False cigarette detection and information intelligent auditing system based on cigarette-related law enforcement

Abstract

The invention discloses a false cigarette detection and information intelligent auditing system based on cigarette-related law enforcement, and relates to the technical field of false cigarette detection.A cigarette case quality detection module verifies a bar code through code scanning and judges whether a cigarette case meets authenticity identification requirements, if yes, a combustion behavior characteristic model analysis module collects combustion behavior parameters, and if not, the combustion behavior parameters are analyzed; calculating combustion rate variance and ash layering coefficient, comparing the combustion rate variance and the ash layering coefficient with a standard threshold value to evaluate whether the cigarette accords with the characteristics of the false cigarette, acquiring multi-modal parameters such as cigarette case image variation coefficient by a comprehensive false cigarette probability evaluation module when an evaluation result is doubted, calculating a comprehensive false cigarette probability score, and outputting the comprehensive false cigarette probability score. And verifying an evaluation result, and if the evaluation result is not accurate, simulating a high-simulation false cigarette sample through a generative adversarial network by utilizing a misjudgment sample, optimizing a combustion behavior characteristic model and improving false cigarette judgment accuracy, so that complex conditions in false cigarette detection work are effectively handled.

Description

Technical Field

[0001] This invention relates to the field of counterfeit cigarette detection technology, specifically to a counterfeit cigarette detection and intelligent information verification system for tobacco-related law enforcement. Background Technology

[0002] The current counterfeit cigarette market is rife with problems, which not only harms consumers' rights but also seriously affects the healthy development of the tobacco industry. As counterfeiting technology continues to upgrade, traditional counterfeit cigarette detection methods are unable to guarantee accuracy and efficiency when faced with complex, diverse, and highly realistic counterfeit cigarettes. In order to effectively address these challenges, a counterfeit cigarette detection and information intelligent review system for tobacco law enforcement has emerged to meet the urgent need for accurate and efficient detection of counterfeit cigarettes.

[0003] Existing technology, such as the invention patent application with publication number CN113627951A, discloses an internet-based intelligent verification system for detecting counterfeit cigarettes for tobacco law enforcement, relating to the field of tobacco law enforcement technology. This system includes an APP platform module, a cloud database module, a data processing module, and a data transmission module. The APP platform module includes a communication unit, a legal document unit, an evidence storage unit, a detection and identification unit, and a customizable verification process. The detection and identification unit includes a photo-taking unit and an image-text recognition unit. These units photograph and identify the inkjet and barcode codes on the cigarette pack to generate text. The generated inkjet and barcode text is then compared with information in the inkjet and barcode information databases until the corresponding cigarette information is found. By comparing the information corresponding to the two types of text, it can be determined whether the cigarette is counterfeit, ensuring the accuracy of the detection results and making this system worthy of widespread promotion.

[0004] The above solutions have at least the following technical problems: 1. The above solutions lack the collection and analysis of cigarette combustion behavior parameters, which makes it impossible to judge the authenticity of cigarettes from the perspective of combustion characteristics. During the combustion process, there are often differences between counterfeit and genuine cigarettes in combustion behavior parameters such as the linear displacement velocity per second, the total combustion time, and the gray value of the combustion image. The above solutions do not involve the collection and analysis of these parameters, so the system can only rely on the cigarette box inkjet code and barcode information to judge authenticity. For counterfeit cigarettes manufactured by tampering with barcodes or using genuine cigarette inkjet codes, it is impossible to effectively identify them from the combustion essence, which greatly weakens the system's ability to detect counterfeit cigarettes and easily leads to missed detection of counterfeit cigarettes.

[0005] 2. The above-mentioned scheme lacks a multimodal parameter acquisition and comprehensive verification process, which leads to the system relying solely on single-dimensional information comparison when judging the authenticity of cigarettes. When encountering counterfeit cigarettes with high degree of counterfeiting and difficult-to-distinguish inkjet and barcode information, it is unable to conduct comprehensive verification through multimodal parameters such as the coefficient of variation of cigarette box image, coefficient of variation of burning rate, and coefficient of variation of ash image. This makes it difficult to accurately judge the authenticity of cigarettes, reduces the system's ability to deal with complex counterfeit cigarette situations, and fails to meet the increasingly diverse needs of counterfeit cigarette detection.

[0006] 3. The above scheme lacks an optimization mechanism for the counterfeit cigarette detection feature model. As counterfeit cigarette manufacturing technology continues to upgrade, the system will be unable to adapt to newly emerging counterfeit cigarette features in a timely manner when detecting new types of counterfeit cigarettes. Since there is no mechanism to feed back misjudged samples to optimize the model, new counterfeit cigarette features discovered by law enforcement officers during the detection process cannot be learned and absorbed by the system. The system will always use fixed judgment standards for detection, causing the detection accuracy to gradually decline. It will be difficult to effectively cope with the challenges of new counterfeit cigarettes and cannot guarantee the long-term effectiveness and accuracy of counterfeit cigarette detection work. Summary of the Invention

[0007] The purpose of this invention is to provide a counterfeit cigarette detection and information intelligent review system for tobacco-related law enforcement, which solves the problems existing in the background technology.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides a counterfeit cigarette detection and information intelligent review system based on tobacco-related law enforcement, including: a cigarette box quality detection module, used to mark a specified cigarette as a cigarette to be detected, and analyze whether the cigarette box corresponding to the cigarette to be detected meets the requirements for authenticity identification.

[0009] The combustion behavior characteristic model analysis module is used to collect the combustion behavior parameters of the cigarettes to be tested when the cigarette box meets the requirements for authenticity identification, and then evaluate whether the cigarettes to be tested meet the characteristics of counterfeit cigarettes.

[0010] The comprehensive counterfeit cigarette probability assessment module is used to obtain the multimodal parameters of the cigarette to be tested when the cigarette to be tested does not meet the characteristics of a counterfeit cigarette, and then verify whether the assessment result of whether the cigarette to be tested meets the characteristics of a counterfeit cigarette is accurate.

[0011] The feature model optimization module is used to optimize the accuracy of counterfeit cigarette judgment during the counterfeit cigarette detection process when the evaluation result of whether the cigarette to be detected meets the characteristics of counterfeit cigarette is inaccurate.

[0012] The beneficial effects of the present invention are as follows: 1. The counterfeit cigarette detection and information intelligent review system based on tobacco-related law enforcement provided by the embodiments of the present invention compares the barcode on the surface of the cigarette to be tested with the database through the mobile phone scanning function during the cigarette box quality inspection process. This facilitates the rapid determination of whether the cigarette box meets the requirements for authenticity identification, helps to screen out cigarettes with abnormal barcodes in the early stage of detection, avoids the need for subsequent complex testing processes for cigarettes that cannot be properly identified, saves testing time and resources, and also improves the overall efficiency and targeting of the detection, ensuring that subsequent testing is based on valid cigarette boxes that meet the identification conditions.

[0013] 2. In the process of assessing whether a cigarette is counterfeit, this invention calculates the combustion rate variance and ash stratification coefficient and compares them with standard thresholds. This facilitates accurate judgment from the key dimension of cigarette combustion characteristics. Since counterfeit and genuine cigarettes have different physicochemical properties during combustion, this process, by quantifying these parameters, can capture subtle differences that are difficult to detect with the naked eye. For counterfeit cigarettes that are difficult to distinguish from appearance alone, this provides a scientific and reliable basis for judgment, greatly improving the accuracy of counterfeit cigarette detection and reducing the risk of misjudgment.

[0014] 3. In the process of verifying the accuracy of the evaluation results, the embodiments of the present invention calculate a comprehensive fake cigarette probability score and compare it with a preset threshold. This is beneficial for in-depth verification by integrating multi-dimensional information. When it is difficult to make an accurate judgment based solely on combustion behavior parameters, the introduction of multi-modal parameters analyzes the characteristics of cigarettes from different angles, avoiding the limitations of a single judgment dimension. This is beneficial for a more comprehensive evaluation of the authenticity of cigarettes and can effectively identify high-quality counterfeit cigarettes. This enhances the system's ability to cope with complex counterfeit cigarette situations and improves the credibility and reliability of the detection results.

[0015] 4. In the process of optimizing the accuracy of counterfeit cigarette detection, this embodiment of the invention feeds back the labels of misjudged samples to the feature model optimization module. When the number of misjudged samples reaches a threshold, a generative adversarial network is used to simulate highly realistic counterfeit cigarette samples. These samples are then weighted and aggregated in conjunction with the training results uploaded by the law enforcement terminal to optimize the combustion behavior feature model. This helps the system adapt to the ever-changing counterfeit cigarette manufacturing technology. As the counterfeit cigarette manufacturing process is upgraded, new counterfeit cigarette features continue to emerge. This optimization mechanism enables the system to continuously learn new features, automatically update the judgment criteria, and always maintain a high accuracy rate in counterfeit cigarette detection. This ensures that the system can effectively cope with the challenges of new counterfeit cigarettes during long-term use and maintain the effectiveness and advancement of counterfeit cigarette detection work. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0018] 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 embodiments of the present invention, and not all embodiments. 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.

[0019] Please see Figure 1 As shown, this invention provides a counterfeit cigarette detection and information intelligent review system based on tobacco-related law enforcement. The system includes: a cigarette box quality detection module, a combustion behavior feature model analysis module, a comprehensive counterfeit cigarette probability assessment module, a feature model optimization module, and a database.

[0020] The cigarette box quality detection module is connected to the combustion behavior feature model analysis module and the database, the combustion behavior feature model analysis module is connected to the comprehensive counterfeit cigarette probability assessment module and the database, and the comprehensive counterfeit cigarette probability assessment module is connected to the feature model optimization module and the database.

[0021] The cigarette box quality inspection module is used to mark specified cigarettes as cigarettes to be inspected and analyze whether the corresponding cigarette boxes meet the requirements for authenticity identification.

[0022] In a specific embodiment, the process of analyzing whether the cigarette box corresponding to the cigarette to be tested meets the requirements for authenticity verification is as follows: When the cigarette to be tested needs to be authenticated, the barcode on the surface of the cigarette to be tested is scanned and verified using the scanning function of a mobile phone. The system compares the scanned barcode on the surface of the cigarette to be tested with the database. If a matching record is found in the database, it indicates that the cigarette box corresponding to the cigarette to be tested meets the requirements for authenticity verification; otherwise, it indicates that the cigarette box corresponding to the cigarette to be tested does not meet the requirements for authenticity verification, and the cigarette to be tested is determined to be counterfeit.

[0023] It should be noted that the specific process of scanning and verifying the barcode on the surface of the cigarette to be tested using the mobile phone's scanning function has been described in detail in the patent with announcement number CN113627951A, and will not be repeated here.

[0024] The combustion behavior characteristic model analysis module is used to collect the combustion behavior parameters of the cigarettes to be tested when the cigarette box meets the requirements for authenticity identification, and then evaluate whether the cigarettes to be tested meet the characteristics of counterfeit cigarettes.

[0025] In a specific embodiment, the process of collecting combustion behavior parameters corresponding to the cigarettes to be tested is as follows: Combustion behavior parameters include the combustion linear displacement velocity per second, the total combustion time, and the grayscale value of the combustion image. When the cigarette box of the cigarettes to be tested meets the requirements for authenticity verification, each cigarette is randomly selected from the cigarettes to be tested. The combustion of each cigarette during the smoking process is simulated using a simulation device. The changes in the combustion line position of each cigarette at each time point are acquired using a high-speed camera and thermal imaging sensor. The time interval corresponding to each time point is set to 1 second. The real-time changes in the corresponding combustion line of each cigarette during the combustion process are determined by the thermal imaging sensor, thereby obtaining the combustion linear displacement velocity v of each cigarette at each time point. i ,t, and then the average burning rate v′ corresponding to each cigarette is calculated by the arithmetic mean. i Where t is the number corresponding to each time point, i is the number corresponding to each cigarette, t = 1, 2, ..., t′, t′ is a positive integer, i = 1, 2, ..., n, n is a positive integer, and the total burning time corresponding to each cigarette is directly read from the high-speed camera equipment.

[0026] Images of the ash formed after each cigarette is burned are captured using a high-definition camera. Image processing techniques are then used to convert these ash images into grayscale images. The grayscale levels of each cigarette's corresponding grayscale image are obtained from the image processing software, leading to the number of rows and columns in the grayscale co-occurrence matrix (GCM) of each cigarette's corresponding grayscale image. This constitutes the GCM for each cigarette's corresponding grayscale image, and the number of rows and columns of this GCM is denoted as x. i and y i x i and y i It is also represented as the gray value of the grayscale image corresponding to the i-th cigarette.

[0027] It should be noted that "simulating the combustion of each cigarette during the smoking process through simulation equipment" refers to using a device similar to a "manual smoking machine" or "standardized inhalation simulator" to set parameters such as frequency, air flow rate, and inhalation duration to simulate the actual inhalation process of a human smoking a cigarette. For example, it can be set to inhale once every 30 seconds, inhaling 35 milliliters of gas each time, lasting for 2 seconds, and then restoring after an interval, realistically reproducing the cigarette combustion process caused by human smoking.

[0028] It should also be noted that the combustion line displacement velocity is obtained by recording the distance the combustion line of the cigarette moves forward in each second using a thermal imaging sensor. For example, if it moves forward by 0.8 mm between the 1st and 2nd second, the velocity is 0.8 mm per second. The average displacement velocity per second is used to obtain the average combustion rate of the cigarette throughout the entire combustion process.

[0029] In a specific embodiment, the process of evaluating whether the cigarette to be tested meets the characteristics of counterfeit cigarettes is as follows: Based on the combustion behavior parameters corresponding to the cigarette to be tested, the combustion rate variance α and ash stratification coefficient β corresponding to the cigarette to be tested are calculated. The standard combustion rate variance threshold α′ and standard ash stratification coefficient threshold β′ corresponding to genuine cigarettes of the same type as the cigarette to be tested are obtained from the database. When α is less than or equal to α′ and β is less than or equal to β′, it indicates that the cigarette to be tested does not meet the characteristics of counterfeit cigarettes. Otherwise, it indicates that the cigarette to be tested meets the characteristics of counterfeit cigarettes.

[0030] It should be noted that, in addition to these cases, there are also cases where α is less than or equal to α′ and β is greater than β′, α is greater than α′ and β is less than or equal to β′, and α is greater than α′ and β is greater than β′.

[0031] In a specific embodiment, the calculation of the combustion rate variance and ash stratification coefficient of the cigarette to be tested is carried out as follows: The calculation formula is as follows:

[0032] The variance α of the combustion rate corresponding to the cigarette to be tested is obtained.

[0033] Calculation formula based on ash layering:

[0034] The ash stratification coefficient β corresponding to the cigarette to be tested is obtained, where P(x i ,y i ) represents the frequency of occurrence of pixels with gray values ​​x and y corresponding to the i-th cigarette.

[0035] It should be noted that, for example, if a pixel in a grayscale image has a grayscale value of 100 and its adjacent pixel has a grayscale value of 110, then P(100,110) represents the frequency of the pixel pair with grayscale values ​​of 100 and 110 in the grayscale image.

[0036] In assessing whether a cigarette is counterfeit, this invention calculates the combustion rate variance and ash stratification coefficient and compares them with standard thresholds. This facilitates accurate judgment from the key dimension of cigarette combustion characteristics. Since counterfeit and genuine cigarettes differ in their physicochemical properties during combustion, this process, by quantifying these parameters, can capture subtle differences that are difficult to detect with the naked eye. For counterfeit cigarettes that are difficult to distinguish from appearance alone, this provides a scientific and reliable basis for judgment, greatly improving the accuracy of counterfeit cigarette detection and reducing the risk of misjudgment.

[0037] The comprehensive counterfeit cigarette probability assessment module is used to obtain the multimodal parameters of the cigarette to be tested when the cigarette to be tested does not meet the characteristics of a counterfeit cigarette, and then verify whether the assessment result of whether the cigarette to be tested meets the characteristics of a counterfeit cigarette is accurate.

[0038] In a specific embodiment, the process of obtaining the multimodal parameters corresponding to the cigarette to be detected is as follows: the multimodal parameters include the coefficient of variation of the cigarette box image, the coefficient of variation of the burning rate, and the coefficient of variation of the ash image. The coefficient of variation of the cigarette box image is obtained by analyzing the fluctuation of the image features, the coefficient of variation of the burning rate is obtained by analyzing the fluctuation of the burning rate, and the coefficient of variation of the ash image is obtained by analyzing the fluctuation of the spectral data.

[0039] It should be noted that, taking the coefficient of variation of burning rate as an example, it is calculated by "standard deviation ÷ mean" and is used to measure the degree of fluctuation of this parameter at different points in time. For example, if the burning rate of a cigarette is 0.75 mm / s, 0.80 mm / s, 0.70 mm / s, 0.85 mm / s, and 0.90 mm / s, with an average of 0.80 mm / s and a standard deviation of approximately 0.07 mm / s, then the coefficient of variation is 0.07 divided by 0.80, which equals 0.0875, indicating that the burning rate fluctuates by approximately 8.75%. The process of obtaining the coefficient of variation of the cigarette pack image and the coefficient of variation of the ash image is the same as that of obtaining the coefficient of variation of burning rate, and will not be elaborated further here.

[0040] In a specific embodiment, the process for verifying the accuracy of the assessment result regarding whether the cigarette to be detected meets the characteristics of a counterfeit cigarette is as follows: Based on the multimodal parameters corresponding to the cigarette to be detected, and obtaining preset baseline weight factors corresponding to the coefficient of variation of the cigarette box image, the coefficient of variation of the burning rate, and the coefficient of variation of the ash image from the database, and denoted as μ1, μ2, and μ3, the dynamic weight factors corresponding to each multimodal parameter are calculated. Based on the dynamic weight factors corresponding to each item in the multimodal parameters, the comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is calculated. A preset comprehensive counterfeit cigarette probability score threshold is obtained from the database. The comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is compared with the preset comprehensive counterfeit cigarette probability score threshold. If the comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is greater than the preset comprehensive counterfeit cigarette probability score threshold, it indicates that the cigarette to be detected is a counterfeit cigarette, and the assessment result regarding whether the cigarette to be detected meets the characteristics of a counterfeit cigarette is inaccurate. Conversely, if the score is less than the threshold, it indicates that the cigarette to be detected is a genuine cigarette, and the assessment result regarding whether the cigarette to be detected meets the characteristics of a counterfeit cigarette is accurate.

[0041] It should be noted that the benchmark weight factor refers to the reference value used to measure the importance of different modal features, such as cigarette box images, burning rates, and ash images, in the overall discrimination process when comprehensively evaluating the probability score of counterfeit cigarettes. It is equivalent to assigning a "weight reference standard" to each feature, which determines the influence of the feature on the final score. For example, if the fluctuation of the burning rate is found to contribute the most to the identification of counterfeit cigarettes in historical detection, then the benchmark weight factor of the coefficient of variation of the burning rate will be set high.

[0042] It should also be noted that the preset process for the benchmark weight factors is set through large-scale sample data analysis. For example, multimodal parameter data of 1000 genuine cigarettes and 1000 counterfeit cigarettes are collected, and then the discrimination accuracy of the coefficient of variation of image, rate, and ash features is calculated. Suppose the analysis results show that the probability of correctly identifying counterfeit cigarettes by the coefficient of variation of the cigarette box image is 60%, the coefficient of variation of the burning rate is 80%, and the coefficient of variation of the ash image is 70%, then the benchmark weight factors are set as follows: image 0.6, rate 0.8, and ash 0.7. Then the benchmark weight factors are stored in the database as a reference for dynamic weight adjustment when the cigarettes to be tested are comprehensively scored. The values ​​of the benchmark weight factors corresponding to the coefficient of variation of the cigarette box image, the coefficient of variation of the burning rate, and the coefficient of variation of the ash image are all greater than 0 and less than 1. The coefficient of variation is a statistical indicator describing the degree of dispersion of data.

[0043] In a specific embodiment, the calculation of the dynamic weight factors corresponding to the multimodal parameters is carried out as follows: The dynamic weight factors corresponding to the coefficient of variation of the cigarette box image, the coefficient of variation of the burning rate, and the coefficient of variation of the ash image are respectively denoted as... and Let q1, q2, and q3 be the coefficients of variation for the cigarette box image, the burning rate, and the ash image, respectively. Where p is the index of each item in the multimodal parameters, p = 1, 2, 3 correspond to the first item (cigarette box image variation coefficient), the second item (burning rate variation coefficient), and the third item (ash image variation coefficient) in the multimodal parameters, respectively; μ1 is the baseline weighting factor corresponding to the cigarette box image in the multimodal parameters; and q1 represents the cigarette box image variation coefficient in the multimodal parameters. p Let q be the baseline weighting factor corresponding to the p-th term in the multimodal parameters. p This represents the specific value corresponding to the p-th term in the multimodal parameters.

[0044] It should be noted that, and The values ​​of are all greater than 0 and less than 1. and The calculation process and The calculation process is the same, so I will not go into details here.

[0045] In a specific embodiment, the calculation of the comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is carried out as follows: The comprehensive probability score is calculated using the following formula: The comprehensive counterfeit cigarette probability score Pf corresponding to the cigarette to be tested is obtained, where S p The standardized score for the analysis process corresponds to the p-th parameter in the multimodal parameters. It is represented as the dynamic weight factor corresponding to the p-th term in the multimodal parameters.

[0046] It should be noted that S p It is used to reflect the degree of anomaly or the probability of counterfeit cigarettes in this modal dimension of the cigarette being tested, and is calculated using the following formula: Obtain the standardized score S corresponding to the p-th term in the multimodal parameters. p H p H′ represents the modal parameter value of the p-th term in the multimodal parameters. p and ζ p These are the mean and standard deviation of the p-th term in the multimodal parameters corresponding to the real cigarette sample. For example, suppose the coefficient of variation of the burning rate of the current cigarette being tested is 0.42, while the mean of the coefficient of variation of the burning rate of real cigarettes in the database is 0.30 and the standard deviation is 0.05. Then the standardized score of the coefficient of variation of the burning rate is:

[0047] In the process of verifying the accuracy of the evaluation results, this invention calculates a comprehensive fake cigarette probability score and compares it with a preset threshold. This facilitates in-depth verification by integrating multi-dimensional information. When it is difficult to make an accurate judgment based solely on combustion behavior parameters, the introduction of multi-modal parameters analyzes cigarette characteristics from different angles, avoiding the limitations of a single judgment dimension. This allows for a more comprehensive evaluation of the authenticity of cigarettes and can effectively identify even highly counterfeit cigarettes. This enhances the system's ability to cope with complex fake cigarette situations and improves the credibility and reliability of the detection results.

[0048] The feature model optimization module is used to optimize the accuracy of counterfeit cigarette judgment during the counterfeit cigarette detection process when the evaluation result of whether the cigarette to be detected meets the characteristics of counterfeit cigarette is inaccurate.

[0049] In a specific embodiment, the process of optimizing the accuracy of counterfeit cigarette detection is as follows: When the evaluation result of whether the cigarette to be detected meets the characteristics of counterfeit cigarette is inaccurate, the cigarette to be detected is recorded as a misjudged sample, and the misjudged sample is marked and fed back to the feature model optimization module. When the number of misjudged samples is greater than or equal to the preset threshold for the number of misjudged samples, based on each misjudged sample, a highly similar counterfeit cigarette sample is simulated by a generative adversarial network. The law enforcement terminal uploads the training results of the combustion behavior feature model of each misjudged sample on the newly added sample to the cloud, performs weighted aggregation according to the proportion of the number of misjudged samples, calculates the global model parameters, and finally realizes the adaptive optimization of the combustion behavior feature model for the new counterfeit cigarette features within a set range, thereby achieving the purpose of optimizing the accuracy of counterfeit cigarette detection.

[0050] It should be noted that the preset threshold for the number of misjudged samples serves as the basis for evaluating whether to begin simulating highly similar counterfeit cigarette samples using generative adversarial networks (GANs). For example, when an enforcement terminal in a certain area detects a batch of cigarettes, the system finds that 5 cigarettes are misjudged as genuine, but subsequent manual verification confirms they are counterfeit. These are recorded as "misjudged samples." If the preset threshold for the number of misjudged samples is 5, once the number of misjudged samples reaches this value, the system feeds each misjudged sample back to the feature model optimization module. The GAN is then used to generate highly similar counterfeit cigarette samples with similar characteristics to each misjudged sample. Each enforcement terminal then uses each new sample to retrain its own combustion behavior feature model and uploads the updated model parameters to the cloud. The system weights the parameters according to the proportion of new samples added by each terminal and synthesizes new global model parameters, thereby achieving adaptive learning and optimization of new counterfeit cigarette features. The preset threshold for the number of misjudged samples is usually determined by statistically analyzing the misjudgment frequency based on historical detection data. For example, experience shows that if 5 cigarettes are misjudged out of 1000 cigarettes detected, it is considered highly representative, so the threshold is set to 5.

[0051] In optimizing the accuracy of counterfeit cigarette detection, this invention feeds back misjudged samples to the feature model optimization module. When the number of misjudged samples reaches a threshold, a generative adversarial network is used to simulate highly realistic counterfeit cigarette samples. These samples are then weighted and aggregated with training results uploaded from law enforcement terminals to optimize the combustion behavior feature model. This allows the system to adapt to constantly evolving counterfeit cigarette manufacturing technologies. As counterfeit cigarette manufacturing processes upgrade and new counterfeit cigarette features emerge, this optimization mechanism enables the system to continuously learn new features, automatically update judgment criteria, and maintain a high accuracy rate in counterfeit cigarette detection. This ensures that the system can effectively cope with the challenges of new counterfeit cigarettes during long-term use, maintaining the effectiveness and advancement of counterfeit cigarette detection.

[0052] The database stores the barcode information of genuine cigarettes when they leave the factory, as well as the standard burning rate variance threshold and standard ash stratification coefficient threshold corresponding to genuine cigarettes of the same type as the cigarettes to be tested. It also stores the preset baseline weight factors corresponding to the coefficient of variation of cigarette box images, the coefficient of variation of burning rate, and the coefficient of variation of ash images, and the preset comprehensive counterfeit cigarette probability score threshold.

[0053] The counterfeit cigarette detection and intelligent information verification system for tobacco-related law enforcement provided in this invention compares the barcode on the surface of the cigarette to be tested with a database through a mobile phone scanning function during the cigarette box quality inspection process. This facilitates a quick determination of whether the cigarette box meets the requirements for authenticity identification, helps to screen out cigarettes with abnormal barcodes in the early stages of inspection, avoids complicated subsequent inspection processes for cigarettes that cannot be properly identified, saves inspection time and resources, and improves the overall efficiency and targeting of the inspection, ensuring that subsequent inspections are based on valid cigarette boxes that meet the identification conditions.

[0054] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. A counterfeit cigarette detection and intelligent information verification system for tobacco-related law enforcement, characterized in that: include: The cigarette box quality inspection module is used to mark specified cigarettes as cigarettes to be inspected and analyze whether the cigarette boxes corresponding to the cigarettes to be inspected meet the requirements for authenticity identification. The combustion behavior characteristic model analysis module is used to collect the combustion behavior parameters of the cigarettes to be tested when the cigarette box meets the requirements for authenticity identification, and then evaluate whether the cigarettes to be tested meet the characteristics of counterfeit cigarettes. The comprehensive counterfeit cigarette probability assessment module is used to obtain the multimodal parameters of the cigarette to be tested when the cigarette to be tested does not meet the characteristics of counterfeit cigarettes, and then verify whether the assessment result of whether the cigarette to be tested meets the characteristics of counterfeit cigarettes is accurate. The feature model optimization module is used to optimize the accuracy of counterfeit cigarette judgment during the counterfeit cigarette detection process when the evaluation result of whether the cigarette to be detected meets the characteristics of counterfeit cigarette is inaccurate.

2. The intelligent verification system for detecting counterfeit cigarettes used in tobacco-related law enforcement as described in claim 1, characterized in that, The specific process for analyzing whether the cigarette pack corresponding to the cigarette to be tested meets the requirements for authenticity verification is as follows: When a cigarette needs to be authenticated, the barcode on the surface of the cigarette is scanned using a mobile phone. The system compares the scanned barcode with the database. If a matching record is found in the database, the cigarette box meets the authentication requirements; otherwise, the cigarette box does not meet the authentication requirements, and the cigarette is determined to be counterfeit.

3. The counterfeit cigarette detection and intelligent information verification system for tobacco-related law enforcement as described in claim 2, characterized in that, The specific process for collecting the combustion behavior parameters of the cigarette to be tested is as follows: Combustion behavior parameters include the linear displacement velocity per second, total combustion time, and grayscale value of the combustion image. When the cigarette pack meets the requirements for authenticity verification, individual cigarettes are randomly selected from the cigarette pack. The combustion of each cigarette during the smoking process is simulated using a simulation device. The changes in the position of the combustion line of each cigarette at each time point are acquired using a high-speed camera and thermal imaging sensor. The time interval corresponding to each time point is set to 1 second. The real-time changes of the corresponding combustion line of each cigarette during the combustion process are determined by the thermal imaging sensor, thereby obtaining the linear displacement velocity v of the combustion line of each cigarette at each time point. i ,t, and then the average burning rate v′ corresponding to each cigarette is calculated by the arithmetic mean. i Where t is the number corresponding to each time point, i is the number corresponding to each cigarette, t = 1, 2, ..., t′, t′ is a positive integer, i = 1, 2, ..., n, n is a positive integer, and the total burning time corresponding to each cigarette is directly read from the high-speed camera equipment; Images of the ash formed after each cigarette is burned are captured using a high-definition camera. Image processing techniques are then used to convert these ash images into grayscale images. The grayscale levels of each cigarette's corresponding grayscale image are obtained from the image processing software, leading to the number of rows and columns in the grayscale co-occurrence matrix (GCM) of each cigarette's corresponding grayscale image. This constitutes the GCM for each cigarette's corresponding grayscale image, and the number of rows and columns of this GCM is denoted as x. i and y i x i and y i It is also represented as the gray value of the grayscale image corresponding to the i-th cigarette.

4. The counterfeit cigarette detection and intelligent information verification system for tobacco-related law enforcement as described in claim 3, characterized in that, The specific process for assessing whether the cigarette to be tested meets the characteristics of counterfeit cigarettes is as follows: Based on the combustion behavior parameters of the cigarette to be tested, the combustion rate variance α and ash stratification coefficient β of the cigarette to be tested are calculated. The standard combustion rate variance threshold α′ and standard ash stratification coefficient threshold β′ of the same type of genuine cigarette as the cigarette to be tested are obtained from the database. When α is less than or equal to α′ and β is less than or equal to β′, it indicates that the cigarette to be tested does not meet the characteristics of a counterfeit cigarette. Otherwise, it indicates that the cigarette to be tested meets the characteristics of a counterfeit cigarette.

5. The intelligent verification system for detecting counterfeit cigarettes used in tobacco-related law enforcement as described in claim 4, characterized in that, The specific process for calculating the combustion rate variance and ash stratification coefficient of the cigarette to be tested is as follows: Calculation formula: The variance α of the combustion rate corresponding to the cigarette to be tested is obtained; Calculation formula based on ash layering: The ash stratification coefficient β corresponding to the cigarette to be tested is obtained, where P(x i ,y i ) represents the frequency of occurrence of pixels with gray values ​​x and y corresponding to the i-th cigarette.

6. The intelligent verification system for detecting counterfeit cigarettes used in tobacco-related law enforcement as described in claim 5, characterized in that, The specific process for obtaining the multimodal parameters corresponding to the cigarette to be detected is as follows: The multimodal parameters include the coefficient of variation of the cigarette box image, the coefficient of variation of the combustion rate, and the coefficient of variation of the ash image. The coefficient of variation of the cigarette box image is obtained by analyzing the fluctuation of image features, the coefficient of variation of the combustion rate is obtained by analyzing the fluctuation of the combustion rate, and the coefficient of variation of the ash image is obtained by analyzing the fluctuation of spectral data.

7. The intelligent verification system for detecting counterfeit cigarettes used in tobacco-related law enforcement as described in claim 6, characterized in that, The specific process for verifying the accuracy of the assessment results regarding whether the cigarettes to be tested meet the characteristics of counterfeit cigarettes is as follows: Based on the multimodal parameters corresponding to the cigarette to be detected, and retrieving from the database the preset baseline weight factors corresponding to the coefficient of variation of the cigarette box image, the coefficient of variation of the burning rate, and the coefficient of variation of the ash image, denoted as μ1, μ2, and μ3, the dynamic weight factors corresponding to each multimodal parameter are calculated. Based on the dynamic weight factors corresponding to each item in the multimodal parameters, the comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is calculated. The preset comprehensive counterfeit cigarette probability score threshold is retrieved from the database. The comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is compared with the preset comprehensive counterfeit cigarette probability score threshold. If the comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is greater than the preset comprehensive counterfeit cigarette probability score threshold, it indicates that the cigarette to be detected is counterfeit, and the evaluation result of whether the cigarette to be detected meets the characteristics of counterfeit cigarette is inaccurate. Conversely, if the score is less than the threshold, it indicates that the cigarette to be detected is genuine, and the evaluation result of whether the cigarette to be detected meets the characteristics of counterfeit cigarette is accurate.

8. The intelligent verification system for detecting counterfeit cigarettes used in tobacco-related law enforcement as described in claim 7, characterized in that, The specific process for calculating the dynamic weight factors corresponding to the multimodal parameters is as follows: The dynamic weighting factors corresponding to the coefficient of variation of the cigarette box image, the coefficient of variation of the burning rate, and the coefficient of variation of the ash image are respectively denoted as follows: and Let q1, q2, and q3 be the coefficients of variation for the cigarette box image, the burning rate, and the ash image, respectively. Where p is the index of each item in the multimodal parameters, p = 1, 2, 3 correspond to the first item (cigarette box image variation coefficient), the second item (burning rate variation coefficient), and the third item (ash image variation coefficient) in the multimodal parameters, respectively; μ1 is the baseline weighting factor corresponding to the cigarette box image in the multimodal parameters; and q1 represents the cigarette box image variation coefficient in the multimodal parameters. p Let q be the baseline weighting factor corresponding to the p-th term in the multimodal parameters. p This represents the specific value corresponding to the p-th term in the multimodal parameters.

9. The intelligent verification system for detecting counterfeit cigarettes used in tobacco-related law enforcement as described in claim 8, characterized in that, The specific process for calculating the comprehensive counterfeit cigarette probability score corresponding to the cigarette to be detected is as follows: Calculated using the comprehensive probability score formula: The comprehensive counterfeit cigarette probability score Pf corresponding to the cigarette to be tested is obtained, where S p The standardized score for the analysis process corresponds to the p-th parameter in the multimodal parameters. It is represented as the dynamic weight factor corresponding to the p-th term in the multimodal parameters.

10. The counterfeit cigarette detection and intelligent information verification system for tobacco-related law enforcement as described in claim 9, characterized in that, The process for optimizing the accuracy of counterfeit cigarette detection is as follows: When the assessment of whether the cigarette to be detected meets the characteristics of counterfeit cigarettes is inaccurate, the cigarette to be detected is recorded as a misjudged sample, and the misjudged sample is marked and fed back to the feature model optimization module. When the number of misjudged samples is greater than or equal to the preset threshold for the number of misjudged samples, based on each misjudged sample, a highly similar counterfeit cigarette sample is simulated by a generative adversarial network. The law enforcement terminal uploads the training results of the combustion behavior feature model of each misjudged sample on the newly added samples to the cloud, performs weighted aggregation according to the proportion of the number of misjudged samples, calculates the global model parameters, and finally realizes the adaptive optimization of the combustion behavior feature model for the characteristics of new counterfeit cigarettes within a set range, thereby achieving the goal of optimizing the accuracy of the corresponding counterfeit cigarette judgment in the counterfeit cigarette detection process.

Images