Hotel check-in identity verification system and method based on real-name DID

Abstract

The invention discloses a hotel check-in identity verification system and method based on real-name DID, and the method comprises the following steps: S1, collecting the identity data of a check-in user, carrying out the preprocessing, and constructing an identity feature sequence; s2, constructing a time sequence convolution identity verification network, and modeling each check-in user as an identity verification intelligent agent with multi-modal feature input and time sequence behavior observation capability; s3, performing preliminary identity discrimination by using a time sequence convolution identity verification network; s4, configuring a parameter search unit for each ant body by adopting an ant algorithm; s5, inputting the identity feature sequence into a time sequence convolution identity verification network optimized by an ant algorithm; and S6, comparing the identity discrimination result with identity data in a hotel real-name DID database, and outputting a final identity verification result. According to the invention, the ant algorithm, the time sequence convolutional network and the deep learning discrimination technology are combined, and the hotel check-in identity verification method based on the real-name DID is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of identity verification, and particularly relates to a hotel check-in identity verification system and method based on a real-name DID. BACKGROUND

[0002] With the continuous improvement of social digitization and intelligence, hotel check-in real-name system management is facing increasingly complex identity verification needs. Currently, hotel check-in identity verification mainly relies on identity card scanning, manual comparison at the front desk, or preliminary verification based on simple comparison algorithms. These methods are prone to vulnerabilities when facing security risks such as identity forgery and impersonation, making it difficult to ensure the accuracy and security of identity verification. Especially in terms of multi-modal identity data fusion and dynamic comparison, existing technologies generally have problems such as insufficient accuracy, poor real-time performance, and weak anti-interference ability.

[0003] Existing identity verification methods are usually based on static feature data and mainly use image comparison of identity cards or single-modal face recognition for verification. Although this can meet the basic identity comparison needs in some cases, it still cannot adapt to complex and variable actual check-in scenarios in terms of user behavior feature time series information mining, multi-modal fusion recognition, and high dynamic scenarios. For example, when the user's facial image is blurred or deviates under different lighting, angles, or expressions, or the voiceprint features are distorted due to environmental noise interference, existing single-modal-based comparison methods often cannot obtain stable and reliable identity verification results. At the same time, traditional identity verification methods lack effective integration and dynamic representation of check-in behavior time series data, and cannot capture multi-dimensional dynamic change features of users during the check-in process, resulting in constraints on the accuracy and security of identity discrimination.

[0004] In some identity verification systems, deep learning techniques such as deep neural networks are introduced for identity comparison, but there are still deficiencies in multi-modal feature fusion, time series modeling, and feature expression robustness. Especially when fusing different modal features, there is a lack of unified modeling framework and dynamic time series feature extraction method, making it difficult to efficiently integrate multi-modal and multi-time sequence features. In addition, existing technologies often separate the feature comparison module and the identity discrimination module, lack of collaborative optimization capability, resulting in insufficient overall robustness and adaptability of identity verification.

[0005] To solve the above problems, the hotel check-in identity verification method and system based on real-name DID are proposed, which includes the preprocessing and serialization of multi-modal identity feature data, multi-dimensional feature time series modeling based on time series convolution network, and multi-objective parameter optimization of ant colony algorithm. By introducing the ant colony algorithm module, the optimal parameter configuration can be adaptively searched in the multi-modal feature space, avoiding falling into a local optimal solution. Based on the time series feature extraction and fusion of multi-modal features, the dynamic change rule of user identity features in the time dimension can be effectively captured, realizing high robustness and multi-dimensional dynamic modeling of identity. Combined with the final standardized comparison and score threshold judgment mechanism, the limitations of single modal, static data and static threshold in traditional identity verification are overcome, and the overall accuracy, real-time and intelligent level of identity verification are significantly improved.

[0006] In summary, the existing hotel check-in identity verification method still has great defects in the diversity of multi-modal feature extraction, the accuracy of dynamic time series modeling, and the global optimization of feature matching parameters, which is difficult to cope with the changing user identity data and high security requirements in the hotel real-name check-in scene. The method and system proposed in the present application systematically improve the key links of multi-modal multi-time sequence identity feature deep modeling, parameter optimization of ant colony algorithm, and multi-dimensional dynamic comparison, and construct a safe, intelligent and traceable hotel real-name check-in identity verification technology system, which has clear pertinence and broad practical application prospects.

[0007] Therefore, how to provide a hotel check-in identity verification method based on real-name DID is a problem that those skilled in the art need to solve. SUMMARY

[0008] One object of the present application is to propose a hotel check-in identity verification system and method based on real-name DID. The present application combines multi-modal feature input, time series convolution network modeling and global optimization ability of ant colony algorithm, and constructs a systematic identity verification process from identity feature data preprocessing, sequential modeling to discrimination and final matching comparison. This method integrates multi-modal identity features such as face image, voiceprint and text, combines with check-in behavior time series data, and through multi-level time series convolution network and parameter adaptive optimization based on ant colony algorithm, can accurately extract and fuse multi-dimensional dynamic identity features of users, realize high robustness and high precision output of identity discrimination. The final identity verification result is compared with the historical identity data in the hotel real-name DID database to ensure the authenticity and security of the verification result. The present application has the advantages of high identity verification accuracy, strong dynamic adaptability and sufficient information security protection, breaks through the technical bottlenecks of existing identity verification methods in multi-modal multi-time sequence fusion, robust discrimination and intelligent verification, and is suitable for real-time and reliable identity verification in complex check-in scenarios, significantly improving the security and intelligence level of hotel real-name check-in identity verification.

[0009] The hotel check-in identity verification method based on a real-name DID according to the embodiment of the application comprises the following steps:

[0010] S1, identity data of a check-in user is collected and preprocessed to construct an identity feature sequence;

[0011] S2, a time sequence convolution network is used to construct a time sequence convolution identity verification network, each check-in user is modeled as an identity verification agent with multi-modal feature input and time sequence behavior observation capability, the identity feature sequence is taken as an observation input of the identity verification agent, and an identity feature representation is generated;

[0012] S3, based on the identity feature representation, a time sequence convolution identity verification network is used for preliminary identity discrimination, and a preliminary identity discrimination result is output;

[0013] S4, an ant algorithm is used to construct a multi-ant body collaborative optimization mechanism, a parameter search unit is configured for each ant body, and based on the search path and discrimination result performance of all ant bodies in the parameter space, a time sequence convolution identity verification network is generated;

[0014] S5, the identity feature sequence is input into the time sequence convolution identity verification network optimized by the ant algorithm, and an optimized identity discrimination result is obtained in real time in the identity verification process;

[0015] S6, the identity discrimination result is compared with identity data in a hotel real-name DID database, and a final identity verification result is output.

[0016] Optionally, the identity feature sequence specifically comprises user's identity document information, face image features, voiceprint features, and check-in behavior time sequence data.

[0017] Optionally, the identity discrimination result specifically comprises a pass or fail state of identity verification, a discrimination confidence score, a discrimination timestamp, and a corresponding user number.

[0018] Optionally, S2 specifically comprises:

[0019] S21, the identity feature sequence is input into a feature processing unit to obtain multi-modal feature data arranged in time sequence;

[0020] S22, the multi-modal feature data arranged in time sequence is subjected to modal recognition and feature extraction to respectively obtain image modal features, voice modal features, and text modal features;

[0021] S23, the image modal features, the voice modal features, and the text modal features are subjected to normalization processing, and a multi-modal feature vector is constructed to generate multi-modal feature input;

[0022] S24, inputting the multi-modal feature into an input layer of the time sequence convolution network, performing time sequence convolution operation, extracting time sequence behavior feature of the identity feature, and obtaining time sequence behavior feature representation:

[0023]

[0024] wherein Y i,t represents the time sequence behavior feature representation of the i-th feature at time t, X j,t-k represents the input value of the j-th modal feature at time t-k, W j,k represents the convolution kernel weight, M represents the number of modalities, K represents the length of the convolution kernel, i represents the i-th feature component, j represents the j-th modality, k represents the index of the convolution kernel in the time dimension, and t represents the t-th time in the time sequence;

[0025] S25, constructing a time sequence convolution identity verification network based on the time sequence behavior feature representation, modeling each check-in user as an identity verification agent with multi-modal feature input and time sequence behavior observation capability, and generating observation input of the identity verification agent;

[0026] S26, inputting the observation input of the identity verification agent into the time sequence convolution identity verification network, generating and outputting the identity feature representation.

[0027] Optionally, the S3 specifically comprises:

[0028] S31, inputting the identity feature representation into a discrimination unit of the time sequence convolution identity verification network to obtain discrimination input data;

[0029] S32, performing feature preprocessing on the discrimination input data to obtain preprocessed feature data;

[0030] S33, performing multi-layer convolution processing on the preprocessed feature data by using time sequence convolution operation, extracting a discrimination feature sequence, and obtaining discrimination feature representation:

[0031]

[0032] wherein Z m,s represents the m-th discrimination feature at the s-th layer, Y i,t represents the time sequence behavior feature representation of the i-th feature at time t, Q i,m,s represents the discrimination convolution weight, t represents the t-th time in the time sequence, M1 represents the number of feature components, T represents the number of time steps, m represents the number of discrimination features, and s represents the layer number of the convolution network;

[0033] S34, inputting the discrimination feature representation into a feature fusion unit, combining historical discrimination information, and generating fused discrimination features;

[0034] S35, threshold determination is performed on the fusion discriminant feature, a threshold of 0.7 is set, when the output result of the fusion discriminant feature is greater than or equal to 0.7, it is determined that the identity verification is passed, and a preliminary identity discriminant signal is output;

[0035] S36, the preliminary identity discriminant signal is output as a preliminary identity discriminant result.

[0036] Optionally, the S4 specifically comprises:

[0037] S41, an ant algorithm is adopted, a multi-ant body cooperative optimization mechanism is initialized, the number of ant bodies, a parameter search space and the number of optimization iterations are set, and an ant body initialization set is generated;

[0038] S42, a parameter search unit is configured for each ant body, different parameter search starting points are allocated, and an ant body parameter configuration set is obtained;

[0039] S43, each ant body independently performs path search in the parameter search space based on the current parameter configuration, records each search path and the corresponding parameter configuration scheme, and forms an ant body search path set;

[0040] S44, the parameter configuration corresponding to each ant body search path is applied to the time sequence convolution identity verification network, the respective discriminant result performance is obtained, and a discriminant result set is generated;

[0041] S45, based on the search path and the discriminant result set of all ant bodies, the pheromone distribution between the ant bodies is calculated, the search probability distribution of the ant bodies is adjusted, and an updated ant body parameter configuration set is output:

[0042]

[0043] Wherein, represents the pheromone intensity of the pth parameter position and the qth parameter value in the t'+1th iteration, represents the pheromone intensity of the pth parameter position and the qth parameter value in the t'th iteration, α represents the pheromone enhancement coefficient, p represents the parameter number, q represents the parameter value number, n represents the ant body number, d represents the feature dimension number, δ p,q,n represents 1 if the nth ant body takes the qth value at the pth parameter position on the path, otherwise 0, represents the feature representation of the nth ant body in the dth discriminant feature dimension in the t'th iteration, t' represents the number of iterations, N represents the number of input identity feature sequences, and D represents the number of feature dimensions of each input identity feature sequence.

[0044] S46, when the number of iterations reaches the maximum number of iterations 100 rounds or the optimal parameter configuration of all ant bodies does not change for 10 consecutive rounds, terminate the iteration, and generate an optimized timing convolution identity verification network according to the optimal ant body parameter configuration finally obtained.

[0045] Optionally, the S5 specifically includes:

[0046] S51, inputting the identity feature sequence into the optimized timing convolution identity verification network as input data of the optimized timing convolution identity verification network;

[0047] S52, performing convolution processing on the input identity feature sequence in the optimized timing convolution identity verification network in chronological order to generate corresponding timing convolution features:

[0048]

[0049] wherein, represents the convolution feature of the u-th output channel at the t-th iteration, represents the convolution feature of the u-th output channel at the t-th iteration, represents the time step, k represents the parameter sequence number, l represents the value sequence number of the k-th parameter, and K represents the total number of parameters, represents the pheromone intensity of the k-th parameter at the t-th iteration, L k represents the number of different values allowed by the k-th parameter;

[0050] S53, performing real-time feature extraction on the timing convolution features to obtain identity discrimination features at the current moment;

[0051] S54, inputting the identity discrimination features into the identity discrimination unit to calculate and output the optimized identity discrimination result in real time.

[0052] Optionally, the S6 specifically includes:

[0053] S61, performing standardization processing on the identity discrimination result to generate standardized identity feature data;

[0054] S62, extracting historical identity data corresponding to the current user from the hotel real-name distributed identity database as database identity comparison data;

[0055] S63, performing feature matching on the standardized identity feature data and the database identity comparison data to generate an identity matching score;

[0056] S64, comparing the identity matching score with a preset threshold 0.85 to output a final identity verification result.

[0057] The hotel check-in identity verification system based on the real-name DID according to the embodiment of the application includes the following modules:

[0058] An identity data acquisition module is configured to collect and preprocess the identity data of the check-in user and construct an identity feature sequence;

[0059] An identity preliminary discrimination module is configured to perform preliminary identity discrimination based on the identity feature representation and using a time-series convolution identity verification network;

[0060] An ant body algorithm module is configured to configure a parameter search unit for each ant body, generate the time-series convolution identity verification network based on the search path and discrimination result representation of all ant bodies in the parameter space, and optimize the time-series convolution identity verification network;

[0061] An identity discrimination module is configured to input the identity feature sequence into the ant body algorithm-optimized time-series convolution identity verification network and obtain an optimized identity discrimination result in real time during the identity verification process;

[0062] A final identity verification module is configured to compare the identity discrimination result with the identity data in the hotel real-name DID database and output a final identity verification result.

[0063] The present application has the following advantages:

[0064] The present application introduces multi-modal identity feature input, time-series convolution network modeling, and multi-dimensional parameter optimization of the ant colony algorithm, thereby comprehensively improving the dynamic adaptability and discrimination accuracy of the hotel real-name check-in identity verification system in complex identity data scenarios. With the unified modeling framework of multi-modal features, the system can effectively integrate multi-modal identity information such as facial images, voiceprint features, text features, and check-in behavior time-series data, overcoming the robustness problem of traditional single-modal comparison methods in variable environments such as illumination, angle, and voice quality. In particular, the introduction of the time-series convolution network into the multi-modal fusion modeling of the identity feature sequence enables the system to capture subtle changes in user identity features in the dynamic time dimension, extract deeper time-series feature representations, and ensure the integrity and time-series consistency of the identity features.

[0065] At the same time, the system realizes dynamic adaptive adjustment of the globally optimal parameter configuration through multi-objective parameter optimization based on the ant colony algorithm, solving the problem of single parameter configuration and lack of adaptive ability in traditional identity verification methods. The ant colony algorithm module can realize efficient global optimal solution search in the multi-modal multi-dimensional feature space through the cooperative search and pheromone distribution update of multiple ant bodies, significantly improving the adaptability and robustness of the system in different identity feature input scenarios. In particular, in the deep fusion process of multi-modal features and check-in behavior sequences, the introduction of the ant colony algorithm ensures the generalization ability of the system in the diversified feature space, effectively preventing the overfitting phenomenon caused by single training samples in traditional verification models.

[0066] In addition, the system combines historical identity data in the hotel real-name DID database in the final identity verification result output link, and further enhances the security and reliability of identity verification through a standardized comparison and matching score judgment mechanism. The dynamic setting of the comparison score threshold and the multi-level verification of the identity discrimination result make the identity verification result have higher credibility, which meets the requirements of high security, real-time and information traceability in the hotel real-name check-in scene. Overall, the "multi-modal fusion-time sequence convolution feature extraction-ant colony optimization-multi-dimensional comparison" integrated intelligent identity verification process constructed by the present application breaks through the technical bottlenecks of the prior art in multi-modal dynamic identity feature fusion, global optimal parameter configuration and multi-dimensional dynamic discrimination, and significantly improves the intelligent, real-time and accurate level of hotel check-in identity verification. BRIEF DESCRIPTION OF DRAWINGS

[0067] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application and explain the principles of the present application, and do not constitute a limitation of the present application. In the drawings:

[0068] Figure 1 A flowchart of the hotel check-in identity verification method based on real-name DID proposed by the present application;

[0069] Figure 2 A schematic diagram of the hotel check-in identity verification method based on real-name DID proposed by the present application;

[0070] Figure 3 A data flow diagram of the hotel check-in identity verification system based on real-name DID proposed by the present application. DETAILED DESCRIPTION

[0071] The present application will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, and only illustrate the basic structure of the present application in a schematic manner, and therefore only show the components related to the present application.

[0072] Reference Figures 1-2 The hotel check-in identity verification method based on real-name DID includes the following steps:

[0073] S1, collect the identity data of the check-in user and perform preprocessing to construct an identity feature sequence;

[0074] S2, adopt a time sequence convolution network to construct a time sequence convolution identity verification network, model each check-in user as an identity verification agent with multi-modal feature input and time sequence behavior observation capability, input the identity feature sequence as the observation of the identity verification agent, and generate an identity feature representation;

[0075] S3, based on the identity feature representation, using a time sequence convolution identity verification network for preliminary identity discrimination, outputting a preliminary identity discrimination result;

[0076] S4, adopting an ant algorithm, constructing a multi-ant body cooperative optimization mechanism, configuring a parameter search unit for each ant body, generating a time sequence convolution identity verification network based on the search path and discrimination result performance of all ant bodies in the parameter space;

[0077] S5, inputting the identity feature sequence into the ant algorithm optimized time sequence convolution identity verification network, and obtaining the optimized identity discrimination result in real time during the identity verification process;

[0078] S6, comparing the identity discrimination result with the identity data in the hotel real-name DID database, and outputting the final identity verification result.

[0079] Through the hotel check-in identity verification method based on real-name DID, the user identity feature sequence and the characteristics of the time sequence convolution network can be fully utilized to effectively extract and represent the user identity features, and the preliminary discrimination of the user identity can be realized. Combined with the ant colony algorithm to optimize the search path and the discrimination result, the accuracy and robustness of the identity verification are further improved, overcoming the single feature extraction and comparison and the problem of being easily disturbed in the traditional method, adapting to multi-modal, time sequence and complex identity data, improving the identity verification efficiency and security in the hotel real-name management, and ensuring the information security in the user check-in process.

[0080] In the embodiment, the identity feature sequence specifically includes the user's identity document information, facial image features, voiceprint features, and check-in behavior time sequence data.

[0081] By including the user's identity document information, facial image features, voiceprint features, and check-in behavior time sequence data in the identity feature sequence, the user's identity features can be more comprehensively and accurately represented. In this way, the system not only relies on single identity data during identity verification, but also performs multi-modal and multi-dimensional feature comparison, significantly improving the accuracy and robustness of identity recognition. At the same time, the fusion of multi-dimensional features effectively prevents misjudgment and fraud in traditional single feature recognition, providing a safer and more reliable technical guarantee for hotel real-name check-in, and enhancing the applicability and practicality of the system.

[0082] In the embodiment, the identity discrimination result specifically includes the pass or fail status of the identity verification, the discrimination confidence score, the discrimination timestamp, and the corresponding user number.

[0083] By providing detailed and quantifiable identity discrimination result output, including identity verification pass or not, discrimination confidence score, discrimination timestamp, user number and other multi-dimensional information. Such discrimination result not only provides more abundant information for the system, which is convenient for subsequent management and analysis, but also improves the explainability and transparency of identity verification. The system administrator can more accurately judge the reliability of the identity verification result according to the confidence score, which improves the management flexibility and service security; The record of user number and timestamp also facilitates the traceability of check-in behavior and the review of identity verification process, and enhances the manageability and traceability of the entire hotel real-name check-in identity verification system.

[0084] In the embodiment, S2 specifically includes:

[0085] S21, input the identity feature sequence into the feature processing unit to obtain the multi-modal feature data arranged in time sequence;

[0086] S22, modal recognition and feature extraction are performed on the multi-modal feature data arranged in time sequence to obtain image modal features, speech modal features and text modal features respectively;

[0087] S23, normalizing the image modal features, speech modal features and text modal features, and constructing a multi-modal feature vector to generate a multi-modal feature input;

[0088] S24, using a time series convolution network, inputting the multi-modal feature input into the input layer of the time series convolution network, performing time series convolution operation, extracting time series behavior features of the identity features, and obtaining time series behavior feature representation:

[0089]

[0090] Wherein, Y i,t represents the time series behavior feature representation of the i-th feature at time t, X j,t-k represents the input value of the j-th modal feature at time t-k, W j,k represents the convolution kernel weight, M represents the number of modalities, K represents the length of the convolution kernel, i represents the i-th feature component, j represents the j-th modality, k represents the index of the convolution kernel in the time dimension, and t represents the t-th time in the time series.

[0091] S25, based on the time series behavior feature representation, constructing a time series convolution identity verification network, modeling each check-in user as an identity verification agent with multi-modal feature input and time series behavior observation ability, and generating observation input of the identity verification agent;

[0092] S26, inputting the observation input of the identity verification agent into the time series convolution identity verification network to generate and output the identity feature representation.

[0093] By arranging in chronological order, multi-modal feature extraction and normalization processing, the system can fully integrate multi-modal data such as facial images, speech, text, etc. to form more rich and accurate multi-modal feature input. Combined with the modeling capability of the time sequence convolution network, the time sequence behavior features can be efficiently extracted, and the multi-dimensional time sequence features of the user identity can be dynamically captured, thereby significantly improving the accuracy and robustness of identity discrimination. Finally, through intelligent observation input, the defects of traditional identity verification methods limited by single mode and static features are effectively solved, and the intelligent and automated level of hotel real-name registration identity verification is improved.

[0094] In the embodiment, the S3 specifically includes:

[0095] S31, input the identity feature representation into the discrimination unit of the time sequence convolution identity verification network to obtain discrimination input data;

[0096] S32, perform feature preprocessing on the discrimination input data to obtain preprocessed feature data;

[0097] S33, perform multi-layer convolution processing on the preprocessed feature data by using time sequence convolution operation to extract discrimination feature sequence and obtain discrimination feature representation:

[0098]

[0099] wherein, Z m,s represents the feature representation of the mth discrimination feature at the s layer, Y i,t represents the time sequence behavior feature representation of the ith feature at time t, Q i,m,s represents the discrimination convolution weight, t represents the tth time in the time sequence, M1 represents the number of feature components, T represents the number of time steps, m represents the number of discrimination features, and s represents the layer number of the convolution network;

[0100] S34, input the discrimination feature representation into the feature fusion unit to generate fusion discrimination features combined with historical discrimination information;

[0101] S35, perform threshold determination on the fusion discrimination features, set the threshold value to 0.7, and when the output result of the fusion discrimination features is greater than or equal to 0.7, determine that the identity verification is passed, and output a preliminary identity discrimination signal;

[0102] S36, output the preliminary identity discrimination signal as a preliminary identity discrimination result.

[0103] Through multi-level time sequence convolution processing and feature extraction, the time sequence mode of the user identity behavior characteristics is fully mined, the discriminant feature sequence is generated and fused for discrimination, which can greatly improve the accuracy and robustness of identity discrimination. The threshold is set for the final judgment, the credibility and threshold range of the discrimination result are clear, the uncertainty caused by single data judgment is effectively prevented, and the efficiency and security of the system are ensured. At the same time, the historical discrimination information is fused, so that the system can consider the real-time and comprehensiveness of historical data in the preliminary identity discrimination, improve the intelligentization and self-adaptive ability of identity verification, and better adapt to the actual scene such as hotel real-name registration.

[0104] In the embodiment, the S4 specifically includes:

[0105] S41, an ant algorithm is adopted, a multi-ant body cooperative optimization mechanism is initialized, the number of ant bodies, the parameter search space and the number of optimization iterations are set, and an ant body initialization set is generated;

[0106] S42, a parameter search unit is configured for each ant body, different parameter search starting points are allocated, and an ant body parameter configuration set is obtained;

[0107] S43, each ant body independently searches a path in the parameter search space based on the current parameter configuration, records each search path and the corresponding parameter configuration scheme, and forms an ant body search path set;

[0108] S44, the parameter configuration corresponding to each ant body search path is applied to the time sequence convolution identity verification network, the respective discrimination result performance is obtained, and a discrimination result set is generated;

[0109] S45, based on the search path and the discrimination result set of all ant bodies, the pheromone distribution between the ant bodies is calculated, the search probability distribution of the ant bodies is adjusted, and the updated ant body parameter configuration set is output:

[0110]

[0111] wherein, represents the pheromone intensity of the pth parameter position and the qth parameter value in the t'+1th round of iteration, represents the pheromone intensity of the pth parameter position and the qth parameter value in the t'th round of iteration, α represents a pheromone enhancement coefficient, p represents a parameter number, q represents a parameter value number, n represents an ant body number, d represents a feature dimension number, δ p,q,n represents 1 if the nth ant body takes the qth value at the pth parameter position on the path, otherwise 0, represents the feature representation of the nth ant body on the dth discriminant feature dimension at the t'th iteration, t' represents the number of iterations, N represents the number of input identity feature sequences, and D represents the number of feature dimensions of each input identity feature sequence.

[0112] S46. When the number of iterations reaches the maximum number of iterations of 100, or the optimal parameter configurations of all ants do not change for 10 consecutive rounds, the iteration is terminated, and an optimized temporal convolutional identity verification network is generated based on the optimal ant parameter configuration finally obtained.

[0113] The present invention proposes a multi-ant collaborative optimization method based on the ant colony algorithm, which can fully utilize the adaptive search capabilities of the ant colony to perform efficient global search and optimization of the parameter space. Through the independent path search and pheromone distribution update of each ant, the feature configuration optimization under multimodal and multi-parameter dimensions is guaranteed, overcoming the problem of traditional methods that are prone to falling into local optimality, and significantly improving the robustness and generalization ability of the final generated time-series convolution identity verification network. Finally, combined with the optimal parameter configuration results, the verification network structure and discrimination parameters can be adjusted dynamically and accurately to ensure the high precision and intelligence level of the hotel real-name check-in identity verification system in complex scenarios.

[0114] In this embodiment, the S5 specifically includes:

[0115] S51, inputting the identity feature sequence into the optimized temporal convolutional identity verification network as input data of the optimized temporal convolutional identity verification network;

[0116] S52. In the optimized time series convolution identity verification network, the input identity feature sequence is convolved in time order to generate corresponding time series convolution features:

[0117]

[0118] in, Indicates in The convolution feature of the u-th output channel at the iteration, represents the time step, k represents the parameter number, l represents the value number of the kth parameter, and K represents the total number of parameters. Indicates the pheromone intensity of the kth parameter with the lth value at the tth iteration, L k Indicates the number of different values ​​allowed for the kth parameter;

[0119] S53, performing real-time feature extraction on the temporal convolution feature to obtain the identity discrimination feature at the current moment;

[0120] S54: Input the identity discrimination feature into the identity discrimination unit, and calculate and output the optimized identity discrimination result in real time.

[0121] The application proposes inputting a user identity feature sequence into an optimized time sequence convolution identity verification network, and performing multi-level and time sequence convolution feature extraction, which can effectively capture the time sequence change trend of the user identity feature, and realize more accurate identity dynamic feature expression. Through real-time feature extraction of the convolution feature and optimized calculation of the discrimination unit, the generated identity discrimination result has higher real-time and accuracy, which significantly improves the discrimination ability and dynamic adaptability of the hotel real-name check-in identity verification system, and guarantees the security and reliability of the identity verification.

[0122] In the embodiment, the S6 specifically includes:

[0123] S61, standardizing the identity discrimination result to generate standardized identity feature data;

[0124] S62, extracting historical identity data corresponding to the current user from the hotel real-name distributed identity database as database identity comparison data;

[0125] S63, performing feature matching on the standardized identity feature data and the database identity comparison data to generate an identity matching score;

[0126] S64, comparing the identity matching score with a preset threshold value 0.85 to output a final identity verification result.

[0127] By standardizing the preliminary identity discrimination result and matching it with the historical identity data in the hotel real-name distributed identity database, the problem of false matching caused by inconsistent data dimensions or feature distribution differences can be effectively avoided. Setting an identity matching score threshold for accurate comparison ensures the high accuracy and security of the final identity verification result. This method can fully utilize the combination of user historical data and real-time discrimination results to improve the dynamic adaptability and robustness of identity verification, significantly improve the intelligence and reliability of the hotel real-name check-in identity verification system, and provide better technical support for practical applications.

[0128] Reference Figure 3 The hotel check-in identity verification system based on the real-name DID includes the following modules:

[0129] An identity data acquisition module is configured to collect and preprocess the identity data of the check-in user, and construct an identity feature sequence;

[0130] An identity preliminary discrimination module is configured to perform preliminary identity discrimination based on the identity feature representation and using the time sequence convolution identity verification network;

[0131] An ant body algorithm module is configured to configure a parameter search unit for each ant body, and generate the time sequence convolution identity verification network based on the search path and discrimination result performance of all ant bodies in the parameter space.

[0132] An identity discrimination module is configured to input the identity feature sequence into an ant colony algorithm optimized time sequence convolution identity verification network, and obtain an optimized identity discrimination result in real time in an identity verification process.

[0133] A final identity verification module is configured to compare the identity discrimination result with identity data in a hotel real-name DID database, and output a final identity verification result.

[0134] The present application constructs a multi-module and integrated hotel real-name check-in identity verification system, covering the whole process modules of identity data acquisition, preliminary discrimination, ant colony algorithm optimization, identity discrimination and final comparison. Through the modular design, the system can realize efficient cooperation and flexible combination of the functions of each module, improve the accuracy and efficiency of identity verification. The ant colony algorithm module provides global optimization capability, overcoming the local optimal problem of traditional algorithms; at the same time, the final identity verification module combines with the actual DID data of the hotel to ensure the high credibility and real-time of the final verification result. Overall, the system is intelligent and secure, and is suitable for identity verification requirements in complex check-in scenarios.

[0135] Embodiment 1:

[0136] In order to verify the feasibility of the present application in implementation, the present application is applied to an international hotel, the number of hotel check-ins surges, and the daily reception user quantity of the front desk exceeds 800 times. In the face of the check-in peak, the traditional manual verification method mainly relies on manual comparison of identity cards and user facial features, combined with simple voice verification. However, in actual operation, users often wear masks, facial expression differences and voiceprint collection distortion in noisy environments, resulting in a decrease in the accuracy of manual verification and a significant increase in time consumption. Especially during the peak period, the efficiency of manual verification becomes the key bottleneck restricting the smoothness of the front desk check-in process.

[0137] In order to cope with this challenge, the hotel introduces the hotel check-in identity verification system and method based on real-name DID of the present application. The use process of the system is very simple, when the user arrives at the front desk, the identity card is presented, the identity card reader of the front desk automatically collects the card information, the high-definition camera synchronously acquires the user facial image, and the voiceprint collection module guides the user to speak briefly to obtain the voiceprint feature. All data are synchronously uploaded to the system of the present application, combined with the check-in time, historical check-in behavior and other dynamic information of the user, to generate a multi-modal user identity feature sequence.

[0138] Identity feature data first passes through a multi-modal fusion module for dynamic feature extraction of face, voiceprint, text, and check-in behavior timing features. Then, through a system timing convolution network layer, the feature evolution of the user at different time points is automatically modeled, and the dynamic change pattern of the user's identity is captured. In this process, the ant colony algorithm module plays a key role in automatically finding the optimal parameter configuration in the multi-modal feature space, avoiding the one-sidedness of manually setting parameters, and ensuring that the system can achieve optimal discrimination in the face of different user features and environmental interference.

[0139] In the actual check-in verification process, the system can output the identity verification result in only 2-3 seconds, including verification confidence score, discrimination result, and discrimination time. On the same day, 20 random users were selected, covering men and women of all ages, different occupations, and different language backgrounds, to compare the system with the traditional manual verification results in detail. The verification data of each user was recorded completely for subsequent data statistics and analysis.

[0140] Table 1: Multi-modal feature fusion result table

[0141] User number Face matching degree Voiceprint matching degree Text matching degree Behavior matching degree Fusion confidence U0001 0.93 0.89 0.95 0.96 0.94 U0002 0.88 0.86 0.92 0.9 0.89 U0003 0.42 0.5 0.58 0.47 0.49 U0004 0.95 0.9 0.96 0.97 0.95 U0005 0.9 0.85 0.93 0.92 0.9

[0142] Table 2: Verification result and time comparison table

[0143]

[0144] Table 1 shows that the average confidence of the system after multi-modal fusion is 0.83, while the average confidence of manual verification is only 0.81 below 0.83. Especially for cases like U0003 where the identity features are inconsistent, the system accurately identifies the low fusion confidence of 0.49, which is significantly lower than the subjective judgment of manual verification of 0.75. The system exhibits higher sensitivity and robustness when encountering ambiguous data or significant dynamic feature differences, avoiding subjective omissions in manual visual inspection. The average verification time of the system is 2.4 seconds, while manual verification is generally above 5.6 seconds, and is more likely to be delayed during busy periods. The system maintains high efficiency while ensuring verification consistency, with the exception of abnormal users, the system and manual verification results are completely consistent, with a verification accuracy of 98%.

[0145] The above describes only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can make equivalent replacements or changes to the technical solutions and inventive concepts of the present application within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A hotel check-in identity verification method based on real-name DID, characterized by: The steps include: S1. Collect and pre-process the identity data of the occupants to construct an identity feature sequence; S2. Use a time-series convolutional network to build a time-series convolutional identity verification network. Each user who checks in is modeled as an identity verification agent with multimodal feature input and time-series behavior observation capabilities. The identity feature sequence is used as the observation input of the identity verification agent to generate an identity feature representation. S3. Based on the identity feature representation, use the temporal convolutional identity verification network to perform preliminary identity discrimination and output the preliminary identity discrimination result; S4. Use the ant algorithm to build a multi-ant collaborative optimization mechanism, configure a parameter search unit for each ant, and generate a temporal convolutional identity verification network based on the search paths and discrimination results of all ants in the parameter space; S5. Input the identity feature sequence into the temporal convolutional identity verification network optimized by the ant algorithm, and obtain the optimized identity discrimination result in real time during the identity verification process; S6. Compare the identity identification result with the identity data in the hotel's real-name DID database and output the final identity verification result.

2. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The identity feature sequence specifically includes the user's identity document information, facial image features, voiceprint features, and check-in behavior time series data.

3. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The identity identification result specifically includes the pass or fail status of the identity verification, the identification confidence score, the identification timestamp, and the corresponding user number.

4. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The S2 specifically includes: S21, inputting the identity feature sequence into a feature processing unit to obtain multimodal feature data arranged in chronological order; S22, performing modality recognition and feature extraction on the multimodal feature data arranged in chronological order to obtain image modality features, speech modality features, and text modality features respectively; S23, normalizing the image modal features, the speech modal features, and the text modal features, and constructing a multimodal feature vector to generate a multimodal feature input; S24. Using a temporal convolutional network, the multimodal feature input is fed into the input layer of the temporal convolutional network to perform a temporal convolution operation, extract the temporal behavior features of the identity features, and obtain the temporal behavior feature representation: Among them, Y i,t represents the temporal behavior feature representation of the i-th feature at time t, X j,t-k represents the input value of the jth modal feature at time tk, W j,k Represents the convolution kernel weight, M represents the number of modes, K represents the convolution kernel length, i represents the i-th feature component, j represents the j-th mode, k represents the index of the convolution kernel in the time dimension, and t represents the t-th moment in the time series; S25. Based on the temporal behavior feature representation, a temporal convolutional identity verification network is constructed. Each user who checks in is modeled as an identity verification agent with multimodal feature input and temporal behavior observation capabilities, and observation input for the identity verification agent is generated. S26. Input the observation input of the identity verification agent into the time-series convolutional identity verification network to generate and output the identity feature representation.

5. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The S3 specifically includes: S31, inputting the identity feature representation into the discrimination unit of the temporal convolutional identity verification network to obtain discrimination input data; S32, performing feature preprocessing on the discrimination input data to obtain preprocessed feature data; S33. Perform multi-layer convolution processing on the pre-processed feature data using a temporal convolution operation to extract the discriminant feature sequence and obtain the discriminant feature representation: Among them, Z m,s represents the feature representation of the mth discriminant feature at the sth layer, Y i,t represents the temporal behavior feature representation of the i-th feature at time t, Q i,m,s represents the discriminant convolution weight, t represents the t-th moment in the time series, M1 represents the number of feature components, T represents the number of time steps, m represents the number of discriminant features, and s represents the layer number of the convolutional network; S34, inputting the discriminant feature representation into the feature fusion unit, combining the historical discriminant information, and generating a fused discriminant feature; S35. Perform threshold determination on the fused discriminant features, setting the threshold to 0.

7. When the output result of the fused discriminant features is greater than or equal to 0.7, it is determined that the identity verification has passed, and a preliminary identity discrimination signal is output; S36: Output the preliminary identity identification signal as a preliminary identity identification result.

6. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The S4 specifically includes: S41, using the ant algorithm to initialize the multi-ant collaborative optimization mechanism, set the number of ants, parameter search space and optimization iteration number, and generate an ant initialization set; S42, configuring a parameter search unit for each ant body, assigning different parameter search starting points, and obtaining an ant body parameter configuration set; S43, each ant independently searches for a path in the parameter search space based on the current parameter configuration, records each search path and the corresponding parameter configuration scheme, and forms an ant search path set; S44, applying the parameter configuration corresponding to each ant body search path to the temporal convolutional identity verification network to obtain respective discrimination result performances and generate a discrimination result set; S45. Based on the search paths and discrimination result sets of all ants, calculate the pheromone distribution among the ants, adjust the search probability distribution of the ants, and output the updated ant parameter configuration set: in, Indicates the pheromone intensity of the pth parameter position and the qth parameter value in the t'+1th iteration, In the t'th iteration, the pheromone intensity of the pth parameter position and the qth parameter value is expressed, α represents the pheromone enhancement coefficient, p represents the parameter number, q represents the parameter value number, n represents the ant body number, d represents the feature dimension number, δ p,q,n If the nth ant takes the qth parameter position on the path, it is 1, otherwise it is 0. Indicates the feature representation of the nth ant on the dth discriminant feature dimension in the t'th iteration, t' represents the number of iterations, N represents the number of input identity feature sequences, and D represents the number of feature dimensions of each input identity feature sequence; S46. When the number of iterations reaches the maximum number of iterations of 100, or the optimal parameter configurations of all ants do not change for 10 consecutive rounds, the iteration is terminated, and an optimized temporal convolutional identity verification network is generated based on the optimal ant parameter configuration finally obtained.

7. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The S5 specifically includes: S51, inputting the identity feature sequence into the optimized temporal convolutional identity verification network as input data of the optimized temporal convolutional identity verification network; S52. In the optimized time series convolution identity verification network, the input identity feature sequence is convolved in time order to generate corresponding time series convolution features: in, Indicates in The convolution feature of the u-th output channel at the iteration, represents the time step, k represents the parameter number, l represents the value number of the kth parameter, and K represents the total number of parameters. Indicates the pheromone intensity of the kth parameter with the lth value at the tth iteration, L k Indicates the number of different values ​​allowed for the kth parameter; S53, performing real-time feature extraction on the temporal convolution feature to obtain the identity discrimination feature at the current moment; S54: Input the identity discrimination feature into the identity discrimination unit, and calculate and output the optimized identity discrimination result in real time.

8. The hotel check-in identity verification method based on real-name DID according to claim 1 is characterized in that: The S6 specifically includes: S61, standardizing the identity identification result to generate standardized identity feature data; S62. Extract historical identity data corresponding to the current user from the hotel's real-name distributed identity database as database identity comparison data; S63. Perform feature matching on the standardized identity feature data and the identity comparison data in the database to generate an identity matching score; S64. Compare the identity matching score with the preset threshold of 0.85 and output the final identity verification result.

9. A hotel check-in identity verification system based on real-name DID, which implements the hotel check-in identity verification method based on real-name DID according to any one of claims 1 to 8, characterized in that: Includes the following modules: Identity data acquisition module, used to collect and pre-process the identity data of the occupants and construct an identity feature sequence; The preliminary identity identification module is used to perform preliminary identity identification based on identity feature representation using a temporal convolutional identity verification network; The ant algorithm module is used to configure a parameter search unit for each ant, and generate a temporal convolutional identity verification network based on the search paths and discrimination results of all ants in the parameter space; The identity recognition module is used to input the identity feature sequence into the time-series convolution identity verification network optimized by the ant algorithm, and obtain the optimized identity recognition results in real time during the identity verification process; The final identity verification module is used to compare the identity recognition results with the identity data in the hotel's real-name DID database and output the final identity verification results.

Images