A pattern recognition-based anti-counterfeiting verification method and system

Abstract

This invention discloses an anti-counterfeiting verification method and system based on pattern recognition, relating to the fields of information security and image recognition technology. The method includes the following steps: establishing a nanosecond-level unified illumination timescale and imaging energy datum; acquiring the exposure phase trajectory during the ticket imaging process through continuous sampling; extracting energy density distribution information based on the exposure phase trajectory; and identifying the saturation initiation kernel and persistence window caused by illumination flicker based on the energy density distribution, thus constructing a dynamic illumination command surface. This invention achieves highly robust recognition and reliable repair of anti-counterfeiting patterns through the construction of illumination timescales and energy trajectories, counterfactual replay, anomaly identification, polarization resampling, and dynamic control mechanisms, thereby improving the system's adaptability and security to illumination interference.

Description

Technical Field

[0001] This invention relates to the fields of information security and image recognition technology, specifically to an anti-counterfeiting verification method and system based on pattern recognition. Background Technology

[0002] "Pattern recognition-based anti-counterfeiting verification" refers to the technical process of using image recognition and pattern matching technologies to digitally collect and extract features from specific anti-counterfeiting patterns (such as microstructure textures, dynamic watermarks, hidden codes, random dot matrix, etc.) printed on lottery tickets. Algorithms are then used to compare and analyze the similarity and consistency with the original authorized template, thereby determining the authenticity of the lottery ticket. This method scans the lottery ticket surface pattern using a camera or mobile terminal, utilizing feature point matching, grayscale distribution analysis, and frequency domain feature alignment to identify minute differences, forgery reconstructions, or copying traces in the ticket pattern, achieving automated and intelligent anti-counterfeiting verification. Unlike traditional QR codes, barcodes, or manual ticket verification methods, pattern recognition-based anti-counterfeiting verification can identify subtle deviations in the printed microstructure, possessing a higher level of anti-counterfeiting and stronger anti-copying capabilities, making it particularly suitable for lottery tickets where authenticity, uniqueness, and concealment are all crucial.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, lottery anti-counterfeiting verification systems typically rely on cameras to capture images of tickets and use pattern recognition algorithms to compare authenticity. However, when the ticket verification camera is exposed to dynamic environments such as sudden light flickering or laser illumination, the image sensor is prone to saturation within a very short time, causing the instantaneous loss of key frame information in the captured image. Since existing technologies mostly employ automatic frame loss compensation and frame interpolation smoothing algorithms to maintain video stream continuity, such anomalies are often misjudged by the system as normal sampling behavior and cannot be identified during the detection phase. Although these short-term saturation events are extremely brief, they create a recognition "window" in the anti-counterfeiting feature area, causing subsequent recognition models to make judgments based on missing data, failing to accurately reflect the ticket's pattern features. Attackers can exploit this window by controlling the timing of light flickering to selectively interfere with the system, causing counterfeit tickets to be identified as genuine even when the anti-counterfeiting area is obscured. This leads to an increased pass rate for counterfeit tickets and a huge risk of false payouts, seriously affecting the anti-counterfeiting security and verification reliability of lottery issuing institutions.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a pattern recognition-based anti-counterfeiting verification method and system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] On the one hand, a pattern recognition-based anti-counterfeiting verification method is provided, including the following steps:

[0009] S1. Establish a nanosecond-level unified illumination timescale and imaging energy datum. Obtain the exposure phase trajectory during the ticket imaging process through continuous sampling. Extract energy density distribution information based on the exposure phase trajectory. Identify the saturation initiation kernel and continuous window caused by illumination flicker based on the energy density distribution. Construct a dynamic illumination command surface.

[0010] S2, under the constraint of dynamic illumination command surface, constructs a counterfactual playback chain, performs time inversion calculation on the exposure trajectory within the saturation window, replays the image energy flow change process based on the time inversion result, and generates statistical fingerprint information of key frames to form a flicker recovery baseline;

[0011] S3. Train a cross-domain anomaly discriminator based on the statistical fingerprint information in the flicker recovery baseline, use the frequency domain sparse mapping method and the extreme value continuity verification method to identify abnormal information fragments in the anti-counterfeiting verification area, mark the real missing certificate fragments and generate a time-frequency compensation reference surface.

[0012] S4, under the constraint of time-frequency compensation reference plane, performs multi-view polarization resampling, performs double-threshold amplitude limiting back-write calculation based on the calibrated missing certificate fragment, reconstructs the reliable mapping of anti-counterfeiting texture, and realizes the continuous restoration of ticket pattern information;

[0013] S5, based on the stable output of reliable anti-counterfeiting texture mapping, performs dual-track dynamic control of phase conjugate breathing traction and spectral reverse diffusion. It achieves synchronous gating of exposure control, shutter control and gain control through time-frequency self-suppression interference field, eliminates the blank area caused by light flicker, and constructs a steady-state anti-counterfeiting identification closed loop.

[0014] Optionally, step S1 includes:

[0015] During image acquisition, a light monitoring channel covering the sampling field of view is constructed by simultaneously deploying a high-speed photosensitive detection array and imaging sensing components in the imaging path, and a nanosecond-level light time scale is formed by using a unified crystal oscillator pulse drive.

[0016] Under illumination time-scale constraints, an imaging energy datum is established through continuous sampling, and the exposure phase trajectory of each pixel is extracted;

[0017] Based on the energy datum, the synchronic point of energy abrupt change is identified as the starting kernel of illumination flicker, and the energy change trend before and after the starting kernel is extracted and the duration window is calibrated.

[0018] In a unified illumination timescale and energy datum, spatial and temporal regions are dynamically labeled to construct a dynamic illumination command surface, which is used to identify pixel units and time nodes in the energy saturation state during image sampling.

[0019] Optionally, step S2 includes:

[0020] Under the constraint of dynamic illumination command surface, the calibrated saturation persistence window is extracted as the target processing region, and counterfactual trajectories are constructed to recover the energy evolution path;

[0021] Based on the counterfactual trajectory, time inversion calculation is performed, and a joint inversion matrix is ​​constructed by combining the trajectories of neighboring pixels to complete the reverse evolution of the energy propagation process;

[0022] After completing the energy trajectory inversion, image energy flow replay is performed, and regional energy blocks are generated and continuous maps are output through spatial-temporal fusion.

[0023] Brightness response, energy distribution, and texture orientation features are extracted from the replay image to form a statistical fingerprint set and a flicker recovery baseline is constructed to provide input data for subsequent identification chains.

[0024] Optionally, during the construction of the statistical fingerprint set, region overlap analysis and full-image fusion processing are performed by combining the brightness response amplitude, energy distribution center and texture direction consistency index to improve the local stability and overall recognition accuracy of the statistical fingerprint.

[0025] Optionally, step S3 includes:

[0026] Statistical fingerprint sets of keyframe images are extracted from the flicker recovery baseline to establish indices for luminance response, energy density, texture continuity, and spatial frequency response.

[0027] Based on the statistical fingerprint set, cross-domain feature alignment is performed, and anomaly identification classification boundary is constructed according to the offset metric to identify suspected missing certificate areas;

[0028] In the identified suspected missing certificate areas, frequency domain sparse mapping and extreme value continuity verification operations are performed to determine the actual missing certificate segments and correct the boundaries;

[0029] Based on the correction results, a time-frequency compensation reference surface containing spatial coordinates, temporal distribution, and spectral intensity is constructed to provide navigation and constraint basis for subsequent image restoration.

[0030] Optionally, during the construction of the time-frequency compensation reference surface, frequency bandwidth diffusion analysis and extreme value jump detection are performed on the spectral region corresponding to the missing certificate segment, and the low-scoring region is completed through spatial interpolation and spectral smoothing to improve the accuracy of structural identification of the missing certificate region and the continuity of the time-frequency compensation reference surface.

[0031] Optionally, step S4 includes:

[0032] Under the constraint of time-frequency compensation reference surface, multi-view polarization resampling is performed to obtain the texture response results of the target image region under different incident characteristics;

[0033] Based on the sampling results, texture orientation consistency calculation and vector registration are performed to form a texture fusion source set;

[0034] On the texture fusion source set, a dual-threshold amplitude limiting write-back is performed on the missing evidence fragments, and structural amplitude limiting constraints and brightness amplitude limiting constraints are executed in sequence to generate image compensation data that meets the requirements of continuity and consistency.

[0035] By integrating image compensation data with the original image, a reliable anti-counterfeiting texture mapping map is constructed.

[0036] Optionally, after constructing the anti-counterfeiting texture trusted mapping map, a global brightness equalization operation and frequency bandwidth matching process are performed. By unifying the tone mapping relationship and the spectral density constraint function, the color shift and texture frequency fluctuation of the compensation area are adjusted to improve the overall consistency and stability of the image.

[0037] Optionally, step S5 includes:

[0038] After the output of the anti-counterfeiting texture trusted mapping, the texture response surface is constructed based on the image spatial energy density and structural continuity information, and a synchronous phase evolution path is established to generate a phase conjugate mapping map;

[0039] Based on phase conjugate traction, spectral inverse diffusion processing is introduced to achieve signal energy stabilization compensation through reverse propagation;

[0040] Under the dual-track control, a time-frequency self-suppressive interference field is formed, which performs synchronous gating operation on exposure control, shutter control and gain control;

[0041] After the gating operation is completed, a steady-state anti-counterfeiting identification closed loop is established to achieve parameter self-adjustment and light-tolerant identification output.

[0042] On the other hand, a pattern recognition-based anti-counterfeiting verification system is provided, including an illumination timescale construction module, an exposure inversion reconstruction module, an anomaly identification and calibration module, a texture reconstruction and repair module, and an exposure synchronization control module:

[0043] The illumination timescale construction module establishes a nanosecond-level unified illumination timescale and imaging energy datum. It obtains the exposure phase trajectory during the ticket imaging process through continuous sampling, extracts energy density distribution information based on the exposure phase trajectory, and identifies the saturation initiation kernel and continuous window caused by illumination flicker based on the energy density distribution, thus constructing a dynamic illumination command surface.

[0044] The exposure inversion and reconstruction module constructs a counterfactual playback chain under the constraint of dynamic illumination command surface, performs time inversion calculation on the exposure trajectory within the saturation window, reconstructs the image energy flow change process based on the time inversion results, and generates statistical fingerprint information of key frames to form a flicker recovery baseline.

[0045] The anomaly identification and calibration module trains a cross-domain anomaly discriminator based on the statistical fingerprint information in the flicker recovery baseline, and uses the frequency domain sparse mapping method and the extreme value continuity verification method to identify abnormal information fragments in the anti-counterfeiting verification area, calibrate the real missing certificate fragments, and generate a time-frequency compensation reference surface.

[0046] The texture reconstruction and repair module performs multi-view polarization resampling under the constraint of time-frequency compensation reference plane, performs double-threshold amplitude limiting back-write calculation based on the calibrated missing certificate fragments, reconstructs the reliable mapping of anti-counterfeiting texture, and realizes the continuous repair of ticket pattern information;

[0047] The exposure synchronization control module, based on the stable output of the reliable mapping of anti-counterfeiting texture, performs dual-track dynamic regulation of phase conjugate breathing traction and spectral reverse diffusion. It achieves synchronous gating of exposure control, shutter control and gain control through time-frequency self-suppressed interference field, eliminates the blank area caused by light flicker, and constructs a steady-state anti-counterfeiting identification closed loop.

[0048] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0049] This invention establishes nanosecond-level illumination timescales and imaging energy datums, quantifying exposure behavior during image sampling into traceable phase trajectories for the first time. This enables accurate identification of the illumination disturbance initiation kernel and the continuous window. Furthermore, combined with a counterfactual playback mechanism, it achieves reverse replay of keyframe image energy flow and structural feature recovery, constructing a cross-domain anomaly discriminator with discriminative capabilities based on missing document fragment identification. Through multi-view polarization resampling and dual-threshold amplitude limiting write-back, it successfully completes the reliable reconstruction of anti-counterfeiting textures, ensuring the integrity and continuity of pattern information. Finally, by employing a dual-track dynamic control strategy of phase conjugation and spectral inverse diffusion, it forms a time-frequency interference linkage mechanism for exposure, shutter, and gain control, breaking the dependence of traditional identification processes on the integrity of continuous images. This significantly enhances the anti-counterfeiting system's adaptability and robustness to complex dynamic illumination interference, thereby effectively reducing the pass rate of counterfeit tickets and improving the identification accuracy and security stability of the lottery anti-counterfeiting verification system in high-risk environments. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0051] Figure 1 This is a flowchart of an anti-counterfeiting verification method based on pattern recognition according to the present invention;

[0052] Figure 2 This is a schematic diagram of a pattern recognition-based anti-counterfeiting verification system according to the present invention. Detailed Implementation

[0053] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0054] This invention provides, for example Figure 1 The pattern recognition-based anti-counterfeiting verification method shown includes the following steps:

[0055] S1. Establish a nanosecond-level unified illumination timescale and imaging energy datum. Obtain the exposure phase trajectory during the ticket imaging process through continuous sampling. Extract energy density distribution information based on the exposure phase trajectory. Identify the saturation initiation kernel and continuous window caused by illumination flicker based on the energy density distribution. Construct a dynamic illumination command surface.

[0056] To address the problem of image sensor saturation caused by illumination flicker, which leads to the failure of pattern recognition anti-counterfeiting measures, an image acquisition method based on dynamic illumination analysis is proposed. This method constructs a dynamic illumination command surface during the exposure process, providing high-precision spatiotemporal boundary information for subsequent image reconstruction and feature recovery. The specific steps are as follows:

[0057] During the image acquisition phase, a high-speed photosensitive detection array and imaging sensing components are simultaneously deployed along the imaging path to construct an illumination monitoring channel covering the sampling field of view. This channel monitors the incident light intensity in real time with a nanosecond-level response speed. In practice, for each frame of image exposure, the time axis of the photosensitive detection array and the main imaging component are uniformly calibrated to form a single time-stamped reference. Through this reference, all subsequent energy sampling data and frame image sampling data have strict temporal consistency. This time-stamped reference is achieved through crystal oscillator synchronization pulse drive, using equally spaced time-series nodes across the entire time domain to collect changes in light intensity during the exposure period. Each sampling node is marked with its corresponding physical time and includes a sampled light energy intensity value, thus forming a continuous, high-precision time series. This time series is used to support the dynamic reconstruction of image exposure behavior and provides a temporal basis for subsequent analysis.

[0058] Based on a unified timescale, the imaging energy flow state within the image acquisition area is constructed, and an imaging energy datum is established through continuous sampling. The imaging energy datum refers to the set of projection trajectories of the light intensity received by each pixel unit on the time axis under the temporal sampling dimension. In practice, for each pixel block of the acquired image, its light energy change curve within one exposure cycle is extracted, and this curve is mapped to the corresponding pixel's exposure phase trajectory. The exposure phase trajectory describes the evolution trend of light intensity throughout the entire exposure cycle, possessing fine-grained temporal continuity. To ensure the comparability of the trajectories, the exposure phase trajectories of all pixels are standardized to a unified energy unit range and linearly interpolated according to timescale nodes. This imaging energy datum, based on two-dimensional spatial location and extended by the time axis, constructs a three-dimensional dynamic energy view, providing a complete data matrix for subsequent density analysis.

[0059] After the imaging energy datum is constructed, the energy density distribution across the entire field of view is extracted based on the aforementioned exposure phase trajectories. This extraction process uses the changing gradient within the trajectory set as a feature to identify nonlinear perturbation points in each trajectory. By scanning each trajectory sequentially and comparing the differentials between trajectories, time points where the mutation rate exceeds a set threshold are extracted and spatial aggregation analysis is performed. This analysis process identifies simultaneous energy mutation points in multiple pixel regions; these simultaneous points are suspected illumination flicker initiation kernels. Simultaneously, to verify the persistence of the mutation, the energy change rate of a local region is extracted from each trajectory, centered on the initiation kernel, by sliding windows to the left and right, constructing a local trend curve. The presence of a period exceeding the saturation threshold is then determined. If the condition is met, it is identified as a valid saturation window, and its start and end times are recorded, thus defining the duration of illumination flicker within the entire frame's exposure behavior. This duration window has both spatial and temporal coordinate information, used to define the boundary of the impact of abnormal energy perturbations on the image frame.

[0060] After identifying the saturation initiation kernel and its corresponding duration window, a unified illumination timescale and energy datum are used to perform region annotation processing on the entire image sampling process within the same spatiotemporal coordinate system, thereby constructing a dynamic illumination command surface. The dynamic illumination command surface refers to overlaying a dynamic marker layer on the original exposure energy datum based on the identified abrupt change nodes and duration intervals. Each marker element represents an illumination anomaly state in a specific time slice and spatial segment. This command surface uses two-dimensional pixel coordinates as the horizontal and vertical axes, and the exposure time node as the vertical dimension, precisely describing in a three-dimensional annotation format which spatial pixel units and time periods the image sampling is in an energy saturation state. This command surface not only has dense structure marking capabilities but also high-efficiency dynamic update characteristics, which can be used to drive subsequent image backtracking, anomaly compensation, and identification filtering processes in real time, providing a precise foundation for illumination disturbance identification for the entire image anti-counterfeiting processing chain.

[0061] S2, under the constraint of dynamic illumination command surface, constructs a counterfactual playback chain, performs time inversion calculation on the exposure trajectory within the saturation window, replays the image energy flow change process based on the time inversion result, and generates statistical fingerprint information of key frames to form a flicker recovery baseline;

[0062] To compensate for and restore image information loss caused by illumination flicker interference, image energy inversion and playback, along with statistical feature reconstruction, are performed under the constraints of the constructed dynamic illumination command surface. This aims to restore the structural information of key frames and reconstruct the anti-spoofing data foundation. The specific steps are as follows:

[0063] Based on the constructed dynamic illumination command surface, all calibrated saturation duration windows are extracted as the target processing region. Within each duration window, normal exposure trajectory segments from the preceding and following time periods are selected as boundary conditions for replay calculations. For each exposure trajectory, a bidirectional interpolation sequence is constructed before and after its saturation segment, and various dynamic feature values ​​such as energy change rate, trend slope, and abrupt boundary are extracted from adjacent time periods. Using these features as input conditions, the trajectory change pattern missing within the saturation segment is reconstructed. To ensure temporal consistency, the reconstructed trajectory must be extrapolated under a unified illumination timescale and maintain continuity, energy conservation, and abrupt response matching with the original trajectory, thereby restoring the true energy evolution path lost under short-term intense light interference. This process, called counterfactual trajectory construction, is the core of the initial calculation in this step.

[0064] After constructing the counterfactual trajectories, a time-reversal calculation is performed on the reconstructed trajectory group. Time-reversal is not a simple flipping of the trajectory time axis, but rather a reverse evolution based on imaging physics modeling, simulating the energy propagation direction of the natural imaging process. Specifically, in the time-reversal stage, the energy increment direction of each trajectory is reverse-binded to the time progression direction, gradually tracing the origin and evolution of each energy peak. Simultaneously, a joint inversion matrix is ​​constructed using neighboring pixel trajectories as a collaborative benchmark, and spatial compensation calculations are performed on each target trajectory. This inversion calculation uses the energy resonance characteristics within the pixel time window as a constraint to ensure that the inversion process does not introduce abnormal energy jumps. During this process, edge slope fluctuations and local symmetry imbalance regions appearing during the reverse evolution of the energy trajectory are monitored simultaneously, preparing for subsequent structural information identification.

[0065] After obtaining the counterfactual trajectory set corresponding to all saturation windows, an image energy flow replay operation is performed on the energy evolution process within it. The core objective of this operation is to recover the spatial distribution pattern and temporal feature response path of the image within the anomalous segment. During execution, all inverted trajectories are first restored to the standard exposure energy datum. Using a spatial-temporal bidirectional fusion method, the overlapping areas of inter-pixel trajectories are integrated into regional energy blocks, and then a time-progression sequence is generated on a regional basis. In this sequence, spatial energy maps at different time nodes are extracted, and the brightness gradient map and texture response map of the image segment at a specific time slice are derived accordingly. By integrating the above map data frame by frame, the continuous energy flow structure of the entire frame image is restored. This energy flow replay process not only compensates for missing image information but also provides structurally complete carrier data for the next stage of feature extraction.

[0066] After image energy flow reconstruction is completed, statistical fingerprint information is constructed for each restored keyframe image. Statistical fingerprint information is a set of structural descriptive values ​​that comprehensively characterize the local and global spatial energy features of an image. Based on the image reconstruction results, this process extracts statistical quantities from multiple dimensions, such as brightness response amplitude, energy distribution center, boundary derivative continuity, and spatial texture orientation stability, pixel by pixel. Overlap analysis is performed in multiple local regions of the image to obtain local fingerprint maps. Then, at the global scale, a full-image statistical structure map is synthesized using a region fusion method, and various auxiliary maps, such as texture orientation consistency matrix, frequency distribution intensity map, and local extremum connectivity map, are superimposed to construct a complete statistical fingerprint set. This set not only possesses high discriminative power and stability but also provides a basis for determining whether the image has been affected by external interference, ultimately forming a flicker recovery baseline for subsequent recognition and model training, serving as the first layer of reliable input in the subsequent recognition chain.

[0067] S3. Train a cross-domain anomaly discriminator based on the statistical fingerprint information in the flicker recovery baseline, use the frequency domain sparse mapping method and the extreme value continuity verification method to identify abnormal information fragments in the anti-counterfeiting verification area, mark the real missing certificate fragments and generate a time-frequency compensation reference surface.

[0068] To accurately identify and distinguish image information loss phenomena within areas of abnormal illumination interference, based on the image energy flow obtained from energy trajectory inversion reconstruction, cross-scale feature learning and frequency domain dynamic mapping processing are further performed on the statistical fingerprint set. This identifies abnormal information fragments hidden in the image structure, and, based on multidimensional analysis, identifies the true missing information regions, forming a complete reference structure for subsequent restoration. The specific steps are as follows:

[0069] The statistical fingerprint set of all keyframe images extracted from the generated flicker recovery baseline is used as the input data source. The statistical fingerprint set includes the brightness response features, energy density distribution, boundary structure strength, texture continuity index, spatial frequency response, and brightness extreme value connectivity index of each region in the image. In practice, each frame image is first divided into several equal-area grid regions, and the aforementioned statistical indicators are extracted from each grid to form a vectorized description. Then, the statistical features of all regions are normalized in a unified coordinate system to ensure feature comparability between different frames. Based on this, a feature similarity matrix is ​​constructed to analyze the repetitive distribution pattern of each image region across multiple frames, thereby deriving the statistical fingerprint evolution model of the image under non-interference conditions, which serves as a subsequent judgment criterion.

[0070] Based on a statistical fingerprint evolution model, a cross-domain anomaly detection model was trained on the entire image sequence. During training, a set of high-fidelity images from unperturbed lighting conditions were selected as reference domain samples, and image frames from the flicker recovery baseline were used as input to the target domain, performing cross-domain feature alignment. Specifically, firstly, the texture structure of the reference domain image was mapped to the target domain image using a spatial structure mapping method. Then, the offset metric of each grid region in the target image on the statistical fingerprint was calculated, including mean deviation, standard deviation increase, structural continuity decrease, and texture direction difference. These offset metrics were used as training data input to construct the classification boundary for identifying anomalous regions. After model training, for each frame in the flicker recovery baseline, statistical fingerprint projection and boundary distance determination operations were performed to identify image regions deviating from the normal model threshold range, which were initially identified as suspected missing evidence regions with structural anomalies.

[0071] After identifying suspected missing regions, further precise identification and boundary correction of abnormal segments are achieved through frequency domain sparse mapping and extreme value continuity verification. Frequency domain sparse mapping refers to mapping a selected region of the image frame in the statistical fingerprint space to the frequency domain, obtaining its spectral structure features through Fourier transform or wavelet transform, and analyzing the distribution density of high-frequency components, bandwidth diffusion, and the continuity and energy concentration of spectral lines. The spectrum of normal regions usually shows structural frequency arrangement and continuous spectral energy changes, while the spectral features of missing regions are highly discrete and abrupt. Based on this, the positions of extreme value jumps in the spectrum are mapped back to the original image space, and extreme value continuity verification is performed in the corresponding regions to analyze whether there are spatial anomalies such as edge breaks, structural interruptions, or abrupt changes in brightness in these regions. By combining the spatial verification and spectral mapping results, the true missing image segments are accurately identified, and their boundary positions are corrected at the pixel level.

[0072] After identifying and calibrating the missing document fragments, a time-frequency compensation reference surface is constructed based on their spatial coordinates, temporal distribution, and spectral structure. This reference surface is a multi-dimensional structure with the horizontal dimension representing the image spatial coordinates, the vertical dimension representing the temporal sampling nodes, and the vertical dimension representing the frequency domain spectral intensity distribution. In this three-dimensional structure, each cell represents the image state integrity score at a specific time, location, and frequency. During the construction process, the previously calibrated missing document areas are mapped to low-score areas, while normal areas are mapped to high-score continuous bands. Through spatial interpolation, spectral smoothing, and temporal stretching operations, the reference surface is completed, enabling it to guide the selection of sampling strategies, boundary constraints, and texture prediction path generation in subsequent image restoration processes. The generation of this time-frequency compensation reference surface not only provides navigation information for image structure restoration but also constitutes the basic support surface for subsequent multi-view resampling and amplitude-limiting reconstruction operations, making it one of the key components for ensuring the continuity of image anti-counterfeiting.

[0073] S4, under the constraint of time-frequency compensation reference plane, performs multi-view polarization resampling, performs double-threshold amplitude limiting back-write calculation based on the calibrated missing certificate fragment, reconstructs the reliable mapping of anti-counterfeiting texture, and realizes the continuous restoration of ticket pattern information;

[0074] To further repair structural defects in images caused by illumination interference, under the constraint of a time-frequency compensation reference surface constructed based on frequency domain sparse mapping and extreme value continuity verification, a multi-view polarization resampling and dual-threshold amplitude-limiting write-back mechanism is used to reconstruct the structural texture of the missing regions in the image, thereby restoring the overall continuity and verifiability of the image in both spatial and frequency dimensions. The specific steps are as follows:

[0075] After obtaining the complete time-frequency compensation reference surface, the image resampling input path is planned and configured based on the spatial anomaly region distribution map and frequency response missing segments marked in the reference surface. To achieve directional diversity in information recovery, the target image needs to be repeatedly sampled from multiple perspectives. These perspectives include not only changes in the image acquisition angle but also changes in the incident light direction and polarization angle. Specifically, an adjustable polarization filter component is mounted on the sampling device. By rotating the polarization filter axis to switch at different angles, the reflection texture information of the same ticket area under different polarization states is acquired. In actual operation, the target image is first divided into multiple resampling sub-regions according to the boundary structure of the missing ticket area. The response results of each sub-region under different perspectives and polarization angles are completely recorded and synchronously integrated according to a unified time reference. The obtained sampling results constitute a multi-dimensional image information block, which contains the differences in image performance at the same location under different incident characteristics, providing redundant structural references for subsequent texture reconstruction.

[0076] After completing multi-view polarization resampling, the sampled information blocks are mapped to the image regions corresponding to the missing fragments, and preliminary texture fusion is performed within the regions. During the fusion process, texture direction consistency calculation is first performed on the image sampling results at each polarization angle. By analyzing the distribution of the main texture direction, a local texture direction field is constructed. Then, vector registration processing is performed on each image sample based on the direction field to eliminate image misalignment caused by different viewpoints or polarization angles. Based on this, the average brightness and local contrast index of the same region under different sampling conditions are further analyzed. Abnormally bright or low-light artifact images are removed, and image fragments with high structural fidelity are retained to form a texture fusion source set. A pixel-by-pixel weighted averaging method is used during the fusion process, and a boundary smoothing operator is introduced to optimize the transition at the edges of the fusion region, ensuring that the image structure does not experience breaks or strong abrupt changes in the transition area, maintaining spatial continuity.

[0077] After the texture fusion source set is formed, a write-back calculation operation based on dual-threshold limiting constraints is performed on the identified missing fragments to achieve reliable backfilling of texture data in the target region. Dual-threshold limiting write-back is an image restoration mechanism based on dual standards of content integrity and authenticity, specifically including two layers of processing: structural limiting constraints and brightness limiting constraints. In the structural limiting constraint stage, the similarity between the target pixel block and adjacent normal regions in texture direction, edge structure, and frequency response is judged. If the similarity exceeds a set structural deviation threshold, the write-back pixel is canceled to prevent erroneous structures from entering the target region. In the brightness limiting constraint stage, the difference between the brightness value of the target pixel and the brightness distribution of historical normal regions is calculated. If the difference exceeds the upper and lower limits, brightness correction is performed to ensure that the image is consistent with the overall environment in terms of brightness. The combined effect of these two limiting mechanisms ensures that the generated write-back pixels meet the requirements of image structural continuity while avoiding abnormal jumps in brightness or contrast, thereby constructing a highly reliable and visually consistent image compensation dataset.

[0078] After completing the amplitude-limited write-back, all repaired areas are integrated and merged with the original image to construct the final reliable anti-counterfeiting texture mapping map. This mapping map not only includes the real image data of the undisturbed areas in the original image, but also the repaired data of the missing areas generated by multi-view resampling, polarization analysis, structural registration, texture fusion, and amplitude-limited write-back. To ensure the overall consistency and stability of the mapping map, a global brightness equalization operation and frequency bandwidth matching processing are performed on the entire image. By constructing a unified tone mapping relationship and spectral density constraint function, the color shift and texture frequency fluctuation of local areas are uniformly adjusted to the optimal distribution state of the entire image. The final output image mapping result not only presents a highly natural pattern continuity at the perceptual level, but also meets multiple constraints of texture consistency, brightness balance, and structural integrity in terms of technical indicators, providing a reliable image input foundation for subsequent anti-counterfeiting feature extraction and pattern consistency verification.

[0079] S5, based on the stable output of reliable anti-counterfeiting texture mapping, performs dual-track dynamic control of phase conjugate breathing traction and spectral reverse diffusion. It achieves synchronous gating of exposure control, shutter control and gain control through time-frequency self-suppression interference field, eliminates the blank area caused by light flicker, and constructs a steady-state anti-counterfeiting identification closed loop.

[0080] After completing the reliable mapping output of the anti-counterfeiting texture, to avoid the impact of windowing caused by illumination flicker during future imaging processes, a dual-track dynamic control mechanism is further introduced. This mechanism, driven by a combination of phase conjugate breathing traction and spectral anti-diffusion, achieves dynamic and stable control in both time and frequency dimensions during imaging exposure, ultimately constructing a steady-state anti-counterfeiting identification closed-loop structure with adaptive control capabilities. The specific steps are as follows:

[0081] Based on the stable output of the anti-counterfeiting texture mapping results, a high-precision texture response surface is constructed using the spatial energy density distribution and structural continuity information in the mapping map. This response surface serves as the dynamic control benchmark, establishing a set of phase evolution paths synchronized with the imaging process. In practical operation, an optical response feedback device is introduced to the photosensitive surface of the image acquisition device. This device is used to capture the phase interference difference between the incident light field and the texture structure in real time. By conjugate pairing the energy response trajectory corresponding to each pixel of the image with the phase change of the light field, the interference perturbation function caused by the interaction between the light field and the texture is extracted, thereby generating a phase conjugate mapping map. Based on this map, a traction curve operating in a "breathing rhythm" manner is constructed. This curve simulates the periodic fluctuations in the optical breathing process and applies dynamic fine-tuning to the exposure time axis of the imaging window, ensuring that the imaging behavior closely follows the true phase response of the image rather than external abrupt interference, thus establishing the basis for breathing traction control in the image acquisition stage.

[0082] After the phase conjugate traction mechanism is constructed, a frequency-based inverse diffusion processing strategy is further introduced to stabilize and compensate for the signal energy distribution through dynamic spectrum reconstruction. The core idea of ​​spectral inverse diffusion is to concentrate the signal energy during the imaging process and propagate it in reverse from high frequency to low frequency, avoiding frequency band drift caused by sudden changes in illumination due to high-frequency disturbances. In specific implementation, continuous frame spectrum analysis is performed on the acquired images to extract the main frequency band at each moment and determine its frequency density change trend. When a frequency rise phenomenon is detected in a sudden change region, spectral inverse diffusion processing is immediately triggered to gradually expand the signal in that region in time and perform low-pass recycling in frequency to suppress crosstalk to the surrounding areas. This processing relies on the reference rhythm output from the previous phase traction step to dynamically align the diffusion window with the phase breathing rhythm, thereby ensuring that spectral control and phase response are synchronized, realizing a signal stabilization mechanism with dual coordination in the frequency and time domains.

[0083] After establishing the dual-track control mechanism of phase conjugation and spectral inverse diffusion, a time-frequency self-suppressive interferometric field is constructed based on this mechanism to drive the three-element synchronous gating operation of exposure control, shutter control, and gain control during the imaging process. This interferometric field is essentially a field intensity control mechanism based on a dynamic response function. During image acquisition, it continuously calculates the phase perturbation degree and frequency offset amplitude of the current frame and adjusts the imaging parameters according to the interference intensity. Specifically, exposure control is based on energy density distribution; when the interference field shows an increasing trend, the exposure duration is automatically reduced. Shutter control uses the phase perturbation gradient as an indicator; when the interference slope approaches a threshold, the image sampling time window is forcibly shortened. Gain control dynamically adjusts the sensor sensitivity based on the frequency diffusion intensity to prevent overexposure or underexposure. In the linkage of these three parameters, the time-frequency interferometric field serves as a unified scheduling basis, ensuring that each control objective always operates collaboratively within the same response framework, no longer executing independently or sequentially, thereby fundamentally eliminating image voids caused by control delays or response deviations.

[0084] After completing the ternary synchronous gating operation, a steady-state anti-counterfeiting identification process with closed-loop self-adjustment capability is formed. The core of the closed-loop structure lies in the real-time feedback connection between image acquisition output and image evaluation input, forming a continuous and iterative image quality detection and control link. Specifically, after each frame of image is acquired and texture mapped, it immediately enters the evaluation path, re-executes statistical fingerprint analysis and frequency feature comparison operations, and makes a quantitative evaluation of its structural integrity and energy distribution rationality. If the evaluation value is lower than the preset stable standard, the exposure threshold, shutter response timing, and gain coefficient in the next frame of image acquisition process are automatically corrected, thus forming a feedback loop of "image evaluation - parameter control - image re-acquisition". After the feedback mechanism is running stably, the image acquisition behavior has high adaptability and illumination fault tolerance. Even under complex lighting environments or external interference conditions, it can continuously output image data with complete anti-counterfeiting texture features and statistically reliable fingerprints, ultimately realizing a practically valuable steady-state anti-counterfeiting identification closed loop.

[0085] This invention establishes nanosecond-level illumination timescales and imaging energy datums, quantifying exposure behavior during image sampling into traceable phase trajectories for the first time. This enables accurate identification of the illumination disturbance initiation kernel and the continuous window. Furthermore, combined with a counterfactual playback mechanism, it achieves reverse replay of keyframe image energy flow and structural feature recovery, constructing a cross-domain anomaly discriminator with discriminative capabilities based on missing document fragment identification. Through multi-view polarization resampling and dual-threshold amplitude limiting write-back, it successfully completes the reliable reconstruction of anti-counterfeiting textures, ensuring the integrity and continuity of pattern information. Finally, by employing a dual-track dynamic control strategy of phase conjugation and spectral inverse diffusion, it forms a time-frequency interference linkage mechanism for exposure, shutter, and gain control, breaking the dependence of traditional identification processes on the integrity of continuous images. This significantly enhances the anti-counterfeiting system's adaptability and robustness to complex dynamic illumination interference, thereby effectively reducing the pass rate of counterfeit tickets and improving the identification accuracy and security stability of the lottery anti-counterfeiting verification system in high-risk environments.

[0086] The present invention also provides, for example Figure 2 The anti-counterfeiting verification system based on pattern recognition shown includes an illumination timescale construction module, an exposure inversion reconstruction module, an anomaly identification and calibration module, a texture reconstruction and repair module, and an exposure synchronization control module.

[0087] The illumination timescale construction module establishes a nanosecond-level unified illumination timescale and imaging energy datum. It obtains the exposure phase trajectory during the ticket imaging process through continuous sampling, extracts energy density distribution information based on the exposure phase trajectory, and identifies the saturation initiation kernel and continuous window caused by illumination flicker based on the energy density distribution, thus constructing a dynamic illumination command surface.

[0088] The exposure inversion and reconstruction module constructs a counterfactual playback chain under the constraint of dynamic illumination command surface, performs time inversion calculation on the exposure trajectory within the saturation window, reconstructs the image energy flow change process based on the time inversion results, and generates statistical fingerprint information of key frames to form a flicker recovery baseline.

[0089] The anomaly identification and calibration module trains a cross-domain anomaly discriminator based on the statistical fingerprint information in the flicker recovery baseline, and uses the frequency domain sparse mapping method and the extreme value continuity verification method to identify abnormal information fragments in the anti-counterfeiting verification area, calibrate the real missing certificate fragments, and generate a time-frequency compensation reference surface.

[0090] The texture reconstruction and repair module performs multi-view polarization resampling under the constraint of time-frequency compensation reference plane, performs double-threshold amplitude limiting back-write calculation based on the calibrated missing certificate fragments, reconstructs the reliable mapping of anti-counterfeiting texture, and realizes the continuous repair of ticket pattern information;

[0091] The exposure synchronization control module, based on the stable output of the reliable mapping of anti-counterfeiting texture, performs dual-track dynamic regulation of phase conjugate breathing traction and spectral reverse diffusion. It achieves synchronous gating of exposure control, shutter control and gain control through time-frequency self-suppressed interference field, eliminates the blank area caused by light flicker, and constructs a steady-state anti-counterfeiting identification closed loop.

[0092] The present invention provides a pattern recognition-based anti-counterfeiting verification method, which is implemented through the above-mentioned pattern recognition-based anti-counterfeiting verification system. For details of the specific method and process of the pattern recognition-based anti-counterfeiting verification system, please refer to the above-mentioned embodiment of the pattern recognition-based anti-counterfeiting verification method, which will not be repeated here.

[0093] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A pattern recognition based anti-counterfeiting verification method, characterized in that, Includes the following steps: S1. Establish a nanosecond-level unified illumination timescale and imaging energy datum. Obtain the exposure phase trajectory during the ticket imaging process through continuous sampling. Extract energy density distribution information based on the exposure phase trajectory. Identify the saturation initiation kernel and continuous window caused by illumination flicker based on the energy density distribution. Construct a dynamic illumination command surface. S2, under the constraint of dynamic illumination command surface, constructs a counterfactual playback chain, performs time inversion calculation on the exposure trajectory within the saturation window, replays the image energy flow change process based on the time inversion result, and generates statistical fingerprint information of key frames to form a flicker recovery baseline; Step S2 includes: Under the constraint of dynamic illumination command surface, the calibrated saturation persistence window is extracted as the target processing region, and counterfactual trajectories are constructed to recover the energy evolution path; Based on the counterfactual trajectory, time inversion calculation is performed, and a joint inversion matrix is ​​constructed by combining the trajectories of neighboring pixels to complete the reverse evolution of the energy propagation process; After completing the energy trajectory inversion, image energy flow replay is performed, and regional energy blocks are generated and continuous maps are output through spatial-temporal fusion. Brightness response, energy distribution and texture orientation features are extracted from the replay map to form a statistical fingerprint set and construct a flicker recovery baseline to provide input data for subsequent identification chains. In the process of constructing the statistical fingerprint set, the consistency index of brightness response amplitude, energy distribution center and texture direction is combined to perform regional overlap analysis and full image fusion processing to improve the local stability and overall recognition accuracy of the statistical fingerprint. S3. Train a cross-domain anomaly discriminator based on the statistical fingerprint information in the flicker recovery baseline, use the frequency domain sparse mapping method and the extreme value continuity verification method to identify abnormal information fragments in the anti-counterfeiting verification area, mark the real missing certificate fragments and generate a time-frequency compensation reference surface. Step S3 includes: The statistical fingerprint set of all keyframe images is extracted from the generated flicker recovery baseline and used as the input data source. The statistical fingerprint set includes the brightness response features, energy density distribution, boundary structure strength, texture continuity index, spatial frequency response, and brightness extreme value connectivity index of each region in the image. In practice, each frame image is first divided into several grid regions of equal area, and statistical indicators are extracted for each grid to form a vectorized description. Then, the statistical features of all regions are normalized in a unified coordinate system to ensure feature comparability between different frames. Based on this, by constructing a feature similarity matrix, the repetitive distribution pattern of each image region in multiple frames is analyzed, thereby deriving the statistical fingerprint evolution model of the image under non-interference conditions. Based on the statistical fingerprint evolution model, a cross-domain anomaly detection model is trained on the entire image sequence. During training, a set of high-fidelity images from unperturbed lighting conditions are selected as reference domain samples, and image frames from the flicker recovery baseline are used as input to the target domain, performing cross-domain feature alignment. Specifically, firstly, the texture structure of the reference domain image is mapped to the target domain image using a spatial structure mapping method. Then, the offset measure of each grid region of the target image on the statistical fingerprint is calculated, including mean deviation, standard deviation increase, structural continuity decrease, and texture direction difference. These offset indicators are used as training data input to construct the classification boundary for identifying anomalous regions. After the model training is completed, for each frame of the flicker recovery baseline, statistical fingerprint projection and boundary distance determination operations are performed to identify image regions that deviate from the normal model threshold range, initially determining them as suspected missing evidence regions with structural anomalies. In the identified suspected missing certificate areas, frequency domain sparse mapping and extreme value continuity verification operations are performed to determine the actual missing certificate segments and correct the boundaries; Based on the correction results, a time-frequency compensation reference surface containing spatial coordinates, temporal distribution, and spectral intensity is constructed to provide navigation and constraint basis for subsequent image restoration; S4, under the constraint of time-frequency compensation reference surface, performs multi-view polarization resampling, performs double-threshold amplitude-limited write-back calculation based on the calibrated missing certificate fragments, and reconstructs the reliable mapping of anti-counterfeiting texture; Step S4 includes: Under the constraint of time-frequency compensation reference surface, multi-view polarization resampling is performed to obtain the texture response results of the target image region under different incident characteristics; Based on the sampling results, texture orientation consistency calculation and vector registration are performed to form a texture fusion source set; On the fusion source set, a dual-threshold amplitude limiting write-back is performed on the missing evidence segments, and structural amplitude limiting constraints and brightness amplitude limiting constraints are executed in sequence to generate image compensation data that meets the requirements of continuity and consistency. The compensation data is integrated with the original image to construct a reliable anti-counterfeiting texture mapping map; S5, based on the stable output of reliable anti-counterfeiting texture mapping, performs dual-track dynamic control of phase conjugate breathing traction and spectral reverse diffusion. It achieves synchronous gating of exposure control, shutter control and gain control through time-frequency self-suppression interference field, eliminates the blank area caused by light flicker, and constructs a steady-state anti-counterfeiting identification closed loop. Step S5 includes: After the output of the anti-counterfeiting texture trusted mapping, the texture response surface is constructed based on the image spatial energy density and structural continuity information, and a synchronous phase evolution path is established to generate a phase conjugate mapping map; Based on phase conjugate traction, spectral inverse diffusion processing is introduced to achieve signal energy stabilization compensation through reverse propagation; Under the dual-track control, a time-frequency self-suppressive interference field is formed, which performs synchronous gating operation on exposure control, shutter control and gain control; After gating is completed, an image acquisition feedback closed loop is established to achieve parameter self-adjustment and illumination fault-tolerant recognition output.

2. The anti-counterfeiting verification method based on pattern recognition according to claim 1, characterized in that, Step S1 includes: During image acquisition, a light monitoring channel covering the sampling field of view is constructed by simultaneously deploying a high-speed photosensitive detection array and imaging sensing components in the imaging path, and a nanosecond-level light time scale is formed by using a unified crystal oscillator pulse drive. Under illumination time-scale constraints, an imaging energy datum is established through continuous sampling, and the exposure phase trajectory of each pixel is extracted; Based on the energy datum, the synchronic point of energy abrupt change is identified as the starting kernel of illumination flicker, and the energy change trend before and after the starting kernel is extracted and the duration window is calibrated. In a unified illumination timescale and energy datum, spatial and temporal regions are dynamically labeled to construct a dynamic illumination command surface, which is used to identify pixel units and time nodes in the energy saturation state during image sampling.

3. The anti-counterfeiting verification method based on pattern recognition according to claim 1, characterized in that, In the process of constructing the time-frequency compensation reference surface, frequency bandwidth diffusion analysis and extreme value jump detection are performed on the spectral region corresponding to the missing certificate segment. Spatial interpolation and spectral smoothing are used to complete the low-scoring region, thereby improving the accuracy of structural identification of the missing certificate region and the continuity of the compensation reference surface.

4. The anti-counterfeiting verification method based on pattern recognition according to claim 1, characterized in that, After constructing the anti-counterfeiting texture trusted mapping map, a global brightness equalization operation and frequency bandwidth matching process are performed. By unifying the tone mapping relationship and the spectral density constraint function, the color shift and texture frequency fluctuation of the compensation area are adjusted to improve the overall consistency and stability of the image.

5. A pattern recognition-based anti-counterfeiting verification system, used to implement the pattern recognition-based anti-counterfeiting verification method according to any one of claims 1-4, characterized in that, It includes a lighting timescale construction module, an exposure inversion and reconstruction module, an anomaly identification and calibration module, a texture reconstruction and repair module, and an exposure synchronization control module: The illumination timescale construction module establishes a nanosecond-level unified illumination timescale and imaging energy datum. It obtains the exposure phase trajectory during the ticket imaging process through continuous sampling, extracts energy density distribution information based on the exposure phase trajectory, and identifies the saturation initiation kernel and continuous window caused by illumination flicker based on the energy density distribution, thus constructing a dynamic illumination command surface. The exposure inversion and reconstruction module constructs a counterfactual playback chain under the constraint of dynamic illumination command surface, performs time inversion calculation on the exposure trajectory within the saturation window, reconstructs the image energy flow change process based on the time inversion results, and generates statistical fingerprint information of key frames to form a flicker recovery baseline. The anomaly identification and calibration module trains a cross-domain anomaly discriminator based on the statistical fingerprint information in the flicker recovery baseline, and uses the frequency domain sparse mapping method and the extreme value continuity verification method to identify abnormal information fragments in the anti-counterfeiting verification area, calibrate the real missing certificate fragments, and generate a time-frequency compensation reference surface. The texture reconstruction and repair module performs multi-view polarization resampling under the constraint of time-frequency compensation reference surface, performs double-threshold amplitude-limited write-back calculation based on the calibrated missing certificate fragments, and reconstructs the reliable mapping of anti-counterfeiting texture; The exposure synchronization control module, based on the stable output of the reliable mapping of anti-counterfeiting texture, performs dual-track dynamic regulation of phase conjugate breathing traction and spectral reverse diffusion. It achieves synchronous gating of exposure control, shutter control and gain control through time-frequency self-suppressed interference field, eliminates the blank area caused by light flicker, and constructs a steady-state anti-counterfeiting identification closed loop.

Images