Intelligent inspection system and method for over-packaged products

Abstract

The invention discloses an intelligent inspection system and method for an over-packaged product, and the method comprises the steps: S1, constructing an intelligent recognition model through the combination of a generative adversarial network and multi-modal neural symbol learning, and generating a robustness detection result for carrying out the dynamic semantic segmentation of a complex form of a packaged product; s2, generating a multi-dimensional key feature matrix of the packaging structure and the material by adopting a multi-dimensional data optimization technology assisted by quantum computing; s3, based on a block chain distributed storage technology, generating a multi-level credible data chain according to the detection result of the packaging material and the structure; s4, carrying out nondestructive testing on the spectral response characteristics of the packaging material by utilizing a self-adaptive spectral analysis technology; s5, performing multi-scale quantitative analysis on the defect area, the redundant structure and the unnecessary material part in the packaging product; and S6, generating a package optimization evaluation model, and providing a green design improvement scheme and a resource utilization rate optimization suggestion. The method has the advantages of high detection precision, high resource utilization rate and comprehensive optimization scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of product packaging inspection, and in particular to an intelligent inspection system and method for over-packaged products. Background Art

[0002] With growing global environmental awareness, green design and resource utilization optimization of packaging materials have become key concerns in the industrial manufacturing and consumer goods industries. In particular, in the product packaging sector, excessive packaging not only wastes resources and increases costs, but also poses a serious challenge to environmental protection and sustainable development. Intelligent inspection technology for over-packaged products aims to identify redundant design, defective areas, and non-essential components in packaging materials through advanced detection and optimization methods, thereby providing green improvement recommendations to achieve efficient resource utilization and environmental protection goals.

[0003] In existing technologies, traditional over-packaging detection methods mainly rely on rule-based manual detection systems or static data analysis mechanisms. These methods often show obvious limitations when faced with complex and changing packaging forms and material properties, including the following aspects:

[0004] 1. Insufficient detection accuracy: Traditional methods rely primarily on fixed rules and single-dimensional data detection methods, making it difficult to identify microscopic components and analyze material thickness in complex packaging shapes. Consequently, the detection results lack accuracy and credibility.

[0005] 2. Inefficient recognition: Faced with diverse packaging materials and designs, traditional systems are inefficient when processing large-scale, multimodal data, making it difficult to quickly extract key features and achieve real-time detection.

[0006] 3. Lack of data credibility: Due to the lack of a reliable data traceability and authentication mechanism, traditional testing data is susceptible to human intervention or tampering, resulting in insufficient reliability of optimization recommendations.

[0007] 4. Insufficient technology integration capabilities: Existing solutions are often limited to single technical means, such as optical detection or rule matching. They lack the integration and application of advanced technologies such as generative adversarial networks, quantum computing, and blockchain, and are unable to provide comprehensive green optimization solutions.

[0008] 5. Limited optimization capabilities: Traditional methods focus on detecting packaging defects while ignoring in-depth analysis of resource utilization and design optimization, and are unable to effectively generate specific solutions for material substitution and functional design simplification.

[0009] Therefore, how to provide an intelligent inspection system and method for over-packaged products is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0010] One purpose of the present invention is to propose an intelligent inspection system and method for over-packaged products based on generative adversarial networks, quantum computing, adaptive spectral analysis and blockchain authentication. The present invention fully integrates multimodal data processing, dynamic optimization algorithms and resource utilization analysis technology, and describes in detail the technical solutions for realizing non-destructive testing of packaging materials, packaging defect identification and green design optimization. It has the advantages of high detection accuracy, strong data credibility, high resource utilization and comprehensive optimization solutions.

[0011] An intelligent inspection system and method for over-packaged products according to an embodiment of the present invention includes the following steps:

[0012] S1. Combining generative adversarial networks with multimodal neural symbolic learning to build an intelligent recognition model. Through deep feature extraction and interactive optimization of multi-source data, robust detection results for dynamic semantic segmentation of complex packaging product forms are generated.

[0013] S2. Using quantum computing-assisted multi-dimensional data optimization technology, we conduct nonlinear analysis of high-dimensional detection data based on the parallel processing capability of quantum superposition states to generate a multi-dimensional key feature matrix of packaging structure and materials.

[0014] S3. Based on blockchain distributed storage technology, the encrypted authentication and full-process traceability mechanism of the test data is used to generate a decentralized, tamper-proof, multi-level trusted data chain for the test results of packaging materials and structures;

[0015] S4. Utilize adaptive spectral analysis technology, combined with real-time spectral characteristic modeling methods, to conduct non-destructive testing of the spectral response characteristics of packaging materials to identify their microscopic composition, material thickness, and compliance with usage regulations.

[0016] S5. Use a semantic segmentation algorithm based on dynamic optimization and an efficient rule learning mechanism to perform multi-scale quantitative analysis of defective areas, redundant structures, and non-essential materials in packaging products;

[0017] S6. Integrate the inspection results based on intelligent identification, quantum computing optimization and blockchain authentication to generate a packaging optimization evaluation model, and provide green design improvement solutions and resource utilization optimization suggestions, including material substitution, functional design simplification and production cost analysis.

[0018] Optionally, the S1 specifically includes:

[0019] S11. Construct a feature generation model based on a bidirectional generative adversarial network. The generator functions G1(w,θ1) and G2(v,θ2) take the random noise vector w and the multidimensional perturbation vector v as input respectively, and use the parameters θ1 and θ2 to generate the high-dimensional feature distribution of the packaging product form. The discriminators D1(u,φ1) and D2(t,φ2) perform dual discrimination between the generated sample u and the real sample t through the dynamically adjusted weight parameters φ1 and φ2;

[0020] S12. Optimize the cross-adversarial loss function L of the bidirectional generative adversarial network ADV and feature consistency correction loss L COR :

[0021]

[0022]

[0023] Among them, μ and ν are weight factors used to balance the optimization objectives between generating adversarial and feature correction;

[0024] S13, adopting a multimodal fusion mechanism based on symbolic learning, using the symbolic reasoning module to jointly represent different modal features and generate a high-dimensional feature matrix H ij , where H ij represents the fused feature vector of the i-th row and j-th column;

[0025] S14. Based on the dynamic semantic segmentation algorithm, combined with the spatial variation characteristics of packaging morphology, the modal fusion features are converted into the semantic segmentation result matrix S through a high-order multi-scale optimization strategy. kl , where S kl is the dynamic segmentation label of the kth row and lth column;

[0026] S15, adopt nonlinear boundary correction mechanism, use robust constraint function to refine and adjust the complex boundary area in the segmentation result, and generate a multi-dimensional segmentation result matrix T mn , where T mn Represents the segmentation area weight of the coordinate point (m,n).

[0027] Optionally, the S2 specifically includes:

[0028] S21. Construct a multimodal high-dimensional data representation model based on quantum superposition state, and initialize the packaging detection data into a feature vector set Q = {q a ,q b ,…,q z}, through the quantum state represents the multimodal nature of the packaging data, where γ k is the eigenvector q k The associated weight coefficient satisfies the normalization condition

[0029] S22. Quantum Optimization Operator U via Phase Modulation and Entanglement Enhancement phase , optimize the high-dimensional feature correlation of the initial quantum state |Φ0> and generate the optimized state |Φ opt >:

[0030] |Φ opt >=U phase |>

[0031] Among them, U phase It is formed by combining a phase modulation matrix and a characteristic coherence enhancement matrix;

[0032] S23, introduce dynamic adaptive Hamiltonian H adapt (t), for the optimized state |Φ opt >Perform time evolution processing to generate dynamic characteristic states |Φ t >, the dynamic characteristic state reflects the characteristic changes of packaging materials in the time series dimension:

[0033]

[0034] S24. Using high-dimensional quantum measurement operator M uv , perform projection measurement on the dynamic characteristic state and extract the characteristic matrix R ij :

[0035] R ij =| i ||q j >| 2 ;

[0036] Among them, the matrix R ij Represents the quantitative relationship between different feature dimensions in packaging inspection data;

[0037] S25, feature matrix R ij Perform nonlinear dimensionality reduction processing to construct the key feature matrix M xy , where M xy Indicates the core characteristics of packaging materials and structures.

[0038] Optionally, the S3 specifically includes:

[0039] S31. Establish a distributed storage architecture based on blockchain and represent the packaging inspection data as an initial data block set P = {p1, p2, ..., p m}, through the one-way hash function H1(p i ) For each data block p i Generate a unique identification value h i , where h i =H1(p​i ), used to verify the integrity of the data block;

[0040] S32, using an encryption algorithm based on elliptic curve cryptography to encrypt the data block set P, generating a ciphertext data set C = {c1, c2, ..., c m}, each encrypted data block is K e Represents the data block encryption key;

[0041] S33, build the chain data structure of the blockchain, each block B k Including data content c k 、The hash value of the previous block H2(B k-1 ) and timestamp T k :

[0042] B k ={c k ,H2(B k-1 ),T k};

[0043] Where H2 is the cryptographic hash function used for blockchain operations;

[0044] S34. Verify the new block B through the distributed Byzantine fault-tolerant consensus algorithm k The effectiveness of the consensus function F is defined consensus (Q k ,M),Q k Represents the set of nodes participating in consensus verification, M is the global verification matrix of the blockchain network, and the function outputs the consensus state F consensus (Q k ,M)=Valid when the new block is added to the blockchain;

[0045] S35. Define the index function S(x) to retrieve a specific data block p through the traceability retrieval mechanism based on blockchain. x , to locate and verify the integrity of the data stored on the chain, and apply the retrieval results to the test result report and traceability analysis module:

[0046] p x =S(x),x∈{1,2,…,m}.

[0047] Optionally, the S4 specifically includes:

[0048] S41, construct a method based on adaptive spectral characteristic modeling, collect the spectral response data of the packaging material into an initial data matrix Q = {q ab},q ab Indicates band λ a and the spectral intensity at the sample position b, through the real-time dynamic spectral sampling function Λdyn (ξ,ζ), adjust the effective spectrum range to [λ ξ ,λ ζ ];

[0049] S42. Decompose the data matrix Q into feature submatrices through the multimodal spectral decomposition mechanism Indicates the wavelength range λ c to λ d The characteristic intensity under mode k is normalized by the function N spec (T k ), generate a standardized feature matrix

[0050] S43, based on the nonlinear spectrum fitting function F adaptive (λ; ψ) for the standardized feature matrix R k Perform fitting by minimizing E k Optimization parameter ψ k , capturing the complex variation of the spectral response:

[0051]

[0052] S44, based on the high-dimensional spectrum deconstruction algorithm, the standardized feature matrix R k Decomposition into material component matrix Indicates the wavelength range λ h to λ i The characteristic contribution value of the specific component g;

[0053] S45. Using the thickness modeling method based on spectral differential analysis, the material thickness H(λ) is expressed as:

[0054]

[0055] Where η is the thickness adjustment coefficient, which is used to calculate the material thickness distribution characteristics;

[0056] S46, combined with the normative matching model, the spectral deconstruction and thickness analysis results are compared with the target template G m Perform comparison and match the discriminant function D match (ω) is:

[0057]

[0058] Among them, δ is the allowable matching error threshold, outputs the compliance judgment result and generates a test report.

[0059] Optionally, the S5 specifically includes:

[0060] S51. Construct a multimodal feature adjustment mechanism based on dynamic optimization, and express the multidimensional data matrix of the packaging product as Z = {z ij}, where z ij Represents the value of sample i in feature dimension j; through dynamic weight optimization function T dyn (Z; α) optimizes the feature weight α of the matrix Z to generate the adjusted feature matrix Z′={z′ ij};

[0061] S52, use the multi-scale feature extraction algorithm to process the optimized matrix Z' to generate a multi-scale feature map set

[0062]

[0063] in, represents the response value of the feature position (k, l) under scale s, is a convolution kernel of scale s, which extracts multi-scale feature information of packaging materials;

[0064] S53, based on the rule learning mechanism, define the rule base R = {r n}, where r n Represents the nth rule, by matching the feature map F s and rule base R to generate classification results Y n , and use the optimization objective function L rule Update rule parameters:

[0065]

[0066] S54, through the dynamic semantic segmentation model, the multi-scale feature map set F s Perform semantic classification and generate semantic segmentation result matrix S = {s pq}, s pq Represent the semantic category of the coordinate point (p,q) and refine the segmentation result through the boundary optimization algorithm;

[0067] S55, according to the semantic segmentation results, the defect area, redundant structure and non-essential material part are mapped into a quantization result set D = {d i}、E={e j} and N={n k}, output the corresponding area, distribution and proportion analysis results.

[0068] Optionally, the S6 specifically includes:

[0069] S61, integrating the segmentation result matrix S generated by intelligent recognition = {s uv}、The characteristic matrix M generated by quantum computing optimization is M={m ij} and the trusted detection data set D stored in the blockchain = {d k}, construct the joint feature set F = {S, M, D}, and use the normalization function G norm (F,θ) normalizes the joint features and generates a standardized feature matrix F′={f′ ab}, where θ is the normalization parameter;

[0070] S62, construct a feature optimization module, using a dynamic feature weighting function W opt (F′; β) assigns weight β to each dimension feature of the standardized feature matrix F′, and generates a weighted optimization matrix O={o xy}:

[0071]

[0072] Among them, β z is the feature dimension weight, Normalized eigenvalues for specific dimensions in optimization calculations;

[0073] S63. Based on the weighted optimization matrix O, a green design generation module is established, and the resource conservation objective function R eco (μ,ν,η), by minimizing the multi-objective function L design Optimize design parameters α, γ, and λ to generate a green design solution that includes material substitution, design simplification, and energy consumption control:

[0074] L design =μ·α+ν·γ+η·λ;

[0075] Among them, μ represents the amount of material used, ν represents the design complexity, and η represents the production energy consumption;

[0076] S64, combining the optimization matrix O and the green design results, constructing a resource utilization analysis model, generating a resource utilization matrix U = {u kl}, through the resource cost optimization function C resource Calculate resource cost distribution across regions;

[0077]

[0078] Among them, ρ kl is the unit resource cost, u kl represents the utilization efficiency of resource type k in region l;

[0079] S65. Integrate the optimization analysis results to generate a packaging optimization assessment report, including material substitution suggestions, design streamlining solutions, resource utilization indicators, and production cost distribution. The report is encrypted and stored through the blockchain authentication module, and a blockchain-based verification certificate is generated.

[0080] The beneficial effects of the present invention are:

[0081] (1) By combining generative adversarial networks (GANs) with quantum computing technology, this paper provides in-depth extraction and dynamic optimization analysis of multimodal features of packaging materials and structures. This enables the system to accurately identify defective areas, redundant structures, and non-essential components in overpackaging, effectively addressing the challenges of detecting complex and diverse packaging formats and their microscopic components. The use of multidimensional data optimization technology based on quantum computing significantly improves detection efficiency and accuracy, particularly in highly complex packaging scenarios.

[0082] (2) This invention uses adaptive spectral analysis and blockchain distributed storage technology, combined with real-time dynamic modeling and a trusted authentication mechanism, to achieve nondestructive testing of the spectral response characteristics of packaging materials and trusted traceability of test data. The thickness and composition distribution model constructed through spectral characteristic analysis provides a scientific basis for the precise detection and analysis of packaging materials, while blockchain technology is used to ensure the integrity and immutability of the data.

[0083] (3) This invention integrates intelligent identification, quantum computing optimization, and blockchain authentication technologies to construct a packaging optimization and evaluation model, providing green design improvement solutions and resource utilization optimization recommendations. This model not only achieves packaging material substitution, functional design simplification, and production cost optimization, but also constructs a comprehensive solution for resource conservation and cost optimization through a multi-objective optimization method. This significantly improves the greenness of packaging design and resource utilization efficiency, providing comprehensive technical support for environmental protection and sustainable development in multiple industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0085] Figure 1 This is a flow chart of an intelligent inspection system and method for over-packaged products proposed by the present invention. DETAILED DESCRIPTION

[0086] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0087] refer to Figure 1 , an intelligent inspection system and method for over-packaged products, comprising the following steps:

[0088] S1. Combining generative adversarial networks with multimodal neural symbolic learning to build an intelligent recognition model. Through deep feature extraction and interactive optimization of multi-source data, robust detection results for dynamic semantic segmentation of complex packaging product forms are generated.

[0089] In this embodiment, S1 specifically includes:

[0090] S11. Construct a feature generation model based on a bidirectional generative adversarial network. The generator functions G1(w,θ1) and G2(v,θ2) take the random noise vector w and the multidimensional perturbation vector v as input respectively, and use the parameters θ1 and θ2 to generate the high-dimensional feature distribution of the packaging product form. The discriminators D1(u,φ1) and D2(t,φ2) perform dual discrimination between the generated sample u and the real sample t through the dynamically adjusted weight parameters φ1 and φ2;

[0091] S12. Optimize the cross-adversarial loss function L of the bidirectional generative adversarial network ADV and feature consistency correction loss L COR :

[0092]

[0093] Among them, μ and ν are weight factors used to balance the optimization objectives between generating adversarial and feature correction;

[0094] S13, adopting a multimodal fusion mechanism based on symbolic learning, using the symbolic reasoning module to jointly represent different modal features and generate a high-dimensional feature matrix H ij , where H ij represents the fused feature vector of the i-th row and j-th column;

[0095] S14. Based on the dynamic semantic segmentation algorithm, combined with the spatial variation characteristics of packaging morphology, the modal fusion features are converted into the semantic segmentation result matrix S through a high-order multi-scale optimization strategy. kl , where S kl is the dynamic segmentation label of the kth row and lth column;

[0096] S15, adopt nonlinear boundary correction mechanism, use robust constraint function to refine and adjust the complex boundary area in the segmentation result, and generate a multi-dimensional segmentation result matrix T mn , where T mn Represents the segmentation area weight of the coordinate point (m,n).

[0097] S2. Using quantum computing-assisted multi-dimensional data optimization technology, based on the parallel processing capability of quantum superposition states, nonlinear analysis of high-dimensional detection data is performed to generate a multi-dimensional key feature matrix of packaging structure and materials.

[0098] In this embodiment, S2 specifically includes:

[0099] S21. Construct a multimodal high-dimensional data representation model based on quantum superposition state, and initialize the packaging detection data into a feature vector set Q = {q a ,q b ,…,q z}, through the quantum state represents the multimodal nature of the packaging data, where γ k is the eigenvector q k The associated weight coefficient satisfies the normalization condition

[0100] S22. Quantum Optimization Operator U via Phase Modulation and Entanglement Enhancement phase , optimize the high-dimensional feature correlation of the initial quantum state |Φ0> and generate the optimized state |Φ opt >:

[0101] |Φ opt >=U phase |>

[0102] Among them, U phase It is formed by combining a phase modulation matrix and a characteristic coherence enhancement matrix;

[0103] S23, introduce dynamic adaptive Hamiltonian H adapt (t), for the optimized state |Φ opt >Perform time evolution processing to generate dynamic characteristic states |Φ t >, the dynamic characteristic state reflects the characteristic changes of packaging materials in the time series dimension:

[0104]

[0105] S24. Using high-dimensional quantum measurement operator M uv , perform projection measurement on the dynamic characteristic state and extract the characteristic matrix R ij :

[0106] R ij =| i ||q j >| 2 ;

[0107] Among them, the matrix R ij Represents the quantitative relationship between different feature dimensions in packaging inspection data;

[0108] S25, feature matrix R ij Perform nonlinear dimensionality reduction processing to construct the key feature matrix M xy , where M xy Indicates the core characteristics of packaging materials and structures.

[0109] ​S3. Based on blockchain distributed storage technology, the encrypted authentication and full-process traceability mechanism of the test data is used to generate a decentralized, tamper-proof, multi-level trusted data chain for the test results of packaging materials and structures;

[0110] In this embodiment, S3 specifically includes:

[0111] S31. Establish a distributed storage architecture based on blockchain and represent the packaging inspection data as an initial data block set P = {p1, p2, ..., p m}, through the one-way hash function H1(p i ) For each data block p i Generate a unique identification value h i , where h i =H1(p i ), used to verify the integrity of the data block;

[0112] S32, using an encryption algorithm based on elliptic curve cryptography to encrypt the data block set P, generating a ciphertext data set C = {c1, c2, ..., c m}, each encrypted data block is K e Represents the data block encryption key;

[0113] S33, build the chain data structure of the blockchain, each block B k Including data content c k 、The hash value of the previous block H2(B k-1 ) and timestamp T k :

[0114] B k ={c k ,H2(B k-1 ),T k};

[0115] Where H2 is the cryptographic hash function used for blockchain operations;

[0116] S34. Verify the new block B through the distributed Byzantine fault-tolerant consensus algorithm k The effectiveness of the consensus function F is defined consensus (Q k ,M),Q k Represents the set of nodes participating in consensus verification, M is the global verification matrix of the blockchain network, and the function outputs the consensus state F consensus (Q k ,M)=Valid when the new block is added to the blockchain;

[0117] S35. Define the index function S(x) to retrieve a specific data block p through the traceability retrieval mechanism based on blockchain.x , to locate and verify the integrity of the data stored on the chain, and apply the retrieval results to the test result report and traceability analysis module:

[0118] p x =S(x),x∈{1,2,…,m}.

[0119] S4. Utilize adaptive spectral analysis technology, combined with real-time spectral characteristic modeling methods, to conduct non-destructive testing of the spectral response characteristics of packaging materials to identify their microscopic composition, material thickness, and compliance with usage regulations.

[0120] In this embodiment, S4 specifically includes:

[0121] S41, construct a method based on adaptive spectral characteristic modeling, collect the spectral response data of the packaging material into an initial data matrix Q = {q ab},q ab Indicates band λ a and the spectral intensity at the sample position b, through the real-time dynamic spectral sampling function Λ dyn (ξ,ζ), adjust the effective spectrum range to [λ ξ ,λ ζ ];

[0122] S42. Decompose the data matrix Q into feature submatrices through the multimodal spectral decomposition mechanism Indicates the wavelength range λ c to λ d The characteristic intensity under mode k is normalized by the function N spec (T k ), generate a standardized feature matrix

[0123] S43, based on the nonlinear spectrum fitting function F adaptive (λ; ψ) for the standardized feature matrix R k Perform fitting by minimizing E k Optimization parameter ψ k , capturing the complex variation of the spectral response:

[0124]

[0125] S44, based on the high-dimensional spectrum deconstruction algorithm, the standardized feature matrix R k Decomposition into material component matrix Indicates the wavelength range λ h to λ i The characteristic contribution value of the specific component g;

[0126] S45. Using the thickness modeling method based on spectral differential analysis, the material thickness H(λ) is expressed as:

[0127]

[0128] Where η is the thickness adjustment coefficient, which is used to calculate the material thickness distribution characteristics;

[0129] S46, combined with the normative matching model, the spectral deconstruction and thickness analysis results are compared with the target template G m Perform comparison and match the discriminant function D match (ω) is:

[0130]

[0131] Among them, δ is the allowable matching error threshold, outputs the compliance judgment result and generates a test report.

[0132] S5. Use a semantic segmentation algorithm based on dynamic optimization and an efficient rule learning mechanism to perform multi-scale quantitative analysis of defective areas, redundant structures, and non-essential materials in packaging products;

[0133] In this embodiment, S5 specifically includes:

[0134] S51. Construct a multimodal feature adjustment mechanism based on dynamic optimization, and express the multidimensional data matrix of the packaging product as Z = {z ij}, where z ij Represents the value of sample i in feature dimension j; through dynamic weight optimization function T dyn (Z; α) optimizes the feature weight α of the matrix Z to generate the adjusted feature matrix Z′={z′ ij};

[0135] S52, use the multi-scale feature extraction algorithm to process the optimized matrix Z' to generate a multi-scale feature map set

[0136]

[0137] in, represents the response value of the feature position (k, l) under scale s, is a convolution kernel of scale s, which extracts multi-scale feature information of packaging materials;

[0138] S53, based on the rule learning mechanism, define the rule base R = {r n}, where r n Represents the nth rule, by matching the feature map F s and rule base R to generate classification results Y n , and use the optimization objective function Lrule Update rule parameters:

[0139]

[0140] S54, through the dynamic semantic segmentation model, the multi-scale feature map set F s Perform semantic classification and generate semantic segmentation result matrix S = {s pq}, s pq Represent the semantic category of the coordinate point (p,q) and refine the segmentation result through the boundary optimization algorithm;

[0141] S55, according to the semantic segmentation results, the defect area, redundant structure and non-essential material part are mapped into a quantization result set D = {d i}、E={e j} and N={n k}, output the corresponding area, distribution and proportion analysis results.

[0142] S6. Integrate the inspection results based on intelligent identification, quantum computing optimization and blockchain authentication to generate a packaging optimization evaluation model, and provide green design improvement solutions and resource utilization optimization suggestions, including material substitution, functional design simplification and production cost analysis.

[0143] In this embodiment, S6 specifically includes:

[0144] S61, integrating the segmentation result matrix S generated by intelligent recognition = {s uv}、The characteristic matrix M generated by quantum computing optimization is M={m ij} and the trusted detection data set D stored in the blockchain = {d k}, construct the joint feature set F = {S, M, D}, and use the normalization function G norm (F,θ) normalizes the joint features and generates a standardized feature matrix F′={f′ ab}, where θ is the normalization parameter;

[0145] S62, construct a feature optimization module, using a dynamic feature weighting function W opt (F′; β) assigns weight β to each dimension feature of the standardized feature matrix F′, and generates a weighted optimization matrix O={o xy}:

[0146]

[0147] Among them, β z is the feature dimension weight, Normalized eigenvalues for specific dimensions in optimization calculations;

[0148] S63. Based on the weighted optimization matrix O, a green design generation module is established, and the resource conservation objective function R eco (μ,ν,η), by minimizing the multi-objective function L design Optimize design parameters α, γ, and λ to generate a green design solution that includes material substitution, design simplification, and energy consumption control:

[0149] L design =μ·α+ν·γ+η·λ;

[0150] Among them, μ represents the amount of material used, ν represents the design complexity, and η represents the production energy consumption;

[0151] S64, combining the optimization matrix O and the green design results, constructing a resource utilization analysis model, generating a resource utilization matrix U = {u kl}, through the resource cost optimization function C resource Calculate resource cost distribution across regions;

[0152]

[0153] Among them, ρ kl is the unit resource cost, u kl represents the utilization efficiency of resource type k in region l;

[0154] S65. Integrate the optimization analysis results to generate a packaging optimization assessment report, including material substitution suggestions, design streamlining solutions, resource utilization indicators, and production cost distribution. The report is encrypted and stored through the blockchain authentication module, and a blockchain-based verification certificate is generated.

[0155] Example 1:

[0156] To verify the feasibility of this invention, we applied it to Company B, a domestic packaging manufacturer. The company handles the production and inspection of approximately 500,000 packages per month, involving a variety of packaging materials and complex packaging designs. Due to increasing customer demand for environmental protection and resource conservation in recent years, Company B urgently needed an efficient and intelligent overpackaging detection and optimization solution to reduce material waste and improve resource utilization.

[0157] In this scenario, Company B faced key challenges: inefficient manual inspections for packaging testing, complex packaging design with a lack of green design optimization solutions, insufficient test data reliability, and test results that failed to meet the requirements of diverse packaging formats. The intelligent over-packaged product inspection system and method, utilizing intelligent recognition, quantum computing optimization, adaptive spectral analysis, and blockchain authentication technologies, provides Company B with a comprehensive packaging inspection and optimization solution.

[0158] The system's application begins with the inspection process, using a generative adversarial network (GAN) to dynamically segment the complex packaging form, extracting defective areas and redundant components. Subsequently, quantum computing-assisted analysis optimizes and calculates key characteristics of the packaging material, generating quantified thickness and composition characteristics. An adaptive spectral analysis module non-destructively tests the spectral response characteristics of the packaging material. Using blockchain technology, each test data is recorded to ensure the credibility and traceability of the test results. Based on this, the system comprehensively evaluates the packaging's resource utilization and green design solutions, generating an optimization report and proposing recommendations for material substitution and design streamlining.

[0159] In practice, the system tested and optimized 100 different packaging types produced by Company B for three consecutive months. Table 1 shows the comparison of key indicators before and after optimization:

[0160] Table 1 Comparison of packaging testing and optimization results of Company B

[0161]

[0162]

[0163] As shown in Table 1, before optimization, Company B faced problems such as low inspection efficiency and excessive material waste during packaging inspection. Through the implementation of this invention, the inspection time for a single package was reduced from 45 seconds to 12 seconds, and inspection accuracy was improved to 97.6%. Furthermore, monthly material waste was significantly reduced from 15.6 tons to 2.8 tons, and resource utilization increased to 91.8%. The system records the data of each inspection via blockchain, and the credibility score has increased from 68 to 99, ensuring data security and traceability. In terms of green design optimization, the proportion of customers adopting the optimization solution reached 87%, further demonstrating the practicality and value of the system.

[0164] In one typical case, the system detected that the thickness of the plastic material used in a food packaging exceeded the required thickness by 20%. Through spectral analysis and quantum computing optimization, the system proposed a more efficient material replacement solution, replacing the original plastic with a stronger, lightweight composite material and reducing the thickness of the redundant part. This is expected to save Company B approximately 2 million yuan in raw material costs annually and reduce carbon emissions by approximately 500 tons.

[0165] As demonstrated in the preceding examples, the intelligent over-packaged product inspection system and method of the present invention significantly improves the efficiency and accuracy of packaging inspection and optimization, promoting the achievement of green design and environmental protection goals. The system integrates multiple cutting-edge technologies, providing reliable technical support and practical solutions for the packaging industry.

[0166] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. An intelligent inspection system and method for over-packaged products, characterized in that: The steps include: S1. Combining generative adversarial networks with multimodal neural symbolic learning to build an intelligent recognition model. Through deep feature extraction and interactive optimization of multi-source data, robust detection results for dynamic semantic segmentation of complex packaging product forms are generated. S2. Using quantum computing-assisted multi-dimensional data optimization technology, we conduct nonlinear analysis of high-dimensional detection data based on the parallel processing capability of quantum superposition states to generate a multi-dimensional key feature matrix of packaging structure and materials. S3. Based on blockchain distributed storage technology, the encrypted authentication and full-process traceability mechanism of the test data is used to generate a decentralized, tamper-proof, multi-level trusted data chain for the test results of packaging materials and structures; S4. Utilize adaptive spectral analysis technology, combined with real-time spectral characteristic modeling methods, to conduct non-destructive testing of the spectral response characteristics of packaging materials to identify their microscopic composition, material thickness, and compliance with usage regulations. S5. Use a semantic segmentation algorithm based on dynamic optimization and an efficient rule learning mechanism to perform multi-scale quantitative analysis of defective areas, redundant structures, and non-essential materials in packaging products; S6. Integrate the inspection results based on intelligent identification, quantum computing optimization and blockchain authentication to generate a packaging optimization evaluation model, and provide green design improvement solutions and resource utilization optimization suggestions, including material substitution, functional design simplification and production cost analysis.

2. The intelligent inspection system and method for over-packaged products according to claim 1 is characterized in that: Said S1 specifically includes: S11. Construct a feature generation model based on a bidirectional generative adversarial network. The generator functions G1(w,θ1) and G2(v,θ2) take the random noise vector w and the multidimensional perturbation vector v as input respectively, and use the parameters θ1 and θ2 to generate the high-dimensional feature distribution of the packaging product form. The discriminators D1(u,φ1) and D2(t,φ2) perform dual discrimination between the generated sample u and the real sample t through the dynamically adjusted weight parameters φ1 and φ2; S12. Optimize the cross-adversarial loss function L of the bidirectional generative adversarial network ADV and feature consistency correction loss L COR : Among them, μ and ν are weight factors used to balance the optimization objectives between generating adversarial and feature correction; S13, adopting a multimodal fusion mechanism based on symbolic learning, using the symbolic reasoning module to jointly represent different modal features and generate a high-dimensional feature matrix H ij , where H ij represents the fused feature vector of the i-th row and j-th column; S14. Based on the dynamic semantic segmentation algorithm, combined with the spatial variation characteristics of packaging morphology, the modal fusion features are converted into the semantic segmentation result matrix S through a high-order multi-scale optimization strategy. kl , where S kl is the dynamic segmentation label of the kth row and lth column; S15, adopt nonlinear boundary correction mechanism, use robust constraint function to refine and adjust the complex boundary area in the segmentation result, and generate a multi-dimensional segmentation result matrix T mn , where T mn Represents the segmentation area weight of the coordinate point (m,n).

3. The intelligent inspection system and method for over-packaged products according to claim 1 is characterized in that: The S2 specifically includes: S21. Construct a multimodal high-dimensional data representation model based on quantum superposition state, and initialize the packaging detection data into a feature vector set Q = {q a ,q b ,…,q z }, through the quantum state represents the multimodal nature of the packaging data, where γ k is the eigenvector q k The associated weight coefficient satisfies the normalization condition S22. Quantum Optimization Operator U via Phase Modulation and Entanglement Enhancement phase , optimize the high-dimensional feature correlation of the initial quantum state |Φ0> and generate the optimized state |Φ opt >: |Φ opt >=U phase |>; Among them, U phase It is formed by combining a phase modulation matrix and a characteristic coherence enhancement matrix; S23, introduce dynamic adaptive Hamiltonian H adapt (t), for the optimized state |Φ opt >Perform time evolution processing to generate dynamic characteristic states |Φ t >, the dynamic characteristic state reflects the characteristic changes of packaging materials in the time series dimension: S24. Using high-dimensional quantum measurement operator M uv , perform projection measurement on the dynamic characteristic state and extract the characteristic matrix R ij : R ij =|<q i ||q j >| 2 ; Among them, the matrix R ij Represents the quantitative relationship between different feature dimensions in packaging inspection data; S25, feature matrix R ij Perform nonlinear dimensionality reduction processing to construct the key feature matrix M xy , where M xy Indicates the core characteristics of packaging materials and structures.

4. The intelligent inspection system and method for over-packaged products according to claim 1 is characterized in that: The S3 specifically includes: S31. Establish a distributed storage architecture based on blockchain and represent the packaging inspection data as an initial data block set P = {p1, p2, ..., p m }, through the one-way hash function H1(p i ) For each data block p i Generate a unique identification value h i , where h i =H1(p i ), used to verify the integrity of the data block; S32, using an encryption algorithm based on elliptic curve cryptography to encrypt the data block set P, generating a ciphertext data set C = {c1, c2, ..., c m }, each encrypted data block is K e Represents the data block encryption key; S33, build the chain data structure of the blockchain, each block B k Including data content c k 、The hash value of the previous block H2(B k-1 ) and timestamp T k : B k ={c k ,H2(B k-1 ),T k }; Where H2 is the cryptographic hash function used for blockchain operations; S34. Verify the new block B through the distributed Byzantine fault-tolerant consensus algorithm k The effectiveness of the consensus function F is defined consensus (Q k ,M),Q k Represents the set of nodes participating in consensus verification, M is the global verification matrix of the blockchain network, and the function outputs the consensus state F consensus (Q k ,M)=Valid when the new block is added to the blockchain; S35. Define the index function S(x) to retrieve a specific data block p through the traceability retrieval mechanism based on blockchain. x , to locate and verify the integrity of the data stored on the chain, and apply the retrieval results to the test result report and traceability analysis module: p x =S(x),x∈{1,2,…,m}。 5. The intelligent inspection system and method for over-packaged products according to claim 1 is characterized in that: The S4 specifically includes: S41, construct a method based on adaptive spectral characteristic modeling, collect the spectral response data of the packaging material into an initial data matrix Q = {q ab },q ab Indicates band λ a and the spectral intensity at the sample position b, through the real-time dynamic spectral sampling function Λ dyn (ξ,ζ), adjust the effective spectrum range to [λ ξ ,λ ζ ]; S42. Decompose the data matrix Q into feature submatrices through the multimodal spectral decomposition mechanism Indicates the wavelength range λ c to λ d The characteristic intensity under mode k is normalized by the function N spec (T k ), generate a standardized feature matrix S43, based on the nonlinear spectrum fitting function F adaptive (λ; ψ) for the standardized feature matrix R k Perform fitting by minimizing E k Optimization parameter ψ k , capturing the complex variation of the spectral response: S44, based on the high-dimensional spectrum deconstruction algorithm, the standardized feature matrix R k Decomposition into material component matrix Indicates the wavelength range λ h to λ i The characteristic contribution value of the specific component g; S45. Using the thickness modeling method based on spectral differential analysis, the material thickness H(λ) is expressed as: Where η is the thickness adjustment coefficient, which is used to calculate the material thickness distribution characteristics; S46, combined with the normative matching model, the spectral deconstruction and thickness analysis results are compared with the target template G m Perform comparison and match the discriminant function D match (ω) is: Among them, δ is the allowable matching error threshold, outputs the compliance judgment result and generates a test report.

6. The intelligent inspection system and method for over-packaged products according to claim 1 is characterized in that: The S5 specifically includes: S51. Construct a multimodal feature adjustment mechanism based on dynamic optimization, and express the multidimensional data matrix of the packaging product as Z = {z ij }, where z ij Represents the value of sample i in feature dimension j; through dynamic weight optimization function T dyn (Z; α) optimizes the feature weight α of the matrix Z to generate the adjusted feature matrix Z′={z′ ij }; S52, use the multi-scale feature extraction algorithm to process the optimized matrix Z' to generate a multi-scale feature map set in, represents the response value of the feature position (k, l) under scale s, is a convolution kernel of scale s, which extracts multi-scale feature information of packaging materials; S53, based on the rule learning mechanism, define the rule base R = {r n }, where r n Represents the nth rule, by matching the feature map F s and rule base R to generate classification results Y n , and use the optimization objective function L rule Update rule parameters: S54, through the dynamic semantic segmentation model, the multi-scale feature map set F s Perform semantic classification and generate semantic segmentation result matrix S = {s pq }, s pq Represent the semantic category of the coordinate point (p,q) and refine the segmentation result through the boundary optimization algorithm; S55, according to the semantic segmentation results, the defect area, redundant structure and non-essential material part are mapped into a quantization result set D = {d i }、E={e j } and N={n k }, output the corresponding area, distribution and proportion analysis results.

7. The intelligent inspection system and method for over-packaged products according to claim 1 is characterized in that: The S6 specifically includes: S61, integrating the segmentation result matrix S generated by intelligent recognition = {s uv }、The characteristic matrix M generated by quantum computing optimization is M={m ij } and the trusted detection data set D stored in the blockchain = {d k }, construct the joint feature set F = {S, M, D}, and use the normalization function G norm (F,θ) normalizes the joint features and generates a standardized feature matrix F′={f′ ab }, where θ is the normalization parameter; S62, construct a feature optimization module, using a dynamic feature weighting function W opt (F′; β) assigns weight β to each dimension feature of the standardized feature matrix F′, and generates a weighted optimization matrix O={o xy }: Among them, β z is the feature dimension weight, Normalized eigenvalues for specific dimensions in optimization calculations; S63. Based on the weighted optimization matrix O, a green design generation module is established, and the resource conservation objective function R eco (μ,ν,η), by minimizing the multi-objective function L design Optimize design parameters α, γ, and λ to generate a green design solution that includes material substitution, design simplification, and energy consumption control: L design =μ·α+ν·γ+η·λ; Among them, μ represents the amount of material used, ν represents the design complexity, and η represents the production energy consumption; S64, combining the optimization matrix O and the green design results, constructing a resource utilization analysis model, generating a resource utilization matrix U = {u kl }, through the resource cost optimization function C resource Calculate resource cost distribution across regions; Among them, ρ kl is the unit resource cost, u kl represents the utilization efficiency of resource type k in region l; S65. Integrate the optimization analysis results to generate a packaging optimization assessment report, including material substitution suggestions, design streamlining solutions, resource utilization indicators, and production cost distribution. The report is encrypted and stored through the blockchain authentication module, and a blockchain-based verification certificate is generated.

Images