second line. When the first line and the second line overlap, the boundary points of the first line and the boundary points of the second line are considered corresponding points. To facilitate subsequent analysis, the boundary points of the second line can be directly used as corresponding points of the boundary points of the first line, thereby obtaining corresponding points for each boundary point of the point cloud region before expansion.

[0092] Generally speaking, the boundary of the point cloud area after curvature expansion will be larger than the boundary before curvature expansion, so each boundary point of the point cloud area before expansion has its corresponding corresponding point. In extreme cases, some boundary points of the point cloud area before expansion have no corresponding points. At this time, these boundary points will not participate in the subsequent calculation of consistency indicators.

[0093] S322 , determining a consistency index of the point cloud feature range before and after the expansion based on the distance between each boundary point of the point cloud area before the expansion and its corresponding point.

[0094] In this embodiment, when the distance change between each boundary point and the corresponding point of the point cloud area before expansion is more consistent during a single expansion, the possibility that the point cloud area before and after expansion is the same feature range is greater. Therefore, the consistency characteristics of the point cloud feature range before and after expansion can be quantitatively analyzed by the distance between each boundary point and its corresponding point.

[0095] The first step is to construct a distance fluctuation curve with the number of each boundary point as the horizontal coordinate and the distance of each boundary point as the vertical coordinate.

[0096] In this embodiment, the setting of the boundary point numbers is based on the distribution of the boundary points, and the setting results of the boundary point numbers are not unique; the distance of each boundary point refers to the distance between the boundary point of the expanded point cloud area and its corresponding point. Under the premise that the coordinate positions of the two points are known, the distance between the two points can be directly calculated. This is an existing implementation process and will not be elaborated here.

[0097] The second step is to perform dimension reduction segmentation processing on the vertical coordinate data of the distance fluctuation curve to obtain various dimension reduction segments.

[0098] In this embodiment, the ordinate data (distance data) of the distance fluctuation curve is subjected to dimensionality reduction processing using piecewise aggregate approximation (PAA) to obtain reduced-dimensionality segments. Piecewise aggregate approximation is a prior art technique and is outside the scope of this invention, so it will not be elaborated on in detail here.

[0099] It should be noted that the consistency index based on dimensionality reduction segmentation can seek the optimal balance between information compression and feature retention. Compared with other existing distance change similarity quantification methods, it suppresses high-frequency noise through segmented smoothing, which can greatly reduce the computational cost.

[0100] In the third step, the consistency index of the point cloud feature range before and after expansion is determined based on the ratio of the maximum number of curve data points in all dimensionality reduction segments to the total number of boundary points in the expanded point cloud area, as well as the distance between the center points of each two adjacent dimensionality reduction segments.

[0101] The consistency index of the point cloud feature range before and after expansion is determined by judging the proportion of the number of relatively consistent lengths. It needs to consider two factors: the similarity of the distance between the center points of each two adjacent dimensionality reduction segments and the proportion of the maximum number of curve data points of the dimensionality reduction segment to the total number of boundary points of the expanded point cloud area. The more similar the distance and the larger the number of curve data points of the dimensionality reduction segment, the greater the consistency index of the point cloud feature range before and after expansion.

[0102] As an example, the calculation formula for the consistency index of the point cloud feature range before and after the g-th expansion can be: ; Where, Represents the consistency index of the point cloud feature range before and after the g-th expansion, represents the number of curve data points of the vth dimensionality reduction segment, It represents the total number of boundary points in the expanded point cloud area before and after the g-th expansion, that is, the number of boundary points corresponding to the distance fluctuation curve. It represents the ratio of the number of curve data points of the vth dimensionality reduction segment to the total number of boundary points of the expanded point cloud area. The number of curve data points of the vth dimensionality reduction segment is the maximum number of curve data points of all dimensionality reduction segments. , It represents the cumulative value of the distance difference between the center points of each two adjacent dimensionality reduction segments before and after the g-th expansion, V represents the number of dimensionality reduction segments, Represents the distance between the center points of the vth and v-1th dimensionality reduction segments.

[0103] In the calculation formula of consistency index, The larger it is, the ratio of the number of curve data points of the vth dimensionality reduction segment to the number of boundary points corresponding to the expanded point cloud area is. The larger it is, the smaller the distance between the center points of adjacent segments is. At this time, the distance change is more consistent, and the point cloud area before and after the g-th expansion is more in the same feature point cloud range.

[0104] By referring to the process of determining the consistency index of the point cloud feature range before and after the g-th expansion, the consistency index of the point cloud feature range before and after each expansion can be obtained.

[0105] It is worth noting that each dimensionality reduction segment can represent a stage of distance change of a type. The numerical accuracy of the consistency index determined by each dimensionality reduction segment is stronger because it can ignore small fluctuations within the segment and focus on the overall trend.

[0106] S33, obtaining consistency indicators of several curvature expansions, selecting the optimal secondary curvature expansion according to the difference between the consistency indicators of every two adjacent expansions, and using the point cloud area corresponding to the optimal secondary curvature expansion as the updated point cloud area of ​​the first curvature level.

[0107] Here, two adjacent expansions mean that the next expansion is based on the previous expansion.

[0108] For any point cloud area of ​​any first curvature level, specifically, according to the method for obtaining the consistency index of the point cloud feature range before and after the g-th expansion, the consistency index of several curvature expansions is obtained. The number of curvature expansions can be set according to the distribution of smaller curvatures around the minimum curvature in the point cloud area of ​​the first curvature level, and is not specifically limited here; then the difference between the consistency indexes of each two adjacent expansions is calculated, that is, the consistency index of the latter expansion minus the consistency index of the previous expansion, and then the maximum consistency index difference is selected as the denominator of the ratio. The difference between the consistency indexes of each two adjacent expansions is used as the numerator of the ratio in the order before and after the expansion, and the comparison value is normalized to obtain the feature range conformity; when the feature range conformity corresponding to two expansions is greater than the preset conformity threshold, the empirical value can be 0.7, then the latter expansion of the two expansions is used as the optimal secondary curvature expansion, and the point cloud area corresponding to the optimal secondary curvature expansion is obtained as the updated point cloud area of ​​the first curvature level, and the updated point cloud area of ​​the first curvature level is obtained.

[0109] With reference to the above-mentioned method for determining the updated point cloud region of the first curvature level, the updated point cloud region of each first curvature level can be obtained.

[0110] There are special cases. When the feature range conformity corresponding to the first and second curvature expansions of a certain first curvature level is less than the preset conformity threshold, the directly corresponding original point cloud area can be used as the updated point cloud area. A certain first curvature level may also have multiple feature range conformities that meet the preset conformity threshold requirements. At this time, the last expansion corresponding to the maximum feature range conformity can be used as the optimal secondary curvature expansion.

[0111] So far, this embodiment has obtained updated point cloud regions of various first curvature levels.

[0112] S4, determining the first merge bias value based on the spatial values ​​and distribution range proportions of several selected point cloud areas of the second curvature level to determine whether the second curvature level is updated to the matching curvature level. If updated, obtaining each point cloud area corresponding to the matching curvature level.

[0113] Here, the second curvature level is a curvature level lower than the preset level, and the preset level may be set to 3. For each curvature level, an example is given as follows: currently there are 10 curvature levels, the maximum curvature level is 10, and the levels are 9, 8, 7, 6, 5, 4, 3, 2, and 1 in order.

[0114] In the oral point cloud model, there is a situation where an area with smaller curvature is surrounded by multiple areas with larger curvature. When the spatial value of the area with smaller curvature is small, in order to emphasize the surrounding areas with larger curvature, the feature space range with larger curvature can be larger and more obvious when the point cloud is thinned. After the feature space range is thinned, the greater the degree of retention of the feature space range features in the obtained point cloud area, the curvature level merging operation can be performed on the curvature level with smaller curvature to obtain each matching curvature level.

[0115] The first merged deviation value is determined based on the spatial values ​​and distribution range proportions of the selected point cloud areas of the second curvature level, which can be obtained by Figure 4 The steps S41 to S43 shown implement: S41 : For any second curvature level, point cloud areas of other curvature levels that are adjacent to the point cloud area of ​​the second curvature level and have a level greater than the second curvature level are used as comparison point cloud areas.

[0116] S42 , determining the spatial value of the point cloud region of the second curvature level, and further determining the ratio of the number of discrete points in the point cloud region of the second curvature level to the number of discrete points in each comparison point cloud region.

[0117] S43: Determine a first combined deflection value of the second curvature level and each other curvature level according to the spatial value and each ratio.

[0118] In this embodiment, when the spatial value of the point cloud area of ​​the second curvature level with a smaller level is smaller and the area range ratio of the point cloud area and the comparison point cloud area is larger, the point cloud area of ​​the second curvature level should be merged into the point cloud area of ​​the curvature level of the comparison point cloud area.

[0119] As an example, the calculation formula for the first combined deflection value of the f-th second curvature level and the s-th other curvature level may be: ; Where, represents the first combined bias value of the fth second curvature level and the sth other curvature level, The ratio of the number of discrete points in the f-th second curvature level point cloud area to the number of discrete points in the s-th comparison point cloud area, It represents the spatial value of the point cloud area of ​​the second curvature level f. The calculation of the spatial size of the point cloud area is an existing technology and will not be described in detail. In general, there is no possibility that the spatial value is zero. If there is an extreme case, Add a non-zero constant to the denominator of , and take the empirical value as 0.01.

[0120] After obtaining the first combined deflection value of the second curvature level and each other curvature level, it is determined whether the second curvature level is updated to the matching curvature level. If updated, each point cloud area corresponding to the matching curvature level is obtained, including: If the first merging deviation value of a certain other curvature level is greater than the merging threshold and is the maximum, then the other curvature level is used as the matching curvature level of the second curvature level, and the second curvature level is updated to the matching curvature level, and each point cloud area corresponding to the matching curvature level is obtained; if the first merging deviation values ​​of all other curvature levels are not greater than the merging threshold, then the second curvature level is not updated, and the original point cloud area corresponding to the second curvature level can be kept unchanged, and the implementation process of step S5 can be continued.

[0121] The empirical value of the merge threshold is 0.7. Implementers can set it according to the specific curvature level merging requirements without specific restrictions. In addition to the point cloud area corresponding to the matching curvature level itself, there are also point cloud areas corresponding to the second curvature level. Therefore, the matching curvature level corresponds to multiple point cloud areas. Of course, there is also a case where the matching curvature level is the first curvature level, that is, the curvature range corresponding to the matching curvature level contains a curvature value greater than the curvature threshold of 0.6. In this case, in addition to the updated point cloud area corresponding to the matching curvature level itself, the matching curvature level also includes the point cloud areas corresponding to the second curvature level. These are collectively referred to as the point cloud areas corresponding to the matching curvature level.

[0122] Among them, for the setting of all threshold values ​​in this embodiment, the implementer can set them according to the specific actual situation and no specific limitation is made here.

[0123] By referring to the above process of determining whether the f-th second curvature level is updated to the matching curvature level, the point cloud regions corresponding to all matching curvature levels can be obtained.

[0124] So far, this embodiment has obtained each point cloud area corresponding to each matching curvature level.

[0125] S5. Determine the thinned voxel size based on the spatial size and level of the updated point cloud area of ​​each first curvature level, the point cloud areas corresponding to each matching curvature level, and the point cloud areas of the remaining curvature levels except the first curvature level and the matching curvature level, so as to perform thinning processing on the point cloud areas of the corresponding curvature levels and obtain the thinned oral point cloud model.

[0126] It should be noted that different tissues or different locations of tissues in the oral cavity may have different curvature changes. In order to facilitate the judgment of the patient's oral condition based on the oral point cloud model and reduce the storage size of the oral point cloud model, the point cloud model should be thinned while retaining the features of each part of the oral point cloud model to obtain a more clear-cut, thinned oral point cloud model. In order to better retain the features of each part of the oral point cloud model, different thinning methods should be used for different curvature feature expression positions, that is, the thinning voxel size is adaptively determined for each curvature range, and the point cloud areas with different curvature feature expressions are thinned by thinning voxel size.

[0127] The above-mentioned determination of the thinned voxel size based on the spatial size and level of the updated point cloud area of ​​each first curvature level, the point cloud areas corresponding to each matching curvature level, and the point cloud areas of the remaining curvature levels other than the first curvature level and the matching curvature level can be achieved through steps S51 to S53 (not shown in the figure): S51: For any undetermined curvature level, determine the ratio of the level of the undetermined curvature level to the maximum level.

[0128] S52, calculating the product of the spatial value and the level ratio of the curvature level to be determined, performing negative correlation normalization processing on the product of the spatial value and the level ratio, and obtaining the size correction coefficient of the curvature level to be determined.

[0129] S53: Correct the initial thinned voxel size using the size correction coefficient to determine the thinned voxel size of the to-be-determined curvature level.

[0130] Here, the undetermined curvature level is the first curvature level, the matching curvature level, or the remaining curvature levels except the first curvature level and the matching curvature level.

[0131] In this embodiment, when the curvature level is higher, the features of the point cloud area corresponding to the curvature level in the oral space are more obvious, and when the proportion of all point cloud areas corresponding to the curvature level in the overall range of the oral cavity is smaller, the features of the point cloud area corresponding to the curvature level are more special, and they should be retained more. Therefore, when thinning the point cloud area corresponding to the curvature level, a smaller thinning degree should be used. The control of the thinning degree mainly depends on the size of the voxel. When the voxel size is larger, the thinning degree of the point cloud area through voxel filtering is greater, and the fewer discrete points in the final thinned point cloud model.

[0132] As an example, the calculation formula for the thinned voxel size of the hth undetermined curvature level can be: ; Where, Indicates the thinned voxel size of the hth undetermined curvature level, Y indicates the initial thinned voxel size, which can be set to 0.5mm. Represents the hyperbolic tangent function, which is used to achieve normalization processing. Represents the total spatial value of all point cloud areas corresponding to the hth undetermined curvature level, Indicates the level of the hth undetermined curvature level. The larger the serial number value of the curvature level, the higher the corresponding level. Indicates the maximum level, Represents the size correction coefficient of the hth undetermined curvature level.

[0133] For each curvature level, the thinning voxel size is determined. Using voxel filtering, the point cloud regions corresponding to the curvature level are thinned using different voxel sizes, ultimately resulting in a thinned oral point cloud model. Point cloud regions with the same curvature level have the same thinning voxel size and the same degree of thinning.

[0134] After obtaining the thinned oral point cloud model, this embodiment further includes: visually displaying the thinned oral point cloud model, and marking point cloud regions with different curvature levels with different colors.

[0135] In this embodiment, the position coordinates of each discrete point in the thinned oral point cloud model and the corresponding curvature level are first stored accordingly, and then the position coordinates of each discrete point are visualized on the computer screen, and the discrete points of different curvature levels are marked with different colors to facilitate the staff's subsequent treatment judgment operations.

[0136] The embodiments described above 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned 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 scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A method for oral 3D scanning and data reconstruction based on laser diffraction, characterized in that: The following steps are involved: Performing laser scanning on the oral cavity to obtain an initial oral cavity point cloud model; performing nearest neighbor search on the initial oral cavity point cloud model with several search radii to obtain a curvature ratio change curve for each search radius; Determining a target curvature threshold according to fluctuations of each of the curvature ratio change curves; Using the target curvature threshold, all curvature values ​​corresponding to the target search radius are graded to obtain various curvature grades; Performing consistency analysis on regional features of the point cloud regions before and after curvature expansion corresponding to the selected first curvature levels to update the point cloud regions of the first curvature levels, thereby obtaining updated point cloud regions for each first curvature level; wherein the first curvature level is a curvature level having a curvature value greater than a curvature threshold within a curvature range; Determining a first merging bias value based on the spatial values ​​and distribution range proportions of the selected point cloud areas of the second curvature level to determine whether the second curvature level is updated to a matching curvature level, and if so, obtaining each point cloud area corresponding to the matching curvature level; wherein the second curvature level is a curvature level lower than a preset level; The thinned voxel size is determined according to the spatial size and level of the updated point cloud area of ​​each first curvature level, the point cloud area corresponding to each matching curvature level, and the point cloud area of ​​the remaining curvature levels except the first curvature level and the matching curvature level, so as to perform thinning processing on the point cloud area of ​​the corresponding curvature level and obtain the thinned oral point cloud model.

2. The method for oral three-dimensional scanning and data reconstruction based on laser diffraction according to claim 1, characterized in that: The initial oral cavity point cloud model is subjected to a neighbor search with several search radii to obtain a curvature ratio change curve for each search radius, including: For any search radius, performing a nearest neighbor search on the initial oral point cloud model using the search radius to obtain a set of nearest neighbor points for each discrete point; Based on the set of nearest neighbor points, a covariance matrix is ​​constructed using a covariance calculation method; eigenvalues ​​of the covariance matrix are obtained, and based on the eigenvalues, a curvature value of each discrete point within the search radius is obtained; A first proportion of the number of discrete points corresponding to each curvature value in the total number of all discrete points is determined, and a curvature proportion change curve of the search radius is constructed based on each curvature value and the first proportion corresponding to each curvature value.

3. The method for oral cavity three-dimensional scanning and data reconstruction based on laser diffraction according to claim 1, characterized in that: Determining the target curvature threshold according to the fluctuation of each curvature ratio change curve includes: Determine the minimum and maximum points in each of the curvature ratio change curves, and segment the curvature ratio change curves using the minimum values ​​as segmentation points to obtain initial curvature segments; For any initial curvature segment and any adjacent initial curvature segment, determine a second merging bias value according to the curvature difference between the maximum value points in the two adjacent initial curvature segments; Performing a merging analysis on each initial curvature segment using the second merging bias value to obtain each merged curvature segment; Determine the curvature values ​​corresponding to the two endpoints of each merged curvature segment as suspected curvature thresholds, and obtain each suspected curvature threshold corresponding to each search radius; Determine the selection compliance based on the second proportion of the number of occurrences of each suspected curvature threshold in the total number of all curvature proportion change curves and the first proportion of the number of discrete points corresponding to different search radii in the total number of all discrete points; All the suspected curvature thresholds are screened by the selection conformity to obtain a plurality of target curvature thresholds.

4. The method for oral 3D scanning and data reconstruction based on laser diffraction according to claim 3, characterized in that: Determining the second merging bias value according to the curvature difference between the maximum value points in two adjacent initial curvature segments includes: Determine the absolute value of the difference in curvature between the maximum points in two adjacent initial curvature segments to obtain a first merging bias factor; Obtaining the absolute value of the difference between the first proportion values ​​corresponding to the curvature maximum points in two adjacent initial curvature segments, and using the ratio of the absolute value of the difference between the first proportion values ​​to the absolute value of the maximum proportion difference as a second merging bias factor; wherein the absolute value of the maximum proportion difference is the maximum value among the absolute values ​​of the differences between all the first proportion values; The first merging bias factor and the second merging bias factor are combined to determine a second merging bias value.

5. The method for oral three-dimensional scanning and data reconstruction based on laser diffraction according to claim 4, characterized in that: The target curvature threshold is used to perform grading processing on all curvature values ​​corresponding to the target search radius to obtain various curvature levels, including: Obtain the number of identical values ​​between the suspected curvature threshold and the target curvature threshold corresponding to each search radius, and use the search radius corresponding to the largest of the identical values ​​as the target search radius; The target curvature thresholds are arranged in order, and all curvature values ​​corresponding to the target search radius are graded using the arranged target curvature thresholds to obtain each curvature level.

6. The method for oral three-dimensional scanning and data reconstruction based on laser diffraction according to claim 1, characterized in that: The consistency analysis of the regional features of the point cloud regions before and after the curvature expansion corresponding to the selected first curvature levels is performed to update the point cloud regions of the first curvature level to obtain an updated point cloud region of each first curvature level, including: For any first curvature level, any point cloud area of ​​the first curvature level is obtained as the point cloud area before expansion; any curvature value in the initial oral point cloud model that is adjacent to the minimum curvature value in the curvature range of the first curvature level and is smaller than the minimum curvature value is updated to the curvature range of the first curvature level, and the point cloud area of ​​the updated curvature range is obtained as the point cloud area after expansion; Determining a corresponding point of each boundary point of the point cloud region before expansion in the point cloud region after expansion; determining a consistency index of the point cloud feature range before and after expansion based on the distance between each boundary point of the point cloud region before expansion and the corresponding point; Obtain the consistency index during several curvature expansions, select the optimal secondary curvature expansion based on the difference between the consistency index during each two adjacent expansions, and use the point cloud area corresponding to the optimal secondary curvature expansion as the updated point cloud area of ​​the first curvature level; wherein the two adjacent expansions indicate that the next expansion is based on the previous expansion.

7. The method for oral cavity three-dimensional scanning and data reconstruction based on laser diffraction according to claim 6, characterized in that: The determining of a corresponding point of each boundary point of the point cloud region before expansion in the point cloud region after expansion includes: Determine the center point of the point cloud area before expansion, and then obtain the connection line between each boundary point of the point cloud area before and after expansion and the center point; When the connecting line of any boundary point of the point cloud region before expansion coincides with the connecting line of a certain boundary point of the point cloud region after expansion, the certain boundary point of the point cloud region after expansion is used as the corresponding point of the certain boundary point.

8. The method for oral cavity three-dimensional scanning and data reconstruction based on laser diffraction according to claim 6, characterized in that: Determining the consistency index of the point cloud feature range before and after the expansion based on the distance between each boundary point of the point cloud area before the expansion and its corresponding point includes: Constructing a distance fluctuation curve with the number of each boundary point as the horizontal coordinate and the distance of each boundary point as the vertical coordinate; performing dimensionality reduction segmentation processing on the vertical coordinate data of the distance fluctuation curve to obtain various dimensionality reduction segments; The consistency index of the point cloud feature range before and after expansion is determined based on the ratio of the maximum number of curve data points of all dimensionality reduction segments to the total number of boundary points of the expanded point cloud area, and the distance between the center points of each two adjacent dimensionality reduction segments.

9. The method for oral cavity three-dimensional scanning and data reconstruction based on laser diffraction according to claim 1, characterized in that: The determining of the first merged bias value according to the spatial values ​​and distribution range proportions of the selected point cloud areas of the second curvature level includes: For any second curvature level, point cloud areas of other curvature levels that are adjacent to the point cloud area of ​​the second curvature level and have a level greater than the second curvature level are used as comparison point cloud areas; Determining a spatial value of a point cloud region of a second curvature level, and further determining a ratio of the number of discrete points in the point cloud region of the second curvature level to the number of discrete points in each comparison point cloud region; According to the spatial value and each of the ratios, a first combined deviation value of the second curvature level and each of the other curvature levels is determined; wherein the spatial value is negatively correlated with the first combined deviation value, and the ratio is positively correlated with the first combined deviation value.

10. The method for oral cavity three-dimensional scanning and data reconstruction based on laser diffraction according to claim 1, characterized in that: The step of determining the thinning voxel size according to the spatial size and level of the updated point cloud area of ​​each first curvature level, the point cloud areas corresponding to each matching curvature level, and the point cloud areas of the remaining curvature levels except the first curvature level and the matching curvature level includes: For any undetermined curvature level, determining a ratio of the level of the undetermined curvature level to the maximum level; wherein the undetermined curvature level is the first curvature level, the matching curvature level, or the remaining curvature levels except the first curvature level and the matching curvature level; Calculating the product of the spatial value and the level ratio of the undetermined curvature level, performing negative correlation normalization processing on the product of the spatial value and the level ratio, and obtaining a size correction coefficient of the undetermined curvature level; The initial thinned-out voxel size is corrected using the size correction coefficient to determine the thinned-out voxel size of the undetermined curvature level.

Images