Method, device and computer program product for segmenting three-dimensional particles

Abstract

The application discloses a three-dimensional particle segmentation method, device and computer program product. The segmentation method comprises: receiving a tomographic image of a three-dimensional particle; generating a particle voxel map based on the image; converting the voxel map into a point cloud represented by an N*3-dimensional matrix, N representing the number of points in the point cloud; marking a connected domain in the point cloud by a multi-neighborhood method to obtain a point cloud block representing a plurality of particles, some of which are adhered to each other and form an adhered particle group, each particle group containing two or more particles; for each particle group, receiving an indication of the number of particles contained in the particle group by a user; and based on the number indicated by the indication, using a clustering function to cluster the particle group, so as to segment the particle group into the number of particles. The application realizes accurate and efficient segmentation by receiving the indication of the number of particles contained in the particle group by the user, and clustering the particle group according to the indicated number using a suitable clustering function.

Description

Technical Field

[0001] This application relates to the field of image processing, specifically to a method, apparatus, and computer program product for segmenting three-dimensional particles. Background Technology

[0002] Graphite particles are a crucial component of the battery anode. Currently, software such as Avizo is often used to analyze the three-dimensional morphology and characteristic parameters of these particles. However, these software programs cannot perform fine-grained particle segmentation, and their particle characteristics are limited. Without particle characteristics, it is impossible to predict the state and impact on graphite electrodes under different processes and different states of health (SOH) through simulation calculations.

[0003] For other three-dimensional particles besides graphite, there is also a technical problem of not being able to perform fine-grained segmentation of the particles. Summary of the Invention

[0004] In view of the above problems, this application provides a method, apparatus and computer program product for segmenting three-dimensional particles, which can efficiently and accurately segment three-dimensional particles.

[0005] In a first aspect, this application provides a method for segmenting three-dimensional particles, characterized by comprising: receiving a tomographic image of three-dimensional particles; generating a voxel map of the three-dimensional particles based on the tomographic image; converting the voxel map into a point cloud represented by an N×3 dimensional matrix, where N represents the number of points in the point cloud; labeling connected components in the point cloud using a multi-neighborhood method to obtain a point cloud block representing multiple particles, some of the multiple particles being adhered to each other to form one or more adhered particle groups, each adhered particle group containing two or more adhered particles; for each adhered particle group, receiving an indication from a user of the number of particles contained in the adhered particle group; and clustering the adhered particle group using a clustering function based on the number of particles indicated by the indication, thereby segmenting the adhered particle group into the stated number of particles.

[0006] In the technical solution of this application embodiment, by receiving the user's instruction on the number of particles contained in the adhesive particle group, and using an appropriate clustering function to cluster the adhesive particle group according to the indicated number, the segmentation of adhesive particles can be achieved accurately and efficiently.

[0007] In a second aspect, this application provides a computer program product characterized by comprising instructions that, when executed by one or more processors, cause the one or more processors to perform the method as described above.

[0008] In a third aspect, this application provides an apparatus for segmenting three-dimensional particles, characterized in that the apparatus comprises: a memory storing instructions thereon; and a processor configured to execute the instructions stored in the memory to: receive a tomographic image of the three-dimensional particles; generate a voxel map of the three-dimensional particles based on the tomographic image; convert the voxel map into a point cloud represented by an N×3 dimensional matrix, where N represents the number of points in the point cloud; label connected components in the point cloud using a multi-neighborhood method to obtain a point cloud block representing a plurality of particles, some of the plurality of particles being adhered to each other to form one or more adhered particle groups, each adhered particle group containing two or more adhered particles; for each adhered particle group, receive a user indication of the number of particles contained in the adhered particle group; and cluster the adhered particle groups using a clustering function based on the number of particles indicated by the indication, thereby segmenting the adhered particle groups into the stated number of particles.

[0009] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0010] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0011] Figure 1 A schematic diagram of an image processing procedure for obtaining a three-dimensional multi-particle voxel image according to an embodiment of the present disclosure is shown.

[0012] Figure 2 An example of triggering a clustering process in a GUI according to one embodiment of the present disclosure is shown.

[0013] Figure 3 The effects of a multi-particle adhesion and segmentation process according to an embodiment of the present disclosure are shown from multiple perspectives.

[0014] Figure 4 The surface voxels of a three-dimensional particle obtained according to an embodiment of the present disclosure are shown in the front-back, left-right, and up-down directions. Detailed Implementation

[0015] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0017] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0018] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0019] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0020] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0021] In the description of the embodiments of this application, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this application and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0022] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0023] This disclosure relates to segmenting three-dimensional particles in images of such particles as graphite grains. In existing technologies, after CT 3D reconstruction of three-dimensional particles, software such as Avizo is used to segment the particles on two-dimensional slices based on a threshold. However, this threshold requires manual trial and error to determine, and it cannot be applied to every slice, resulting in multiple particles adhering together in 3D space. Avizo's 3D particle segmentation function cannot distinguish between multiple adhering particles. Furthermore, Avizo's statistical analysis function only performs statistical analysis on simple physical quantities such as volume, which is insufficient to meet the simulation calculation requirements for digitally describing the morphology of three-dimensional particles.

[0024] This disclosure is proposed to accurately and efficiently segment three-dimensional particles in an image.

[0025] In one embodiment of this disclosure, a method for segmenting three-dimensional particles is proposed, comprising: receiving a tomographic image of three-dimensional particles; generating a voxel map of the three-dimensional particles based on the tomographic image; converting the voxel map into a point cloud represented by an N×3 dimensional matrix, where N represents the number of points in the point cloud; labeling connected components in the point cloud using a multi-neighborhood method to obtain a point cloud block representing multiple particles, some of the particles being adhered to each other to form one or more adhered particle groups, each adhered particle group containing two or more adhered particles; for each adhered particle group, receiving a user indication of the number of particles contained in the adhered particle group; and clustering the adhered particle group using a clustering function based on the number of particles indicated by the indication, thereby segmenting the adhered particle group into the stated number of particles.

[0026] Tomographic images, also known as computed tomography (CT) images, are obtained by performing CT (Computed Tomography) 3D reconstruction on the 3D particles. This reconstruction process includes:

[0027] Step 1: Data Acquisition. X-rays emitted from the X-ray source pass through the rotating sample and are received by the detector. Projection data from multiple angles are acquired. A sine curve is formed, which is a two-dimensional matrix with angles and detector channel as independent variables and the detector received signal as the dependent variable.

[0028] Step 2: Use ramp filtering to process the projection data, enhance high-frequency components, and reduce blurring in the subsequent back projection process.

[0029] Step 3: Back-projection image reconstruction. The filtered projection data is back-projected into the image space. Through the back-projection process, the filtered projection data at all angles and projection positions are accumulated to form the final tomographic image.

[0030] After obtaining tomographic images, voxel maps of three-dimensional particles can be generated based on these images. For example, three-dimensional particles can be extracted from the images using software such as Avizo. Specifically, CT three-dimensional reconstruction slices of the sample are assembled using software such as Avizo, and the specific three-dimensional particles are extracted. This process includes, for example, the following:

[0031] Step 1: Import the CT 3D reconstructed slice data into software such as Avizo. Manually set the threshold segmentation based on the pixel value distribution to separate the background from the grain.

[0032] Step 2: Use image opening operation, i.e., erosion followed by dilation, to remove small particles or noise from the image.

[0033] Step 3: Remove the small pixel spots to obtain the sample tube.

[0034] Step 4: Call the arithmetic module, set it to subtraction, and subtract the sample tube from the threshold segmentation map obtained in Step 1 to obtain a two-dimensional particle slice map.

[0035] Step 5: Using 3D volumetric methods, small voxel spots are removed to obtain a three-dimensional multi-particle voxel map.

[0036] Figure 1 A schematic diagram of the image processing procedure for obtaining the above-mentioned three-dimensional multi-particle voxel image is shown.

[0037] After obtaining the three-dimensional multi-particle voxel map, the three-dimensional particles are segmented.

[0038] When segmenting multi-scale particles using methods based on morphological erosion and expansion and data downsampling and upsampling, severe particle distortion can occur.

[0039] This disclosure employs connected component labeling and clustering algorithms to achieve accurate and efficient segmentation within a GUI (Graphical User Interface) program without losing shape information. The specific steps are as follows:

[0040] Step 1: Convert the 3D multi-particle voxel (image) obtained using software such as Avizo into a point cloud, and label connected components using a multi-neighborhood (e.g., 26-neighborhood) approach, resulting in multiple (e.g., 214 or more) point cloud blocks. These point cloud blocks contain multiple particles clustered together. A voxel, similar to a pixel, can be represented as a function f(x,y,z) with a specific range in xyz coordinates. The point cloud is an N*3 matrix, where N represents the number of points. The range can be, for example, [0, V], where V is a positive integer. For example, in one embodiment, V could be 255, in which case the range is an integer from 0 to 255.

[0041] In another embodiment, the value range can be (0, 1), that is, the value of a voxel or point can be 0 or 1, that is, 0 indicates no voxel or point, and 1 indicates the presence of a voxel or point.

[0042] Step 2: In the point cloud clumps generated as described above, some particles have already separated from the other particles. However, some particles are often stuck together, forming one or more groups of stuck particles, each group containing two or more particles stuck together.

[0043] Furthermore, for each group of adherent particles, the system can receive a user's instruction regarding the number of particles contained in that group. More specifically, the user can view the number of adherent particles (i.e., the number of particles stuck together) in the GUI and provide an instruction regarding that number. For example, this instruction can be given by entering the number in the GUI, by voice input, or in any other way.

[0044] Furthermore, based on the number of particles indicated by this indicator, a clustering function is used to cluster the adhered particle group, thereby segmenting the adhered particle group into the stated number of particles with an accuracy of up to 97.5%. Clustering the adhered particle group using a clustering function may include, for example, clustering the point cloud representing the adhered particle group. Therefore, the clustered point cloud can be saved in a GUI program or software, thus saving the particle segmentation result.

[0045] Since most particles are already separated when the point cloud is obtained, with only a few particles adhering to each other, the above method ensures efficient and accurate particle segmentation. More specifically, by receiving the user's indication of the number of particles contained in the few adhering particle groups, and clustering the adhering particle groups using an appropriate clustering function based on the indicated number, accurate and efficient segmentation of adhering particles can be achieved.

[0046] In the above embodiments, the clustering function can be one clustering function or multiple clustering functions. For example, in one embodiment, the clustering function can vary depending on the indicated number of particles, thus there can be multiple different clustering functions. In one embodiment, the clustering function can be the same clustering function, and the number of particles can be a parameter of that clustering function.

[0047] In one embodiment, the method for segmenting three-dimensional particles may further include displaying a graphical representation of the clustering results for groups of adherent particles in a graphical user interface. For example, the clustering results may be displayed in a GUI program or software.

[0048] Figure 2 An example of triggering a clustering process in a GUI according to one embodiment of the present disclosure is shown. Figure 3 The effects of a multi-particle adhesion and segmentation process according to an embodiment of the present disclosure are shown from multiple perspectives.

[0049] In one embodiment, clustering a group of adherent particles using a clustering function includes using the number of particles as one of a plurality of parameters of the clustering function. In this embodiment, clustering may be performed using only one clustering function.

[0050] After segmenting the particles, it is also desirable to understand their morphological characteristics.

[0051] Due to the irregularity of the surface of real three-dimensional particles and the influence of noise, the surface is relatively rough, making it difficult to generate two-dimensional surfaces using techniques such as Poisson reconstruction. However, two-dimensional surfaces are important for mining the morphological features of three-dimensional particles, such as sphericity, concavity, and roundness.

[0052] This disclosure presents a method for solving the surface area of ​​three-dimensional particles.

[0053] In one embodiment, the method of this disclosure further includes: for each particle, by translating the set of points representing the particle by one unit in one of the directions of up, down, left, right, front, and back in the coordinate system representing the point cloud, performing a logical AND operation on the set before the translation and the set after the translation, and subtracting the result of the logical AND operation from the set before the translation, to obtain all surface elements of the particle's surface facing the one direction.

[0054] Specifically, irregular three-dimensional particles can be viewed as being composed of multiple voxels (for example, a unit cube in a coordinate system). By translating the stacked voxels (three-dimensional particles) as a whole up, down, left, right, forward, and backward, the surface element facing one of the directions up, down, left, right, forward, and backward can be obtained. From this, the surface area facing that direction can be easily calculated, thereby achieving accurate calculation of the surface area of ​​the non-smooth particles.

[0055] A set of points (or voxels) of stacked voxels (3D particles) can be denoted as A. After shifting set A down by one unit, the resulting set of points (or voxels) can be denoted as B. Performing a logical AND operation on A and B, equivalent to taking the intersection of sets A and B, results in A&B (or AUB). In this disclosure, if both points or voxels have non-zero values ​​(e.g., 1, 2... or 255), the result of the logical AND operation on these two points or voxels is considered 1; if one of the two points or voxels has a value of zero, the result of the logical AND operation on these two points or voxels is considered 0.

[0056] Furthermore, calculate AA&B, that is, subtract the result of the above logical AND operation from set A, thereby obtaining all points or voxels of the above three-dimensional particle with upward-facing face elements. Count the number of these points or voxels, denoted as n, and assume that the area of ​​each face element (i.e., one of the six surfaces of the unit cube representing a voxel) is area. Then, the sum of the areas of all points or voxels of the above three-dimensional particle with upward-facing face elements is n*area, which gives the area of ​​the upper surface of the three-dimensional particle.

[0057] Similarly, the areas of the bottom, front, back, left, and right surfaces can be calculated. Adding these six surface areas together gives the total surface area of ​​the three-dimensional particle.

[0058] That is, in one embodiment, the method of this disclosure may further include: obtaining the surface area of ​​the particle's surface facing the one direction by summing the surface areas of all facets facing the one direction.

[0059] In one embodiment, the method of this disclosure may further include: obtaining the surface area of ​​the particle by summing the surface areas of the particle in each of the directions of up, down, left, right, front, and back.

[0060] Figure 4 The surface voxels of a three-dimensional particle obtained according to an embodiment of the present disclosure are shown in the front-back, left-right, and up-down directions. Figure 4 In the diagram, green represents the surface voxels of the particles, and red represents the internal voxels of the particles.

[0061] In one embodiment, the method of this disclosure may further include: for the morphological characteristics of all particles, employing a Gaussian mixture clustering model, selecting the optimal number of cluster categories based on the Bayesian information criterion, and classifying all particles into the optimal number of cluster categories. The morphological characteristics may be, for example, one of volume, sphericity, concavity, roundness, average axial ratio, axial ratio, average particle size, flatness, thinness, and principal dimensions.

[0062] Three-dimensional particles exhibit numerous morphological features, and mining these features holds significant physical meaning. Features that can be used as input parameters for simulation calculations include volume, sphericity, concavity, roundness, and average axial ratio. Except for volume, sphericity, concavity, roundness, and average axial ratio all exhibit Gaussian mixture distribution characteristics. Therefore, a Gaussian mixture clustering model can be employed, and the optimal number of clusters can be determined using the Bayesian Information Criterion (BIC) value. After determining the optimal number of clusters, a corresponding Gaussian mixture model can be further determined, where overfitting is avoided by balancing the model's goodness of fit and complexity. Thus, unsupervised classification of specific morphological features of three-dimensional particles can be achieved with relatively low complexity while ensuring goodness of fit. For example, with an optimal number of clusters of 3 and volume as the morphological feature, the volume of all three-dimensional particles can be classified into three categories: large, medium, and small.

[0063] Here, the Bayesian Information Criterion (BIC) is used to select the optimal model from multiple candidate Gaussian mixture models. Different clustering categories k constitute the candidate set of Gaussian mixture models. The BIC method is used to balance the goodness of fit and complexity of the model. Specifically, BIC avoids overfitting by maximizing the model's likelihood function while penalizing excessive parameters in the model.

[0064] The formula for calculating BIC is:

[0065] BIC = -2∙ln(L) + p∙ln(n).

[0066] Where L is the maximum likelihood estimate of the Gaussian mixture model, p is the number of parameters in the Gaussian mixture model, and n is the number of samples. -2∙ln(L) measures the goodness of fit of the Gaussian mixture model to the data; the larger ln(L) is, the smaller the BIC and the better the fit. p∙ln(n) is a penalty term for the complexity of the Gaussian mixture model; the more parameters there are, the larger the BIC value and the greater the penalty, thus inhibiting the replication of the model.

[0067] Therefore, this disclosure determines the optimal number of clusters based on the minimum BIC value, that is, it determines and selects the number of clusters that minimizes the BIC value as the optimal number of clusters, and then determines the corresponding Gaussian mixture model.

[0068] In one embodiment, this disclosure provides a computer program product characterized by comprising instructions that, when executed by one or more processors, cause the one or more processors to perform the method according to any of the above embodiments.

[0069] In one embodiment, this disclosure provides an apparatus for segmenting three-dimensional particles, characterized in that the apparatus includes: a memory storing instructions thereon; and a processor configured to execute the instructions stored in the memory to: receive a tomographic image of the three-dimensional particles; generate a voxel map of the three-dimensional particles based on the tomographic image; convert the voxel map into a point cloud represented by an N×3 dimensional matrix, where N represents the number of points in the point cloud; label connected components in the point cloud using a multi-neighborhood approach to obtain a point cloud block representing a plurality of particles, some of which are adhered to each other to form one or more adhered particle groups, each adhered particle group containing two or more adhered particles; for each adhered particle group, receive a user indication of the number of particles contained in the adhered particle group; and cluster the adhered particle groups using a clustering function based on the number of particles indicated by the indication, thereby segmenting the adhered particle groups into the stated number of particles.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application. In particular, as long as there is no structural conflict, the various technical features mentioned in the embodiments can be combined in any way. This application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for segmenting three-dimensional particles, characterized in that, include: Receive tomographic images of three-dimensional particles; A three-dimensional voxel map of particles is generated based on the tomographic image; The voxel image is converted into a point cloud represented by an N×3 matrix, where N represents the number of points in the point cloud; By labeling the connected components in the point cloud using a multi-neighbor method, a point cloud block representing multiple particles is obtained. Some of the particles are stuck together to form one or more sticky particle groups, and each sticky particle group contains two or more particles that are stuck together. For each group of adhering particles, receive an instruction from the user regarding the number of particles contained in the group of adhering particles; as well as Based on the number of particles indicated by the indication, the group of adherent particles is clustered using a clustering function, thereby dividing the group of adherent particles into the number of particles.

2. The method according to claim 1, characterized in that, Also includes: A graphical representation of the clustering results for the group of adherent particles is displayed in the graphical user interface.

3. The method according to claim 1, characterized in that, Clustering the group of adherent particles using a clustering function includes using the number of particles as one of the multiple parameters of the clustering function.

4. The method according to claim 1, characterized in that, Also includes: For each particle, by translating the set of points representing the particle by one unit in one of the directions up, down, left, right, front, and back in the coordinate system representing the point cloud, performing a logical AND operation on the set before and after the translation, and subtracting the result of the logical AND operation from the set before the translation, all surface elements of the particle facing the one direction are obtained.

5. The method according to claim 4, characterized in that, Also includes: The surface area of ​​the particle's surface facing the one direction is obtained by summing the surface areas of all face elements facing the one direction.

6. The method according to claim 5, characterized in that, Also includes: The surface area of ​​the particle is obtained by summing the surface areas of the particle in each of the directions of up, down, left, right, front, and back.

7. The method according to claim 1, characterized in that, Also includes: Based on the morphological characteristics of all particles, a Gaussian mixture clustering model is adopted, and the optimal number of cluster categories is selected based on the Bayesian information criterion, classifying all particles into the optimal number of cluster categories.

8. The method according to claim 7, characterized in that, in, The optimal number of cluster categories minimizes the Bayesian information criterion value.

9. The method according to claim 7, characterized in that, The morphological features are one of the following: volume, sphericity, concavity, roundness, average axial ratio, axial ratio, average grain size, flatness, fineness, and principal dimension.

10. A computer program product, characterized in that, The method includes instructions that, when executed by one or more processors, cause the one or more processors to perform the method according to any one of claims 1 to 8.

11. An apparatus for segmenting three-dimensional particles, characterized in that, The device includes: A memory, on which instructions are stored; and The processor is configured to execute instructions stored in the memory to: Receive tomographic images of three-dimensional particles; A three-dimensional voxel map of particles is generated based on the tomographic image; The voxel image is converted into a point cloud represented by an N×3 matrix, where N represents the number of points in the point cloud; By labeling the connected components in the point cloud using a multi-neighbor method, a point cloud block representing multiple particles is obtained. Some of the particles are stuck together to form one or more sticky particle groups, and each sticky particle group contains two or more particles that are stuck together. For each group of adhering particles, the system receives a user instruction regarding the number of particles contained in that group; and Based on the number of particles indicated by the indication, the group of adherent particles is clustered using a clustering function, thereby dividing the group of adherent particles into the number of particles.

Images