True and false cigarette risk assessment method

Abstract

The invention provides a true and false cigarette risk assessment method, and belongs to the technical field of information, and the method comprises the steps: carrying out the image collection of the end face of a cigarette, obtaining the flatness and surface roughness of a cut, carrying out the pixel-level segmentation of a collected image, and obtaining the edge contour and texture distribution of a cut region; for a suspected sample of mechanical finishing, the finishing intervention strength and the finishing coverage area are identified through the compression depth and the section smoothness of a blunt extrusion fracture, and the intervention degree of manual intervention is evaluated; according to the intervention degree of manual intervention, the tobacco shred filling density and the tobacco shred color uniformity in the inferior tobacco shred sample library are combined, the evaluation standard of true and false tobacco identification is adjusted, and a corrected true and false judgment basis is obtained; comprehensively evaluating the cut flatness, the surface roughness, the burr distribution density, the fracture type and the mechanical finishing trace of the cigarette according to the corrected authenticity judgment basis to obtain a final authenticity identification result; and false cigarettes which are camouflaged and delicate can be effectively identified.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for assessing the risk of counterfeit cigarettes. Background Technology

[0002] In the tobacco industry, distinguishing genuine from counterfeit cigarettes is a crucial task, directly impacting market order and protecting consumer rights. Research in this area not only ensures product quality but also plays a vital role in combating illegal production and distribution. As a core component, genuine-counterfeit identification bears the heavy responsibility of maintaining industry standards, and its accuracy and reliability are a focal point of industry attention. However, current mainstream identification methods have significant limitations, relying primarily on the smoothness of the cigarette's cut surface for visual judgment, assuming that a smoother cut indicates higher processing quality and thus a genuine cigarette. This surface-feature-oriented method is highly susceptible to failure after counterfeiters employ advanced finishing techniques, as counterfeit cigarettes can now closely mimic the appearance of genuine cigarettes, leading to inaccurate identification. A deeper technical challenge lies in identifying and analyzing the key factor of tobacco fiber breakage morphology. Fiber breakage morphology reflects the physical forces exerted on the cigarette during processing, such as natural tearing or mechanical compression. There is a complex and contradictory relationship between the smoothness of the cut and the morphology of fiber breakage. A seemingly smooth cut may conceal traces of fibers being forcibly squeezed and broken. These traces are precisely key evidence that counterfeit cigarettes use mechanical finishing to conceal inferior materials. If the specific morphology of fiber breakage and its relationship to the smoothness of the cut cannot be accurately identified, it is difficult to expose the camouflage methods of counterfeit cigarettes. Specifically, during actual identification, staff found that the end-face cut was smooth as a mirror, with uniform thickness, meeting the standards of genuine cigarettes. However, upon magnification, the fiber bundles at the ends of the tobacco were clearly flattened, and the broken ends did not naturally disperse but were neatly compressed into blocks. This indicates that the cigarette had undergone non-standard mechanical intervention, possibly artificial finishing, to conceal the rough nature of the internal materials. This inconsistency between appearance and internal characteristics greatly complicates the determination of authenticity. Therefore, accurately identifying the camouflage tactics used by counterfeit cigarettes to conceal inferior materials through surface finishing has become a key issue in the risk assessment of counterfeit cigarettes. Summary of the Invention

[0003] This invention provides a method for assessing the risk of counterfeit cigarettes, mainly including: Images of the cigarette end face are acquired to obtain the flatness of the cut and the surface roughness. The acquired images are then segmented at the pixel level to obtain the edge contour and texture distribution of the cut area. Based on the edge contour and texture distribution of the cut area, a depth scan of the tobacco cross-section is performed to obtain the distribution density of burrs at the end of the tobacco and the fiber fracture morphology, and to identify whether the fiber fracture type is a blunt extrusion fracture or a natural tear fracture. Based on the fiber fracture morphology and cut smoothness, the combination features of smoothness and fracture type are identified. When the cut smoothness is high and the fracture type is blunt extrusion fracture, it is marked as a sample suspected of mechanical trimming. When the cut smoothness is moderate and the fracture type is natural tear fracture, it is marked as a sample with natural cutting characteristics. Samples with natural cutting characteristics are directly entered into the regular genuine and counterfeit cigarette identification process for evaluation. For samples suspected of mechanical trimming, the intensity of trimming intervention and the trimming coverage area are identified by the compression depth and surface smoothness of the blunt extrusion fracture surface, and the degree of human intervention is assessed. Based on the degree of human intervention, and combined with the tobacco filling density and color uniformity in the inferior tobacco sample library, the evaluation criteria for identifying genuine and counterfeit cigarettes are adjusted to obtain the revised criteria for determining authenticity. By using the revised criteria for determining authenticity, a comprehensive evaluation is conducted on the cigarette's cut smoothness, surface roughness, burr distribution density, fracture type, and mechanical finishing marks to obtain the final result for authenticity identification.

[0004] Furthermore, the process of acquiring images of the cigarette end face to obtain the flatness and surface roughness of the cut, and performing pixel-level segmentation on the acquired images to obtain the edge contour and texture distribution of the cut area, includes: Within a preset focal length range, the cigarette end face is scanned layer by layer to obtain the image sharpness value of each focal plane. The focal plane with the highest sharpness value is selected as the optimal imaging plane, and a panoramic image of the cigarette end face is acquired on the optimal imaging plane. For the panoramic image, the micro-concavity and convexity depth of the end face is extracted to determine the roughness index. The roughness index is multiplied by a preset benchmark value to obtain an adaptive segmentation threshold. Threshold segmentation processing is performed on the panoramic image to obtain the pixel density distribution of the binarized image. Based on the pixel density distribution, a continuous coordinate point sequence of the cut edge is extracted as the edge contour. The texture contrast and correlation are calculated using the gray-level co-occurrence matrix as the texture distribution.

[0005] Furthermore, the step of extracting the micro-uneven depth of the end face and determining the roughness index for the panoramic image includes: projecting illumination beams from at least three different angles for the panoramic image, obtaining the grayscale gradient value of the edge of the end face at each angle, determining the micro-uneven depth of the end face based on the variation of the grayscale gradient value, and calculating the roughness index by the standard deviation of the uneven depth.

[0006] Furthermore, the step of performing a depth scan on the tobacco cross-section based on the edge contour and texture distribution of the cut area to obtain the distribution density of burrs at the tobacco ends and the fiber breakage morphology, and identifying the fiber fracture type as a blunt compression fracture or a natural tear fracture, includes: Based on the edge contour and texture distribution of the cut area, the cross-section of the tobacco shreds is collected layer by layer within a preset scanning depth range to obtain the burr height distribution map of each layer. The burr distribution density is calculated by the ratio of the number of burrs to the scanned area in the burr height distribution map. For areas where the burr distribution density exceeds a preset density threshold, the fracture angle and tear length of the fibers in that area are extracted. If the fracture angle is less than a preset angle and the tear length exceeds a preset length, the area is determined to be a natural tear fracture. If the burr distribution density is less than a preset density threshold, the presence of compression marks on the cross-section is detected. The area is determined to be a blunt compression fracture by measuring the width and depth of the compression marks.

[0007] Furthermore, regarding the fiber fracture morphology and cut smoothness, the combination of smoothness and fracture type is identified. When the cut smoothness is high and the fracture type is a blunt compression fracture, it is marked as a sample suspected of mechanical trimming. When the cut smoothness is moderate and the fracture type is a natural tear fracture, it is marked as a sample with natural cutting characteristics. Samples with natural cutting characteristics directly enter the conventional genuine and counterfeit cigarette identification process for evaluation, including: Based on the fiber fracture morphology and cut smoothness, the standard deviation is calculated by the deviation distance of the edge contour relative to the baseline. When the standard deviation is lower than the preset smoothness threshold, it is judged as high smoothness; when it is higher than the preset smoothness threshold but lower than the preset roughness threshold, it is judged as moderate smoothness. A combined feature code is generated based on the smoothness level and the fracture type. According to the combined feature code, when the code represents a combination of high smoothness and blunt extrusion fracture, the cigarette is marked as a suspected mechanically trimmed sample; when the code represents a combination of moderate smoothness and natural tear fracture, it is marked as a naturally cut feature sample.

[0008] Furthermore, for samples suspected of mechanical trimming, the intensity and coverage area of ​​trimming intervention are identified by the compression depth and surface smoothness of the blunt extrusion fracture surface, and the degree of human intervention is assessed, including: For samples suspected of mechanical finishing, the blunt extrusion fracture surface is scanned point by point to obtain the compression depth value of each point on the fracture surface relative to the reference plane. The rate of change of the compression depth value from the center point to the edge point is calculated as the indentation gradient. Based on the indentation gradient and the compression depth value sequence, the root mean square value of the height difference between adjacent scan points is calculated as the cross-sectional smoothness. When the root mean square value is lower than a preset threshold, it is determined that the area has undergone fine grinding. The boundary coordinates of the grinding area are extracted and the grinding trace features are recorded. Using the grinding trace features and boundary coordinates, the ratio of the texture roughness inside and outside the boundary is calculated to determine the finishing intensity. The percentage of the area of ​​the finished area to the entire cross-section is calculated using the finishing intensity and boundary coordinates as the finishing coverage area. The degree of human intervention is obtained by multiplying the finishing intensity and the coverage area.

[0009] Furthermore, based on the degree of human intervention and combined with the tobacco filling density and color uniformity in the inferior tobacco sample library, the evaluation criteria for identifying genuine and counterfeit cigarettes are adjusted to obtain the revised criteria for determining authenticity, including: Based on the degree of human intervention, the distribution range of tobacco filling density and color uniformity parameters corresponding to the intervention level are extracted from the inferior tobacco sample library. The density deviation rate is obtained by dividing the difference between the filling density and the standard genuine cigarette density by the standard density. The color difference rate is obtained by dividing the difference between the color uniformity and the standard color parameter by the standard parameter. The density correction value is obtained by multiplying the density deviation rate by the first weighting coefficient, and the color correction value is obtained by multiplying the color difference rate by the second weighting coefficient. A comprehensive score is determined based on the density correction value, the color correction value, and the intervention level assessment value. The original judgment threshold is adjusted based on the comprehensive score to obtain the corrected judgment threshold.

[0010] Furthermore, the modified authenticity determination criteria are used to comprehensively evaluate the cigarette's cut smoothness, surface roughness, burr distribution density, fracture type, and mechanical finishing marks to obtain the final authenticity identification result, including: Using the revised criteria for determining authenticity, the smoothness of the cigarette cut, surface roughness, burr distribution density, fracture type, and mechanical finishing marks are scored on a 100-point scale. Each score is multiplied by its corresponding weighting coefficient and then summed to obtain a comprehensive evaluation score. The result of identifying genuine and counterfeit cigarettes is determined based on the comprehensive evaluation score.

[0011] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for risk assessment of genuine and counterfeit cigarettes. Addressing the common practice in counterfeit cigarette production of mechanically trimming the end face to conceal rough cuts and inferior tobacco, the method uses a high-magnification imaging module to acquire images of the cigarette end face and performs pixel-level segmentation. It extracts features of cut smoothness and surface roughness, and then uses a micro-scanning probe to deeply scan the tobacco cross-section based on edge contours and texture distribution. This obtains burr distribution density and fiber fracture morphology, identifying blunt compression fractures or natural tear fractures. When the cut smoothness is high and it is a blunt compression fracture, it is marked as a suspected mechanically trimmed sample. Further evaluation of compression depth, smoothness, and trimmed coverage area quantifies the degree of human intervention. The filling density and color uniformity evaluation standards are adjusted based on a database of inferior tobacco samples to achieve a corrected comprehensive judgment. Samples with natural cutting features directly enter the conventional genuine / counterfeit identification process. This invention accurately distinguishes between mechanical trimming traces and natural cutting features through the microscopic morphology of the end face, effectively identifying sophisticatedly disguised counterfeit cigarettes, improving the accuracy of genuine / counterfeit identification and anti-counterfeiting capabilities, generating detailed test reports, and storing indicators for subsequent verification. Attached Figure Description

[0012] Figure 1This is a flowchart of a method for assessing the risk of counterfeit cigarettes according to the present invention.

[0013] Figure 2 This is a schematic diagram of a method for assessing the risk of counterfeit cigarettes according to the present invention.

[0014] Figure 3 This is another schematic diagram of a method for assessing the risk of counterfeit cigarettes according to the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] like Figures 1-3 This embodiment of a method for assessing the risk of counterfeit cigarettes may specifically include: Step S101: Image acquisition is performed on the end face of the cigarette to obtain the flatness of the cut and the surface roughness. Pixel-level segmentation processing is performed on the acquired image to obtain the edge contour and texture distribution of the cut area.

[0017] The cigarette end face is scanned layer by layer within a preset focal length range using a high-magnification imaging module. The image sharpness value of each focal plane is obtained, and the focal plane with the highest sharpness value is selected as the optimal imaging plane. A panoramic image of the cigarette end face is then acquired on the optimal imaging plane. For the panoramic image, illumination beams are projected from at least three different angles, and the grayscale gradient value of the end face edge is obtained at each angle. The microscopic unevenness depth of the end face is determined based on the variation of the grayscale gradient value, and the roughness index is calculated using the standard deviation of the unevenness depth. The dimensionless roughness index R is obtained by dividing the roughness index by the reference depth value. R is multiplied by a preset benchmark value of 0.8 to 1.2 to obtain an adaptive segmentation threshold. Threshold segmentation is performed on the panoramic image to obtain the pixel density distribution of the binarized image. The continuous coordinate point sequence of the cut edge is extracted as the edge contour based on the pixel density distribution. The texture contrast and correlation are calculated as the texture distribution through the gray-level co-occurrence matrix. The gray-level co-occurrence matrix is ​​constructed with a gray level of 256 and a distance parameter of 1 in the panoramic image. The contrast is the matrix variance, and the correlation is the ratio of the product of the matrix covariance and the standard deviation, which is used to quantify the uniformity of the texture distribution.

[0018] In one embodiment, the cigarette end-face image acquisition system achieves precise imaging and analysis of the cigarette cut through a high-magnification imaging module. This system integrates an adjustable focal length lens, a multi-angle illumination device, and an image processing unit, enabling it to acquire high-resolution images of the cigarette end-face and extract key feature parameters.

[0019] Specifically, the focal plane scanning process employs a step-by-step focusing mechanism. Starting from the closest focusing distance, the lens advances step-by-step to the furthest focusing distance according to a preset step size. At each focal plane position, an image frame is acquired and its sharpness value is calculated. This sharpness value is obtained by calculating the variance of the image's Laplacian operator; a larger variance indicates sharper image edges and more accurate focusing. The sharpness values ​​of all focal planes are compared, and the focal plane with the highest value is selected as the optimal imaging plane. At this point, the break marks and cut details of the tobacco fibers are at their clearest.

[0020] It should be noted that multi-angle lighting is achieved through a ring-shaped LED array, which contains at least three independently controlled light sources located at 30, 60, and 90 degrees relative to the normal direction of the cigarette end face. When each light source illuminates the cigarette individually, the concave and convex structure of the end face produces different shadow distributions, forming a unique grayscale gradient pattern.

[0021] Preferably, the roughness index is calculated based on the statistical characteristics of the microscopic unevenness depth of the end face. By analyzing the variation amplitude of grayscale gradient values ​​under different illumination angles, the relative height of each pixel is calculated, forming a three-dimensional contour map of the end face. The roughness index is equal to the standard deviation of the height values ​​of all pixels, reflecting the overall roughness of the end face.

[0022] In one embodiment, the adaptive threshold determination process multiplies the roughness index by a preset benchmark value, which is set based on statistical data from a real cigarette sample library and typically ranges from 0.8 to 1.2. A larger roughness index indicates significant unevenness on the end face, requiring a higher segmentation threshold to distinguish texture details; conversely, a smaller roughness index lowers the threshold to preserve subtle texture information. The binarized image obtained after threshold segmentation clearly shows the distribution pattern of tobacco fibers. The system extracts continuous coordinate sequences of edge pixels to form edge contours and simultaneously constructs a gray-level co-occurrence matrix to calculate texture contrast and correlation parameters, comprehensively describing the texture feature distribution of the end face.

[0023] Step S102: Based on the edge contour and texture distribution of the cut area, perform a depth scan on the cross-section of the tobacco shreds to obtain the distribution density of burrs at the end of the tobacco shreds and the fiber fracture morphology, and identify whether the fiber fracture type is a blunt extrusion fracture or a natural tear fracture.

[0024] Based on the edge contour and texture distribution of the cut area, a micro-scanning probe is used to collect data layer by layer from the tobacco cross-section within a preset scanning depth range, obtaining a burr height distribution map for each layer. The burr distribution density is calculated by the ratio of the number of burrs to the scanned area in the burr height distribution map. For areas where the burr distribution density exceeds a preset density threshold, the fracture angle and tear length of the fibers in that area are extracted. If the fracture angle is less than 45 degrees and the tear length exceeds three times the fiber diameter, the area is identified as a natural tear feature area, and its texture interval parameters are recorded. If the burr distribution density is less than the preset density threshold, the presence of compression marks on the cross-section is detected. By measuring the width and depth of the compression marks, if the width is greater than twice the fiber diameter, the area is identified as a blunt compression feature area, and its cross-sectional roughness is recorded. Based on the texture interval parameters or cross-sectional roughness, combined with the area ratio of the natural tear feature area and the blunt compression feature area, when the area ratio of the natural tear feature area exceeds 70%, the fiber fracture type is identified as a natural tear fracture; when the area ratio of the blunt compression feature area exceeds 70%, the fiber fracture type is identified as a blunt compression fracture. When the proportions of both characteristic areas are less than 70% and the proportion of the natural tear characteristic area is greater than the proportion of the blunt extrusion characteristic area, it is identified as a mixed fracture surface dominated by natural tearing; when the proportions of both characteristic areas are less than 70% and the proportion of the blunt extrusion characteristic area is greater than the proportion of the natural tear characteristic area, it is identified as a mixed fracture surface dominated by blunt extrusion; when the proportions of both characteristic areas are equal and both are less than 70%, it is identified as an uncertain type of fracture surface.

[0025] In one embodiment, the microscopic morphology of the tobacco cross-section is detected through a fine-grained analysis using a micro-scanning probe. The micro-scanning probe is equipped with a high-resolution depth sensor and a laser rangefinder, enabling it to acquire three-dimensional structural information of the tobacco fibers within a millimeter-level range, thus achieving accurate identification of the fracture morphology.

[0026] Specifically, the working principle of the micro-scanning probe is based on laser triangulation. A laser beam emitted by the probe illuminates the cross-section of the tobacco at a specific angle. When the laser encounters burrs or uneven structures, it scatters, and the scattered light is received by a photoelectric sensor inside the probe. Based on the geometric relationship between the laser emission angle, the reception angle, and the sensor position, the height value of each scanning point is calculated. The scanning probe advances layer by layer within a preset depth range at micrometer-level steps, with each layer covering the entire cross-sectional area. At each scanning layer, the probe moves along a spiral path to ensure no blind spots. During the scanning process, the three-dimensional coordinates and reflected light intensity of each sampling point are recorded, forming point cloud data. By meshing the point cloud data, a burr height distribution map for each layer is constructed, where the grayscale value of each pixel corresponds to the burr height at that location.

[0027] It should be noted that the generation of the burr height distribution map uses an interpolation algorithm to process the discrete point cloud into a continuous shape. The scanned area is divided into a regular grid. For point cloud data falling into each grid cell, its average height is calculated as the height value of that grid. For grids without point cloud data, bilinear interpolation is performed using the height values ​​of neighboring grids to ensure the continuity and integrity of the height distribution map.

[0028] Preferably, the calculation of burr distribution density involves two stages: burr identification and statistics. In the burr identification stage, edge detection is performed on the height distribution map, and continuous areas with height values ​​exceeding a preset height threshold on the reference plane are marked as burr regions. In the statistics stage, the total number of all burr regions is calculated and divided by the cross-sectional area covered by the scan to obtain the burr distribution density value. The preset density threshold is determined based on statistical data from a real cigarette sample library and is typically set to 5 to 8 burrs per square millimeter.

[0029] For example, the determination of the natural tear feature area is based on the mechanical properties of fiber breakage. When tobacco is subjected to tearing force, the fibers gradually separate along the weakest direction, forming an irregular fracture profile. The fracture angle characterizes the angle between the fiber fracture surface and the cutting plane. The fracture angle produced by natural tearing is usually less than 45 degrees, exhibiting an oblique cut. The tear length reflects the extension distance of fiber separation. During natural tearing, the fibers gradually peel off, and the tear length often exceeds three times the fiber diameter. The fracture profile is extracted using an image recognition algorithm, the angle between the profile and the reference plane is calculated to obtain the fracture angle, and the curve length of the profile is measured to obtain the tear length. When both conditions are met simultaneously, the area is determined to be a natural tear feature area, and its texture interval parameter, i.e., the distance between adjacent torn fibers, is recorded.

[0030] In one possible implementation, the identification of the blunt compression feature area focuses on the indentation characteristics of the cross-section. Blunt compression applies pressure perpendicular to the cross-section to the fiber, flattening and deforming it, leaving obvious compression marks on the cross-section. The location of the compression marks is identified by analyzing the recessed areas in the height distribution map. Compression marks are characterized by a local area with a significantly lower height than the surrounding area, forming a groove-like structure. The width and depth of the groove are measured; when the width is greater than twice the fiber diameter, it indicates that the fiber has been subjected to strong compression and severe deformation.

[0031] Understandably, surface roughness is obtained by calculating the root mean square deviation of the height values ​​within the extrusion region. The height values ​​of all pixels within the extrusion feature area are extracted, their deviation from the average height is calculated, and the root mean square value is used as the roughness index to reflect the degree of surface unevenness caused by extrusion.

[0032] For example, the comprehensive determination of fracture type uses the area proportion method. The areas of the natural tear feature area and the blunt compression feature area are counted separately, and their respective percentages of the total cross-sectional area are calculated. When the area of ​​the natural tear feature area exceeds 70%, the fracture surface exhibits natural tear characteristics; when the area of ​​the blunt compression feature area exceeds 70%, the fracture surface exhibits blunt compression characteristics. Furthermore, this identification method can accurately distinguish the processing characteristics of genuine and counterfeit cigarettes, avoiding being misled by surface finishing techniques and improving the reliability of cigarette authenticity identification.

[0033] Step S103: Based on the fiber fracture morphology and the flatness of the cut, identify the combination features of flatness and fracture type. When the flatness of the cut is high and the fracture type is blunt extrusion fracture, it is marked as a sample suspected of mechanical trimming. When the flatness of the cut is moderate and the fracture type is natural tear fracture, it is marked as a sample with natural cutting features. Samples with natural cutting features directly enter the conventional genuine and counterfeit cigarette identification process for evaluation.

[0034] For fiber fracture morphology and cut smoothness, a standard deviation is calculated based on the deviation of the edge contour from the baseline. When the standard deviation is below a preset smoothness threshold, it is considered high smoothness; when it is above the preset smoothness threshold but below a preset roughness threshold, it is considered moderate smoothness. A combined feature code is generated based on the smoothness level and fracture type. According to the combined feature code, when the code represents a combination of high smoothness and blunt extrusion fracture, the presence of parallel lines or uniformly distributed indentations left by machining on the cross-section is detected. If present, the cigarette is marked as a suspected mechanically modified sample. When the combined feature code represents a combination of moderate smoothness and natural tear fracture, the cutting morphology conforms to natural cutting characteristics and is marked as a natural cutting feature sample. This natural cutting feature sample directly enters the conventional genuine / counterfeit cigarette identification process for evaluation.

[0035] In one implementation, the combined feature identification of cigarette cuts is achieved through quantitative evaluation of flatness and fracture type matching. A feature coding system is established by comprehensively analyzing the geometric features and fracture morphology of the cuts for rapid classification and determination.

[0036] Specifically, the measurement of edge contour deviation distance is based on the coordinate sequence of the cut edge obtained through image processing. First, a reference straight line for the edge contour is fitted using the least squares method, and the perpendicular distance from each edge point to the reference straight line is calculated, forming a deviation distance sequence. The standard deviation is calculated using statistical methods: first, the average of all deviation distances is calculated; then, the sum of the squares of the differences between each distance value and the average is calculated, divided by the sample size, and the square root is taken to obtain the standard deviation value. The standard deviation value directly reflects the smoothness of the cut edge; the smaller the value, the smoother the edge and the finer the cutting process. The preset smoothness threshold is determined according to the standards for genuine cigarette production, typically set at 0.5 mm, and the preset roughness threshold is set at 1.5 mm.

[0037] It should be noted that the flatness grade is determined using a range division method. When the standard deviation is less than 0.5 mm, it is judged as high flatness; when the standard deviation is between 0.5 and 1.5 mm, it is judged as moderate flatness; and when it exceeds 1.5 mm, it falls into the low flatness category.

[0038] Preferably, the combined feature encoding adopts a binary encoding method, with high flatness encoded as 10, moderate flatness encoded as 01, blunt extrusion fracture encoded as 100, and natural tear fracture encoded as 010, forming a six-bit code through bit operations.

[0039] In one possible implementation, the detection of machining marks focuses on the distribution characteristics of parallel lines and indentations. Parallel lines are analyzed using Fourier transform to examine the frequency domain characteristics of the cross-sectional image; the presence of obvious periodic peaks in the spectrum indicates the existence of regular parallel textures. Uniform indentation distribution is determined by calculating the standard deviation of the distance between adjacent indentations; a standard deviation less than a preset value is considered a uniform distribution, a typical characteristic of machining.

[0040] For example, when a combination of high flatness and blunt extrusion fracture is detected, and parallel lines or uniform indentations are also found, it is determined that the cigarette has undergone mechanical trimming and is marked as a suspected mechanical trimming sample, which then enters a special in-depth testing process; while a combination of moderate flatness and natural tear fracture is in line with normal production characteristics and directly enters the routine testing stage.

[0041] Step S104: For suspected mechanically repaired samples, the intensity of repair intervention and the repair coverage area are identified by the compression depth and surface smoothness of the blunt extrusion fracture surface, and the degree of human intervention is assessed.

[0042] For samples suspected of mechanical finishing, a laser rangefinder is used to scan the blunt extrusion fracture surface point by point, obtaining the compression depth value of each point on the fracture surface relative to a reference plane. The rate of change of the compression depth value from the center point to the edge point is calculated as the indentation gradient. Based on the indentation gradient and compression depth value sequence, the root mean square value of the height difference between adjacent scan points is calculated as the cross-sectional smoothness. When the root mean square value is lower than a preset threshold, the area is determined to have undergone fine grinding. The boundary coordinates of the grinding area are extracted and the grinding mark features are recorded. Using the grinding mark features and boundary coordinates, the ratio R of the texture roughness outside the boundary to the texture roughness inside the boundary is calculated to determine the finishing intensity, where... Ro represents the external roughness of the boundary, and Ri represents the internal roughness of the boundary. When R is greater than 3, it is considered heavy intervention; R between 2 and 3 is considered moderate intervention; and R less than 2 is considered light intervention. The percentage of the area of ​​the trimmed region relative to the entire cross-section is calculated using the trimming intensity and boundary coordinates as the trimmed coverage area. The intervention level assessment value is obtained by multiplying the trimming intensity by the coverage area. When the assessment value exceeds a preset intervention threshold, it is considered artificial deep trimming.

[0043] In one implementation, the depth detection of suspected mechanically repaired samples is achieved through multi-dimensional quantitative evaluation. A quantitative evaluation system for the degree of human intervention is established by comprehensively analyzing multiple characteristic parameters such as compression depth, cross-sectional smoothness, and repair intensity.

[0044] Specifically, laser ranging scanning employs the principle of triangulation to achieve high-precision depth measurement. A laser emitter projects a point-like laser beam onto the surface of the tobacco fracture. The laser beam is at a fixed angle to the normal to the fracture surface. When the laser irradiates the surface at different depths, the position of the reflected light shifts. A photoelectric sensor receives the reflected light and records its position on the sensor array, calculating the precise depth value of that point based on trigonometric relationships. The scanning process uses a grid-like path, with the laser beam scanning the entire fracture surface point by point according to a preset row and column spacing. The coordinates and depth value of each scanned point are recorded to form three-dimensional point cloud data. First, the lowest point of the fracture is identified as the origin of the reference plane. Then, the vertical distance of all scanned points relative to the reference plane is calculated, i.e., the compressed depth value. By interpolating the point cloud data, a continuous depth distribution map is constructed, where the grayscale value of each pixel in the map corresponds to the compressed depth at that location.

[0045] It should be noted that the compression depth value sequence reflects the undulation characteristics of the fracture surface. The depth values ​​of the central and edge regions of the fracture are extracted, and their average values ​​are calculated. The difference between the two values ​​reflects the attenuation trend of the compression degree from the center to the edge.

[0046] Preferably, the indentation gradient is calculated using a piecewise linear fitting method. A sequence of depth values ​​is extracted along a radial path from the center to the edge, the path is divided into multiple segments, and a linear fit is performed on the depth values ​​within each segment. The slope of the fitted line is the local gradient of that segment. The average of the gradients from all segments is taken as the overall indentation gradient, reflecting the spatial distribution characteristics of the compressive force.

[0047] In one possible implementation, the smoothness of the cross-section is evaluated based on the statistical characteristics of the surface's micro-undulations. A statistical analysis is performed on the height differences between adjacent scan points. First, the height difference for each pair of adjacent points is calculated, forming a height difference sequence. The sum of squares of this sequence is calculated, divided by the sequence length, and then the square root is taken to obtain the root mean square (RMS) value. The smaller the RMS value, the smoother the height change between adjacent points, and the smoother the cross-section. A preset threshold is typically set at 0.1 mm; surfaces below this value are considered finely polished.

[0048] For example, the identification of the polished area is achieved through an edge detection algorithm. The smoothness distribution map is binarized, and areas with smoothness below a threshold are marked as polished areas. The outer contour of the polished area is extracted as the boundary coordinates, and the directional features of the polishing marks are recorded.

[0049] Understandably, the determination of finishing intensity is based on a comparative analysis of texture roughness. The surface roughness of the interior and exterior of the polished area is calculated separately, with roughness obtained by calculating the standard deviation of depth values ​​within the local area. The ratio of the interior to exterior roughness reflects the degree to which the finishing process alters the surface texture. When the ratio is greater than 3, it indicates that the texture of the polished area has been severely flattened, classified as heavy intervention; a ratio between 2 and 3 indicates that some texture is preserved, classified as moderate intervention; and a ratio less than 2 indicates that the texture is largely preserved, classified as light intervention. This grading standard is based on statistical analysis of a large number of genuine and counterfeit cigarette samples and can accurately reflect the intensity level of mechanical finishing.

[0050] For example, the trimmed coverage area is calculated using pixel statistics, which involves counting the number of pixels in the trimmed area to the total number of pixels in the entire cross-section; the ratio of these two values ​​represents the coverage percentage. The intervention level assessment value is equal to the product of the trimming intensity value and the coverage percentage. Furthermore, this assessment method can quantitatively describe the impact of mechanical trimming, providing a reliable quantitative basis for counterfeit cigarette identification.

[0051] Step S105: Based on the degree of human intervention, and combined with the tobacco filling density and color uniformity in the inferior tobacco sample library, adjust the evaluation criteria for identifying genuine and counterfeit cigarettes to obtain the revised criteria for determining authenticity.

[0052] Based on the degree of human intervention, the tobacco filling density distribution range and color uniformity parameters corresponding to the intervention level assessment value are extracted from a database of inferior tobacco samples. The difference between the filling density and the standard genuine cigarette density is calculated and divided by the standard density to obtain the density deviation rate. The difference between the color uniformity and the standard color parameter is calculated and divided by the standard parameter to obtain the color difference rate. The density deviation rate is multiplied by a first weighting coefficient to obtain a density correction value, and the color difference rate is multiplied by a second weighting coefficient to obtain a color correction value. The density correction value, color correction value, and intervention level assessment value are added together to obtain a comprehensive score. If the comprehensive score exceeds a preset counterfeit cigarette threshold, the strictness of the authenticity determination is increased. The original judgment threshold is adjusted based on the comprehensive score. The corrected judgment threshold is obtained by adding the comprehensive score and the adjustment coefficient to the original threshold. When the weighted score of the five indicators of the cigarette detection parameters exceeds the corrected judgment threshold, it is judged as a counterfeit cigarette, thus obtaining the corrected criteria for authenticity determination.

[0053] In one implementation, the evaluation criteria for identifying genuine and counterfeit cigarettes are dynamically adjusted based on the degree of human intervention. A multi-dimensional correction mechanism is established by combining characteristic parameters from a database of inferior tobacco samples to achieve accurate identification of mechanically processed counterfeit cigarettes.

[0054] Specifically, the inferior tobacco sample library stores sample characteristics categorized by the degree of intervention. Mildly intervened samples typically have a filling density between 0.65 and 0.75 g / cm³, and a color uniformity parameter between 70 and 80; moderately intervened samples have a filling density between 0.55 and 0.65 g / cm³, and a color uniformity between 60 and 70; severely intervened samples have a filling density below 0.55 g / cm³, and a color uniformity below 60. Based on the degree of intervention of the cigarette to be tested, the corresponding parameter range is extracted from the sample library. The filling density of standard genuine cigarettes is set at 0.85 g / cm³, and the standard color parameter is 90; these standard values ​​are derived statistically from a large number of genuine cigarette samples.

[0055] It should be noted that the density deviation rate is calculated by dividing the difference between the measured density and the standard density by the standard density, reflecting the degree of deviation in the filler density. The color difference rate is calculated using a similar method, by dividing the difference between the measured color parameter and the standard parameter by the standard parameter.

[0056] Preferably, the weighting coefficients are dynamically set according to different levels of intervention. The first weighting coefficient controls the degree of influence of density deviation, set to 0.6 for heavy intervention, 0.4 for moderate intervention, and 0.2 for mild intervention. The second weighting coefficient controls the degree of influence of color difference, set to 0.4, 0.3, and 0.2 respectively. The correction values ​​are obtained by multiplying the weighting coefficients by the deviation rate.

[0057] In one possible implementation, the comprehensive score is obtained by adding the density correction value, color correction value and intervention degree assessment value to form a unified quantitative index, and the preset fake cigarette threshold is usually set to 2.5.

[0058] For example, if the original judgment threshold is 3.0, and the comprehensive score is 2.8, with an adjustment coefficient of 0.3, then the corrected judgment threshold is equal to 3.0 plus the product of 2.8 and 0.3, which is 3.84. When the cigarette detection parameter exceeds 3.84, the cigarette is judged to be counterfeit, thus realizing the function of dynamically adjusting the judgment standard according to the degree of mechanical repair.

[0059] Step S106: Using the revised authenticity criteria, comprehensively evaluate the flatness of the cigarette cut, surface roughness, burr distribution density, fracture type, and mechanical trimming marks to obtain the final authenticity identification result.

[0060] Using the revised authenticity criteria, the cigarette's cut smoothness, surface roughness, burr distribution density, fracture type, and mechanical trimming marks are scored on a 100-point scale. Each score is multiplied by its corresponding weighting coefficient and summed to obtain a comprehensive evaluation score. The original values ​​of each test item are recorded. Based on the comprehensive evaluation score, the identification result is determined: a score above 80 is considered genuine, below 60 is considered counterfeit, and scores between 60 and 80 are marked as samples requiring re-inspection. The identification result code, intervention intensity value, and trimming coverage percentage are concatenated in a fixed format to form a test report number. Using this test report number as the primary key, a test report file containing the test time, batch number, five test indicator values, comprehensive evaluation score, and identification result is created. The report is stored in a designated table in the database, and an association index with historical test records is established for subsequent comparison and verification queries.

[0061] In one implementation, the comprehensive evaluation of cigarette authenticity identification employs a multi-dimensional quantitative scoring mechanism. Five detection indicators are assigned scores, and a weighted sum is used to obtain a comprehensive evaluation score, achieving accurate authenticity determination.

[0062] Specifically, the percentage-based scoring system establishes scoring standards based on the characteristics of each test item. Cut smoothness is scored according to edge deviation: deviation less than 0.5 mm scores 90-100 points, 0.5-1 mm scores 70-90 points, and greater than 1 mm scores below 70 points. Surface roughness is scored based on the root mean square value; the smaller the value, the higher the score. Burr distribution density is based on the number of burrs per square millimeter; density within the standard range scores high, and deviations from the standard range result in decreasing scores. Fracture type is scored by category: natural tear fracture scores 80-100 points, blunt extrusion fracture scores 40-60 points, and mixed fracture scores 60-80 points. Mechanical finishing marks are scored inversely based on the visibility of the marks; no marks receive full marks, and more noticeable marks result in higher deductions. Each score corresponds to a specific numerical range to ensure the objectivity and repeatability of the scoring.

[0063] It should be noted that the weighting coefficients are set according to the contribution of each test item to the determination of authenticity. The fracture type has the highest weight, set at 0.3; mechanical finishing marks have a weight of 0.25; cut smoothness and burr distribution density each have a weight of 0.2; and surface roughness has a weight of 0.15.

[0064] Preferably, the three-level judgment criteria adopt a dual-threshold classification method. A comprehensive evaluation score above 80 points is judged as genuine cigarettes, indicating that all indicators of the sample meet the characteristics of genuine cigarettes; a score below 60 points is judged as counterfeit cigarettes, indicating obvious mechanical trimming or inferior material characteristics; samples with scores between 60 and 80 points are marked as samples to be re-inspected, requiring further manual review or precision instrument testing.

[0065] In one possible implementation, the test report number adopts a format of timestamp plus test result code, such as "20240315143025-T85-H2-C15", where T85 indicates that the cigarette was genuine and the score is 85, H2 indicates severe intervention, and C15 indicates that the coverage area was adjusted to 15%.

[0066] For example, the database uses a relational structure to store test reports. The main table contains basic information such as report number, test time, and batch number, while the slave tables store detailed values ​​for various test indicators. Data integrity is achieved through foreign key relationships.

[0067] This method acquires the flatness and surface roughness of the cut by acquiring images of the cigarette end face, and obtains the edge contour and texture distribution of the cut area through pixel-level segmentation. Based on depth scanning analysis of the distribution density of burrs at the end of the tobacco shreds and the fiber breakage morphology, the fracture type is identified as a blunt compression fracture or a natural tear fracture. When the cut flatness is high and the fracture type is a blunt compression fracture, it is marked as a sample suspected of mechanical trimming; when the cut flatness is moderate and the fracture type is a natural tear fracture, it is marked as a sample with natural cutting features. Samples with natural cutting features directly enter the conventional authenticity identification process, while samples suspected of mechanical trimming are assessed for the intensity of human intervention by analyzing the compression depth and cross-sectional smoothness of the blunt compression fracture, and the identification criteria are dynamically adjusted by combining the filling density and color uniformity of the inferior tobacco sample library. Finally, by comprehensively considering multi-dimensional features such as cut flatness, surface roughness, burr distribution density, fracture type, and mechanical trimming traces, accurate authenticity identification results are obtained.

[0068] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and additions without departing from the principle of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. A method for assessing the risk of counterfeit cigarettes, characterized in that, The method includes: Images of the cigarette end face are acquired to obtain the flatness of the cut and the surface roughness. The acquired images are then segmented at the pixel level to obtain the edge contour and texture distribution of the cut area. Based on the edge contour and texture distribution of the cut area, a depth scan of the tobacco cross-section is performed to obtain the distribution density of burrs at the end of the tobacco and the fiber fracture morphology, and to identify whether the fiber fracture type is a blunt extrusion fracture or a natural tear fracture. Based on the fiber fracture morphology and cut smoothness, the combination features of smoothness and fracture type are identified. When the cut smoothness is high and the fracture type is blunt extrusion fracture, it is marked as a sample suspected of mechanical trimming. When the cut smoothness is moderate and the fracture type is natural tear fracture, it is marked as a sample with natural cutting characteristics. Samples with natural cutting characteristics are directly entered into the regular genuine and counterfeit cigarette identification process for evaluation. For samples suspected of mechanical trimming, the intensity of trimming intervention and the trimming coverage area are identified by the compression depth and surface smoothness of the blunt extrusion fracture surface, and the degree of human intervention is assessed. Based on the degree of human intervention, and combined with the tobacco filling density and color uniformity in the inferior tobacco sample library, the evaluation criteria for identifying genuine and counterfeit cigarettes are adjusted to obtain the revised criteria for determining authenticity. By using the revised criteria for determining authenticity, a comprehensive evaluation is conducted on the cigarette's cut smoothness, surface roughness, burr distribution density, fracture type, and mechanical finishing marks to obtain the final result for authenticity identification.

2. The method for risk assessment of genuine and counterfeit cigarettes according to claim 1, characterized in that, The process of acquiring images of the cigarette end face to obtain the flatness and surface roughness of the cut, and performing pixel-level segmentation on the acquired images to obtain the edge contour and texture distribution of the cut area, includes: Within a preset focal length range, the cigarette end face is scanned layer by layer to obtain the image sharpness value of each focal plane. The focal plane with the highest sharpness value is selected as the optimal imaging plane, and a panoramic image of the cigarette end face is acquired on the optimal imaging plane. For the panoramic image, the micro-concavity and convexity depth of the end face is extracted to determine the roughness index. The roughness index is multiplied by a preset benchmark value to obtain an adaptive segmentation threshold. Threshold segmentation processing is performed on the panoramic image to obtain the pixel density distribution of the binarized image. Based on the pixel density distribution, a continuous coordinate point sequence of the cut edge is extracted as the edge contour. The texture contrast and correlation are calculated using the gray-level co-occurrence matrix as the texture distribution.

3. The method for risk assessment of genuine and counterfeit cigarettes according to claim 2, characterized in that, The step of extracting the micro-uneven depth of the end face and determining the roughness index for the panoramic image includes: projecting illumination beams from at least three different angles for the panoramic image, obtaining the gray-scale gradient value of the edge of the end face at each angle, determining the micro-uneven depth of the end face based on the variation of the gray-scale gradient value, and calculating the roughness index by the standard deviation of the uneven depth.

4. The method for risk assessment of genuine and counterfeit cigarettes according to claim 1, characterized in that, The process involves performing a depth scan of the tobacco cross-section based on the edge contour and texture distribution of the cut area to obtain the distribution density of burrs at the tobacco ends and the fiber breakage morphology, identifying the fiber breakage type as a blunt compression fracture or a natural tear fracture, including: Based on the edge contour and texture distribution of the cut area, the cross-section of the tobacco shreds is collected layer by layer within a preset scanning depth range to obtain the burr height distribution map of each layer. The burr distribution density is calculated by the ratio of the number of burrs to the scanned area in the burr height distribution map. For areas where the burr distribution density exceeds a preset density threshold, the fracture angle and tear length of the fibers in that area are extracted. If the fracture angle is less than a preset angle and the tear length exceeds a preset length, the area is determined to be a natural tear fracture. If the burr distribution density is less than a preset density threshold, the presence of compression marks on the cross-section is detected. The area is determined to be a blunt compression fracture by measuring the width and depth of the compression marks.

5. The method for risk assessment of genuine and counterfeit cigarettes according to claim 1, characterized in that, The method identifies the combination of flatness and fracture type based on fiber fracture morphology and cut smoothness. Samples with high cut smoothness and blunt compression fracture are marked as potentially mechanically trimmed. Samples with moderate cut smoothness and natural tear fracture are marked as naturally cut samples. Naturally cut samples directly enter the standard genuine / counterfeit cigarette identification process for evaluation, including: Based on the fiber fracture morphology and cut smoothness, the standard deviation is calculated by the deviation distance of the edge contour relative to the baseline. When the standard deviation is lower than the preset smoothness threshold, it is judged as high smoothness; when it is higher than the preset smoothness threshold but lower than the preset roughness threshold, it is judged as moderate smoothness. A combined feature code is generated based on the smoothness level and the fracture type. According to the combined feature code, when the code represents a combination of high smoothness and blunt extrusion fracture, the cigarette is marked as a suspected mechanically trimmed sample; when the code represents a combination of moderate smoothness and natural tear fracture, it is marked as a naturally cut feature sample.

6. The method for risk assessment of genuine and counterfeit cigarettes according to claim 1, characterized in that, For samples suspected of mechanical trimming, the intensity and coverage area of ​​trimming intervention are identified by the compression depth and surface smoothness of the blunt extrusion fracture surface, and the degree of human intervention is assessed, including: For samples suspected of mechanical finishing, the blunt extrusion fracture surface is scanned point by point to obtain the compression depth value of each point on the fracture surface relative to the reference plane. The rate of change of the compression depth value from the center point to the edge point is calculated as the indentation gradient. Based on the indentation gradient and the compression depth value sequence, the root mean square value of the height difference between adjacent scan points is calculated as the cross-sectional smoothness. When the root mean square value is lower than a preset threshold, it is determined that the area has undergone fine grinding. The boundary coordinates of the grinding area are extracted and the grinding trace features are recorded. Using the grinding trace features and boundary coordinates, the ratio of the texture roughness inside and outside the boundary is calculated to determine the finishing intensity. The percentage of the area of ​​the finished area to the entire cross-section is calculated using the finishing intensity and boundary coordinates as the finishing coverage area. The degree of human intervention is obtained by multiplying the finishing intensity and the coverage area.

7. The method for risk assessment of genuine and counterfeit cigarettes according to claim 1, characterized in that, The evaluation criteria for identifying genuine and counterfeit cigarettes are adjusted based on the degree of human intervention, combined with the tobacco filling density and color uniformity in the inferior tobacco sample library, resulting in a revised basis for determining authenticity, including: Based on the degree of human intervention, the distribution range of tobacco filling density and color uniformity parameters corresponding to the intervention level are extracted from the inferior tobacco sample library. The density deviation rate is obtained by dividing the difference between the filling density and the standard genuine cigarette density by the standard density. The color difference rate is obtained by dividing the difference between the color uniformity and the standard color parameter by the standard parameter. The density correction value is obtained by multiplying the density deviation rate by the first weighting coefficient, and the color correction value is obtained by multiplying the color difference rate by the second weighting coefficient. A comprehensive score is determined based on the density correction value, the color correction value, and the intervention level assessment value. The original judgment threshold is adjusted based on the comprehensive score to obtain the corrected judgment threshold.

8. The method for risk assessment of genuine and counterfeit cigarettes according to claim 1, characterized in that, The modified authenticity determination criteria comprehensively evaluate the cigarette's cut smoothness, surface roughness, burr distribution density, fracture type, and mechanical finishing marks to obtain the final authenticity identification result, including: Using the revised criteria for determining authenticity, the smoothness of the cigarette cut, surface roughness, burr distribution density, fracture type, and mechanical finishing marks are scored on a 100-point scale. Each score is multiplied by its corresponding weighting coefficient and then summed to obtain a comprehensive evaluation score. The result of identifying genuine and counterfeit cigarettes is determined based on the comprehensive evaluation score.

Images