An intelligent access control management system

Abstract

The present invention relates to the technical field of intelligent access control management, and particularly to an intelligent access control management system, comprising: a data acquisition module that real-time acquires the out-of-store activity video, in-store activity video, and current opening person image of an unmanned store; a coordination judgment module that judges whether the opening action of a customer is coordinated according to the out-of-store activity video; a type judgment module that judges the customer type according to the out-of-store activity video, historical opening person images, and the current opening person image; a behavior judgment module that determines the customer consumption behavior of each customer according to the real-time in-store activity video, and constructs a customer behavior label in combination with the customer type judgment result, coordination judgment result, and historical in-store activity video; and an access control module that determines an access control switch strategy according to the customer type and the customer behavior label and updates the customer type. The present invention can improve the accuracy of abnormal behavior recognition and enhance the security of access control management.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent access control management, and particularly to an intelligent access control management system. Background Art

[0002] In order to address problems such as low security and poor management efficiency of traditional access control systems, intelligent access control management systems use diverse identity recognition technologies such as card swiping, fingerprint, and face recognition to verify personnel permissions and prevent illegal intrusion. At the same time, through automated management and remote control functions, manual intervention is reduced and access efficiency is improved. In addition, the system can completely record access information, facilitating the tracing of security incidents and management decision-making, and comprehensively ensuring regional security.

[0003] A prior art discloses an artificial intelligence access control management system (CN119477221B), which includes: supervising and processing and analyzing the single-opening operation data of an unmanned access control each time it runs, obtaining the single-opening operation state and digital data of the unmanned access control opening operation, implementing abnormal impact tracking and evaluation in different aspects on the single-opening operation abnormal state obtained from the supervision and processing of the single-opening operation of the unmanned access control, and adaptively implementing targeted experience optimization management and risk prevention management on the unmanned access control according to the evaluation results, and adaptively implementing management measures in different aspects.

[0004] Although the above technical solution can achieve autonomous security supervision of unmanned access control, there are still the following problems: the above access control management system relies on the single-opening operation data of the access control and lacks the correlation analysis with customer behavior or historical data. Summary of the Invention

[0005] Therefore, the present invention provides an intelligent access control management system to overcome the problems in the prior art that the access control management system relies on the single-opening operation data of the access control and lacks the correlation analysis with customer behavior or historical data.

[0006] To achieve the above object, the present invention provides an intelligent access control management system, including: An intelligent access control management system includes:

[0007] A data acquisition module for real-time acquisition of the out-of-store activity video, in-store activity video, and current opening portrait of an unmanned store;

[0008] A coordination judgment module connected to the data acquisition module for judging whether the customer's opening action is coordinated according to the out-of-store activity video;

[0009] A type judgment module respectively connected to the data acquisition module and the coordination judgment module for judging the customer type according to the out-of-store activity video, historical opening portrait, and current opening portrait;

[0010] Among them, the customer types include low-risk customers, medium-risk customers, and high-risk customers;

[0011] The behavior judgment module is respectively connected to the data collection module, the type judgment module, and the coordination judgment module, and is used to analyze the customer consumption behaviors of each customer through artificial intelligence based on the in-store activity video, and construct customer behavior tags in combination with the customer type judgment result, the coordination judgment result, and the historical in-store activity video;

[0012] Among them, the customer consumption behaviors include shopping duration, shopping product type, shopping product quantity, single consumption record, real-time activity path, and real-time activity behavior;

[0013] The access control module is respectively connected to the type judgment module and the behavior judgment module, and is used to determine the access control switch strategy according to the customer type and the customer behavior tag and update the customer type.

[0014] As an optimal technical solution of the intelligent access control management system, the type judgment module respectively determines the in-store face data set and the face data set passing by the store according to the historical opening portraits and the historical out-of-store activity videos, and determines the customer type of the corresponding customer according to the current opening portrait, specifically including:

[0015] Determine whether the customer type is a low-risk customer according to the comparison result between the current opening portrait and the in-store face data set, and determine the customer type according to the judgment result of non-low-risk customers in combination with the face data set passing by the store;

[0016] Among them, the face data set passing by the store includes a regular face data set passing by the store and an irregular face data set passing by the store.

[0017] As an optimal solution of the intelligent access control management system, the type judgment module determines the face data set passing by the store according to the judgment result of non-low-risk customers in combination with the out-of-store activity videos within the historical period, and determines the trajectory representation state of the corresponding customer according to the out-of-store activity time period and the out-of-store activity date of a single face data passing by the store to determine the face data attribute of the corresponding customer passing by the store, and determines the customer type according to the face data attribute.

[0018] As an optimal technical solution of the intelligent access control management system, the coordination judgment module extracts the key frames of the opening action of a single opening customer from the out-of-store activity video, determines the arm movement amplitude and the body rotation amplitude of the customer according to the key frames of the opening action to determine the movement amplitude ratio to judge whether the opening action of the customer is coordinated.

[0019] As an optimal technical solution of the intelligent access control management system, the behavior judgment module constructs customer behavior tags based on the historical in-store activity videos and customer consumption behaviors according to the low-risk customer determination results, specifically including:

[0020] Determine the historical shopping duration based on the historical in-store activity video of a single low-risk customer, and determine the customer behavior tag according to the shopping duration of this customer and the historical shopping duration;

[0021] Among them, the customer behavior tags include normal customers, violating customers, and customers with abnormal stays.

[0022] As an optimal technical solution of the intelligent access control management system, the behavior judgment module constructs customer behavior tags based on the medium-risk customer determination results according to the customer consumption behaviors, specifically including:

[0023] Determine the regular shopping duration based on the historical shopping durations of all low-risk customers when they first enter the store, and determine the customer behavior tag of the medium-risk customer based on the current shopping duration of the medium-risk customer and the regular shopping duration.

[0024] As an optimal technical solution of the intelligent access control management system, the behavior judgment module constructs customer behavior tags based on the high-risk customer determination results according to the coordination judgment results and customer consumption behaviors, specifically including:

[0025] Preliminarily judge whether the customer behavior is abnormal according to the coordination judgment result, determine the customer behavior tag according to the customer consumption behavior based on the normal customer behavior determination result, or lock the access control and trigger an alarm based on the abnormal customer behavior determination result.

[0026] As an optimal technical solution of the intelligent access control management system, the behavior judgment module constructs customer behavior tags based on the normal customer behavior determination results according to the customer consumption behaviors, specifically including:

[0027] Determine the payment status of a high-risk customer according to the type of purchased goods, the quantity of purchased goods, and the single consumption record of the customer, determine the stay status according to the real-time activity path of this customer and the preset sensitive area, determine the activity behavior type of the customer according to the real-time activity behavior, and construct the customer behavior tag according to the payment status, the stay status, and the activity behavior type.

[0028] As an optimal technical solution of the intelligent access control management system, the behavior judgment module determines the stay status by determining the number of times and the stay time that the customer passes through any preset sensitive area according to the real-time activity path;

[0029] Among them, the stay status includes continuous stay status and temporary stay status.

[0030] As an optimal technical solution of the intelligent access control management system, the access control module determines the access control switch strategy according to the customer behavior label, including:

[0031] If the customer behavior label is a normal customer, incorporate the in-store activity video of this time into the historical in-store activity videos and do not update the customer type;

[0032] If the customer behavior label is a violating customer, lock the access control and trigger an alarm and update the customer type;

[0033] If the customer behavior label is an abnormal stay customer, determine the access control switch strategy according to the stay time and the preset observation period and update the customer type.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows. In the above embodiments, the intelligent access control management system of the present invention realizes the refined control of the security and operation of the unmanned store through the coordinated operation of multiple modules. The data acquisition module captures the in-store and out-of-store videos and the opening portrait in real time, providing detailed data for subsequent analysis; the type judgment module combines the opening portrait and the out-of-store activity video to accurately classify the customer risk level and early warn of potential threats; the coordination judgment module screens abnormal behaviors from the opening link to prevent problems in advance; the behavior judgment module deeply analyzes the customer consumption behavior to generate behavior labels and comprehensively master the customer dynamics; the access control module dynamically adjusts the access control strategy according to the customer type and behavior label to achieve differentiated management. The whole process monitoring not only improves the accuracy and timeliness of abnormal behavior recognition, but also effectively enhances the security and intelligent level of the unmanned store access control management, ensures the operation order and property safety of the store, provides data support for the subsequent operation analysis of the unmanned store, and helps to optimize the store management strategy.

[0035] In particular, the type judgment module of the present invention judges the behavior trajectory law of pedestrians passing by the unmanned store according to the out-of-store activity time period and out-of-store activity date of any pedestrian in the out-of-store activity video within the historical period, and marks the customer type according to the behavior trajectory law, which helps to optimize the access control management strategy and enhance the security and intelligent level of the unmanned store access control management.

[0036] In particular, the present invention comprehensively judges the customer risk level through the recognition of the customer face features and the analysis of their behavior trajectories, and at the same time allocates a differentiated behavior label judgment strategy based on the customer risk level, saving computing resources while facilitating the further differentiated access control management and monitoring of different types of customers, realizing the unity of operation efficiency and security management, and providing reliable support for the intelligent and refined management of the access control system of the unmanned store.

[0037] In particular, through the cross-verification and comprehensive consideration of multi-dimensional data, the present invention accurately classifies customers into normal customers, customers who have underpaid, or customers who violate regulations, thereby providing a reliable basis for access control, security warning, and subsequent management decisions, effectively ensuring the property safety and normal operation order of unmanned stores.

[0038] In particular, the present invention can accurately identify the action characteristics of customers in unmanned stores. The determination of the staying state can monitor and warn against the long-term staying or abnormal entry and exit behaviors of customers in sensitive areas (such as the cashier area and the area of valuable goods), promptly detect potential violations such as theft and damage, and contribute to the subsequent implementation of differential access control management for customers with different risk levels, making the access control management more scientific and reasonable and improving the intelligent level of unmanned stores. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a structural block diagram of the intelligent access control management system according to an embodiment of the present invention;

[0040] Figure 2 is a determination diagram for determining the customer type according to an embodiment of the present invention;

[0041] Figure 3 is a step diagram for determining whether the opening action of a customer is coordinated according to an embodiment of the present invention;

[0042] Figure 4 is a step diagram for determining the customer behavior label of high-risk customers according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to make the objectives and advantages of the present invention clearer, the present invention will be further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and do not limit the protection scope of the present invention.

[0045] It should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, they can be fixedly connected, detachably connected, or integrally connected; they can be mechanically connected or electrically connected; they can be directly connected or indirectly connected through an intermediate medium, and they can be the communication inside two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0046] Please refer to Figure 1As shown, it is a structural block diagram of the intelligent access control management system according to an embodiment of the present invention; specifically, the present invention provides an intelligent access control management system, including:

[0047] A data acquisition module, which is used to collect the out-of-store activity video, in-store activity video and current opening portrait of the unmanned store in real time;

[0048] A coordination judgment module, which is connected to the data acquisition module and is used to judge whether the opening action of the customer is coordinated according to the out-of-store activity video;

[0049] A type judgment module, which is respectively connected to the data acquisition module and the coordination judgment module, and is used to judge the customer type according to the out-of-store activity video, historical opening portrait and current opening portrait;

[0050] Among them, the customer types include low-risk customers, medium-risk customers and high-risk customers;

[0051] A behavior judgment module, which is respectively connected to the data acquisition module, the type judgment module and the coordination judgment module, and is used to analyze the customer consumption behaviors of each customer through artificial intelligence based on the in-store activity video, and construct a customer behavior label of the current in-store customers in combination with the customer type judgment result, the coordination judgment result and the historical in-store activity video;

[0052] Among them, the customer consumption behaviors include shopping duration, shopping commodity type, shopping commodity quantity, single consumption record, real-time activity path and real-time activity behavior;

[0053] An access control module, which is respectively connected to the type judgment module and the behavior judgment module, and is used to determine the access control switch strategy according to the customer type and the customer behavior label and update the customer type.

[0054] Specifically, due to the lack of on-site supervision by store clerks and reliance on self-service payment transactions, there are many potential safety hazards and management blind spots in unmanned stores. On the one hand, unattended operation makes it difficult to detect risks such as violent intrusion and theft in real time. Judging whether the customer's opening behavior is coordinated and whether it is a violent behavior can identify abnormalities in time before the personnel enter the store and prevent security threats; on the other hand, by monitoring the in-store activity video images and payment links of customers in real time to determine the customer behavior label and judge whether there are abnormal behaviors of customers, it can effectively prevent phenomena such as malicious damage, theft, wrong payment and missed payment, so as to construct the decision-making of the access control system, comprehensively protect the interests of the store, and ensure the safe and orderly operation of the unmanned store.

[0055] In implementation, the person opening the door image and the video of the activities outside the store can be recorded by the same camera. Through image recognition and analysis technology processing of the video of the activities outside the store, the person's door-opening behavior can be accurately determined, and during the door-opening action, image acquisition technology is used to collect the person opening the door image in real time.

[0056] In the above embodiments, the intelligent access control management system of the present invention realizes refined control over the security and operation of unmanned stores through the coordinated operation of multiple modules. The data acquisition module captures the in-store and out-of-store videos and the person opening the door image in real time, providing detailed data for subsequent analysis; the type judgment module combines the person opening the door image and the video of the activities outside the store to accurately classify the customer risk level and early warning of potential threats; the coordination judgment module screens abnormal behaviors from the door-opening link to prevent problems in advance; the behavior judgment module deeply analyzes the customer consumption behavior to generate behavior tags and comprehensively grasps the customer dynamics; the access control module dynamically adjusts the access control strategy according to the customer type and behavior tags to achieve differential management. The full-process monitoring not only improves the accuracy and timeliness of abnormal behavior recognition, but also effectively enhances the security and intelligent level of the unmanned store access control management, ensures the operation order and property safety of the store, provides data support for subsequent unmanned store operation analysis, and helps to optimize the store management strategy.

[0057] Please refer to Figure 2 as shown Figure 2 which is the determination diagram for determining the customer type in the embodiment of the present invention; specifically, the type judgment module determines the in-store face data set and the face data set passing through the store according to the historical person opening the door image and the historical video of the activities outside the store, and determines the customer type of the corresponding customer according to the current person opening the door image, specifically including:

[0058] Determine whether the customer type is a low-risk customer according to the comparison result between the current person opening the door image and the in-store face data set, and determine the customer type according to the determination result of the non-low-risk customer combined with the face data set passing through the store;

[0059] Among them, the face data set passing through the store includes the regular face data set passing through the store and the irregular face data set passing through the store.

[0060] Specifically, based on the above embodiments, by comparing the historical door-opening portrait with the current door-opening portrait, it is determined whether the customer has ever had a normal shopping behavior record in the store. The system believes that such customers have a relatively high credibility and determines them as low-risk customers. When the comparison is unsuccessful, it is further compared with the store-face dataset. If the comparison is successful, it proves that although the customer has no historical record of consuming in the store, they have appeared near the store. Then, the appearance time period and appearance date of the customer are determined through the historical out-of-store activity videos to judge whether their behavior trajectory is regular. Customers with regular behavior trajectories may be because they work or live near the unmanned store and have not yet had actual shopping behavior in the unmanned store. Whether they will consume in the store in the future and the consumption possibility are uncertain. Therefore, they are determined as medium-risk customers. If the behavior trajectory is irregular or the comparison is unsuccessful, it indicates that the customer's behavior is uncertain, difficult to predict, and has a certain risk. Then, such customers are determined as high-risk customers and need to be focused on and monitored.

[0061] In implementation, the in-store face dataset is constructed based on the historical door-opening portraits, and the store-face dataset is determined according to the historical out-of-store activity videos within the historical period. The store-face dataset is formed by collecting, storing, and organizing the face features in the historical out-of-store activity videos. The value range of the historical period is 20 days to 40 days. Preferably, the value of the historical period is 30 days.

[0062] In implementation, if the comparison between the current door-opening portrait and the in-store face dataset is successful, the customer type is determined as a low-risk customer. If the customer type is not a low-risk customer, the current door-opening portrait is compared with the store-face dataset. If the comparison with the regular store-face dataset is successful, the customer type is determined as a medium-risk customer; if the comparison between the current door-opening portrait and the store-face dataset is unsuccessful, or the comparison with the irregular store-face dataset is successful, the customer type is determined as a high-risk customer.

[0063] Specifically, the type judgment module determines the store-face dataset according to the judgment result of non-low-risk customers in combination with the out-of-store activity videos within the historical period, and determines the trajectory characterization status of the corresponding customer based on the out-of-store activity time period and out-of-store activity date of a single store-face data to determine the face data attribute of the corresponding store customer, and determines the customer type according to the face data attribute.

[0064] Specifically, on the basis of the above embodiments, there may be a person who regularly passes by an unmanned store at fixed elapsed time periods and fixed elapsed dates. Such a person with a relatively regular behavior trajectory can be predicted as a person who stably lives / works in the vicinity. However, since there is no entry behavior, that is, the purchase behavior is unknown, there is still a certain risk. Therefore, such a person is determined as a medium-risk customer. In addition, if the behavior trajectory is irregular, since the system lacks the behavior data of the pedestrian and the pedestrian's behavior cannot be predicted, the person can be determined as a high-risk customer.

[0065] In implementation, the type judgment module can respectively determine the fixed passing time period and the fixed passing date of a single person passing by the store according to the out-of-store activity time period and the out-of-store activity date of the single person passing by the store. By using the algorithms of set operation and interval merging, the out-of-store activity time periods of a single pedestrian within the historical period are integrated, and the overlapping out-of-store activity time periods are merged into a time period interval, thereby obtaining one or more time period intervals, that is, the fixed passing time period. And the out-of-store activity dates of a single pedestrian within the historical period are integrated to form a range of fixed passing dates. In the historical period, according to the number of times the out-of-store activity time period of a single person passing by the store deviates from the fixed passing time period within the fixed passing date, its trajectory characterization state is determined (it can be understood that if the overlapping duration between the activity time period and the fixed passing time period is less than one-fourth to one-fifth of the duration of the fixed passing time period, it is determined that the out-of-store activity time period deviates from the fixed passing time period). If the number of deviations is greater than the preset number of times, it is determined that its trajectory characterization state is an irregular trajectory state and the face data attribute corresponding to the face data of the single person is the out-of-store irregular face data set. If the number of deviations is less than or equal to the preset number of times, the trajectory characterization state is a regular trajectory state and the face data attribute corresponding to the face data of the single person is the out-of-store regular face data set. The value range of the preset number of times is 3 to 8 times. Preferably, the value range of the preset number of times is 6 times.

[0066] It can be understood that when the matching between the in-store customer and the in-store face data set is incorrect, the analysis of the historical out-of-store activity video will be triggered to match the historical out-of-store activity time period and the historical out-of-store activity date of the customer to confirm its trajectory characterization state; for non-in-store customers, the system does not actively analyze the out-of-store video to prevent the leakage of sensitive information such as personal action patterns and living rules due to excessive data collection, and reduce the management loopholes of the access control system caused by too many data processing links and complex logic.

[0067] On the basis of the above effects, the invention type judgment module judges the behavior trajectory law of newly entered customers outside the unmanned store by the out-of-store activity time period and out-of-store activity date of any pedestrian in the out-of-store activity video within the historical period, and marks the customer type according to the behavior trajectory law, which helps to optimize the access control management strategy and enhance the security and intelligence level of the unmanned store access control management.

[0068] Please refer to Figure 3 as shown in Figure 3 It is a step diagram for the embodiment of the present invention to determine whether the customer's door-opening action is coordinated; specifically, the coordination judgment module extracts the key frames of the door-opening action of a single door-opening customer according to the out-of-store activity video, and determines the amplitude of the customer's arm movement and the amplitude of body rotation according to the key frames of the door-opening action to determine the action amplitude ratio to judge whether the customer's door-opening action is coordinated.

[0069] Specifically, on the basis of the above embodiment, when opening the door normally, there is a certain coordination between the movement amplitudes of various parts of the body. For example, there is a certain ratio between the amplitude of arm movement and the amplitude of body rotation. If this proportional relationship in the key frame is seriously out of balance, such as the arm waving greatly while the body hardly rotates, it means that the action is uncoordinated and there is a violent abnormal behavior situation.

[0070] In implementation, a target detection algorithm can be used to process the video of the out-of-store activity, identify the human targets in the video frames, and use a multi-target tracking algorithm to track the human targets. The human skeleton key points are detected using a deep learning-based action recognition model. By analyzing the movement trajectory and changes of the skeleton key points, the customer's door opening action is identified. Once the door opening action is detected, the video frame is marked. In the process of door opening action, video frames are extracted according to preset time intervals, which are the key frames of the door opening action. The value range of the preset time interval is 0.05s to 0.15s. Preferably, the value of the preset time interval is 0.1s. In each key frame of the door opening action, select the arm key points related to the arm, including the shoulder, elbow and wrist key points, determine the pixel coordinates of these key points in each key frame of the door opening action, calculate the displacement of the arm key points between adjacent frames, take the shoulder as the fixed point, calculate the modulus length of the displacement vector of the elbow and wrist key points relative to the shoulder, normalize the corresponding modulus lengths between adjacent frames, and accumulate the normalized modulus lengths between all adjacent key frames of the door opening action to obtain the arm movement amplitude. Select the bone key points related to the body rotation, such as the key points of the neck, waist and hip. In each frame, calculate the orientation angle of the customer's body according to the position of these bone key points, determine the body orientation by calculating the vector angle between the key points, and determine the angle change of the body orientation angle between the key frames of the door opening action, normalize the corresponding angle change between adjacent frames, and accumulate the normalized angle change between all adjacent key frames of the door opening action to obtain the body rotation amplitude. The action amplitude ratio = arm movement amplitude / body rotation amplitude.

[0071] Specifically, whether the customer's door opening action is coordinated is judged based on the motion amplitude ratio and the preset amplitude ratio range, wherein if the motion amplitude ratio meets the preset amplitude ratio range, the door opening action is determined to be coordinated; if the motion amplitude ratio does not meet the preset amplitude ratio range, the door opening action is determined to be uncoordinated.

[0072] In implementation, under normal circumstances, the value range of the preset amplitude ratio range is 1.0 to 3.0. Preferably, the value range of the preset amplitude ratio range is 1.5 to 2.5.

[0073] Specifically, the behavior judgment module constructs a customer behavior tag based on the low-risk customer judgment result according to the historical in-store activity video and customer consumption behavior, specifically including:

[0074] Determine the historical shopping duration based on the historical in-store activity video of a single low-risk customer, and determine the customer behavior label based on the shopping duration of the customer and the historical shopping duration;

[0075] The customer behavior labels include normal customers, illegal customers and abnormally staying customers.

[0076] Specifically, based on the above embodiments, low-risk customers have shown good past behavior and have a relatively high level of credibility. Using the historical shopping duration as a reference to judge customer behavior can save system resources. Under normal circumstances, including in-store activity videos in the historical data for low-risk customers can optimize the judgment benchmark and make the system's judgment of low-risk customers more in line with the actual situation.

[0077] In implementation, if the shopping duration of a low-risk customer is greater than the historical shopping duration, the customer behavior label is an abnormal stay customer. If the shopping duration of a low-risk customer is less than or equal to 2 to 3 times the historical shopping duration, the customer behavior label is a normal customer. The historical shopping duration can be the average shopping duration of each time a customer enters the store within a historical period, and the current in-store activity video of normal customers can be incorporated into the historical in-store activity videos.

[0078] Specifically, the behavior judgment module constructs customer behavior labels based on the determination results of medium-risk customers according to customer consumption behavior, specifically including:

[0079] Determine the regular shopping duration based on the historical shopping durations of all low-risk customers' first store visits, and determine the customer behavior labels of medium-risk customers based on the current shopping duration of medium-risk customers and the regular shopping duration.

[0080] Specifically, based on the above embodiments, medium-risk customers are those who do not enter the store but always pass by outside the unmanned store and have relatively regular behavior trajectories. There is a certain level of security. Therefore, determine the customer behavior labels based on the in-store shopping durations of medium-risk customers and the first shopping durations of existing low-risk customers in the system.

[0081] In implementation, if the shopping duration is greater than the regular shopping duration, the customer behavior label is an abnormal stay customer. If the shopping duration is less than or equal to the regular shopping duration, the customer behavior label is a normal customer. The regular shopping duration is 1 to 2 times the average duration of the historical shopping durations of low-risk customers' first store visits. Since low-risk customers have had normal in-store shopping behavior, the average duration of their historical shopping durations of the first store visit can reflect the time range required for most customers to become familiar with the store environment and complete the regular shopping process. Setting the 1 to 2 times interval can accommodate individual differences among different customers, such as not being familiar with the store layout, spending more time selecting products, comparing prices, etc.

[0082] Specifically, the behavior judgment module constructs customer behavior labels based on the determination results of high-risk customers according to the coordinated judgment results and customer consumption behavior, specifically including:

[0083] Based on the coordinated judgment result, preliminarily judge whether the customer behavior is abnormal. Based on the judgment result of normal customer behavior, determine the customer behavior label according to the customer consumption behavior, or based on the judgment result of abnormal customer behavior, lock the access control and trigger an alarm.

[0084] Specifically, on the basis of the above embodiments, high-risk customers are those who have neither entered the store nor passed outside the store, or those with irregular behavior trajectories outside the store. Given the certain uncertainty in the past behavior of high-risk customers and the relatively high security threat they pose, when evaluating high-risk customers, the door-opening action is taken into consideration. If the door-opening action is coordinated, it indicates that the customer's entry behavior is initially normal, and then the behavior label is determined in combination with the consumption behavior and the historical in-store activity video is updated; if the door-opening action is uncoordinated, such as violent door-opening, abnormal probing and other behaviors, it is directly determined that the customer behavior is abnormal, the access control is immediately locked and an alarm is triggered to block the security threat from the source and maximize the protection of the store's property safety.

[0085] On the basis of the above effects, the present invention comprehensively judges the customer risk level through the recognition of the customer's face features and the analysis of their behavior trajectories, and at the same time allocates a differential behavior label judgment strategy based on the customer risk level, saving computing resources and facilitating further differential access control management and monitoring of different types of customers, realizing the unity of operation efficiency and security management, and providing reliable support for the intelligent and refined management of the access control system of unmanned stores.

[0086] Please refer to Figure 4 as shown in Figure 4 is a step diagram for determining the customer behavior label of high-risk customers in an embodiment of the present invention. Specifically, the behavior judgment module constructs a customer behavior label according to the customer consumption behavior based on the judgment result of normal customer behavior, which specifically includes:

[0087] Determine the payment status of high-risk customers according to the types of shopping goods, the quantity of shopping goods and the single consumption record of the customer, determine the staying status according to the real-time activity path of the customer and the preset sensitive area, determine the activity behavior type of the customer according to the real-time activity behavior, and construct a customer behavior label according to the payment status, the staying status and the activity behavior type.

[0088] Specifically, based on the above embodiments, the payment status is determined by the types and quantities of purchased goods and single consumption records, enabling timely detection of abnormal payment behaviors. For example, there are a large number of purchased goods but only small payment records, or high-value goods are purchased but the payment amount is significantly inconsistent, thus identifying situations such as missed payments and illegal payments. Evaluating the staying status based on the real-time activity path and preset sensitive areas takes into account the customer's behavioral intentions. A complex and abnormal path or staying in sensitive areas (such as goods storage areas and blind spots of monitoring) for a long time may indicate potential risks. The real-time activity behaviors directly reflect the customer's current status. Abnormal behaviors such as wandering around and deliberately covering the face are significantly different from normal shopping behaviors. Combining the payment status, staying status, and activity behavior types to construct customer behavior tags can accurately classify customers.

[0089] In implementation, the single consumption goods and single consumption quantity of the customer are determined according to the customer's single consumption record, and the types and quantities of the customer's purchased goods are determined according to the customer's in-store activity video and then matched. If they are completely matched and the payment amount is consistent with the total price of the goods, it is determined that the customer's payment status is normal payment; if the payment amount is less than the total price of the goods, it is determined that the customer's payment status is abnormal payment. If the payment amount is greater than the total price of the goods, it indicates that there is a situation of repeated scanning for the customer, and the payment status is redundant payment. Customer behavior tags are constructed based on the payment status, the staying status, and the activity behavior type; if the customer's payment status is redundant payment or normal payment, the staying status is a temporary staying status, and the activity behavior type is a normal activity, then the customer is determined to be a normal customer; if the customer's payment status is abnormal payment, the staying status is a temporary staying status, and the activity type is a normal activity, then the customer is determined to be a missed payment customer; if the customer's payment status is abnormal payment, the staying status is a continuous staying status, and the activity type is a disruptive activity, then the customer is determined to be an illegal customer.

[0090] Based on the above effects, the present invention accurately classifies customers as normal customers, missed payment customers, or illegal customers through cross-verification and comprehensive consideration of multi-dimensional data, and further provides a reliable basis for access control, security warning, and subsequent management decisions, effectively ensuring the property safety and normal operation order of unmanned stores.

[0091] Specifically, the behavior judgment module determines the passing times and staying time of the customer passing through any preset sensitive area according to the real-time activity path to determine the staying status;

[0092] Among them, the staying status includes a continuous staying status and a temporary staying status.

[0093] In detail, based on the above embodiments, by analyzing multi-dimensional data such as the customer's real-time activity path and the sensitivity of the staying area, it is possible to more comprehensively and accurately evaluate the customer's behavioral characteristics and potential risks. For example, for customers who stay in sensitive areas for a long time, the access control system can take corresponding strict control measures to achieve differentiated management.

[0094] In implementation, the preset sensitive areas are the intersection of shelf aisles, the cashier, and the place where valuable goods are placed. The location information corresponding to the customer's real-time activity path can be compared with the preset sensitive areas to determine the number of times the customer passes through each preset sensitive area. While counting the number of times the customer passes through, the time when the customer enters and leaves each preset sensitive area is recorded to determine the residence time; if the customer always stays in the same preset sensitive area, the residence time is accumulated based on the number of times the customer passes through to determine the total residence time. If the total residence time of the customer in the same preset sensitive area is greater than or equal to the preset observation period, the residence state is determined to be a continuous residence state; if the total residence time of the customer in the same preset sensitive area is less than the preset observation period, the residence state is determined to be a temporary residence state. The value range of the preset observation period is 3min to 7min, and preferably, the value of the preset observation period is 5min.

[0095] On the basis of the above effects, the present invention can accurately identify the behavior characteristics of customers in unmanned stores and determine their stay status. It can monitor and warn customers' long-term stays or abnormal entry and exit behaviors in sensitive areas (such as cashier areas and valuable goods areas), and promptly discover potential violations such as theft and destruction. This will help to subsequently achieve differentiated access control management for customers of different risk levels, making access control management more scientific and reasonable, and improving the intelligence level of unmanned stores.

[0096] Specifically, the access control module determines the access control switch strategy based on the customer behavior tag, including:

[0097] If the customer behavior label is a normal customer, the in-store activity video will be included in the historical in-store activity videos and the customer type will not be updated;

[0098] If the customer behavior tag is a violating customer, the access control will be locked, the alarm will be triggered, and the customer type will be updated;

[0099] If the customer behavior label is an abnormally lingering customer, the access control switch strategy is determined based on the stay time and the preset observation period and the customer type is updated.

[0100] Specifically, based on the above embodiments, by taking different measures for customers with different behavior tags, precise management of customers can be achieved. For normal customers, the access control can be opened and closed normally, and their in-store activity videos can be included in the historical videos, which helps accumulate data for subsequent analysis of customer behavior patterns and consumption habits. For customers who violate the regulations, the access control is directly locked and an alarm is triggered, which can effectively ensure the safety and normal operation order of the store. For customers who stay abnormally, the access control switch strategy is selected according to the set preset observation period. That is, within the observation period, the access control is switched normally. Beyond the observation period, the customer behavior tag can be re-determined according to the consumption behavior to execute the corresponding access control switch strategy.

[0101] Specifically, based on the above embodiments, the customer behavior tag is determined according to the customer's single in-store behavior to decide whether to update the customer type. This dynamic management method can timely reflect the changes in customer behavior. For example, if a normal customer who has always been identified as low-risk shows a violation behavior during a certain in-store visit, the tag is triggered to be updated, and the customer type can be updated to a medium-risk customer, which helps the store pay key attention to and prevent their subsequent behaviors and reduce potential risks.

[0102] In implementation, the value range of the preset time period is 2 min to 6 min. Preferably, the value of the preset time period is 5 min.

[0103] So far, the technical solution of the present invention has been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

[0104] The above are only the preferred embodiments of the present invention and are not used to limit the present invention; for those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent access control management system, characterized in that, Including: A data acquisition module for real-time acquisition of videos of off-store activities, in-store activities, and current opening portraits of the unmanned store; A coordination judgment module connected to the data acquisition module for judging whether the opening action of a customer is coordinated according to the off-store activity video; A type judgment module respectively connected to the data acquisition module and the coordination judgment module for judging the customer type according to the off-store activity video, historical opening portraits, and current opening portraits; Among them, the type judgment module respectively determines an in-store face data set and a face data set of people passing by the store according to historical opening portraits and historical off-store activity videos, determines whether the customer type is a low-risk customer according to the comparison result between the current opening portrait and the in-store face data set, and determines the customer type according to the judgment result of non-low-risk customers in combination with the face data set of people passing by the store. The face data set of people passing by the store is determined according to the judgment result of non-low-risk customers in combination with off-store activity videos within a historical period; The customer types include low-risk customers, medium-risk customers, and high-risk customers. The face data set of people passing by the store includes a face data set of regular people passing by the store and a face data set of irregular people passing by the store; In a historical period, the trajectory characterization state is determined according to the number of deviations of the off-store activity time period of a single person passing by the store from the fixed passing time period within a fixed passing date. If the number of deviations is greater than a preset number, the trajectory characterization state is determined to be an irregular trajectory state and the face data attribute corresponding to the face data of a single pedestrian is the face data set of irregular people passing by the store. If the number of deviations is less than or equal to the preset number, the trajectory characterization state is a regular trajectory state and the face data attribute corresponding to the face data of a single pedestrian is the face data set of regular people passing by the store; A behavior judgment module respectively connected to the data acquisition module, the type judgment module, and the coordination judgment module for analyzing the customer consumption behaviors of each customer through artificial intelligence based on the in-store activity video, and constructing a customer behavior label of the current in-store customer in combination with the customer type judgment result, the coordination judgment result, and historical in-store activity videos; Among them, the customer consumption behaviors include shopping duration, shopping commodity types, shopping commodity quantities, single consumption records, real-time activity paths, and real-time activity behaviors; An access control module respectively connected to the type judgment module and the behavior judgment module for determining an access control switch strategy according to the customer type and the customer behavior label and updating the customer type.

2. The intelligent access control management system according to claim 1, wherein The type judgment module determines the customer type according to the face data attribute.

3. The intelligent access control management system according to claim 1, wherein The coordination judgment module extracts key frames of the opening action of a single opening customer according to the off-store activity video, and determines the arm movement amplitude and body rotation amplitude of the customer according to the key frames of the opening action to determine the movement amplitude ratio to judge whether the opening action of the customer is coordinated.

4. The intelligent access control management system according to claim 1, wherein The behavior judgment module constructs a customer behavior label based on the historical in-store activity video and customer consumption behaviors according to the low-risk customer judgment result, specifically including: Determine the historical shopping duration based on the historical in-store activity video of a single low-risk customer, and determine the customer behavior label based on the shopping duration of the customer and the historical shopping duration; The customer behavior labels include normal customers, illegal customers and abnormally staying customers.

5. The intelligent access control management system according to claim 4, wherein The behavior judgment module constructs a customer behavior label based on the medium-risk customer judgment result and the customer's consumption behavior, specifically including: The regular shopping duration is determined according to the historical shopping duration of all low-risk customers who enter the store for the first time, and the customer behavior label of the medium-risk customer is determined based on the current shopping duration of the medium-risk customer and the regular shopping duration.

6. The intelligent access control management system according to claim 5, wherein, The behavior judgment module constructs a customer behavior label based on the high-risk customer judgment result according to the coordination judgment result and the customer consumption behavior, specifically including: A preliminary judgment is made as to whether the customer behavior is abnormal based on the coordination judgment result; a customer behavior tag is determined based on the customer consumption behavior based on the normal customer behavior judgment result; or the access control is locked and an alarm is triggered based on the abnormal customer behavior judgment result.

7. The intelligent access control management system according to claim 6, characterized in that The behavior judgment module constructs a customer behavior tag based on the customer consumption behavior based on the normal customer behavior judgment result, specifically including: The payment status of high-risk customers is determined based on the type of shopping items, the quantity of shopping items and single consumption records of the customers, the stay status is determined based on the real-time activity path and preset sensitive areas of the customers, the type of activity behavior of the customers is determined based on the real-time activity behavior, and a customer behavior label is constructed based on the payment status, the stay status and the activity behavior type.

8. The intelligent access control management system according to claim 7, characterized in that, The behavior judgment module determines the number of times the customer passes through any preset sensitive area and the duration of stay according to the real-time activity path to determine the stay state; The stay status includes a continuous stay status and a temporary stay status.

9. The intelligent access control management system according to claim 7, wherein, The access control module determines the access control strategy based on the customer behavior tags, including: If the customer behavior label is a normal customer, the in-store activity video will be included in the historical in-store activity videos and the customer type will not be updated; If the customer behavior tag is a violating customer, the access control will be locked, the alarm will be triggered, and the customer type will be updated; If the customer behavior label is an abnormally lingering customer, the access control switch strategy is determined based on the stay time and the preset observation period and the customer type is updated.

Images