A scenic spot intelligent ticketing gate control system based on multi-modal verification

Abstract

This invention relates to the field of control system technology, specifically disclosing a multimodal verification-based intelligent ticketing gate control system for scenic areas. The system includes a multimodal spatiotemporal synchronous acquisition module, a spatiotemporal alignment and feature normalization module, a dual-link dynamic verification module, a risk-based access decision module, a gate adaptive execution module, and a closed-loop feedback optimization module. This invention adopts a standardized and modular design, adaptable to both large-scale deployment of newly built scenic area gates and upgrades of existing gates without requiring extensive hardware replacement, effectively controlling upgrade costs. It is widely applicable to various cultural and tourism scenarios such as mountainous scenic areas, amusement parks, and museums, comprehensively improving the intelligence level of scenic area ticketing management, optimizing the visitor experience, and possessing significant economic benefits and industry promotion value.

Description

Technical Field

[0001] This invention relates to the field of control system technology, specifically to a smart ticketing gate control system for scenic areas based on multimodal verification. Background Technology

[0002] With the in-depth development of China's cultural tourism industry, the number of tourists visiting scenic spots continues to rise, and during peak holiday periods, entrances face pressure from large crowds and high concurrency in verification and passage. As the core infrastructure for scenic spot entry management, the ticketing gate control system's verification accuracy, passage efficiency, security capabilities, and environmental adaptability directly determine the operational efficiency of the scenic spot and the visitor experience.

[0003] Currently, scenic area ticket gate control systems have gradually evolved from traditional manual verification of paper tickets and electronic ticket scanning verification to intelligent biometric verification. Facial recognition, with its advantages of being contactless and convenient, has become the mainstream verification method. However, in large-scale practical applications in scenic areas, existing technologies still have many core shortcomings: First, the verification modality is too simplistic, resulting in insufficient security and anti-attack capabilities. Existing turnstiles mostly use a "QR code / ID card + single facial recognition" model, which is highly vulnerable to forgery using photos, videos, 3D-printed masks, and other methods. This makes it impossible to prevent identity theft, scalping, and multiple uses of a single ticket. Furthermore, the single-modal verification has a low fault tolerance rate; changes in appearance such as wearing masks, sunglasses, or altering makeup can significantly reduce the verification pass rate, leading to congestion at the park entrance.

[0004] Second, the spatiotemporal misalignment of multimodal data leads to insufficient verification stability. Existing multimodal verification gates generally suffer from asynchronous and spatially mismatched data collection from multiple sources: data collection from faces, irises, ID cards, and tickets has time delays, easily resulting in mismatches between preceding and subsequent visitors; image distortion and spatial misalignment caused by visitor positioning deviations are not considered, making it impossible to guarantee that multimodal data corresponds to the same subject in the same spatiotemporal context, easily leading to misjudgments in feature matching and distorted verification results, frequently resulting in mis-allowing and mis-blocking in high-traffic scenarios.

[0005] Third, the verification weights are fixed, resulting in poor adaptability to all scenarios. The feature weights of existing multimodal verification solutions are all preset fixed values ​​at the factory, which cannot be dynamically adjusted according to real-time working conditions such as ambient light and collection distance. In outdoor environments with strong light, backlight, or low light at night, the verification accuracy will drop sharply. At the same time, existing liveness detection relies on only a single facial feature, which is a single detection dimension and cannot cope with highly realistic forgery attacks, posing a security risk.

[0006] Fourth, the fixed gate operation strategy makes it difficult to balance passage efficiency and security control. The opening and closing time and prompt mode of the existing gates are all fixed settings, which cannot be adaptively adjusted according to the verification risk level. This makes it impossible to achieve rapid passage for low-risk tourists, nor can it provide effective early warning and secondary verification for high-risk verification results. Moreover, the existing anti-tailgating monitoring relies mostly on single-point infrared beam sensors, which have a single detection dimension, are easy to circumvent, and have weak protection capabilities.

[0007] Fifth, the system lacks closed-loop self-optimization capabilities, resulting in high long-term operation and maintenance costs. The existing gate verification model and weight parameters cannot be iteratively optimized based on the actual operating data of the scenic area, abnormal events, and manual review results. As the environment changes and attack methods are upgraded, the verification accuracy continues to decline, requiring on-site manual debugging by technical personnel, which leads to high operation and maintenance costs and makes it difficult to adapt to the long-term dynamic operation needs of the scenic area.

[0008] Based on this, in order to address the aforementioned technical deficiencies of existing scenic area ticket gate control systems, developing an intelligent ticket gate control system capable of multimodal spatiotemporal synchronous verification, dynamic weight adaptation, risk-level control, and closed-loop self-optimization has become an urgent technical problem to be solved in this field. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a multimodal verification-based intelligent ticketing gate control system for scenic areas, which solves the problems mentioned in the background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solution: a scenic area intelligent ticketing gate control system based on multimodal verification, comprising: a multimodal spatiotemporal synchronous acquisition module, a spatiotemporal alignment and feature normalization module, a dual-link dynamic verification module, a risk-level access decision module, a gate adaptive execution module, and a closed-loop feedback optimization module; The multimodal spatiotemporal synchronous acquisition module is used to synchronously trigger the face image acquisition unit, iris acquisition unit, ID card reading unit, electronic ticket verification unit and spatial positioning unit to collect tourists' facial biometrics, iris biometrics, ID card information, electronic ticket voucher information and tourist spatial coordinate information in the same spatiotemporal space. It binds a unified timestamp and spatial anchor point to each set of collected data to construct a multimodal verification dataset with spatiotemporal labels. The spatiotemporal alignment and feature normalization module is used to perform dual alignment of the multimodal verification dataset in the temporal and spatial domains based on a unified timestamp and spatial anchor point, eliminating the acquisition delay and spatial offset of multi-source data. At the same time, the aligned multimodal features are subjected to dimensionless normalization processing to generate a standardized multimodal feature matrix. The dual-link dynamic verification module is used to execute two links—identity compliance verification and liveness verification—based on a standardized multimodal feature matrix, and calculates and generates an identity compliance coefficient. With confidence coefficient of living organism and combined with ticket validity identification Generate a comprehensive verification score ; The risk-based access decision module is used to pre-set multi-level risk thresholds and integrate the verification scores. The risk level of tourist passage is classified by comparing it with multi-level risk thresholds, and corresponding passage control instructions, early warning instructions or secondary verification instructions are generated based on the risk level. The gate adaptive execution module is used to receive passage control commands, adaptively adjust the gate opening duration, passage channel width, and sound and light prompt mode according to the risk level, and execute the corresponding gate opening and closing actions, while collecting passage status data in the channel in real time. The closed-loop feedback optimization module is used to feed back the passage status data, verification result data and abnormal event data to the dual-link dynamic verification module, dynamically adjust the weight parameters of identity compliance verification and liveness verification, and realize the system's self-optimization iteration.

[0011] Preferably, the multimodal spatiotemporal synchronous acquisition module includes a synchronization triggering unit, a biometric acquisition unit, a document and ticket verification unit, and a spatial positioning unit; The synchronous triggering unit is used to send collection commands to the biometric feature collection unit, the document and ticket verification unit, and the spatial positioning unit simultaneously, using the infrared human body sensing signal of the gate channel as the trigger source, to ensure that the triggering time difference of all collection actions does not exceed a preset delay threshold. ; The biometric acquisition unit includes a binocular face acquisition subunit and a near-infrared iris acquisition subunit, which are used to acquire visible light face image sequences and near-infrared iris image sequences of tourists, respectively. The document and ticket verification unit includes an ID card chip reading subunit and a ticket QR code reading subunit, which are used to read the encrypted identity information in the tourist's resident ID card and the ticket validity period, entry permission and personnel binding information in the scenic area's electronic ticket QR code, respectively. The spatial positioning unit includes a depth camera and a millimeter-wave radar, used to collect the real-time three-dimensional spatial coordinates of tourists within the turnstile channel and determine the relative distance between the tourists and the turnstile's data collection terminal. With lateral offset , as a spatial anchor point.

[0012] Preferably, the spatiotemporal alignment and feature normalization module includes a temporal domain alignment unit, a spatial domain alignment unit, and a feature normalization unit; The time domain alignment unit is used to bind a corresponding collection timestamp to each set of collected biometric features, document information, ticketing information, and spatial coordinate data, using the instruction sending time of the synchronization trigger unit as the base time, and to remove data whose difference between the timestamp and the base time exceeds a delay threshold. Invalid data is used to complete time domain alignment; The spatial alignment unit is used to align relative distances collected by the spatial positioning unit. With lateral offset Perspective correction and cropping are performed on facial and iris images to eliminate image distortion caused by tourist positioning deviations. At the same time, spatial subject matching is performed on the ID information, ticket information and corrected biometric images to ensure that all data correspond to the same tourist subject and complete spatial domain alignment. The feature normalization unit is used to extract facial feature vectors from the aligned multimodal data. iris feature vector ID document identity feature vector The L2 normalization process is applied to all feature vectors to unify the magnitude of the feature vectors to 1, thereby generating a standardized multimodal feature matrix.

[0013] Preferably, the dual-link dynamic verification module includes an identity compliance verification unit, a liveness verification unit, and a dynamic weight configuration unit; The identity compliance verification unit is used to process the facial feature vector in the standardized multimodal feature matrix. iris feature vector Respectively with the identity feature vector of the document The similarity of the corresponding pre-stored registered feature vectors is calculated by performing similarity matching. Similarity to iris features and combined with relative distance Generate identity compliance coefficient The identity compliance coefficient Obtain it using the following formula: ; In the formula, Facial feature weights, For iris feature weights, To achieve the optimal sampling distance, This is the distance attenuation coefficient. is the base of the natural logarithm; The dynamic weight configuration unit is used to adjust the ambient light intensity during data acquisition. Dynamic adjustment and The values ​​of , including ambient light intensity The higher, The larger the value, The smaller the value, the greater the value, and vice versa.

[0014] Preferably, the liveness verification unit is used to extract dynamic features from a series of continuously acquired face image sequences and iris image sequences, respectively, to obtain the average amplitude of facial micro-movements. Pupil diameter change rate and the 3D depth variance of human face Based on the above characteristics, a liveness confidence coefficient is calculated. The liveness confidence coefficient Obtain it using the following formula: ; In the formula, The preset liveness feature weight coefficients are given, and satisfy the following conditions: ; This is the Sigmoid normalization function, used to map eigenvalues ​​to the [0,1] interval; Among them, the pupil diameter change rate Obtain it using the following formula: ; In the formula, For the first The pupil diameter detected in the frame of the iris image. This represents the total number of iris image frames acquired continuously.

[0015] Preferably, the dual-link dynamic verification module further includes a comprehensive verification calculation unit, which is used to obtain the ticket validity identifier. And combined with identity compliance coefficient With confidence coefficient of living organism Calculate and generate a comprehensive verification score. The verification comprehensive score Obtain it using the following formula: ; In the formula, To comprehensively verify the weighting coefficients and meet the requirements The ticket validity identifier The value is: when the ticket information is verified (within the validity period, has park entry authority, and is consistent with the identity information), When ticket information verification fails, .

[0016] Preferably, the risk-based access decision module includes a threshold preset unit, a risk level classification unit, and an instruction generation unit; The threshold preset unit is used to preset three levels of risk thresholds, namely, safety thresholds. Warning threshold Interception threshold And satisfy ; The risk level classification unit is used to calculate the overall verification score. The corresponding passage risk level is determined by comparing the data with the Level 3 risk threshold, as detailed below: when At that time, it is classified into security levels; when At that time, it is classified as a level of attention; when At that time, it is classified into risk levels; The instruction generation unit is used to generate corresponding instructions according to the risk level: a safe level generates a pass instruction, a concern level generates a pass instruction with audio and visual prompts and a secondary verification prompt, and a risk level generates an interception instruction and a manual review warning.

[0017] Preferably, the gate adaptive execution module includes a gate drive unit, a passage status monitoring unit, and an adaptive parameter adjustment unit; The gate drive unit is used to receive passage control commands and drive the gate body to perform opening and closing actions. The passage status monitoring unit is used to monitor the passage progress of tourists in real time through infrared photoelectric sensors and depth cameras in the passage, and to determine whether tourists have completely passed through the gate passage. The adaptive parameter adjustment unit is used to adjust the parameters based on the passage risk level and the comprehensive verification score. Dynamically adjust the opening and holding time of the gate. The duration of the open state Obtain it using the following formula: ; In the formula, Based on the base opening duration, This is the maximum adjustable duration increment; when hour, Take the first preset value; when hour, Take the second preset value, and the second preset value is greater than the first preset value.

[0018] Preferably, the scenic area intelligent ticket gate control system based on multimodal verification further includes a channel anti-tailgating monitoring module, which is signal-connected to the risk-level access decision module; The channel anti-tailgating monitoring module is used to collect real-time 3D point cloud data and personnel movement trajectories within the turnstile channel using millimeter-wave radar and a depth camera, and to calculate the effective number of personnel within the channel. Minimum distance between people and the speed of personnel passage ; When detected and Preset spacing threshold, or When the preset speed threshold is reached, the channel anti-tailgating monitoring module immediately sends a tailgating warning signal to the risk-level passage decision module. After receiving the signal, the risk-level passage decision module forcibly switches the current passage control command to an interception command and triggers an audible and visual alarm.

[0019] Preferably, the closed-loop feedback optimization module includes a data acquisition unit, a parameter iteration unit, and a model update unit; The data acquisition unit is used to collect daily verification result data, passage status data, abnormal event data, and manual review result data to construct an optimized dataset; The parameter iteration unit is used to calculate the verification pass rate and false judgment rate under different ambient light intensities and different collection distances based on the optimized dataset, and dynamically adjust the facial feature weights in the identity compliance coefficient calculation formula. Iris feature weights And the weighting coefficient of living features in the formula for calculating the confidence coefficient of living individuals. The iterative formula is as follows: ; ; In the formula, For the first The weight values ​​after the next iteration The first The accuracy rate of face verification and iris verification within each iteration cycle. The learning rate; The model update unit is used to incrementally fine-tune the face feature extraction model and the iris feature extraction model based on the optimized dataset, thereby improving the feature extraction accuracy in complex scenarios.

[0020] This invention addresses the shortcomings of existing technologies by providing a multimodal verification-based intelligent ticketing gate control system for scenic areas. Through a full-link architecture design encompassing multimodal spatiotemporal synchronous data acquisition, spatiotemporal dual alignment, dual-link dynamic verification, risk-based hierarchical decision-making, gate adaptive execution, and closed-loop feedback optimization, it achieves high security, high precision, high adaptability, and high efficiency in the verification process. Compared to existing technologies, it offers the following significant advantages: 1. This invention breaks through the security bottleneck of single-modal verification by fusing multimodal data fusion verification of facial and iris dual biometric features, ID cards, electronic tickets, and spatial positioning; it achieves dual security verification from the two core dimensions of "consistency between person, ID card, and ticket" and "authenticity of real person" through independent dual-link verification of identity compliance and liveness authenticity, fundamentally eliminating illegal activities such as forgery, impersonation, and scalping, while solving the problem of verification failure in scenarios such as tourists wearing masks, thus balancing verification security and pass rate in complex scenarios.

[0021] 2. This invention utilizes a multimodal spatiotemporal synchronous acquisition module, using infrared human body sensing signals as the trigger source, to synchronously trigger multi-unit acquisition actions, strictly controlling acquisition latency differences, and binding a unified timestamp and spatial anchor point to each group of data. Then, through a spatiotemporal alignment and feature normalization module, it completes the removal of invalid data in the time domain and image correction and subject matching in the spatial domain, eliminating acquisition latency and spatial offset of multi-source data, ensuring that all verification data correspond to the same tourist subject in the same spatiotemporal context, avoiding data mismatch and verification distortion from the source, and significantly improving the verification stability in high-traffic scenarios.

[0022] 3. This invention uses a dynamic weight configuration unit to dynamically adjust the verification weights of facial and iris features based on the ambient light intensity. In strong light environments, the weight of facial features is increased, while in weak light environments, the weight of iris features is increased. This fully leverages the advantages of different biometric features in various scenarios and solves the industry pain point of fixed weight schemes experiencing a sharp drop in accuracy under complex lighting conditions. At the same time, the identity compliance coefficient is corrected by a distance attenuation factor to eliminate the impact of collection distance deviation on the verification results, thereby improving the verification accuracy and robustness in different station scenarios.

[0023] 4. This invention uses a preset three-level risk threshold to classify access risk levels based on comprehensive verification scores, generating differentiated control instructions. Simultaneously, through an adaptive gate execution module, parameters such as gate opening duration and audio-visual prompts are dynamically adjusted according to the risk level. Safe-level visitors are quickly allowed passage, improving efficiency; at-risk visitors, audio-visual prompts provide secondary verification reminders; and high-risk visitors are directly intercepted, triggering manual warnings, thus balancing efficiency with tiered safety management. Furthermore, a three-dimensional anti-tailgating monitoring module using millimeter-wave radar and a depth camera can accurately identify the number of people, spacing, and passage speed, completely solving the problems of easy evasion and low accuracy in traditional anti-tailgating methods, thus strengthening the park entry safety defense line.

[0024] 5. This invention, through a closed-loop feedback optimization module, can collect data on passage status, verification results, abnormal events, and manual review in real time, construct an optimized dataset, dynamically iterate and adjust verification weight parameters based on actual operating data, and incrementally fine-tune the feature extraction model. This enables the system to continuously adapt to changes in the scenic area environment, changes in passenger flow characteristics, and upgrades in attack methods, achieving autonomous iterative optimization without the need for on-site manual debugging by technical personnel, significantly reducing long-term operation and maintenance costs, and ensuring the stability of the system throughout its entire lifecycle.

[0025] In summary, this invention adopts a standardized and modular design, which can be adapted to the large-scale deployment of newly built scenic area gates, as well as to the transformation and upgrading of existing scenic area gates without large-scale hardware replacement, effectively controlling transformation costs. It can be widely adapted to various cultural and tourism scenarios such as mountain scenic areas, amusement parks, and museums, comprehensively improving the intelligent level of scenic area ticketing management, optimizing the visitor entry experience, and possessing significant economic benefits and industry promotion value. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the process of a scenic area intelligent ticketing gate control system based on multimodal verification as described in this invention; Figure 2 This is a schematic diagram of the overall principle of a scenic area intelligent ticketing gate control system based on multimodal verification as described in this invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Reference Figure 1 and Figure 2 This invention discloses a multimodal verification-based intelligent ticketing gate control system for scenic areas. Its core components include six modules: a multimodal spatiotemporal synchronous acquisition module, a spatiotemporal alignment and feature normalization module, a dual-link dynamic verification module, a risk-based access decision module, a gate adaptive execution module, and a closed-loop feedback optimization module. It also includes a channel anti-tailgating monitoring module connected to the risk-based access decision module. All modules achieve low-latency communication via an industrial Ethernet network within the scenic area, collaboratively completing multimodal verification, risk-based control, gate adaptive execution, and system closed-loop self-optimization throughout the entire visitor entry process. This addresses the industry pain points of existing scenic area gates, such as low verification accuracy, weak anti-attack capabilities, poor environmental adaptability, rigid control strategies, and lack of autonomous optimization capabilities.

[0029] The multimodal spatiotemporal synchronous acquisition module is the core of the system's data input, used to complete the synchronous acquisition and spatiotemporal tag binding of tourists' multidimensional verification data. Specifically, it includes a synchronous triggering unit, a biometric feature acquisition unit, a document and ticket verification unit, and a spatial positioning unit.

[0030] In practical implementation, the synchronous triggering unit uses the infrared human body sensing signal at the entrance of the turnstile channel as the sole trigger source. When a visitor enters the sensing area of ​​the turnstile channel, the infrared human body sensing sensor is triggered, and the synchronous triggering unit immediately sends collection instructions to the biometric feature collection unit, the document and ticket verification unit, and the spatial positioning unit simultaneously via hard triggering. A preset time delay threshold is used in this embodiment. Setting it to 50ms ensures that the trigger time difference of all acquisition actions does not exceed 50ms, guaranteeing the time synchronization of multi-source data from the hardware level.

[0031] The biometric acquisition unit integrates a binocular face acquisition subunit and a near-infrared iris acquisition subunit. The binocular face acquisition subunit uses a binocular visible light camera with a resolution of 2 megapixels or higher and a frame rate of no less than 30fps to acquire visible light face image sequences of tourists. The near-infrared iris acquisition subunit uses 850nm near-infrared supplementary light combined with a high-definition iris camera to acquire near-infrared iris image sequences of tourists. The supplementary light intensity can be adaptively adjusted according to the ambient light to avoid overexposure in strong light or blurry images in weak light.

[0032] The document and ticket verification unit includes an ID card chip reading subunit and a ticket QR code reading subunit. The ID card chip reading subunit uses a second-generation ID card reader module that conforms to the standards of the Ministry of Public Security. It is used to read the encrypted identity information in the tourist's ID card in a contactless manner, including name, ID number, and facial photo template. The ticket QR code reading subunit uses a megapixel-level QR code scanning module that supports fast reading of screen codes and paper codes. It is used to read the core ticket data in the scenic area's electronic ticket QR code, such as ticket validity period, entry permissions, personnel binding information, and number of entries allowed.

[0033] The spatial positioning unit includes a depth camera and a millimeter-wave radar. The depth camera is a TOF depth camera, and the millimeter-wave radar is a 24GHz range-finding radar. The two work together to collect the real-time three-dimensional spatial coordinates of tourists within the turnstile channel, accurately calculating the relative distance between the tourist and the turnstile's data acquisition terminal. With lateral offset The relative distance and lateral offset are used as spatial anchors for the collected data, providing basic data for subsequent spatial domain alignment.

[0034] After all data collection is completed, the multimodal spatiotemporal synchronous acquisition module binds unified timestamps and spatial anchors to the facial biometrics, iris biometrics, ID card information, electronic ticketing information, and spatial coordinate information collected in this study, constructs a multimodal verification dataset with spatiotemporal labels, and transmits this dataset to the spatiotemporal alignment and feature normalization module.

[0035] The spatiotemporal alignment and feature normalization module is used to complete the spatiotemporal dual alignment and feature standardization of multimodal data, eliminate the acquisition delay and spatial offset of multi-source data, and provide standardized feature data for subsequent verification. Specifically, it includes a time domain alignment unit, a spatial domain alignment unit, and a feature normalization unit.

[0036] The time-domain alignment unit uses the instruction sending time of the synchronization trigger unit as the reference time to bind the corresponding collection timestamp to each set of biometric features, document information, ticketing information, and spatial coordinate data in the multimodal verification dataset. Subsequently, a time-domain filtering operation is performed to remove data whose difference between the collection timestamp and the reference time exceeds the delay threshold. Invalid data (50ms) is discarded, and only valid data that is synchronized with the time is retained to complete the time domain alignment and avoid data mismatch between visitors before and after.

[0037] Spatial domain alignment units are based on relative distances collected by spatial positioning units. With lateral offset First, perspective correction and ROI cropping are performed on the collected face and iris images to eliminate image distortion and feature shift caused by tourists standing too far / too close or shifting left and right, restoring face and iris images from a standard perspective. Then, spatial subject matching is performed on the ID information, ticket information and corrected biometric images. Through initial facial feature matching, it is ensured that the ID information, ticket information and biometric images correspond to the same tourist subject, completing spatial domain alignment and ensuring subject consistency of all verification data.

[0038] The feature normalization unit extracts facial feature vectors from the spatiotemporally aligned multimodal data. iris feature vector ID document identity feature vector The facial feature vector and iris feature vector are extracted through a pre-trained deep convolutional neural network, while the identity document feature vector is extracted from the facial photo template in the ID card. Then, all feature vectors are subjected to L2 normalization to uniformly scale the magnitude of each feature vector to 1, eliminating the influence of the difference in the dimensions of the feature vectors. Finally, a standardized multimodal feature matrix is ​​generated and transmitted to the dual-link dynamic verification module.

[0039] The dual-link dynamic verification module is the core computing unit of the system. Based on a standardized multimodal feature matrix, it executes two independent links in parallel: identity compliance verification and liveness verification. Finally, it generates a comprehensive verification score, which includes an identity compliance verification unit, a liveness verification unit, a dynamic weight configuration unit, and a comprehensive verification calculation unit.

[0040] The identity compliance verification unit is used to complete the identity consistency verification of "person-certificate-ticket". In specific implementation, the facial feature vector in the standardized multimodal feature matrix is ​​used. iris feature vector , respectively with the identity feature vector of the document The cosine similarity matching of the pre-stored registration feature vectors in the corresponding scenic spot ticketing system is used to calculate the facial feature similarity. Similarity to iris features The similarity value ranges from [0,1], with a higher value indicating a higher degree of feature matching.

[0041] Combined with the relative distance collected by the spatial positioning unit The identity compliance verification unit calculates the identity compliance coefficient using the following formula. : ; In the formula, Facial feature weights, These are the weights for iris features, with initial default values ​​of 0.6 and 0.4 respectively; For optimal data collection, this embodiment sets the distance to 80cm. The distance attenuation coefficient is set to 0.02 in this embodiment; This is the base of the natural logarithm. By correcting for the distance attenuation factor, the impact of tourists' sampling distance deviating from the optimal value on identity verification results can be eliminated, improving verification accuracy at different locations.

[0042] The dynamic weight configuration unit is connected to the light sensor signal at the scenic spot to obtain the ambient light intensity in real time. And according to the ambient light intensity Dynamic adjustment and The value of is determined by the ambient light intensity. Visible light facial imaging quality is better at readings above 5000 lux (strong light / backlight environments). The value has been adjusted upwards to 0.7-0.9. The corresponding value should be lowered to 0.3-0.1; when the ambient light intensity... At levels below 100 lux (low light / nighttime environments), near-infrared iris imaging exhibits stronger anti-interference capabilities. The value was lowered to 0.2-0.4. The corresponding weight was adjusted to 0.8-0.6 to achieve the optimal weight configuration under different lighting conditions and fully leverage the scene advantages of the two biological characteristics.

[0043] The liveness verification unit is used to perform liveness detection of real people and resist forgery attacks such as photos, videos, and 3D masks. In specific implementation, dynamic feature extraction is performed on N consecutively acquired face image sequences and iris image sequences (N is 10 in this embodiment) to obtain three major liveness features: the average amplitude of facial micro-movements. Pupil diameter change rate and the 3D depth variance of human face .

[0044] Among them, the average amplitude of facial micro-movements The facial key point displacement is calculated from multiple consecutive frames of facial images, reflecting the natural micro-movements of a living face; the three-dimensional depth variance of the face is also included. The 3D depth data of the face acquired by a binocular camera is used to distinguish between planar forgeries and real 3D faces; pupil diameter change rate. The pupil diameter is calculated from multiple consecutive frames of iris images, using the following formula: ; In the formula, Let represent the pupil diameter detected in the i-th frame of the iris image, and N be the total number of consecutively acquired iris image frames. The pupil diameter change rate reflects the natural scaling characteristics of a living pupil and is a core feature for resisting attacks from highly realistic masks.

[0045] Based on the three main characteristics of liveness mentioned above, the liveness verification unit calculates the liveness confidence coefficient using the following formula. : ; In the formula, , , The preset liveness feature weighting coefficients are initially set to 0.3, 0.4, and 0.3 in this embodiment, and satisfy the following conditions: ; The Sigmoid normalization function is used to map each liveness feature value to the [0,1] interval, resulting in the final liveness confidence coefficient. The value ranges from [0,1], and the higher the value, the higher the authenticity of the living organism.

[0046] The integrated verification calculation unit first obtains the ticket validity identifier from the document and ticket verification unit. The rules for determining the validity of tickets are as follows: when the ticket information passes verification (the ticket is valid, grants entry permission for the day, and is consistent with the identity information), When ticket information verification fails, .

[0047] Combining identity compliance coefficient With confidence coefficient of living organism The comprehensive verification calculation unit generates the comprehensive verification score using the following formula. : ; In the formula, , To comprehensively verify the weighting coefficients, the initial values ​​in this embodiment are 0.5 and 0.5 respectively, which satisfies... The final verification composite score The value range is [0,1], which is the core basis for subsequent risk classification decisions. When ticket verification fails, The overall verification score was 0, which triggered an interception operation.

[0048] The risk grading access decision module is used to classify risk levels based on the comprehensive verification score and generate corresponding control instructions. Specifically, it includes a threshold preset unit, a risk level classification unit, and an instruction generation unit.

[0049] In practical implementation, the threshold preset unit pre-sets three levels of risk thresholds, namely, safety thresholds. Warning threshold Interception threshold In this embodiment, they are respectively set as follows: , , And satisfy The threshold can be flexibly adjusted according to the safety management needs of the scenic area.

[0050] The risk level classification unit will use the comprehensive verification score output by the dual-link dynamic verification module. The risk level is then compared with the Level 3 risk threshold to determine the corresponding passage risk level. The specific classification rules are as follows: when When the verification is completed, it is classified as a safe level, which means that the verification has been completely passed and there is no security risk. when At this time, it is classified as the "attention level," which means that the verification has basically passed, but there is still some uncertainty, and it needs to be closely monitored. when When the time is right, it is classified as a risk level, which means that the verification failed and there is a high security risk.

[0051] The instruction generation unit generates corresponding control instructions based on the risk levels identified: For the security level, a passage instruction is generated and sent to the gate's adaptive execution module; For visitors under the "attention" category, a passage instruction with audio and visual prompts is generated, and a secondary verification prompt is sent to the scenic area's back-end management terminal to remind staff to conduct a manual review of the visitor. Based on the risk level, an interception command is generated, and a manual review warning is sent to the scenic area's back-end management terminal and on-site guard posts, triggering an audible and visual alarm.

[0052] Meanwhile, the risk-based passage decision module and the channel anti-tailgating monitoring module interact in real time. When the channel anti-tailgating monitoring module receives a tailgating warning signal, regardless of the current risk level, the current passage control command is immediately switched to an intercept command, and an audible and visual alarm is triggered.

[0053] The gate anti-tailgating monitoring module uses millimeter-wave radar and depth cameras to collect real-time 3D point cloud data and personnel movement trajectories within the gate channel. Based on point cloud segmentation and target tracking algorithms, it calculates the effective number of personnel within the channel in real time. Minimum distance between people and the speed of personnel passage .

[0054] In this embodiment, the preset spacing threshold is set to 50cm, and the preset speed threshold is set to 2m / s. When any of the following violations are detected: Number of valid personnel in the passage and the minimum distance between people ; Pedestrian traffic speed (Forced gate opening); The channel anti-tailgating monitoring module immediately sends a tailgating warning signal to the risk-level passage decision module, enabling accurate identification and interception of violations such as close-range tailgating, group passage, and forced gate rushing, thus making up for the technical deficiencies of traditional infrared beam anti-tailgating.

[0055] The gate adaptive execution module is used to receive access control commands and complete the gate's adaptive opening and closing and access status monitoring. Specifically, it includes a gate drive unit, an access status monitoring unit, and an adaptive parameter adjustment unit.

[0056] The gate drive unit uses a servo motor to drive the gate body, receives passage control commands sent by the risk classification passage decision module, and drives the gate body to perform corresponding opening and closing actions, including the opening action in the passage mode and the closing and holding action in the interception mode.

[0057] The passage status monitoring unit monitors the passage progress of tourists in real time through multiple sets of infrared beam sensors and depth cameras deployed in the passage, and determines whether tourists have completely passed through the gate passage. When it is detected that tourists have completely passed through, it immediately sends a closing command to the gate drive unit to control the gate to close. When abnormal situations such as people lingering in the passage or limbs obstructing the view are detected, the anti-pinch protection is triggered to keep the gate open and avoid injuring tourists.

[0058] The adaptive parameter adjustment unit is used to adjust parameters based on the passage risk level and the comprehensive verification score. The operating parameters of the turnstile are dynamically adjusted, with the core parameters including the duration of the turnstile gate's open position. Sound and light prompt modes, passage width, and duration of activation. The calculation formula is as follows: ; In the formula, The baseline activation duration is set to 2 seconds in this embodiment; To determine the maximum adjustable duration increment, different preset values ​​are set based on the risk level: when (Security level) The first preset value is 1 second. At this time, the activation duration is the shortest, enabling fast passage and improving traffic efficiency during peak hours; when (Attention level) The second preset value of 3 seconds is taken. Since the second preset value is greater than the first preset value, the gate opening time is extended. At the same time, sound and light prompts are used to remind tourists to cooperate with the secondary verification, so as to balance the efficiency of passage and risk control.

[0059] For high-risk levels, the adaptive parameter adjustment unit triggers a high-frequency audible and visual alarm mode, keeping the gate closed and blocking the passage. At the same time, it can adaptively adjust the gate width according to the needs of tourists, adapting to special passage scenarios such as wheelchairs and strollers.

[0060] The closed-loop feedback optimization module is used to realize the autonomous iterative optimization of the system, which solves the shortcomings of traditional gate parameters that are fixed and cannot adapt to changes in the scene. Specifically, it includes a data acquisition unit, a parameter iteration unit, and a model update unit.

[0061] The data acquisition unit collects daily verification results, passage status data, abnormal event data, and manual review results data on a calendar day basis. This includes verification pass rate, false positive rate, and false negative rate under different light intensities and collection distances, verification results corrected by manual review, and abnormal event data such as tailgating and gate rushing, to construct a system optimization dataset.

[0062] The parameter iteration unit, based on the optimized dataset, statistically analyzes the face verification accuracy under different ambient light intensities and acquisition distances. Accuracy of iris verification The facial feature weights in the identity compliance coefficient calculation formula are dynamically adjusted using the following iterative formula. Weighting of iris features : ; ; In the formula, , Let be the weight value after the t-th iteration. , , respectively, represent the accuracy rates of face verification and iris verification within the t-th iteration period. The learning rate is set to 0.01 in this embodiment to avoid excessive weight iterations that could lead to system instability.

[0063] Meanwhile, the parameter iteration unit dynamically adjusts the liveness feature weight coefficient in the liveness detection formula based on the false positive / false negative data in the optimization dataset. , , We will continue to improve the accuracy of liveness detection and reduce the false positive rate.

[0064] The model update unit performs incremental fine-tuning of the face feature extraction model and the iris feature extraction model based on the optimized dataset. Face and iris images collected on-site in scenic areas under different lighting, angles, and occlusion conditions are added to the model training set for incremental learning, continuously improving the feature extraction accuracy in complex scenic scenes and achieving continuous optimization of the system's verification capabilities.

[0065] The complete workflow of the intelligent ticketing gate control system for scenic spots based on multimodal verification in this embodiment is as follows: When a tourist enters the gate, the infrared human body sensing signal triggers the multimodal spatiotemporal synchronous acquisition module, which simultaneously collects the tourist's face, iris, ID card, ticket, and spatial coordinate data, and binds spatiotemporal tags to construct a multimodal verification dataset; The spatiotemporal alignment and feature normalization module performs dual alignment of the dataset in both the temporal and spatial domains, completes feature normalization processing, and generates a standardized multimodal feature matrix. The dual-link dynamic verification module performs identity compliance verification and liveness verification in parallel, and calculates and generates a comprehensive verification score by combining the ticket validity identifier. The risk-based access decision module classifies risk levels based on the comprehensive verification score and generates corresponding access control instructions. At the same time, the channel anti-tailgating monitoring module monitors violations in the channel in real time and triggers emergency interception. The gate adaptive execution module receives control commands, adaptively adjusts the gate operating parameters, executes opening and closing actions, and monitors the passage status in real time. The closed-loop feedback optimization module continuously collects system operation data, completes weight parameter iteration and model incremental update, and realizes the system's self-optimization iteration.

[0066] It should be understood that any parts not described in detail in this specification belong to the prior art. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention. The thresholds, initial weight values, parameter settings, etc. mentioned in this embodiment can be flexibly adjusted according to the actual scene and management needs of the scenic area, and all fall within the protection scope of the present invention.

Claims

1. A smart ticketing gate control system for scenic areas based on multimodal verification, characterized in that, include: It includes a multimodal spatiotemporal synchronous acquisition module, a spatiotemporal alignment and feature normalization module, a dual-link dynamic verification module, a risk-based access decision module, a gate adaptive execution module, and a closed-loop feedback optimization module; The multimodal spatiotemporal synchronous acquisition module is used to synchronously trigger the face image acquisition unit, iris acquisition unit, ID card reading unit, electronic ticket verification unit and spatial positioning unit to collect tourists' facial biometrics, iris biometrics, ID card information, electronic ticket voucher information and tourist spatial coordinate information in the same spatiotemporal space. It binds a unified timestamp and spatial anchor point to each set of collected data to construct a multimodal verification dataset with spatiotemporal labels. The spatiotemporal alignment and feature normalization module is used to perform dual alignment of the multimodal verification dataset in the temporal and spatial domains based on a unified timestamp and spatial anchor point, eliminating the acquisition delay and spatial offset of multi-source data. At the same time, the aligned multimodal features are subjected to dimensionless normalization processing to generate a standardized multimodal feature matrix. The dual-link dynamic verification module is used to execute two links—identity compliance verification and liveness verification—based on a standardized multimodal feature matrix, and calculates and generates an identity compliance coefficient. With confidence coefficient of living organism and combined with ticket validity identification Generate a comprehensive verification score ; The risk-based access decision module is used to pre-set multi-level risk thresholds and integrate the verification scores. The risk level of tourist passage is classified by comparing it with multi-level risk thresholds, and corresponding passage control instructions, early warning instructions or secondary verification instructions are generated based on the risk level. The gate adaptive execution module is used to receive passage control commands, adaptively adjust the gate opening duration, passage channel width, and sound and light prompt mode according to the risk level, and execute the corresponding gate opening and closing actions, while collecting passage status data in the channel in real time. The closed-loop feedback optimization module is used to feed back the passage status data, verification result data and abnormal event data to the dual-link dynamic verification module, dynamically adjust the weight parameters of identity compliance verification and liveness verification, and realize the system's self-optimization iteration.

2. The intelligent ticketing gate control system for scenic areas based on multimodal verification according to claim 1, characterized in that, The multimodal spatiotemporal synchronous acquisition module includes a synchronous triggering unit, a biometric feature acquisition unit, a document and ticket verification unit, and a spatial positioning unit; The synchronous triggering unit is used to send collection commands to the biometric feature collection unit, the document and ticket verification unit, and the spatial positioning unit simultaneously, using the infrared human body sensing signal of the gate channel as the trigger source, to ensure that the triggering time difference of all collection actions does not exceed a preset delay threshold. ; The biometric acquisition unit includes a binocular face acquisition subunit and a near-infrared iris acquisition subunit, which are used to acquire visible light face image sequences and near-infrared iris image sequences of tourists, respectively. The document and ticket verification unit includes an ID card chip reading subunit and a ticket QR code reading subunit, which are used to read the encrypted identity information in the tourist's resident ID card and the ticket validity period, entry permission and personnel binding information in the scenic area's electronic ticket QR code, respectively. The spatial positioning unit includes a depth camera and a millimeter-wave radar, used to collect the real-time three-dimensional spatial coordinates of tourists within the turnstile channel and determine the relative distance between the tourists and the turnstile's data collection terminal. With lateral offset , as a spatial anchor point.

3. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 2, characterized in that, The spatiotemporal alignment and feature normalization module includes a temporal domain alignment unit, a spatial domain alignment unit, and a feature normalization unit. The time domain alignment unit is used to bind a corresponding collection timestamp to each set of collected biometric features, document information, ticketing information, and spatial coordinate data, using the instruction sending time of the synchronization trigger unit as the base time, and to remove data whose difference between the timestamp and the base time exceeds a delay threshold. Invalid data is used to complete time domain alignment; The spatial alignment unit is used to align relative distances collected by the spatial positioning unit. With lateral offset Perspective correction and cropping are performed on facial and iris images to eliminate image distortion caused by tourist positioning deviations. At the same time, spatial subject matching is performed on the ID information, ticket information and corrected biometric images to ensure that all data correspond to the same tourist subject and complete spatial domain alignment. The feature normalization unit is used to extract facial feature vectors from the aligned multimodal data. iris feature vector ID document identity feature vector The L2 normalization process is applied to all feature vectors to unify the magnitude of the feature vectors to 1, thereby generating a standardized multimodal feature matrix.

4. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 3, characterized in that, The dual-link dynamic verification module includes an identity compliance verification unit, a liveness verification unit, and a dynamic weight configuration unit. The identity compliance verification unit is used to process the facial feature vector in the standardized multimodal feature matrix. iris feature vector Respectively compared with the identity feature vector of the document The similarity of the corresponding pre-stored registered feature vectors is calculated by performing similarity matching. Similarity to iris features and combined with relative distance Generate identity compliance coefficient The identity compliance coefficient Obtain it using the following formula: ； In the formula, Facial feature weights, For iris feature weights, To achieve the optimal sampling distance, This is the distance attenuation coefficient. is the base of the natural logarithm; The dynamic weight configuration unit is used to adjust the ambient light intensity during data acquisition. Dynamic adjustment and The values ​​of , including ambient light intensity The higher, The larger the value, The smaller the value, the greater the value, and vice versa.

5. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 4, characterized in that, The liveness verification unit is used to perform dynamic feature extraction on a series of continuously acquired face image sequences and iris image sequences, respectively, to obtain the average amplitude of facial micro-movements. Pupil diameter change rate and the 3D depth variance of human face Based on the above characteristics, a liveness confidence coefficient is calculated. The live confidence coefficient Obtain it using the following formula: ； In the formula, The preset liveness feature weight coefficients are given, and satisfy the following conditions: ; This is the Sigmoid normalization function, used to map eigenvalues ​​to the [0,1] interval; Among them, the pupil diameter change rate Obtain it using the following formula: ； In the formula, For the first The pupil diameter detected in the frame of the iris image. This represents the total number of iris image frames acquired continuously.

6. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 5, characterized in that, The dual-link dynamic verification module also includes a comprehensive verification calculation unit, which is used to obtain the ticket validity identifier. And combined with identity compliance coefficient With confidence coefficient of living organism Calculate and generate a comprehensive verification score. The verification comprehensive score Obtain it using the following formula: ； In the formula, To comprehensively verify the weighting coefficients and meet the requirements The ticket validity identifier The value is: when the ticket information verification is successful, When ticket information verification fails, .

7. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 6, characterized in that, The risk-level access decision module includes a threshold preset unit, a risk level classification unit, and an instruction generation unit. The threshold preset unit is used to preset three levels of risk thresholds, namely, safety thresholds. Warning threshold Interception threshold And satisfy ; The risk level classification unit is used to calculate the overall verification score. The corresponding passage risk level is determined by comparing the data with the Level 3 risk threshold, as detailed below: when At that time, it is classified into security levels; when At that time, it is classified as a level of attention; when At that time, it is classified into risk levels; The instruction generation unit is used to generate corresponding instructions according to the risk level: a safe level generates a pass instruction, a concern level generates a pass instruction with audio and visual prompts and a secondary verification prompt, and a risk level generates an interception instruction and a manual review warning.

8. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 7, characterized in that, The gate adaptive execution module includes a gate drive unit, a passage status monitoring unit, and an adaptive parameter adjustment unit; The gate drive unit is used to receive access control commands and drive the gate body to perform opening and closing actions. The passage status monitoring unit is used to monitor the passage progress of tourists in real time through infrared photoelectric sensors and depth cameras in the passage, and to determine whether tourists have completely passed through the gate passage. The adaptive parameter adjustment unit is used to adjust the access risk level and the comprehensive verification score. Dynamically adjust the opening and holding time of the gate. The duration of the open state Obtain it using the following formula: ； In the formula, Based on the base opening duration, This is the maximum adjustable duration increment; when hour, Take the first preset value; when hour, Take the second preset value, and the second preset value is greater than the first preset value.

9. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 8, characterized in that, The scenic area intelligent ticketing gate control system based on multimodal verification also includes a channel anti-tailgating monitoring module, which is signal-connected to the risk-level access decision module; The channel anti-tailgating monitoring module is used to collect real-time 3D point cloud data and personnel movement trajectories within the turnstile channel using millimeter-wave radar and a depth camera, and to calculate the effective number of personnel within the channel. Minimum distance between people and the speed of personnel passage ; When detected and Preset spacing threshold, or When the preset speed threshold is reached, the channel anti-tailgating monitoring module immediately sends a tailgating warning signal to the risk-level passage decision module. After receiving the signal, the risk-level passage decision module forcibly switches the current passage control command to an interception command and triggers an audible and visual alarm.

10. A scenic area intelligent ticketing gate control system based on multimodal verification according to claim 9, characterized in that, The closed-loop feedback optimization module includes a data acquisition unit, a parameter iteration unit, and a model update unit. The data acquisition unit is used to collect daily verification result data, passage status data, abnormal event data, and manual review result data to construct an optimized dataset; The parameter iteration unit is used to calculate the verification pass rate and false judgment rate under different ambient light intensities and different collection distances based on the optimized dataset, and dynamically adjust the facial feature weights in the identity compliance coefficient calculation formula. Iris feature weights And the weighting coefficient of living features in the formula for calculating the confidence coefficient of living individuals. The iterative formula is as follows: ； ； In the formula, For the first The weight values ​​after the next iteration The first The accuracy rate of face verification and iris verification within each iteration cycle. The learning rate; The model update unit is used to incrementally fine-tune the face feature extraction model and the iris feature extraction model based on the optimized dataset, thereby improving the feature extraction accuracy in complex scenarios.

Images