A two-dimensional code tracing method and system for a cigarette case generation process

Abstract

The present application relates to machine-readable record carrier and anti-counterfeiting technical field, specifically relates to a kind of cigarette case generation process two-dimensional code traceability method and system, method includes: according to the inverse displacement compensation of substrate Young's modulus and expected compressive stress to code pixel;The dark lines are formed by embedding the edge of the image search pattern in the frequency domain modulation of gilding temperature gradient and die cutting pressure peak value;Based on local information entropy and paper substrate fiber texture gradient, reconstruct the missing pixels of the incomplete code image;The encryption binding factor is generated by nonlinear fusion of process parameters and paper substrate fiber texture in the production time window, and written into database;Synchronous acquisition of visible light and ultraviolet light code image and difference extraction intrinsic dark lines.The present application can improve the reading and tracing reliability of cigarette two-dimensional code in processing deformation and high light environment.

Description

Technical Field

[0001] This invention relates to the field of machine-readable recording media and anti-counterfeiting technology. More specifically, this invention relates to a QR code traceability method and system for cigarette box manufacturing processes. Background Technology

[0002] In the cigarette box manufacturing process, QR codes arranged in a standard orthogonal matrix are typically printed directly onto the paper substrate during the printing stage to support product traceability and anti-counterfeiting verification. However, the non-uniform mechanical stress and thermal effects applied to the substrate by subsequent hot stamping, die-cutting, and embossing processes can cause material deformation, resulting in non-negligible nonlinear displacement of the pre-printed QR code modules. This causes them to deviate from the standard sampling grid, leading to errors in finder image boundary determination and data module misalignment, significantly reducing the decoding success rate. Existing traceability systems mainly rely on network databases to associate code values ​​with production parameters, lacking the ability for the code images to carry their own process information. Once communication is interrupted or backend data is tampered with, the physical packaging loses its self-verification means and is unable to cope with label copying and tampering attacks.

[0003] Under high-speed production line acquisition conditions, issues such as motion blur, ink splatter, and localized overexposure frequently occur. Conventional error correction mechanisms have limited ability to recover from large-area or non-uniform pixel defects, making incomplete QR code images unreliable for reading. Furthermore, digital production parameters are at risk of being cloned or tampered with, while physically unclonable characteristics such as paper fiber texture are not effectively bound to product identification, making it difficult to eliminate the vulnerability of mass copying of anti-counterfeiting codes. Simultaneously, when using highly reflective materials such as aluminum foil and lamination, strong specular reflection can obscure the pixel features of the underlying QR code image, and traditional single visible light acquisition methods cannot effectively remove the reflective components to extract intrinsic information. Therefore, there is an urgent need for a QR code generation and traceability method that can pre-compensate for substrate deformation, solidify process features in the QR code image, stably reconstruct incomplete information, and integrate unclonable physical features with digital parameters for anti-counterfeiting binding, in order to comprehensively improve the robustness of QR code image reading and the reliability of anti-counterfeiting under complex process conditions. Summary of the Invention

[0004] In existing technologies, when using highly reflective materials such as aluminum foil or film, strong specular reflection can obscure the pixel features of the underlying code image. Traditional single visible light acquisition methods cannot effectively strip away the reflective component to extract intrinsic information. To address the above problems, this invention provides solutions in the following aspects.

[0005] In a first aspect, the present invention provides a QR code traceability method for cigarette box manufacturing process, which adopts the following technical solution: when generating the code image, the expected displacement of the pixel is calculated based on the Young's modulus of the substrate and the preset expected compressive stress distribution matrix, and reverse displacement compensation is applied to the pixel. Extract the hot stamping temperature spatial gradient and die-cutting pressure peak generated in real time during the hot stamping and die-cutting processes of the cigarette box. Normalize the hot stamping temperature spatial gradient and die-cutting pressure peak into feature vectors, and then use frequency domain modulation to steganographically write them into the edge brightness of the finder pattern of the code image to form dark patterns. After acquiring the code image, the neighborhood information entropy of the brightness abnormal area is calculated, and the missing pixels are reconstructed by interpolation with the fiber texture gradient of the paper substrate as a constraint. Determine the production time window covering the hot stamping and die-cutting processes, perform non-linear projection fusion of the equipment status parameters within the production time window with the fiber texture of the paper substrate, generate an encrypted binding factor, and write it into the database. During anti-counterfeiting verification, the system simultaneously triggers the acquisition of dual-band spectra using visible and ultraviolet light sources. By suppressing specular reflection components through pixel difference, the system extracts intrinsic dark texture features.

[0006] Preferably, the step of calculating the expected displacement of the pixel based on the Young's modulus of the substrate and a preset expected compressive stress distribution matrix, and applying inverse displacement compensation to the pixel, includes: The positive mechanical compressive stress in the target area is determined based on the expected compressive stress distribution matrix, and the normal distance vector from the current pixel to the center of force is obtained. The reverse displacement vector is calculated based on the substrate stress release coefficient, the positive mechanical compressive stress, the Young's modulus of the substrate, and the normal distance vector. The inverse displacement vector is applied to perform the inverse displacement compensation on the pixel.

[0007] Preferably, the substrate stress relief coefficient is updated in the following manner: The code images under visible light and ultraviolet light, which were collected synchronously during the anti-counterfeiting verification in historical batches, are obtained, and after registration, pixel difference is performed to extract the signal-to-noise ratio of the intrinsic dark pattern features. The stress release coefficient of the substrate is adjusted based on the dynamic feedback of the signal-to-noise ratio.

[0008] Preferably, the step of normalizing the hot stamping temperature spatial gradient and die-cutting pressure peak into a feature vector, and then using frequency domain modulation to steganographically embed it into the edge brightness of the finder pattern of the code image to form a dark pattern, includes: Obtain the preset spatial modulation frequency and dark pattern embedding intensity factor; Determine the relative arc length coordinates of each pixel on the edge contour of the finder pattern of the code image; Based on the normalized feature vector, the dark pattern embedding intensity factor, the spatial modulation frequency, and the relative arc length coordinates, the edge brightness of the finder pattern of the code image is modulated to generate modulated edge brightness carrying dark patterns.

[0009] Preferably, the step of calculating the neighborhood information entropy of the brightness anomaly region and interpolating to reconstruct the missing pixels using the fiber texture gradient of the paper substrate as a constraint includes: Determine the outer annular neighborhood of the region with abnormal brightness; Calculate the local spatial information entropy of the ring neighborhood, and use it as the neighborhood information entropy; Obtain the fiber texture gradient of the paper substrate; Based on the local spatial information entropy and the fiber texture gradient of the paper substrate, the brightness of the effective pixels in the annular neighborhood is weighted and interpolated to reconstruct the grayscale value of the missing pixels.

[0010] Preferably, the equipment status parameters within the production time window are nonlinearly projected and fused with the fiber texture of the paper substrate to generate an encryption binding factor, including: Preset cross-domain fusion balance coefficient and time-series data normalization scaling coefficient; Obtain the production parameter vector formed by the equipment status parameters within the production time window; Obtain the local gradient of the fiber texture of the paper substrate; The nonlinear projection fusion is performed based on the cross-domain fusion balance coefficient, the time-series data normalization scaling coefficient, the production parameter vector, and the local gradient of the paper substrate fiber texture to generate the encryption binding factor.

[0011] Preferably, the simultaneous triggering of visible light and ultraviolet light sources to acquire dual-band spectra, and after registration, suppressing specular reflection components through pixel difference to extract intrinsic dark pattern features, includes: obtaining the visible light band reflectance matrix and the ultraviolet band reflectance matrix from the registered code image; Determine the adaptive weights for specular suppression and the optical minimum constant; Based on the visible light band reflectance matrix and the ultraviolet band reflectance matrix, pixel difference operation is performed to eliminate specular reflection and obtain intrinsic dark texture features.

[0012] Preferably, before generating the code image, the method further includes: acquiring the fiber texture distribution of a specific area of ​​the paper substrate using a high-precision line scan camera, as the fiber texture of the paper substrate.

[0013] Secondly, this invention provides a QR code traceability system for cigarette box manufacturing processes, employing the following technical solution: A QR code traceability system for cigarette box manufacturing processes includes: a QR code generation compensation module, a process dark pattern embedding module, a defective pixel reconstruction module, a binding factor generation and storage module, a traceability database, and a dual-band verification module; the QR code generation compensation module is configured to calculate the expected displacement of pixels based on the Young's modulus of the substrate and the expected compressive stress distribution, and apply reverse displacement compensation to the pixels of the code image to be generated; the process dark pattern embedding module is configured to normalize the hot stamping temperature gradient and die-cutting pressure peak into feature vectors, and embed them into the edge brightness of the code image seeker graphic through frequency domain modulation to form dark patterns; the defective pixel reconstruction module is configured to... After collecting the code image, the neighborhood information entropy of the brightness abnormal area is calculated, and the missing pixels are reconstructed by interpolation with the fiber texture gradient of the paper substrate as a constraint. The binding factor generation and storage module is configured to perform nonlinear projection fusion of the equipment status parameters in the corresponding production time window with the fiber texture of the paper substrate to generate encrypted binding factors and write them into the traceability database. The dual-band verification module is configured to simultaneously collect the code image under visible light and ultraviolet light during verification, and perform pixel difference after registration to eliminate specular reflection and extract intrinsic dark patterns. Among them, the QR code generation compensation module, the process dark pattern embedding module, the missing pixel reconstruction module, the binding factor generation and storage module, and the dual-band verification module work together to execute the QR code traceability method for cigarette box production process described in this application.

[0014] The embodiments of the present invention have at least the following beneficial effects: (1) This application calculates the inverse displacement vector by using the Young's modulus of the substrate, the expected compressive stress distribution and the dynamically updated stress release coefficient, and applies an inverse affine transformation to the pixel to pre-compensate the non-uniform mechanical deformation of the hot stamping and die-cutting processes, so that the position of the QR code module after processing returns to the standard orthogonal grid tolerance range, maintains the integrity of the image seeker structure and data module, and improves the first decoding success rate.

[0015] (2) This application normalizes the hot stamping temperature gradient and the peak of the die-cutting pressure into a process feature vector, and uses frequency domain modulation coding to encode the cosine fluctuation dark pattern of the edge brightness of the finder image, so that the QR code physically carries the process features, provides auxiliary verification under communication-limited conditions, and blocks the counterfeiting behavior of simply copying the code value.

[0016] (3) This application identifies the missing area through joint detection, and reconstructs the missing pixels by weighted interpolation with the paper-based fiber texture gradient as a constraint. It follows the statistical law of the neighborhood and the direction of physical texture, ensures the robustness of high-speed production line reading, and reduces the false rejection rate.

[0017] (4) This application generates a cross-domain encryption binding factor by nonlinearly projecting and fusing the equipment status parameters within the production time window with the local gradient of the paper-based fiber texture, thereby coupling the digital time sequence parameters with the physical non-cloning features, increasing the difficulty of batch copying anti-counterfeiting codes, and enhancing the anti-cloning strength of the anti-counterfeiting system.

[0018] (5) This application acquires dual-band visible and ultraviolet light spectra simultaneously and performs pixel-level differential operations. It utilizes the high reflectivity of the ultraviolet channel to the surface varnish and the difference in absorption of visible light to the underlying ink to suppress specular reflection, extract intrinsic dark texture features, and improve the resolvability of anti-counterfeiting features under extreme reflective conditions. Attached Figure Description

[0019] Figure 1 This is a flowchart of steps S1-S5 in a QR code traceability method for a cigarette box manufacturing process according to an embodiment of this application. Detailed Implementation

[0020] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] S1: When generating the code image, calculate the expected displacement of the pixel based on the Young's modulus of the substrate and the expected compressive stress distribution, and apply reverse displacement compensation to the pixel.

[0023] This step calculates the expected displacement based on the Young's modulus of the substrate and the expected compressive stress, and applies an inverse displacement affine transformation to the QR code pixels to counteract subsequent processing deformation.

[0024] During the post-processing steps of cigarette boxes, such as hot stamping, die-cutting, and embossing, the paper substrate is subjected to local mechanical compressive stress and thermal effects, resulting in varying degrees of elastic deformation and residual plastic deformation.

[0025] If the QR code is printed directly according to the standard orthogonal grid during the coding stage, the deformation of the substrate after processing will cause the printed QR code module to deviate from the original design coordinates.

[0026] The offset of the QR code module near the center of the die-cutting force is particularly noticeable, and the measured value can reach tens to hundreds of micrometers. The tolerance of the position of the QR code module of the standard QR code image sensor is usually within one-fifth of the size of a single module. The deformation deviation is very likely to exceed this tolerance range, resulting in incorrect determination of the image pattern boundary and misalignment of the data module sampling, which in turn causes decoding failure.

[0027] To address the issue of QR code module offset caused by substrate deformation, this step introduces a pre-compensation mechanism for processing deformation during the QR code generation stage.

[0028] Based on the Young's modulus of the current batch of cigarette box base material and the expected compressive stress distribution of the die-cutting equipment, the expected displacement of each pixel of the QR code in the subsequent process is calculated one by one, and an inverse displacement vector with the same magnitude and opposite direction as the expected displacement is applied to the generated coordinates of each target pixel, and the inverse affine transformation of the coordinates is performed.

[0029] After this pre-distortion process, when the substrate is subjected to pressure and undergoes physical deformation in subsequent processes, the deformation displacement will cause the QR code module to return to the theoretical position of the standard orthogonal grid, thereby offsetting the damage to the code image structure caused by the processing deformation.

[0030] When obtaining the basic parameters required for the calculation, first determine the Young's modulus of the substrate in the current batch. .

[0031] Derived from the statistical mean of mechanical tensile or three-point bending tests on batches of samples of the same material, it characterizes the ability of the cigarette box paper substrate to resist elastic deformation, and the unit is Pascal (Pa).

[0032] For common cigarette box white cardboard with a basis weight of 230g / m² to 300g / m², Usually located Pa to Within the Pa interval, this embodiment uses empirical values. Pa is a preset parameter, which can also be adjusted by the implementer based on the actual measured data of the substrate.

[0033] At the same time, the expected compressive stress distribution matrix preset by the die-cutting equipment is read.

[0034] This matrix uses the same spatial coordinate system as the QR code pixel matrix to give the positive mechanical compressive stress at each coordinate position during the die-cutting process. The unit is Pascal (Pa).

[0035] The pressure setting of the die-cutting machine, the contact area of ​​the die plate, and the pressure transmission efficiency are converted, and a compressive stress distribution matrix covering the entire code pattern area is generated by interpolation according to a two-dimensional grid.

[0036] For the local embossing force applied in the embossing process, the corresponding compressive stress matrix is ​​established in the same way and superimposed on the die-cutting compressive stress matrix to obtain the comprehensive compressive stress distribution.

[0037] Subsequently, the pixel index coordinates located in the deformation-sensitive region of the QR code pixel matrix are extracted.

[0038] The basis for determining sensitive areas is the normal distance of a pixel from the center of the die-cutting force. .

[0039] It should be noted that pixel coordinates in a digital image matrix are usually expressed as pixel index coordinates. This means that before calculating the distance, the pixel coordinates need to be determined according to the printing resolution. Convert to physical space coordinates Set the physical coordinates of the die-cutting force center as follows: Then calculate the current pixel. Distance from the center of force The normal distance. Set a distance threshold. In this embodiment, an empirical value of 50 mm is used as a preset parameter, which can also be adjusted by the implementer according to the size of the substrate; when a certain pixel point When this happens, the pixel is added to the set of deformation-sensitive pixels, and inverse displacement compensation is performed; for For each pixel, since the compressive stress decreases inversely with distance within the substrate plane, the expected displacement is already below the code image positioning tolerance, so no compensation is required to reduce the computational load.

[0040] For each pixel in the set of deformation-sensitive pixels, its expected displacement is calculated based on the plane strain assumption of mechanics of materials.

[0041] Under plane strain conditions, the deformation tensor of the stress region of the substrate simplifies to the normal mechanical compressive stress. Young's modulus of the substrate The principal strain components are jointly determined.

[0042] The principal strain characterizes the amount of compression per unit length of the substrate in the direction of compressive stress. Multiplying it by the normal distance of the pixel from the center of force yields the theoretical elastic deformation displacement.

[0043] Based on this, a substrate stress release coefficient is introduced. Correcting the springback and residual stress release effects of the substrate after processing and unloading, the inverse displacement vector is obtained. The calculation formula is as follows: In the formula, This is the inverse displacement vector, in millimeters (mm), representing the amount of inverse spatial displacement applied to the target pixel, with a positive sign. The outward direction ensures that the applied direction is opposite to the expected deformation direction, making the pre-distortion and processing deformation inverse operations; This is the substrate stress relief coefficient, dimensionless, representing the ratio of residual deformation rebound to theoretical elastic deformation after compressive stress unloading. (In cases of initial coding or insufficient historical batch data...) Get experience points Dimensionless, and can also be adjusted by the implementer based on the rebound characteristics of the substrate; during continuous production, the signal-to-noise ratio of the intrinsic dark texture features obtained by dual-band differential extraction of completed batches and the QR code decoding success rate are dynamically adjusted through a preset feedback algorithm. When the signal-to-noise ratio of the intrinsic dark pattern features decreases compared to the previous batch and the decoding success rate decreases, it indicates that the residual deformation has not been fully compensated. Decrease the preset adjustment step size; conversely, maintain or increase it. ; The positive mechanical compressive stress at the target pixel location, in Pascals (Pa), is obtained from the composite compressive stress distribution matrix by pixel index coordinates; This represents the Young's modulus of the substrate, expressed in Pascals (Pa).

[0044] This is the normal distance vector from the current pixel to the center of the die-cutting force, in millimeters (mm). Its direction is from the center of the force to the coordinates of the current pixel, and its magnitude is the straight-line distance from the pixel to the center of the force.

[0045] Obtain the inverse displacement vector of each pixel. Then, during the encoding stage, the original coordinates of this pixel are... Perform the inverse affine transformation.

[0046] The translation component of this affine transformation is... Given the pre-distorted pixel coordinates The calculation is as follows: According to the pre-distortion coordinates The optical density distribution matrix of the generated QR code is used to print the QR code pixels on the substrate with a preset offset that is opposite to the direction of subsequent deformation.

[0047] After the substrate undergoes processes such as die-cutting and embossing, the mechanical compressive stress causes the substrate to compress and deform along the original force direction. The displacement of this compressive deformation is similar in amplitude and opposite in direction to the displacement of the pre-distortion. After the two are superimposed, the actual physical position of the QR code module returns to the coordinates close to the standard orthogonal grid. The boundary of the code image seeker and the data module after processing are still within the expected tolerance range of the reader's sampling grid.

[0048] After this step is completed, the output is a coordinate matrix of the QR code generated after reverse displacement compensation. Each element in the matrix records the precise two-dimensional coordinates of the corresponding pixel in the pre-distortion state, in millimeters (mm).

[0049] The matrix is ​​directly output to the printing plate making or laser marking control unit to generate physical QR codes with deformation pre-compensation capabilities.

[0050] After the above processing, the deviation between the physical position of the processed QR code module and the theoretical position of the standard orthogonal grid is controlled within one-fifth of the size of a single module, and the structural integrity of the finder pattern and data module is maintained.

[0051] S2: Extract the hot stamping temperature gradient and die-cutting pressure peak generated in real time during the hot stamping and die-cutting processes of the cigarette box, normalize them into feature vectors, and then use frequency domain modulation to steganographically write them into the brightness signal of the finder image edge of the code image.

[0052] The hot stamping temperature gradient and die-cutting pressure peak generated in real time during the hot stamping and die-cutting processes of cigarette boxes directly reflect the physical process conditions that the product undergoes.

[0053] If these parameters are only stored in an online database, the QR code itself does not carry process information, and there is a risk of separation between physical packaging and cloud data.

[0054] On the other hand, the lookup image of a QR code consists of a black module against a white background, with sharp brightness abrupt changes at the edges of the module. The human eye is not sensitive to the minute brightness modulations in this edge region, while industrial image sensors can distinguish grayscale differences within a local dynamic range.

[0055] This step normalizes the above process parameters and then uses frequency domain modulation to embed them into the brightness signal at the edge of the finder pattern, so that the code pattern itself carries the process characteristics.

[0056] First, obtain the hot stamping temperature spatial gradient sequence and die-cutting pressure peak value corresponding to the current cigarette box from the production line.

[0057] In the hot stamping process, multiple thermocouple probes are arranged at equal intervals along the axial direction of the hot stamping roller, with a sampling frequency... The system records the temperature values ​​at each measuring point in real time. After each hot stamping cycle, it calculates the gradient value of the temperature distribution along the axial direction and takes the root mean square of all gradients within that cycle to obtain the representative value of the temperature gradient. The unit is degrees Celsius per millimeter (°C / mm).

[0058] peak die-cutting pressure The maximum pressure reading is taken from the hydraulic or piezoelectric sensor of the die-cutting machine during the die-cutting closed-loop operation, and the unit is megapascal (MPa).

[0059] To eliminate dimensional differences, the following normalization process is adopted: Let The maximum value for the temperature gradient reference range is 5.0 ℃ / mm, taken as an empirical value according to the equipment process specifications. This is a preset parameter, but can also be adjusted by the implementer based on actual operating conditions. The rated output pressure of the die-cutting equipment is taken as an empirical value of 20.0 MPa, which is a preset parameter.

[0060] Calculation of dimensionless temperature characteristics and dimensionless pressure characteristics Together, they constitute a two-dimensional process feature vector. .

[0061] Furthermore, the modulation values ​​of the process characteristics are obtained by weighted summation. ,in and The weights for merging temperature and pressure information are all empirical values. , dimensionless, is a preset parameter.

[0062] As the amplitude control factor for frequency domain modulation, its value range is... .

[0063] Then, the edge contour set of each finder graphic in the QR code is locked.

[0064] In 2D barcode standards, the finder pattern has fixed geometric dimensions and shape specifications. Given the pre-distorted coordinate matrix, the precise position sequence of edge pixels can be calculated based on the theoretical outline of the finder pattern.

[0065] For each seeker image, a corner point on the contour is selected as the starting point of the arc length. The Euclidean distance between adjacent edge pixels is accumulated in a counterclockwise direction along the boundary of the image to obtain the relative arc length coordinates of each edge pixel. The unit is millimeters (mm). This arc length coordinate uniquely maps the spatial phase of the edge pixel in the modulated signal.

[0066] After completing the above preparations, perform brightness fine-tuning on each pixel on the edge contour.

[0067] In the original QR code optical density distribution, the standard binarized brightness of the edge pixels of the finder pattern The value is 0 in the black module position and 255 in the white background position, which is a dimensionless gray level.

[0068] This step involves superimposing a cosine-based modulation term onto the original brightness, slightly increasing the edge brightness near the black areas and slightly decreasing it near the white areas, creating a dark pattern that continuously varies along the edge. The modulated grayscale distribution is then halftone screened to ensure that the standard module proportions of the finder pattern are not disrupted. Preferably, the halftone screening frequency is at least 5 times the spatial modulation frequency (i.e., 2.5 mm). -1 and not less than 6 mm -1 (Approximately 150 lpi); The demodulation end uses a low-pass filter with a cutoff frequency of (2~3) times the space modulation frequency to filter out grid noise.

[0069] Modulated pixel brightness Determined by the following formula: In the formula, The modulated pixel brightness is a dimensionless gray level, representing the edge brightness value of the embedded process feature to be output. The original binary brightness, dimensionless gray level; The intensity factor for embedding dark patterns is determined by taking an empirical value. The dimensionless gray level is a preset parameter that controls the maximum brightness shift of the modulation signal in the spatial domain. It can also be adjusted by the implementer according to the tonal characteristics of the substrate ink. The aforementioned process characteristic modulation value is dimensionless and integrates the normalized characteristics of the hot stamping temperature spatial gradient and the die-cutting pressure peak. The preset spatial modulation frequency is taken as an empirical value of 0.5 mm. -1 The preset parameters determine the period of brightness fluctuation along the edge contour of the dark pattern, and can also be adjusted by the implementer according to the size of the image seeker pattern and the resolution of the image sensor. The coordinates of the relative arc length of the edge pixel on the outline of the seeker pattern are given in millimeters (mm).

[0070] For the calculated Clamping is performed according to the grayscale range [0, 255]. Values ​​exceeding the lower limit are set to 0, and values ​​exceeding the upper limit are set to 255.

[0071] After the above processing and halftone halftone conversion, the edge transition area of ​​the finder image is attached with a brightness modulation signal that uniquely corresponds to the process parameters, forming a QR code optical density distribution matrix carrying process characteristics.

[0072] This matrix, in conjunction with the inversely compensated pixel coordinates, is output to the laser marking or offset printing plate-making unit to generate a physical code image. While maintaining the original positioning and graphic recognition capabilities, the seeker image in the code image also contains normalized information of the hot stamping temperature spatial gradient and die-cutting pressure peak value solidified in its edge dark lines, possessing the ability to carry physical process characteristics.

[0073] S3: After acquiring the code image, calculate the neighborhood information entropy for the brightness abnormal area, and reconstruct the missing pixels by interpolation with the fiber texture gradient of the paper substrate as a constraint.

[0074] For the collected incomplete code image, the incomplete region is identified, the local information entropy deviation of the surrounding neighborhood is calculated, and the missing pixels are reconstructed by bidirectional weighted interpolation in combination with the paper-based fiber texture gradient constraint.

[0075] During high-speed production line operation, the cigarette box code images captured by the image sensor often exhibit incomplete areas with abnormal brightness or loss of contrast due to localized surface reflections, ink splatter, or motion blur. The pixels within the incomplete areas cannot provide effective decoding information, directly causing recognition failure. This problem constitutes the focus of this step.

[0076] First, the acquired image of the code to be verified is converted to grayscale to obtain a size of [size missing]. The grayscale matrix has a grayscale range of [0, 255].

[0077] Use one A sliding window scans the entire image, calculating the information entropy of the grayscale values ​​within the window. .

[0078] Set an information entropy anomaly threshold In this embodiment, an empirical value of 4.0 bits is used as a preset parameter, which can also be adjusted by the implementer based on the dynamic range of the image sensor. When the window information entropy is lower than... Furthermore, if the average brightness of the window deviates from the average brightness of the entire image by more than twice the standard deviation, the central area of ​​the window is marked as a brightness abnormality area.

[0079] Subsequently, morphological opening operations and connected component analysis were performed on the labeled results to merge adjacent abnormal pixels and close the holes, forming connected incomplete regions. Each The set of pixel locations to be repaired is defined.

[0080] For each missing area Its outer contour expands outward. Each pixel generates a circular neighborhood. This embodiment expands the width... The default parameter is set to 3 based on experience, but it can also be adjusted by the implementer according to the size of the damaged area.

[0081] Calculate the local spatial information entropy of the gray values ​​of all pixels within the annular neighborhood. This is used as the neighborhood information entropy. Simultaneously, a background region in the code image that is intact and has a smooth texture is selected as a reference, and its grayscale information entropy is calculated. Local information entropy deviation is defined as .

[0082] The degree of grayscale disorder in the incomplete outer region relative to the normal region was quantified, with the unit being bits.

[0083] At this point, the stored fiber texture gradient distribution map is retrieved from the paper-based fiber texture feature database corresponding to the current cigarette box product. This distribution map was generated by acquiring the substrate texture using a high-precision line scan camera and performing gradient calculations before encoding. Each pixel location records the local optical reflectance gradient amplitude caused by the paper-based fiber texture, which is dimensionless. Since the paper-based fiber texture is a physically non-clonable feature, its distribution is spatially unique, and this information provides additional physical constraints for the restoration of incomplete pixels.

[0084] For the incomplete area Each missing pixel in Obtain all valid pixels within its annular neighborhood. , , This represents the total number of valid pixels within the circular neighborhood.

[0085] Extract the grayscale value of each valid pixel. and the gradient magnitude at the corresponding position in the texture gradient map .

[0086] Calculate each valid pixel For missing pixels The weighted interpolation weights are obtained by combining the product of local information entropy bias and paper-based fiber texture gradient using a negative exponential function mapping.

[0087] Specifically, the normalized weights and the grayscale reconstruction values ​​of the missing pixels... Determined by the following formulas: In the formula, To reconstruct pixel brightness, a dimensionless gray level is used, ranging from [0, 255].

[0088] The aforementioned local information entropy deviation, expressed in bits, reflects the grayscale uncertainty of the incomplete neighborhood. Valid pixels within the circular neighborhood The amplitude of the paper-based fiber texture gradient at the location is dimensionless. Effective pixels within the circular neighborhood The original grayscale value is dimensionless; Index Term As an unnormalized weight, this suppresses the contribution of a pixel to missing pixels in neighborhoods with high information entropy deviations or abrupt changes in fiber texture gradients, thus preventing unreliable edge noise information from being transmitted to the missing region. Normalized denominator This ensures that the sum of all weights is 1, forming a complete weighted average.

[0089] The interpolation process is jointly driven by the local information entropy deviation and the physical texture gradient, constraining and reconstructing the missing pixels from multiple directions in the annular neighborhood.

[0090] The above reconstruction operation is performed sequentially on each missing pixel in the incomplete region until the grayscale values ​​of all missing pixels are filled. After reconstruction, the grayscale distribution of pixels in the incomplete region smoothly transitions with the surrounding normal region, restoring a continuous and decodable QR code optical density matrix.

[0091] This step outputs the reconstructed complete code map matrix.

[0092] S4: Determine the production time window covering the hot stamping and die-cutting processes, perform nonlinear projection fusion of the equipment status parameters within the corresponding production time window with the fiber texture of the paper substrate, generate an encrypted binding factor, and write it into the database.

[0093] During cigarette box production, real-time process parameters for hot stamping and die-cutting, including temperature and pressure waveforms, are digitally recorded by the manufacturing execution system. This type of digitized process parameter data is essentially a time-series signal, highly reproducible. If directly used as anti-counterfeiting or traceability evidence, attackers could intercept and replay historical data to achieve replication. Simultaneously, the paper substrate used for cigarette boxes has randomly interwoven fibers during its papermaking process, forming a microscopic undulating texture on the surface. This physical characteristic is non-clonable, but relying solely on texture features for full-image matching requires high computational and storage overhead, and may introduce instability under conditions of high reflectivity or ink coverage.

[0094] Therefore, this step performs cross-domain nonlinear fusion of the timing process parameters and the paper-based fiber texture to generate an encrypted binding factor that simultaneously relies on easily obtainable process digital features and non-replicable physical features. This encrypted binding factor is then stored in the traceability database, giving the product anti-counterfeiting information a dual constraint of digital and physical data.

[0095] Specifically, the first step is to determine the production time window. This is based on the moment when the QR code on the current cigarette pack is completed. As the endpoint, trace back a preset duration. , The empirical value of 5.0, in seconds (s), is a preset parameter, but can be adjusted by the implementer based on production line cycle time and process duration. This production time window covers the critical changes in equipment status during the hot stamping and die-cutting processes for this product.

[0096] Synchronously extract the process parameter matrix within the production time window from the production line data acquisition and monitoring system. The rows of the matrix correspond to different types of process parameters, such as the temperature of the hot stamping roller. The units are degrees Celsius (°C) and die-cutting pressure. The unit is Pa; the columns of the matrix correspond to equally spaced sampling times, and the sampling period in this embodiment is... Take the empirical value of 20, in milliseconds (ms).

[0097] thus, The real-time dynamic waveforms of the above process parameters were fully recorded.

[0098] To extract low-dimensional features that can stably represent the process state from a high-dimensional time series matrix, the statistical features of each row of the matrix, i.e., the time series of each type of process parameter, are calculated.

[0099] Select time series mean With time series standard deviation As basic feature quantities, they constitute the feature vector. , of which Each component ,in and The weights for the mean and standard deviation are both empirical values. , dimensionless, is a preset parameter.

[0100] therefore It is a feature vector that contains the timing characteristics of each process parameter, and each component still retains the original physical quantity dimensions.

[0101] To eliminate dimensional differences and unify numerical scales, a normalization scaling factor vector for time series data is introduced. Its dimensions and same.

[0102] For the One parameter, ,in This is half the width of the historical statistical fluctuation range of this parameter under normal operating conditions, and has the same unit as the original physical quantity; therefore The unit of the product is the reciprocal of the original physical quantity's unit, thus ensuring the inner product... It is a dimensionless scalar.

[0103] Will and By performing an inner product, we obtain the dimensionless comprehensive eigenvalues ​​of the process parameters. This provides a scalar whose value comprehensively reflects the combined fluctuation level of multiple process parameters within the corresponding production time window.

[0104] Simultaneously, the fiber texture features of the paper substrate, acquired by the front-end sensing node, are read. At the front end, before the cigarette box substrate enters the printing unit, a high-precision line scan camera captures a micro-texture image of a predetermined area of ​​the substrate, with a resolution of at least 1200 dpi. After converting the texture image to grayscale, the Sobel operator is applied to calculate the gradient magnitude of each pixel, and the root mean square of the gradient magnitudes across the entire image is calculated to obtain the original texture intensity. The unit is gray level per millimeter.

[0105] To compensate for fluctuations in light intensity and differences in camera system gain, a reference gradient magnitude is introduced. In this embodiment, an empirical value of 200 gray levels per millimeter is used as a preset parameter, but it can also be adjusted by the implementer according to the specific camera parameters.

[0106] Specifically, texture features are normalized into dimensionless local gradient representatives. : It characterizes the degree of spatial variation in optical reflectivity caused by micro-undulations on the paper substrate surface. The higher the value, the richer the texture and the stronger the physical distinguishability.

[0107] In obtaining and Then, cross-domain nonlinear fusion is performed. Specifically, the fusion calculation formula can be: In the formula, This is a dimensionless encryption binding factor, used as the final anti-counterfeiting verification value written to the database. The cross-domain fusion balance coefficient is obtained through offline model training and optimization. It is used to control the weight ratio of time-series digital features and physical space features in the binding factor. In this embodiment, an empirical value of 0.5 is used. It is dimensionless and is a preset parameter, but it can also be adjusted by the implementer according to the application scenario. Using the natural logarithm function right Implement a nonlinear mapping. Since the natural logarithm function can be expanded into a polynomial series, this algebraic projection fusion is essentially a specific form of nonlinear polynomial mapping; When process parameters fluctuate slightly within the normal operating range, logarithmic mapping makes... It produces sensitive discrimination; when parameters deviate significantly due to abnormal events, the logarithmic compression effect can prevent the binding factor from saturating and maintain [the desired performance]. The dynamic range; Linear terms The paper-based fiber texture is directly injected as an unclonable component into the encryption binding factor. This combination method ensures that even if an attacker completely replicates the normal process parameters, as long as the substrate texture is inconsistent with the original registered texture, the result calculated by the above formula will still be valid. This will result in a significant deviation from the database records.

[0108] Finally, the encryption binding factor Associate it with the unique serial number ID of the current cigarette pack product, and include... The records are written to the traceability database. The product serial number can be obtained by decoding the QR code data area or generated by the production system.

[0109] After this step is completed, the encryption binding factor will be output. This completes cross-domain encrypted binding based on process parameters and physical textures.

[0110] S5: During anti-counterfeiting verification, the system simultaneously triggers the acquisition of dual-band spectra using visible light and ultraviolet light sources. It then suppresses specular reflection components through pixel difference and extracts intrinsic dark texture features.

[0111] During the anti-counterfeiting verification stage, the aluminum foil coating or high-gloss varnish layer on the surface of the cigarette box to be tested produces strong specular reflection under normal visible light illumination. Its brightness is much higher than the dark texture feature signal formed by the underlying ink, and the dark texture signal is submerged in the directly acquired image. In order to reliably extract the intrinsic dark texture of the underlying layer, this step uses visible light and ultraviolet light sources to achieve dual-band synchronous acquisition through an optical path beam splitter, or alternately triggers acquisition through a strobe controller within the same exposure cycle and combines motion compensation for registration, and suppresses specular reflection components through pixel-level differential operations.

[0112] The anti-counterfeiting verification data acquisition terminal integrates both white LED and ultraviolet LED light sources, which are alternately illuminated within the same exposure cycle via a synchronous controller. When the visible light source is illuminated, the camera acquires a visible light band reflectance image. The image contains both the specular reflection component of the cigarette box surface and the dark texture information formed by the absorption of the underlying ink. When the ultraviolet light source is illuminated, the camera acquires an image of the ultraviolet reflectance. Because ultraviolet light has high reflectivity to the surface varnish layer and high absorption to the underlying ink, The main focus is on recording the distribution of specular reflections caused by surface varnish and coating. Both images have been converted to grayscale, with a grayscale range of [0, 255], are dimensionless, and have the same image size. .

[0113] Because of slight shifts in the physical positions of the two light sources during the acquisition process, pixel-level misalignment may occur between the two images. Before the differential calculation, the corner points of the QR code image seeker are used as control points to... Perform an affine transformation to make it compatible with Spatial registration and alignment. Registration employs a well-known normalized cross-correlation matching algorithm, which will not be elaborated upon here. After registration, for each pixel position in the image... Pixel difference operations are performed directly to separate intrinsic dark pattern features. , specifically, The calculation formula can be: In the formula, In the intrinsic dark pattern feature diagram The gray value at the location is in dimensionless gray level, representing the optical density distribution characteristics of the underlying ink after filtering out specular reflection components.

[0114] The differential intensity adjustment coefficient is taken as an empirical value in this embodiment. , dimensionless, is a preset parameter used to control the overall amplitude of the ultraviolet channel compensation term to match the actual light intensity ratio of the two light sources; This is for ultraviolet channel compensation. Because ultraviolet light has high reflectivity to surface varnish... It directly characterizes the intensity distribution of the specular reflection component. This is achieved by analyzing visible light images. Subtract the coefficient of longitude from the middle Adjusted UV image It can directly strip away the specular reflection component of the surface and extract the intrinsic dark texture features determined by the underlying ink. .

[0115] right Bright area noise suppression is performed across the entire image. A preset bright area threshold is used. In this embodiment, an empirical value of 240 is used, representing a dimensionless gray level, which is a preset parameter; medium to high The pixel grayscale value is forced to be the local background mean value of the corresponding position in the visible light image to eliminate residual highlight noise caused by uneven light source or local overexposure.

[0116] After the above processing, the complete intrinsic dark fringe feature matrix is ​​finally obtained. This matrix is ​​physically equivalent to the optical density distribution of the underlying ink after removing specular reflection from the surface of a cigarette box.

[0117] This application also discloses a QR code traceability system for cigarette box manufacturing process, including: a QR code generation and compensation module, a process dark pattern embedding module, a defective pixel reconstruction module, a binding factor generation and storage module, a traceability database, and a dual-band verification module.

[0118] The QR code generation compensation module receives the Young's modulus of the substrate, the expected compressive stress distribution matrix, and the pixel coordinate matrix. It calls the stress matrix reading unit, the sensitive pixel filtering unit, and the inverse affine transformation unit to output the QR code generation coordinate matrix after inverse displacement compensation. This output is directly transmitted to the printing plate making or laser marking control unit so that the physical code image returns to the standard orthogonal grid after subsequent hot stamping, die cutting, and embossing.

[0119] The process dark pattern embedding module is connected to the data acquisition interface of the production line to obtain the hot stamping temperature gradient, die-cutting pressure peak and finder pattern edge contour coordinates. After normalizing the above parameters, a process feature vector is formed, and frequency domain modulation is performed on the edge brightness of the finder pattern according to the preset spatial modulation frequency and dark pattern embedding intensity factor.

[0120] The modulated code image output by this module maintains the readable structure of a conventional QR code while retaining dark pattern signals corresponding to the actual process conditions at the edges of the seeker image.

[0121] The incomplete pixel reconstruction module is connected to an industrial camera, receives the code image grayscale matrix collected from the production line, identifies the connected components of the abnormal brightness areas, calculates the local spatial information entropy of the outer ring neighborhood, and reads the paper-based fiber texture gradient of the same cigarette box base material from the traceability database.

[0122] The incomplete pixel reconstruction module performs weighted interpolation based on local spatial information entropy, paper-based fiber texture gradient, and effective neighboring pixel brightness, and outputs the completed code image matrix for subsequent decoding and verification.

[0123] The binding factor generation and storage module receives the equipment status parameter matrix and the local gradient of the paper substrate fiber texture within the production time window. It performs nonlinear projection fusion according to the cross-domain fusion balance coefficient and the time-series data normalization scaling coefficient to generate encrypted binding factors. The encrypted binding factors, product batch number, equipment status summary, fiber texture index and code image identifier are associated and written into the traceability database.

[0124] Through this structure, the QR code pattern, process parameters, and non-clonable paper base texture of a single cigarette box are bound together in a one-to-one relationship.

[0125] The dual-band verification module includes a synchronous trigger control unit, an image registration unit, and a differential extraction unit. The synchronous trigger control unit controls the visible light source and the ultraviolet light source to alternately illuminate within the same exposure cycle. The image registration unit performs affine registration on the visible light code image and the ultraviolet light code image based on the corner points of the finder pattern. The differential extraction unit performs pixel difference operations based on the registered visible light band reflectance matrix and ultraviolet band reflectance matrix to eliminate specular reflection, extract intrinsic dark texture features, and compares the extraction results with dark texture features and encryption binding factors in the traceability database. If the comparison matches, it outputs genuine product information and corresponding production process traceability information; if the comparison does not match, it outputs an abnormal process step or a suspected non-genuine product warning.

[0126] In one embodiment, all modules are deployed in the same industrial edge computing terminal; in another embodiment, the QR code generation and compensation module and the process dark pattern embedding module are deployed at the coding station, the incomplete pixel reconstruction module and the dual-band verification module are deployed at the inspection station, and the traceability database is deployed on a factory server or a cloud server. Regardless of whether a centralized or distributed deployment is adopted, all modules are executed collaboratively in accordance with the data flow sequence of steps S1 to S5, thereby ensuring consistency between the system embodiment and the method embodiment. The above system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0127] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A QR code traceability method for cigarette box manufacturing process, characterized in that, Includes the following steps: When generating the code image, the expected displacement of the pixel is calculated based on the Young's modulus of the substrate and the preset expected compressive stress distribution matrix, and inverse displacement compensation is applied to the pixel. Extract the hot stamping temperature spatial gradient and die-cutting pressure peak generated in real time during the hot stamping and die-cutting processes of the cigarette box. Normalize the hot stamping temperature spatial gradient and die-cutting pressure peak into feature vectors, and then use frequency domain modulation to steganographically write them into the edge brightness of the finder pattern of the code image to form dark patterns. After acquiring the code image, the neighborhood information entropy of the brightness abnormal area is calculated, and the missing pixels are reconstructed by interpolation with the fiber texture gradient of the paper substrate as a constraint. Determine the production time window covering the hot stamping and die-cutting processes, perform non-linear projection fusion of the equipment status parameters within the production time window with the fiber texture of the paper substrate, generate an encrypted binding factor, and write it into the database. During anti-counterfeiting verification, the system simultaneously triggers the acquisition of dual-band spectra using visible and ultraviolet light sources. By suppressing specular reflection components through pixel difference, the system extracts intrinsic dark texture features.

2. The QR code traceability method for a cigarette box manufacturing process according to claim 1, characterized in that, The step of calculating the expected displacement of the pixel based on the Young's modulus of the substrate and a preset expected compressive stress distribution matrix, and applying inverse displacement compensation to the pixel, includes: The positive mechanical compressive stress in the target area is determined based on the expected compressive stress distribution matrix, and the normal distance vector from the current pixel to the center of force is obtained. The reverse displacement vector is calculated based on the substrate stress release coefficient, the positive mechanical compressive stress, the Young's modulus of the substrate, and the normal distance vector. The inverse displacement vector is applied to perform the inverse displacement compensation on the pixel.

3. The QR code traceability method for cigarette box manufacturing process according to claim 2, characterized in that, The substrate stress relief coefficient is updated in the following way: The code images under visible light and ultraviolet light, which were collected synchronously during the anti-counterfeiting verification in historical batches, are obtained, and after registration, pixel difference is performed to extract the signal-to-noise ratio of the intrinsic dark pattern features. The stress release coefficient of the substrate is adjusted based on the dynamic feedback of the signal-to-noise ratio.

4. The QR code traceability method for a cigarette box manufacturing process according to claim 1, characterized in that, The process of normalizing the spatial gradient of hot stamping temperature and the peak value of die-cutting pressure into a feature vector, and then using frequency domain modulation to steganographically embed it into the edge brightness of the finder pattern of the code image to form dark patterns, includes: Obtain the preset spatial modulation frequency and dark pattern embedding intensity factor; Determine the relative arc length coordinates of each pixel on the edge contour of the finder pattern of the code image; Based on the normalized feature vector, the dark pattern embedding intensity factor, the spatial modulation frequency, and the relative arc length coordinates, the edge brightness of the finder pattern of the code image is modulated to generate modulated edge brightness carrying dark patterns.

5. The QR code traceability method for a cigarette box manufacturing process according to claim 1, characterized in that, The step of calculating the neighborhood information entropy of the brightness anomaly region and interpolating and reconstructing the missing pixels using the fiber texture gradient of the paper substrate as a constraint includes: Determine the outer annular neighborhood of the region with abnormal brightness; Calculate the local spatial information entropy of the ring neighborhood, and use it as the neighborhood information entropy; Obtain the fiber texture gradient of the paper substrate; Based on the local spatial information entropy and the fiber texture gradient of the paper substrate, the brightness of the effective pixels in the annular neighborhood is weighted and interpolated to reconstruct the grayscale value of the missing pixels.

6. The QR code traceability method for a cigarette box manufacturing process according to claim 1, characterized in that, The equipment status parameters within the production time window are nonlinearly projected and fused with the fiber texture of the paper substrate to generate an encryption binding factor, including: Preset cross-domain fusion balance coefficient and time-series data normalization scaling coefficient; Obtain the production parameter vector formed by the equipment status parameters within the production time window; Obtain the local gradient of the fiber texture of the paper substrate; The nonlinear projection fusion is performed based on the cross-domain fusion balance coefficient, the time-series data normalization scaling coefficient, the production parameter vector, and the local gradient of the paper substrate fiber texture to generate the encryption binding factor.

7. The QR code traceability method for a cigarette box manufacturing process according to claim 1, characterized in that, The simultaneous triggering of visible and ultraviolet light sources to acquire dual-band spectra, after registration, suppresses specular reflection components through pixel difference and extracts intrinsic dark texture features, including: From the registered code map, obtain the reflectance matrix in the visible light band and the reflectance matrix in the ultraviolet band; Determine the adaptive weights for specular suppression and the optical minimum constant; Based on the visible light band reflectance matrix and the ultraviolet band reflectance matrix, pixel difference operation is performed to eliminate specular reflection and obtain intrinsic dark texture features.

8. The QR code traceability method for a cigarette box manufacturing process according to claim 1, characterized in that, Before generating the code image, the method further includes: acquiring the fiber texture distribution of a specific area of ​​the paper substrate using a high-precision line scan camera, as the fiber texture of the paper substrate.

9. A QR code traceability system for cigarette box manufacturing process, characterized in that, include: The module includes a QR code generation and compensation module, a process dark pattern embedding module, a defective pixel reconstruction module, a binding factor generation and storage module, a traceability database, and a dual-band verification module. The QR code generation compensation module is configured to calculate the expected displacement of pixels based on the Young's modulus of the substrate and the expected compressive stress distribution, and to apply reverse displacement compensation to the pixels of the QR code image to be generated. The process dark pattern embedding module is configured to normalize the hot stamping temperature gradient and the peak of the die-cutting pressure into a feature vector, and embed it into the edge brightness of the code image finder pattern through frequency domain modulation to form a dark pattern. The missing pixel reconstruction module is configured to calculate the neighborhood information entropy of the brightness abnormal area after acquiring the code image, and to reconstruct the missing pixels by interpolation with the fiber texture gradient of the paper substrate as a constraint. The binding factor generation and storage module is configured to perform nonlinear projection fusion of the equipment status parameters within the corresponding production time window and the fiber texture of the paper substrate to generate an encrypted binding factor and write it into the traceability database. The dual-band verification module is configured to simultaneously acquire the code image under visible light and ultraviolet light during verification, perform pixel difference after registration, and eliminate specular reflection to extract intrinsic dark patterns. The QR code generation compensation module, the process dark pattern embedding module, the incomplete pixel reconstruction module, the binding factor generation and storage module, and the dual-band verification module work together to execute the QR code traceability method for cigarette box manufacturing process according to any one of claims 1 to 8.

Images