A method and system for characterizing three-dimensional morphology and evaluating filamentousness of cut tobacco

Abstract

The application discloses a kind of stem filament three-dimensional morphology characterization and filamentous degree evaluation method and system, belong to tobacco processing technical field.The method includes: obtaining stem filament three-dimensional point cloud data, after pre-processing, using cascade segmentation strategy, combined with normal vector distribution entropy criterion and improved watershed algorithm realizes the accurate separation of adherent stem filament;To single stem filament point cloud cluster extraction three-dimensional space skeleton line;Along the skeleton line method plane is analyzed to cross section projection point set, calculates real three-dimensional length, average width and thickness, bending degree, cross section uniformity index and structural integrity coefficient and other basic morphological parameters;On this basis, construct the comprehensive filamentous degree index of fusion multi-attribute penalty function, and calculate thickness unevenness index, structural defect suspicion index and other derivative parameters to assist process diagnosis.The application realizes the objective quantitative evaluation of stem filament "thin, long, straight, even, integral" quality, solves the problem of traditional two-dimensional measurement distortion and subjective evaluation misalignment.

Description

Technical Field

[0001] This invention relates to the field of tobacco processing technology, and in particular to a method and system for three-dimensional morphological characterization and filamentity evaluation of stems. Background Technology

[0002] In cigarette manufacturing, tobacco stems are an important component made from tobacco stems through processes such as pressing, cutting, expansion, and drying. The morphological quality of the tobacco stems (such as length, width, thickness, curvature, and filling value) directly affects the tobacco structure, filling capacity, combustion performance, and sensory comfort of cigarettes. Ideally, tobacco stems should be long, straight, and uniformly thick in sheet or filament form to achieve optimal processing performance.

[0003] Currently, the quality inspection of stem wire relies heavily on the subjective feel, visual inspection, or simple sieve grading of process personnel, lacking quantitative standards. This approach is highly subjective, has poor repeatability, and is inefficient, failing to meet the needs of modern, refined production management. Two-dimensional inspection technology based on machine vision is limited by the projection principle, resulting in distortion in the measurement of the length of spatially curved stem wire and completely failing to obtain crucial thickness information affecting filling and combustion. Although laser three-dimensional scanning technology is becoming increasingly mature, existing three-dimensional morphological characterization methods mostly focus on calculating macroscopic geometric parameters such as the size of the circumscribed cuboid and the projected area. They fail to deeply analyze the internal spatial geometric structure of stem wire as a "flexible slender body," thus making it impossible to accurately define and quantify deep-level morphological attributes closely related to filamentity, such as curvature, cross-sectional uniformity, and microstructural defects.

[0004] In summary, existing technologies suffer from the following bottlenecks: subjective evaluation and two-dimensional measurement cannot reflect the true three-dimensional morphology of the stems, especially in quantifying spatial curvature and cross-sectional variations; there is a lack of objective quantitative indicators that integrate multi-dimensional attributes such as "thinness, length, straightness, uniformity, and integrity"; and the utilization rate of three-dimensional point cloud data is low, mostly limited to macroscopic dimensional measurements, failing to deeply analyze the microscopic geometric structures closely related to filling values ​​and combustibility. Therefore, developing an innovative method and system that can accurately quantify the morphology of stems across all dimensions from the perspective of three-dimensional spatial geometry and scientifically and objectively evaluate their filamentity level is crucial for promoting intelligent and refined quality control in tobacco processing. Summary of the Invention

[0005] To address the above problems, this invention proposes a method and system for three-dimensional morphological characterization and filamentity evaluation of stalks.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] A method for three-dimensional morphological characterization and filamentity evaluation of stalks includes:

[0008] The three-dimensional point cloud data of the filaments is acquired, and preprocessed and segmented to obtain point cloud clusters; the three-dimensional spatial skeleton lines are extracted based on the point cloud clusters.

[0009] Based on the three-dimensional spatial skeleton lines and point cloud clusters, basic morphological parameters and derived parameters are calculated; and based on the basic morphological parameters, the comprehensive filamentity index of the fusion penalty mechanism is calculated.

[0010] Based on the comprehensive filamentity index and derived parameters, a filamentity quality evaluation report of the stem filaments is generated.

[0011] As a preferred embodiment of the present invention, the preprocessing and segmentation include:

[0012] Statistical filtering algorithms are used to process 3D point cloud data to remove discrete noise points;

[0013] For the 3D point cloud data after removing discrete noise points, a clustering algorithm based on Euclidean distance is used to group data points whose 3D Euclidean distance between points is less than or equal to the first threshold into the same initial point cloud cluster, and data points whose 3D Euclidean distance between points is greater than the first threshold into different initial point cloud clusters. If the number of points in an initial point cloud cluster exceeds a preset second threshold, the surface normal vector distribution entropy of the initial point cloud cluster is calculated. If the entropy value of the surface normal vector distribution entropy is lower than a preset third threshold, the initial point cloud cluster is determined to be a cohesive point cloud cluster. For cohesive point cloud clusters, an improved watershed algorithm based on local curvature distance transformation is used for segmentation to obtain the point cloud cluster corresponding to a single strand.

[0014] As a preferred embodiment of the present invention, the step of extracting the three-dimensional spatial skeleton line from the point cloud cluster includes:

[0015] For the point cloud cluster corresponding to a single filament, an iterative calculation is performed using a graph optimization model with Laplacian contraction to shrink the surface of the point cloud cluster toward the center along the normal direction, generating an initial skeleton point set.

[0016] The initial skeleton point set is fitted with a B-spline curve to obtain the master curve. The fitted B-spline curve is then resampled with equal arc length to obtain the three-dimensional spatial skeleton line.

[0017] As a preferred embodiment of the present invention, the basic morphological parameters include the actual three-dimensional length, average width, average thickness, three-dimensional spatial skeleton line curvature, cross-sectional dimension uniformity index, and structural integrity coefficient.

[0018] Among them, the spatial arc length of the three-dimensional skeleton line is calculated as the actual three-dimensional length of the stem;

[0019] Sampling is performed at equal intervals along the three-dimensional spatial skeleton line. Point cloud clusters within a preset neighborhood arc length range before and after the three-dimensional spatial skeleton line are taken as neighborhood point cloud clusters. A normal plane perpendicular to the tangent of the three-dimensional spatial skeleton line is established at each sampling point. The neighborhood point cloud clusters are projected onto this normal plane to obtain a cross-sectional projection point set. The diameter of the minimum circumcircle of the cross-sectional projection point set is calculated as the local width of the sampling point, and the direction of the diameter of the minimum circumcircle is taken as the local width direction. The arithmetic mean of the local widths of all sampling points is calculated to obtain the average width.

[0020] Within the normal plane, the direction perpendicular to the local width direction is determined as the local thickness direction. The coordinate range of the cross-sectional projection point set in the local thickness direction is calculated and used as the local thickness of the sampling point. The arithmetic mean of the local thicknesses of all sampling points is calculated to obtain the average thickness.

[0021] The curvature of the three-dimensional spatial skeleton line is obtained by dividing the actual three-dimensional length by the Euclidean distance between the beginning and end points of the three-dimensional spatial skeleton line.

[0022] Calculate the coefficient of variation (CVW) of the local width at all sampling points to obtain the cross-sectional dimension uniformity index. for ;

[0023] The structural integrity factor is:

[0024]

[0025] In the formula, This represents the actual surface area of ​​the point cloud cluster of filaments. Let V be the volume of the point cloud cluster of the filament. Let Γ be the surface area of ​​an ideal reference cylinder, and let Γ be the structural integrity coefficient. The volume is taken as an ideal reference cylinder.

[0026] As a preferred embodiment of the present invention, the calculation of local width further includes: using the three-dimensional spatial skeleton line vertical distance method, calculating the maximum distance in the vertical direction from the skeleton point to the edge point cloud in the cross-section projection point set, as the local width.

[0027] As a preferred embodiment of the present invention, the derived parameters include: thickness non-uniformity index, structural defect suspicion index, and end face flatness;

[0028] Along the three-dimensional spatial skeleton line, for each sampling point, in the corresponding normal plane, calculate the value of the cross-section projection point set in the local thickness direction, as the local thickness value; calculate the coefficient of variation of the local thickness values ​​corresponding to all sampling points, as the thickness non-uniformity index;

[0029] Statistical risk count: If the ratio of the volume of the point cloud cluster to the volume of the 3D convex hull is less than a preset threshold, then one risk is marked; if along the 3D spatial skeleton line, there are a preset number of consecutive sampling points, and the local cross-sectional area in the normal plane is reduced to less than 50% of the average local cross-sectional area of ​​all sampling points, then another risk is marked; based on the number of risk items, the structural defect suspicion index is determined to be high risk, medium risk, or low risk level;

[0030] Extract the point cloud clusters located at both ends of the three-dimensional spatial skeleton line from the point cloud clusters, fit them to planes respectively, and calculate the angle between the normal vector of each plane and the direction of the three-dimensional spatial skeleton line. Take the average of the two angles as the flatness of the end face.

[0031] As a preferred embodiment of the present invention, the expression for the comprehensive filamentity index of the fusion penalty mechanism is as follows:

[0032]

[0033]

[0034]

[0035] In the formula, To achieve the comprehensive filamentity index, For the actual three-dimensional length, Average width, The average thickness The curvature of the three-dimensional spatial skeleton line. The uniformity index of cross-sectional dimensions The structural integrity coefficient. This is a monotonically increasing penalty function based on the curvature of the three-dimensional skeleton line. This is a joint monotonically increasing penalty function based on the cross-sectional dimension uniformity index and the structural integrity coefficient. , , These are the weighting coefficients. , , This is the penalty weighting coefficient.

[0036] As a preferred embodiment of the present invention, the step of generating a filament quality evaluation report based on a comprehensive filamentity index and derived parameters includes:

[0037] The mean and standard deviation are calculated based on the comprehensive filamentity index of all stems in a batch; and the overall quality grade of the stem batch is determined based on a preset comprehensive threshold range, wherein the overall quality grade of the stem batch includes excellent, good, good, qualified, and unqualified.

[0038] Based on the derived parameters, the defect dimensions and judgment criteria to be diagnosed are determined, including: when the thickness non-uniformity index of a single strand exceeds the upper limit threshold of the process, it is judged as a thickness uniformity defect; when the structural defect suspicion index is at a high risk level, it is judged as a structural integrity defect; when the end face flatness exceeds the process standard threshold, it is judged as an end face shearing quality defect.

[0039] The proportion of the number of strand samples identified as having thickness uniformity defects, structural integrity defects, and end-face shear quality defects in this batch was counted separately to the total number of samples. The proportion of each defect dimension was compared with the preset attention threshold, and the degree of defect for each defect dimension was determined as a significant defect, an item requiring attention, or a normal performance.

[0040] Based on the overall quality grade of the filament batch and the degree of defect in each defect dimension, a filament quality evaluation report is generated.

[0041] As a preferred embodiment of the present invention, the filament quality evaluation report further includes: based on the defect dimensions corresponding to significant defects and items requiring attention, outputting corresponding process parameter adjustment suggestions by querying a pre-established defect dimension-process suggestion mapping table.

[0042] A three-dimensional morphological characterization and filamentity evaluation system for stalks includes:

[0043] The acquisition and preprocessing module is used to acquire the three-dimensional point cloud data of the filaments, and perform preprocessing and segmentation to obtain point cloud clusters; and extract three-dimensional spatial skeleton lines based on the point cloud clusters.

[0044] The comprehensive index module is used to calculate the basic morphological parameters and derived parameters based on the three-dimensional spatial skeleton lines and point cloud clusters; and to calculate the comprehensive filamentity index of the fusion penalty mechanism based on the basic morphological parameters.

[0045] The evaluation report generation module is used to generate a filament quality evaluation report for stems based on the comprehensive filamentity index and derived parameters.

[0046] The beneficial effects of this invention are as follows: By acquiring 3D point cloud data and using Euclidean clustering algorithms, it effectively solves the problem that traditional 2D measurements cannot obtain the true spatial morphology of the stems, especially achieving precise separation of adhered stems, ensuring the independent and accurate extraction of the morphological parameters of individual stems. Based on the extracted 3D spatial skeleton lines, a normal plane section analysis framework with the 3D spatial skeleton lines as a unified reference system is constructed, realizing the precise quantification of multi-dimensional basic morphological parameters such as true 3D length, average width and thickness, 3D spatial skeleton line curvature, section uniformity, and structural integrity, eliminating the interference of stem placement posture and spatial curvature on the measurement results. Unifying the calculation of basic morphological parameters and derived parameters under the same skeleton line sampling framework allows basic morphological characterization and process defect diagnosis to share the same data source. The parameter system is hierarchical and logically consistent, avoiding evaluation contradictions caused by inconsistent parameter benchmarks in traditional methods. An innovative comprehensive filamentity index integrating multiple attributes and penalty mechanisms has been constructed, transforming the quality requirements of "fine, long, straight, uniform, and neat" stem filaments into a single objective quantitative indicator. This overcomes the shortcomings of traditional subjective evaluation, which varies from person to person and is difficult to reproduce, providing a reliable basis for the automatic grading of stem filament quality. Through derived parameters, in-depth process diagnosis and problem tracing are achieved, accurately locating specific process defects such as uneven pressing by the stem-pressing roller and wear of the cutting blade, and outputting targeted process adjustment suggestions. This realizes a closed-loop feedback from "quality evaluation" to "process optimization," powerfully promoting the upgrading of tobacco processing towards intelligence and precision. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0048] Figure 1 This is a flowchart of the method in an embodiment of the present invention;

[0049] Figure 2 This is a system structure diagram in an embodiment of the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0051] like Figure 1As shown, this is an embodiment of the present invention, which provides a method for three-dimensional morphological characterization and filamentity evaluation of filaments, including:

[0052] S1. Obtain the three-dimensional point cloud data of the filaments, and perform preprocessing and segmentation to obtain point cloud clusters; extract the three-dimensional spatial skeleton lines based on the point cloud clusters.

[0053] In this embodiment, a statistical filtering algorithm is used to process the 3D point cloud data to remove discrete noise points;

[0054] For the 3D point cloud data after removing discrete noise points, a clustering algorithm based on Euclidean distance is used to group data points whose 3D Euclidean distance between points is less than or equal to the first threshold into the same initial point cloud cluster, and data points whose 3D Euclidean distance between points is greater than the first threshold into different initial point cloud clusters. If the number of points in an initial point cloud cluster exceeds a preset second threshold, the surface normal vector distribution entropy of the initial point cloud cluster is calculated. If the entropy value of the surface normal vector distribution entropy is lower than a preset third threshold, the initial point cloud cluster is determined to be a cohesive point cloud cluster. For cohesive point cloud clusters, an improved watershed algorithm based on local curvature distance transformation is used for segmentation to obtain the point cloud cluster corresponding to a single strand.

[0055] For the point cloud cluster corresponding to a single filament, an iterative calculation is performed using a graph optimization model with Laplacian contraction to shrink the surface of the point cloud cluster toward the center along the normal direction, generating an initial skeleton point set.

[0056] The initial skeleton point set is fitted with a B-spline curve to obtain the master curve. The fitted B-spline curve is then resampled with equal arc length to obtain the three-dimensional spatial skeleton line.

[0057] Specifically, approximately 5g of stem filaments were randomly sampled from the shredder outlet and gently and evenly spread onto a black, non-reflective velvet cloth, avoiding overlap. A laser 3D scanner was used to scan the filaments, with a scanning resolution of 0.05mm (0.05mm dot pitch) to ensure the capture of detailed features. After scanning, raw 3D point cloud data including the background and the filaments was obtained.

[0058] Statistical filtering preprocessing is performed on the original 3D point cloud data. First, the average distance between each data point in the 3D point cloud data and a specified number of nearest neighbor data points is calculated. Then, the global mean and standard deviation of the average distances of all data points are calculated. Finally, noise is judged according to the three-standard-deviation criterion. Data points whose average distance exceeds three times the upper limit of the global mean are judged as discrete noise points and are removed.

[0059] For the 3D point cloud data after removing discrete noise points, an initial segmentation algorithm based on Euclidean distance is used. Specifically, the first threshold is set to 0.4 mm, slightly larger than the average thickness of the filaments, using the 3D Euclidean distance between points as the criterion. Data points with a 3D Euclidean distance less than or equal to 0.4 mm are considered spatially connected and merged into the same initial point cloud cluster; data points with a 3D Euclidean distance greater than 0.4 mm are considered spatially disconnected and belong to different initial point cloud clusters. This step can separate the 3D point cloud data corresponding to a single filament that is spatially separated into multiple independent initial point cloud clusters.

[0060] For the aforementioned initial point cloud clusters, if the number of data points contained in an initial point cloud cluster exceeds a preset second threshold (500 in this embodiment), it indicates that the initial point cloud cluster may contain multiple strands of adhesive bodies. In this case, the surface normal vector distribution entropy of the initial point cloud cluster is further calculated. Specifically, the surface normal vector of each data point within the initial point cloud cluster is calculated, the distribution of the normal vector directions is statistically analyzed, and the information entropy value is calculated. If the information entropy value of the surface normal vector distribution entropy is lower than a preset third threshold (0.7 in this embodiment), it indicates that the surface normal vector direction distribution of the initial point cloud cluster is concentrated, and there are multiple surfaces with the same orientation inside. The initial point cloud cluster is then determined to be an adhesive point cloud cluster. If the number of data points contained in the initial point cloud cluster is less than or equal to the preset second threshold, and the entropy value of the surface normal vector distribution entropy is greater than or equal to the preset third threshold, then the initial point cloud cluster is directly regarded as a point cloud cluster corresponding to a single strand.

[0061] For the initial point cloud clusters identified as adhering point cloud clusters, an improved watershed algorithm based on local curvature distance transformation is used for further segmentation. The key improvement of the improved watershed algorithm is that, in the traditional watershed algorithm, the terrain height used for segmentation is defined by the Euclidean distance from the data point to the boundary; this invention modifies the terrain height to be defined by the local curvature derivative of the data point, making the watershed segmentation line easier to form along the natural, gently curvatured depressions between the strands, thereby achieving accurate separation of adhering strands and obtaining the point cloud clusters corresponding to individual strands.

[0062] For each point cloud cluster corresponding to a single filament obtained from the above segmentation, perform the following operations:

[0063] First, the Laplace shrinkage algorithm is applied to construct the graph structure of the point cloud cluster corresponding to a single filament. Then, by iteratively solving the Laplace equation, the surface data points of the point cloud cluster shrink towards the center along the normal direction. After 50 iterations, the point cloud cluster shrinks into an initial skeleton point set approximately aligned with the centerline.

[0064] Then, the initial skeleton point set is fitted with a master curve: a cubic B-spline curve is used to fit the initial skeleton point set to obtain a smooth B-spline curve.

[0065] Finally, the smooth B-spline curve is resampled using constant arc lengths: with a fixed arc length step of 0.1 mm, resampling is performed along the B-spline curve to obtain a continuous, smooth, and uniformly dense three-dimensional spatial skeleton line composed of multiple ordered sampling points. This three-dimensional spatial skeleton line serves as the unified spatial reference for all subsequent morphological parameter calculations.

[0066] S2. Based on the three-dimensional spatial skeleton lines and point cloud clusters, calculate the basic morphological parameters and derived parameters; and based on the basic morphological parameters, calculate the comprehensive filamentity index of the fusion penalty mechanism.

[0067] In this embodiment, the basic morphological parameters include the actual three-dimensional length, average width, average thickness, three-dimensional spatial skeleton line curvature, cross-sectional dimension uniformity index, and structural integrity coefficient.

[0068] Among them, the spatial arc length of the three-dimensional skeleton line is calculated as the actual three-dimensional length of the stem;

[0069] Sampling is performed at equal intervals along the three-dimensional spatial skeleton line. Point cloud clusters within a preset neighborhood arc length range before and after the three-dimensional spatial skeleton line are taken as neighborhood point cloud clusters. A normal plane perpendicular to the tangent of the three-dimensional spatial skeleton line is established at each sampling point. The neighborhood point cloud clusters are projected onto this normal plane to obtain a cross-sectional projection point set. The diameter of the minimum circumcircle of the cross-sectional projection point set is calculated as the local width of the sampling point, and the direction of the diameter of the minimum circumcircle is taken as the local width direction. The arithmetic mean of the local widths of all sampling points is calculated to obtain the average width.

[0070] Within the normal plane, the direction perpendicular to the local width direction is determined as the local thickness direction. The coordinate range of the cross-sectional projection point set in the local thickness direction is calculated and used as the local thickness of the sampling point. The arithmetic mean of the local thicknesses of all sampling points is calculated to obtain the average thickness.

[0071] The curvature of the three-dimensional spatial skeleton line is obtained by dividing the actual three-dimensional length by the Euclidean distance between the beginning and end points of the three-dimensional spatial skeleton line.

[0072] Calculate the coefficient of variation (CVW) of the local width at all sampling points to obtain the cross-sectional dimension uniformity index. for ;

[0073] in, , Average width, The coefficient of variation for the local width is 0.18, and the cross-sectional dimension uniformity index is 5.56. A larger value indicates more uniform cross-sectional dimensions along the length of the strand; a smaller value indicates more drastic fluctuations in cross-sectional dimensions. In this embodiment, the coefficient of variation for the local width is 0.18, therefore the cross-sectional dimension uniformity index is 5.56.

[0074] The structural integrity factor is:

[0075]

[0076] In the formula, This represents the actual surface area of ​​the point cloud cluster of filaments. Let V be the volume of the point cloud cluster of the filament. Let Γ be the surface area of ​​an ideal reference cylinder, and let Γ be the structural integrity coefficient. The volume is taken as an ideal reference cylinder.

[0077] in, It can be calculated using the Poisson reconstruction algorithm or the elliptic integral method of sections, such as... , The actual three-dimensional length; This can be calculated by triangulating the point cloud clusters into a mesh. In this embodiment, It is 0.58 mm³. The volume is 8.7 mm². An ideal reference cylinder with the same volume as the point cloud cluster corresponding to the single filament is constructed, and its volume is... = ; Surface area of ​​the ideal reference cylinder for:

[0078]

[0079] In the formula, The radius of the ideal reference cylinder can be obtained by inversely deriving the formula for calculating the volume of the point cloud cluster;

[0080] The closer the structural integrity coefficient is to 1, the smoother and denser the surface of the stem, and the more complete the structure. A coefficient greater than 1 indicates the presence of wrinkles or burrs on the surface. A coefficient less than 1 may indicate the presence of local voids or a loose structure. In this embodiment, the structural integrity coefficient is 1.15, indicating that the surface of the stem has slight wrinkles or burrs.

[0081] In this embodiment, the calculation of local width further includes: using the three-dimensional spatial skeleton line vertical distance method, calculating the maximum distance in the vertical direction from the skeleton point to the edge point cloud in the cross-section projection point set, as the local width.

[0082] Specifically, the spatial arc length of the three-dimensional skeleton line is calculated, that is, the lengths of the line segments between all adjacent ordered sampling points on the three-dimensional skeleton line are summed one by one, and the spatial arc length is taken as the true three-dimensional length of the filament that is not affected by the projection effect. In this embodiment, the true three-dimensional length is taken as 14.3mm.

[0083] Along the 14.3mm long three-dimensional spatial skeleton line, samples were taken at equal intervals of 0.2mm with a preset arc length, resulting in 72 sampling points. For each sampling point, the following operations were performed:

[0084] Centered on the sampling point, the point cloud clusters corresponding to the single filaments within a preset neighborhood arc length of 0.15 mm along the three-dimensional spatial skeleton line are taken as the neighborhood point cloud clusters of the sampling point. A normal plane perpendicular to the tangent of the three-dimensional spatial skeleton line is established at the sampling point, and the neighborhood point cloud clusters are orthogonally projected onto the normal plane to obtain the cross-sectional projection point set. The minimum circumcircle of the cross-sectional projection point set is calculated, which can be implemented using the Megiddo algorithm. The diameter of the minimum circumcircle is taken as the local width of the sampling point, and the direction of the diameter of the minimum circumcircle is recorded as the local width direction. The arithmetic mean of the local widths of all sampling points is calculated, and the average width is 2.2 mm.

[0085] As an alternative, the local width can also be calculated using the vertical distance method of the three-dimensional spatial skeleton line: in the set of projection points of the cross section, the maximum distance from the projection point of the three-dimensional spatial skeleton line to the edge of the set of projection points of the cross section in the direction perpendicular to the three-dimensional spatial skeleton line is calculated, and this distance is taken as the local width of the sampling point.

[0086] Based on the normal plane and cross-sectional projection point set of the same sampling point mentioned above, continue to perform the following operations:

[0087] The direction perpendicular to the local width direction is determined as the local thickness direction of the sampling point; the difference between the maximum and minimum coordinate values ​​of the cross-sectional projection point set in the local thickness direction, i.e., the coordinate range, is calculated as the local thickness of the sampling point; the arithmetic mean of the local thicknesses of all sampling points is calculated to obtain the average thickness; in this embodiment, the average thickness is 0.85 mm.

[0088] In this method, the closer the curvature value of the three-dimensional spatial skeleton line is to 1, the straighter the strands are; the larger the value, the greater the curvature of the strands. In this embodiment, the calculated Euclidean distance between the beginning and end points of the three-dimensional spatial skeleton line is 12.1 mm, and the actual three-dimensional length is 14.3 mm, so the curvature of the three-dimensional spatial skeleton line is 1.18.

[0089] In this embodiment, the derived parameters include: thickness non-uniformity index, structural defect suspicion index, and end face flatness;

[0090] Along the three-dimensional spatial skeleton line, for each sampling point, in the corresponding normal plane, calculate the value of the cross-section projection point set in the local thickness direction, as the local thickness value; calculate the coefficient of variation of the local thickness values ​​corresponding to all sampling points, as the thickness non-uniformity index;

[0091] Statistical risk count: If the ratio of the volume of the point cloud cluster to the volume of the 3D convex hull is less than a preset threshold, then one risk is marked; if along the 3D spatial skeleton line, there are a preset number of consecutive sampling points, and the local cross-sectional area in the normal plane is reduced to less than 50% of the average local cross-sectional area of ​​all sampling points, then another risk is marked; based on the number of risk items, the structural defect suspicion index is determined to be high risk, medium risk, or low risk level;

[0092] Extract the point cloud clusters located at both ends of the three-dimensional spatial skeleton line from the point cloud clusters, fit them to planes respectively, and calculate the angle between the normal vector of each plane and the direction of the three-dimensional spatial skeleton line. Take the average of the two angles as the flatness of the end face.

[0093] Specifically, based on the obtained local thicknesses of all sampling points along the three-dimensional spatial skeleton line, the arithmetic mean and standard deviation of these local thicknesses are calculated. The arithmetic mean is the average thickness; the standard deviation is:

[0094]

[0095] In the formula, The standard deviation of the local thickness. The total number of sampling points. For the first Local thickness at each sampling point This represents the average thickness.

[0096] The coefficient of variation for local thickness was calculated, which is the standard deviation of local thickness divided by the arithmetic mean of local thicknesses, and used as the thickness non-uniformity index. A smaller thickness non-uniformity index indicates a more uniform thickness of the stem along its length; a larger index indicates more drastic thickness fluctuations along the length, suggesting potential issues such as uneven rolling or uneven moisture content during stem pretreatment. In this embodiment, local thickness values ​​were obtained from 72 sampling points along the three-dimensional skeleton line, and the calculated coefficient of variation was 0.25.

[0097] The structural defect suspicion index is comprehensively assessed using the following two risk judgment rules and divided into three levels: high risk, medium risk, or low risk:

[0098] The first risk assessment involves calculating the ratio of the volume of the point cloud cluster corresponding to a single filament to the volume of the 3D convex hull of that point cloud cluster. If this ratio is less than a preset threshold, a risk is identified. In this embodiment, the preset threshold is set to 0.85. The calculated volume of the point cloud cluster is 0.58 mm³, and the volume of the 3D convex hull of the point cloud cluster is 0.70 mm³, resulting in a ratio of 0.83. Since this ratio is less than 0.85, the first risk is identified.

[0099] The second risk assessment involves sequentially examining the local cross-sectional area of ​​the cross-sectional projection points within the normal plane of each sampling point along the three-dimensional spatial skeleton line. If a predetermined number of consecutive sampling points, such as three consecutive points, experience a sudden drop in local cross-sectional area below 50% of the average local cross-sectional area of ​​all sampling points, it indicates a risk of significant local thinning or structural breakage of the filament along its length, thus marking this as another risk. In this embodiment, the local cross-sectional area sequence of all sampling points is examined; if no consecutive sudden drops are found, this risk is not triggered.

[0100] The risk level determination rules are as follows:

[0101] If both risks are triggered, the structural defect suspicion index is determined to be at a high risk level.

[0102] If a risk is triggered, the structural defect suspicion index is determined to be at a medium risk level.

[0103] If no risk is triggered, the structural defect suspicion index is determined to be at a low risk level.

[0104] In this embodiment, since only the first risk is triggered, the structural defect suspicion index is determined to be at a medium risk level.

[0105] Extract the point cloud clusters located at both ends of the three-dimensional spatial skeleton line from the point cloud clusters corresponding to a single strand. Specifically, take the point cloud clusters corresponding to a preset arc length range at each end of the three-dimensional spatial skeleton line, for example, within a range of 0.3 mm from each end point, as the point cloud subsets of the end faces of the strand. Perform plane fitting on the two end face point cloud subsets to obtain two fitting planes and their corresponding plane normal vectors. Calculate the overall direction vector of the three-dimensional spatial skeleton line, that is, the direction of the line connecting the first and last endpoints of the three-dimensional spatial skeleton line. Calculate the angles α and β between the normal vectors of the two fitting planes and the overall direction vector of the three-dimensional spatial skeleton line, and take the arithmetic mean of the two angles as the end face flatness. The smaller the end face flatness, the flatter the shredding end face, and the better the cutting quality. The larger the end face flatness, the more inclined the shredding end face or the existence of tearing deformation, indicating that the sharpness of the shredding blade or the installation angle may be abnormal.

[0106] In this embodiment, the angle α between the normal vector of the first end face fitting plane and the direction of the three-dimensional skeleton line is 12 degrees, and the angle b between the normal vector of the second end face fitting plane and the direction of the three-dimensional skeleton line is 8 degrees. The average value is then taken to obtain an end face flatness of 10 degrees.

[0107] In this embodiment, the expression for the calculated comprehensive filamentity index of the fusion penalty mechanism is:

[0108]

[0109]

[0110]

[0111] In the formula, To achieve the comprehensive filamentity index, For the actual three-dimensional length, Average width, The average thickness The curvature of the three-dimensional spatial skeleton line. The uniformity index of cross-sectional dimensions The structural integrity coefficient. This is a monotonically increasing penalty function based on the curvature of the three-dimensional skeleton line. This is a joint monotonically increasing penalty function based on the cross-sectional dimension uniformity index and the structural integrity coefficient. , , These are the weighting coefficients. , , This is the penalty weighting coefficient.

[0112] Specifically, weighting coefficients and penalty weighting coefficients are set to align with the preference for slenderness in production, and are calibrated with reference to the background data distribution, as follows: =2.0, =1.2, =1.0, =1.5, =0.5, =2.0. Substituting this into the formula for calculating the comprehensive filamentity index of the fusion penalty mechanism, we get:

[0113]

[0114]

[0115]

[0116] According to calculations, the overall filamentity index in this embodiment is 320.

[0117] In the mathematical model of the comprehensive filamentity index: the numerator is a positive power term of the actual three-dimensional length (α>0). The longer the actual three-dimensional length, the larger the comprehensive filamentity index and the better the quality. The denominator is the product of the positive power terms of the average width and average thickness and two penalty functions. The wider the width, the thicker the thickness, the greater the curvature, the more uneven the cross-section, and the less complete the structure, the larger the denominator, the smaller the comprehensive filamentity index, and the worse the quality. The penalty function f(K) adopts an exponential form, so that the penalty intensity increases exponentially when the curvature of the three-dimensional skeleton line deviates from the ideal value of 1. The joint penalty function g(U,Γ) adopts the form of a power function and the product of linear deviation, and applies penalties to cross-sectional unevenness (cross-sectional size uniformity index deviates significantly) and structural incompleteness (structural integrity coefficient deviates from 1), respectively. The weight coefficients α, β, γ and the penalty weight coefficients λ, δ, μ can be calibrated according to the actual process standards and the degree of preference for each attribute.

[0118] The comprehensive filamentity index transforms the multi-dimensional quality requirements of stem fibers—namely, their thinness, length, straightness, uniformity, and neatness—into a single objective quantitative indicator. A higher value indicates superior overall morphological quality, overcoming the drawbacks of traditional subjective evaluations that are subjective and difficult to reproduce. This provides a reliable basis for the automatic grading of stem fiber quality. Based on this comprehensive filamentity index, all individual stem fibers within a batch can be quantitatively ranked and statistically analyzed, providing core data support for subsequent batch quality evaluation.

[0119] S3. Based on the comprehensive filamentity index and derived parameters, generate a filamentity quality evaluation report for the stem filaments.

[0120] In this embodiment, the mean and standard deviation are calculated based on the comprehensive filamentity index of all stems in a batch; and the overall quality grade of the stem batch is determined based on a preset comprehensive threshold range, wherein the overall quality grade of the stem batch includes excellent, good, good, qualified, and unqualified.

[0121] Based on the derived parameters, the defect dimensions and judgment criteria to be diagnosed are determined, including: when the thickness non-uniformity index of a single strand exceeds the upper limit threshold of the process, it is judged as a thickness uniformity defect; when the structural defect suspicion index is at a high risk level, it is judged as a structural integrity defect; when the end face flatness exceeds the process standard threshold, it is judged as an end face shearing quality defect.

[0122] The proportion of the number of strand samples identified as having thickness uniformity defects, structural integrity defects, and end-face shear quality defects in this batch was counted separately to the total number of samples. The proportion of each defect dimension was compared with the preset attention threshold, and the degree of defect for each defect dimension was determined as a significant defect, an item requiring attention, or a normal performance.

[0123] Based on the overall quality grade of the filament batch and the degree of defect in each defect dimension, a filament quality evaluation report is generated.

[0124] In this embodiment, the filamentity quality evaluation report further includes: based on the defect dimensions corresponding to significant defects and items requiring attention, outputting corresponding process parameter adjustment suggestions by querying a pre-established defect dimension-process suggestion mapping table.

[0125] Specifically, 300 filaments are continuously sampled from the current production batch, and the arithmetic mean and standard deviation of the comprehensive filamentity index of the 300 filaments are calculated; the expression is:

[0126]

[0127]

[0128] In the formula, The arithmetic mean of the comprehensive filamentity index is given. The standard deviation of the composite filamentity index, The total number of samples in the stem batch. The sample number. For the first The overall filiformity index of root filaments.

[0129] The overall quality grade of the stem batch is determined based on a preset comprehensive threshold range. In this embodiment, the preset comprehensive threshold range and the corresponding quality grade classification are shown in Table 1:

[0130] Table 1 Scope-Level Mapping Table

[0131]

[0132] In this embodiment, the average value of the comprehensive filamentity index of 300 stems is 305, and the standard deviation is 45. According to the above comprehensive threshold range, the comprehensive filamentity index of 305 falls into the "excellent" grade range, and the overall quality grade of this batch of stems is determined to be "excellent".

[0133] Based on the calculated derived parameters, each stem sample in this batch was assessed for defects. The defect dimensions and assessment criteria are as follows:

[0134] For a single filament, if its thickness non-uniformity index exceeds a preset process upper limit threshold, the filament sample is determined to have a thickness uniformity defect. In this embodiment, the preset process upper limit threshold is 0.3.

[0135] For a single filament, if its structural defect suspicion index is at a high risk level, then the filament sample is determined to have a structural integrity defect.

[0136] For a single strand, if the flatness of its end face exceeds a preset process standard threshold, the strand sample is determined to have an end face shearing quality defect. In this embodiment, the preset process standard threshold is 10 degrees.

[0137] The number of strand samples identified as having the above three defects in this batch was counted separately, and the proportion of each type of defect sample to the total number of samples in the batch M was calculated to obtain the proportion of thickness uniformity defects, structural integrity defects, and end face shear quality defects.

[0138] In this embodiment: 35% of the stem wire samples had a thickness non-uniformity index greater than 0.3, meaning the thickness uniformity defect rate was 35%; 8% of the stem wire samples had a high risk of structural defect, meaning the structural integrity defect rate was 8%; the average flatness of the end face of 300 stem wires was 15°, and the proportion of end face shearing quality defects was statistically calculated based on whether a single stem wire exceeded 10°. In this embodiment, 15% of the stem wire samples had an end face flatness exceeding 10°, meaning the end face shearing quality defect rate was 15%.

[0139] The proportions of each of the above defect dimensions are compared with preset attention thresholds to determine the degree of defect for each defect dimension. In this embodiment, the preset attention thresholds and defect degree determination rules are shown in Table 2:

[0140] Table 2 Defect Judgment Table

[0141]

[0142] In this embodiment, the proportion of thickness uniformity defects is 35%, exceeding the 15% concern threshold, and is therefore judged as a significant defect; the proportion of structural integrity defects is 8%, exceeding the 5% concern threshold, and is also judged as a significant defect; the number of all strands in this batch that are judged to have end-face shear quality defects is counted and divided by the total number of samples participating in the test to obtain the end-face shear quality defect ratio; finally, based on this ratio, the overall degree of end-face shear quality defects of this batch of strands is determined. In this embodiment, the end-face shear quality defect ratio is 15%, which is judged as a significant defect.

[0143] Based on the overall quality grade of the stem batch and the degree of defect in each defect dimension, a filament quality evaluation report is generated. The filament quality evaluation report includes: overall stem quality conclusions, major defect diagnoses, and recommendations for adjusting process parameters; specifically:

[0144] a. Conclusion on the overall quality of the stems

[0145] The overall quality conclusion of the stem fibers is the overall quality grade of the obtained stem fiber batch; in this embodiment, the overall quality grade of this batch of stem fibers is "excellent", with an average value of 305 and a standard deviation of 45 for the comprehensive filamentity index. The relatively large standard deviation indicates that there is some fluctuation in the morphological quality of the stem fibers within the batch, and the process stability needs to be improved.

[0146] b. Diagnosis of major defects

[0147] (1) When the thickness uniformity defect is a significant defect: the proportion of thickness uniformity defects exceeds 15%. In this embodiment, the proportion of thickness uniformity defects is 35%, which indicates that the proportion of samples with excessive thickness non-uniformity index of the tobacco stems in this batch is too high. The thickness fluctuates significantly along the length direction, and the overall uniformity does not meet the process requirements. Therefore, the thickness uniformity defect is judged to be a significant defect. This type of defect is mainly caused by insufficient roller pressing uniformity in the pressing process and uneven distribution of moisture content after the tobacco stems are rehydrated. This causes the stems to be subjected to inconsistent stress and large deformation differences during the rolling process, which ultimately manifests as excessive thickness fluctuation.

[0148] When thickness uniformity defects are a concern: If the proportion of thickness uniformity defects is between 8% and 15%, it indicates that the proportion of samples with excessive thickness non-uniformity index within the tobacco stem batch is within the critical range, and the overall thickness fluctuation trend is increasing, but has not yet reached a significant defect level. Therefore, thickness uniformity defects are considered a concern. This type of defect is mainly caused by minor abnormalities such as slight uneven pressure of the pressure roller, local fluctuations in the moisture content of the tobacco stem, and unstable incoming material conditions. Although it has not caused large-scale quality problems, there is a risk of further deterioration.

[0149] (2) When the structural integrity defect is a significant defect: the proportion of structural integrity defects exceeds 5%. In this embodiment, the proportion of structural integrity defects is 8%, which indicates that the proportion of samples with high risk of structural defects in the batch of stems is relatively high. The stems have obvious internal looseness, voids, local tears or surface burrs, and the structural integrity defect is judged to be a significant defect. This type of defect is mainly caused by insufficient sharpness of the cutting blade, unreasonable matching of cutting parameters, and improper control of expansion drying process, which leads to structural damage, looseness or deformation of the stems during cutting or drying.

[0150] When structural integrity defects are a concern: If the proportion of structural integrity defects is between 2% and 5%, it indicates that the proportion of high-risk structural defect samples in the batch of stem wire is low, but the proportion of medium-risk samples is increasing. There are hidden risks to structural integrity, but they have not reached the level of significant defects. Therefore, structural integrity defects are identified as a concern. These defects are mainly caused by early abnormalities such as slight blade dulling, slight fluctuations in blade gate pressure, and slight deviations in drying parameters. The overall structure is basically stable, but there is a trend of gradual deterioration.

[0151] (3) When the end face shearing quality defect is a significant defect: the proportion of end face shearing quality defects exceeds 10%. In this embodiment, the proportion of end face shearing quality defects is 15%, which indicates that the proportion of samples with an end face flatness angle greater than 10° in the batch of filaments is too high. The cut is tilted and the burrs are obvious. The end face shearing quality does not meet the process requirements. Therefore, the end face shearing quality defect is judged to be a significant defect. This type of defect is mainly caused by the installation angle deviation of the filament cutting blade, uneven blade wear, insufficient clamping force of the blade gate, or vibration during the cutting process, resulting in uneven end face cutting and excessive cutting angle.

[0152] When end-face shearing quality defects are a concern: If the proportion of end-face shearing quality defects is between 5% and 10%, it indicates that the proportion of samples with end-face flatness angles exceeding 10° within the batch of stem wire is close to the critical value, and the end-face quality fluctuation is increasing, but no large-area defects have been formed. Therefore, end-face shearing quality defects are considered a concern. These defects are mainly caused by minor abnormalities such as slight wear of the blade, small deviations in installation accuracy, and unstable blade gate pressure. The cut quality is basically controllable, but there is a risk of further deterioration.

[0153] c. Suggestions for adjusting process parameters

[0154] Based on the above defect diagnosis results, targeted process parameter adjustment suggestions are output by querying the pre-established defect dimension-process suggestion mapping table, as shown in Table 3:

[0155] Table 3 Defect Dimension - Process Recommendation Mapping Table

[0156]

[0157] like Figure 2 As shown, this is the second embodiment of the present invention. This embodiment provides a three-dimensional morphological characterization and filamentity evaluation system for stalks, including:

[0158] The acquisition and preprocessing module is used to acquire the three-dimensional point cloud data of the filaments, and perform preprocessing and segmentation to obtain point cloud clusters; and extract three-dimensional spatial skeleton lines based on the point cloud clusters.

[0159] The comprehensive index module is used to calculate the basic morphological parameters and derived parameters based on the three-dimensional spatial skeleton lines and point cloud clusters; and to calculate the comprehensive filamentity index of the fusion penalty mechanism based on the basic morphological parameters.

[0160] The evaluation report generation module is used to generate a filament quality evaluation report for stems based on the comprehensive filamentity index and derived parameters.

[0161] In summary, this invention effectively solves the problem that traditional two-dimensional measurement cannot obtain the true spatial morphology of filaments by acquiring three-dimensional point cloud data and using a cascaded segmentation strategy. In particular, by using the surface normal vector distribution entropy criterion combined with an improved watershed algorithm based on local curvature distance transformation, it achieves accurate separation of adhered filaments and ensures the independent and accurate extraction of the morphological parameters of individual filaments.

[0162] Based on the extracted three-dimensional spatial skeleton lines, a normal plane section analysis framework with the skeleton lines as a unified spatial reference was constructed. This enabled the accurate quantification of multi-dimensional basic morphological parameters such as true three-dimensional length, average width, average thickness, skeleton line curvature, cross-sectional dimension uniformity index, and structural integrity coefficient. It eliminated the interference of the filament placement posture and spatial curvature on the measurement results. The parameter system is clear in hierarchy and logically consistent.

[0163] By unifying the calculation of basic morphological parameters and derived parameters under the same skeleton line sampling framework, the basic morphological characterization and process defect diagnosis share the same data source, avoiding evaluation contradictions caused by inconsistent benchmarks of different parameters in traditional methods. Among them, derived parameters such as thickness non-uniformity index, structural defect suspicion index, and end face flatness can accurately reflect the local thickness fluctuations of the wire, the looseness of the internal structure, and the quality of the cut wire end face.

[0164] An innovative comprehensive filamentity index integrating multiple attributes and penalty mechanisms was constructed, transforming the multi-dimensional quality requirements of "fine, long, straight, uniform, and neat" stems into a single objective quantitative indicator. This overcomes the drawbacks of traditional subjective evaluation, which varies from person to person and is difficult to reproduce, and provides a reliable basis for the automated and standardized grading of stem quality.

[0165] Based on batch statistics and defect dimension diagnosis of comprehensive filamentity index, the system can automatically generate a filamentity quality evaluation report that includes batch quality grade, defect type and degree, and targeted process adjustment suggestions. This realizes an intelligent closed loop from "3D scanning → parameter quantification → comprehensive evaluation → defect tracing → process feedback", which effectively promotes the upgrading of stem filament processing towards refinement and intelligence.

[0166] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any other combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product, which includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0167] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0168] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for three-dimensional morphological characterization and filamentity evaluation of stalks, characterized in that, include: The three-dimensional point cloud data of the filaments is acquired, and preprocessed and segmented to obtain point cloud clusters; Extract three-dimensional spatial skeleton lines based on the point cloud clusters; Based on the three-dimensional spatial skeleton lines and point cloud clusters, basic morphological parameters and derived parameters are calculated; and based on the basic morphological parameters, the comprehensive filamentity index of the fusion penalty mechanism is calculated. Based on the comprehensive filamentity index and derived parameters, a filamentity quality evaluation report of the stem filaments is generated.

2. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 1, characterized in that, The preprocessing and segmentation include: Statistical filtering algorithms are used to process 3D point cloud data to remove discrete noise points; For the 3D point cloud data after removing discrete noise points, a clustering algorithm based on Euclidean distance is used to group data points whose 3D Euclidean distance between points is less than or equal to the first threshold into the same initial point cloud cluster, and data points whose 3D Euclidean distance between points is greater than the first threshold into different initial point cloud clusters. If the number of points in an initial point cloud cluster exceeds a preset second threshold, the surface normal vector distribution entropy of the initial point cloud cluster is calculated. If the entropy value of the surface normal vector distribution entropy is lower than a preset third threshold, the initial point cloud cluster is determined to be a cohesive point cloud cluster. For cohesive point cloud clusters, an improved watershed algorithm based on local curvature distance transformation is used for segmentation to obtain the point cloud cluster corresponding to a single strand.

3. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 2, characterized in that, The extraction of the three-dimensional spatial skeleton line from the point cloud cluster includes: For the point cloud cluster corresponding to a single filament, an iterative calculation is performed using a graph optimization model with Laplacian contraction to shrink the surface of the point cloud cluster towards the center along the normal direction, generating an initial skeleton point set. The initial skeleton point set is then fitted with a B-spline curve, and the fitted B-spline curve is resampled with equal arc length to obtain the three-dimensional spatial skeleton line.

4. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 1, characterized in that, The basic morphological parameters include the actual three-dimensional length, average width, average thickness, three-dimensional spatial skeleton line curvature, cross-sectional dimension uniformity index, and structural integrity coefficient. Among them, the spatial arc length of the three-dimensional skeleton line is calculated as the actual three-dimensional length of the stem; Sampling is performed at equal intervals along the three-dimensional spatial skeleton line. Point cloud clusters within a preset neighborhood arc length range before and after the three-dimensional spatial skeleton line are taken as neighborhood point cloud clusters. A normal plane perpendicular to the tangent of the three-dimensional spatial skeleton line is established at each sampling point. The neighborhood point cloud clusters are projected onto this normal plane to obtain a cross-sectional projection point set. The diameter of the minimum circumcircle of the cross-sectional projection point set is calculated as the local width of the sampling point, and the direction of the diameter of the minimum circumcircle is taken as the local width direction. The arithmetic mean of the local widths of all sampling points is calculated to obtain the average width. Within the normal plane, the direction perpendicular to the local width direction is determined as the local thickness direction. The coordinate range of the cross-sectional projection point set in the local thickness direction is calculated and used as the local thickness of the sampling point. The arithmetic mean of the local thicknesses of all sampling points is calculated to obtain the average thickness. The curvature of the three-dimensional spatial skeleton line is obtained by dividing the actual three-dimensional length by the Euclidean distance between the beginning and end points of the three-dimensional spatial skeleton line. Calculate the coefficient of variation (CVW) of the local width at all sampling points to obtain the cross-sectional dimension uniformity index. for ; The structural integrity factor is: ； In the formula, This represents the actual surface area of ​​the point cloud cluster of filaments. Let V be the volume of the point cloud cluster of the filament. Let Γ be the surface area of ​​an ideal reference cylinder, and let Γ be the structural integrity coefficient. The volume is taken as an ideal reference cylinder.

5. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 4, characterized in that, The calculation of local width also includes: using the vertical distance method of three-dimensional spatial skeleton line, calculating the maximum distance in the vertical direction from the skeleton point to the edge point cloud in the cross-section projection point set, and using it as the local width.

6. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 4, characterized in that, The derived parameters include: thickness non-uniformity index, structural defect suspicion index, and end face flatness. Along the three-dimensional spatial skeleton line, for each sampling point, in the corresponding normal plane, calculate the value of the cross-section projection point set in the local thickness direction, as the local thickness value; calculate the coefficient of variation of the local thickness values ​​corresponding to all sampling points, as the thickness non-uniformity index; Statistical risk count: If the ratio of the volume of the point cloud cluster to the volume of the 3D convex hull is less than a preset threshold, then one risk is marked; if along the 3D spatial skeleton line, there are a preset number of consecutive sampling points, and the local cross-sectional area in the normal plane is reduced to less than 50% of the average local cross-sectional area of ​​all sampling points, then another risk is marked; based on the number of risk items, the structural defect suspicion index is determined to be high risk, medium risk, or low risk level; Extract the point cloud clusters located at both ends of the three-dimensional spatial skeleton line from the point cloud clusters, fit them to planes respectively, and calculate the angle between the normal vector of each plane and the direction of the three-dimensional spatial skeleton line. Take the average of the two angles as the flatness of the end face.

7. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 6, characterized in that, The expression for the comprehensive filamentity index of the fusion penalty mechanism is as follows: ； ； ； In the formula, To achieve the comprehensive filamentity index, For the actual three-dimensional length, For average width, The average thickness The curvature of the three-dimensional skeleton line. The uniformity index of cross-sectional dimensions The structural integrity coefficient. This is a monotonically increasing penalty function based on the curvature of the three-dimensional skeleton line. This is a joint monotonically increasing penalty function based on the cross-sectional dimension uniformity index and the structural integrity coefficient. , , These are the weighting coefficients. , , This is the penalty weighting coefficient.

8. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 7, characterized in that, The process of generating a filament quality evaluation report based on the comprehensive filamentity index and derived parameters includes: The mean and standard deviation are calculated based on the comprehensive filamentity index of all stems in a batch; and the overall quality grade of the stem batch is determined based on a preset comprehensive threshold range, wherein the overall quality grade of the stem batch includes excellent, good, good, qualified, and unqualified. Based on the derived parameters, the defect dimensions and judgment criteria to be diagnosed are determined, including: when the thickness non-uniformity index of a single strand exceeds the upper limit threshold of the process, it is judged as a thickness uniformity defect; when the structural defect suspicion index is at a high risk level, it is judged as a structural integrity defect; when the end face flatness exceeds the process standard threshold, it is judged as an end face shearing quality defect. The proportion of the number of strand samples identified as having thickness uniformity defects, structural integrity defects, and end-face shear quality defects in this batch was counted separately to the total number of samples. The proportion of each defect dimension was compared with the preset attention threshold, and the degree of defect for each defect dimension was determined as a significant defect, an item requiring attention, or a normal performance. Based on the overall quality grade of the filament batch and the degree of defect in each defect dimension, a filament quality evaluation report is generated.

9. The method for three-dimensional morphological characterization and filamentity evaluation of filaments according to claim 8, characterized in that, The filament quality evaluation report also includes: based on the defect dimensions corresponding to significant defects and items requiring attention, outputting corresponding process parameter adjustment suggestions by querying a pre-established defect dimension-process suggestion mapping table.

10. A three-dimensional morphological characterization and filamentity evaluation system for stem filaments, employing the three-dimensional morphological characterization and filamentity evaluation method for stem filaments as described in claims 1 to 9, characterized in that, include: The acquisition and preprocessing module is used to acquire the 3D point cloud data of the filaments and perform preprocessing and segmentation to obtain point cloud clusters; Extract three-dimensional spatial skeleton lines based on the point cloud clusters; The comprehensive index module is used to calculate the basic morphological parameters and derived parameters based on the three-dimensional spatial skeleton lines and point cloud clusters; and to calculate the comprehensive filamentity index of the fusion penalty mechanism based on the basic morphological parameters. The evaluation report generation module is used to generate a filament quality evaluation report for stems based on the comprehensive filamentity index and derived parameters.

Images