Multi-dimensional compliance detection system and method based on face recognition safety management

Abstract

The invention relates to the technical field of face recognition and safety management, and discloses a multi-dimensional compliance detection system and method based on face recognition safety management. The system comprises an image acquisition module, a feature reconstruction module, a compliance analysis module, a security decision module, a central control unit and the like. The image acquisition module acquires and classifies biological characteristic data streams, the characteristic reconstruction module generates a first characteristic topological graph and a second characteristic behavior chain, the compliance analysis module fuses the data to map the security level, the security decision module calls a security scheme, and the central control unit verifies and generates a final security instruction. The system realizes multi-dimensional biological characteristic acquisition and analysis, accurate evaluation of security level, and efficient making and execution of security decision, also has equipment linkage and data archiving functions, effectively improves the security management level, is suitable for various security management scenes, and ensures the place security.

Description

Technical Field

[0001] The present invention relates to the fields of face recognition technology and security management technology, and specifically to a multi-dimensional compliance detection system and method based on face recognition security management. Background Art

[0002] With the rapid development of the information age, security management is becoming increasingly important across all sectors. Whether in corporate offices, financial institutions, or public spaces, strict requirements exist for personnel identification and security control. Traditional security management methods, such as access cards and password verification, have numerous drawbacks. Access cards are easily lost or stolen, and passwords can be forgotten or cracked. These methods fail to fundamentally guarantee the reliability and accuracy of security management, making them unable to meet today's complex and ever-changing security needs.

[0003] Facial recognition technology, an emerging biometric technology, has gained widespread adoption in recent years. It offers advantages such as uniqueness, resistance to duplication, and ease of identification. However, existing facial recognition security management systems mostly remain limited to simple identity verification and suffer from numerous shortcomings. For one thing, feature collection often relies on a single dimension, often relying solely on basic facial image features while neglecting the integrated use of dynamic features and other biometric characteristics. This results in incomplete information acquisition and an inability to accurately assess a person's true status and potential risks. For example, identifying an individual based solely on static facial images makes it difficult to detect disguises or unusual emotions, creating opportunities for criminals to exploit.

[0004] On the other hand, existing compliance detection capabilities are weak. When addressing security needs in various scenarios, there's a lack of a scientific and rational assessment system. This makes it impossible to conduct multi-dimensional compliance analysis of personnel behaviors and characteristics, making it difficult to accurately define security levels. For example, within financial institutions, it's impossible to effectively assess the authenticity of the identities, compliance of behavior, and potential threats of personnel entering high-risk areas, which can easily lead to security vulnerabilities.

[0005] Furthermore, existing security management systems also suffer from flaws in decision-making and execution. Once an anomaly is detected, the system is unable to quickly and accurately develop and execute effective security plans. The lack of efficient linkage mechanisms between various security devices leads to inefficient security management processes and an inability to promptly respond to and address security incidents, creating potential safety risks.

[0006] The existing system also has shortcomings in data processing and storage. After data collection, there's a lack of effective integration and analysis, and the storage method is relatively simplistic, preventing a complete chain of inspection records and hindering subsequent safety audits and traceability. For example, after a safety incident, it's difficult to quickly obtain comprehensive personnel profile data and handling records through the system, complicating investigations. Summary of the Invention

[0007] The purpose of the present invention is to provide a multi-dimensional compliance detection system and method based on face recognition security management to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional compliance detection system based on face recognition security management, the system comprising:

[0009] An image acquisition module is used to obtain a biometric data stream of a target object and divide the biometric data stream into a static feature data stream and a dynamic feature data stream based on a preset feature classification strategy;

[0010] a feature reconstruction module, configured to perform three-dimensional deformation modeling on the static feature data stream to generate a first feature topology map, and to perform micro-expression tracking on the dynamic feature data stream to generate a second feature behavior chain;

[0011] A compliance analysis module, configured to fuse the first feature topology map and the second feature behavior chain according to a preset compliance matrix, and map the data to a corresponding security level, and use the security level as a detection indicator for the current object;

[0012] A security decision module is used to call a target security solution in a preset strategy database based on the security level and use the target security solution as an execution strategy for the current object;

[0013] The central control unit is used to perform logic verification on the detection indicators and the execution strategy to generate a final security instruction.

[0014] Preferably, the feature reconstruction module performs micro-expression tracking on the dynamic feature data stream, including:

[0015] Dividing a continuous frame sequence in the dynamic feature data stream into a baseline expression group and an abnormal expression group, and performing muscle movement calculation on the abnormal expression group based on a preset optical flow parsing model to generate a dynamic feature set;

[0016] Performing time-axis segmentation processing on pupil change data in the dynamic feature data stream, extracting iris vibration features of each sequence and constructing a behavior map;

[0017] The dynamic feature set is matched with the behavior graph in a temporal and spatial manner to generate the second feature behavior chain.

[0018] Preferably, the preset feature classification strategy includes a basic feature set and an auxiliary feature set; the basic feature set includes a bone structure identifier, a texture feature identifier and a geometric proportion identifier; the auxiliary feature set includes a capillary distribution identifier and a thermal imaging feature identifier, and each identifier corresponds to an independent data analysis channel.

[0019] Preferably, the system further comprises a biometric identification interface, wherein the biometric identification interface is used to realize communication docking between the image acquisition module, the feature reconstruction module, the compliance analysis module and the security decision module and the biosensor network respectively;

[0020] The image acquisition module divides the biometric data stream based on the preset feature classification strategy, including:

[0021] receiving a composite feature packet from a biosensor network in real time via the biometric recognition interface, and matching a protocol tag of the composite feature packet according to an identifier in the basic feature set to separate a basic feature segment;

[0022] traversing the extended tags of the composite feature package according to the identifier in the auxiliary feature set to extract the auxiliary feature segment;

[0023] The basic feature segments and the auxiliary feature segments are aligned according to the acquisition timing and then written into the static feature storage area and the dynamic feature buffer area respectively.

[0024] Preferably, when the preset compliance matrix adopts a static feature model, the security level is an interval mapping result of a weighted fusion value of the first feature topology map and the second feature behavior chain;

[0025] When the preset compliance matrix adopts a dynamic behavior model, the security level is a continuous security set that is dynamically adjusted by an incremental learning algorithm based on the correlation analysis results of the first characteristic topology map and the second characteristic behavior chain.

[0026] Preferably, the system further comprises a device linkage module connected to the central control unit, and the device linkage module is connected to the security device database via the biometric interface;

[0027] The device linkage module is used to filter the list of compatible devices from the security device database according to the device control parameters in the final security instruction, and generate a security execution sequence to optimize the security management process.

[0028] Preferably, the device linkage module generates a security execution sequence including:

[0029] Loading a monitoring area spatial grid model, and locating the spatial coordinate node of each device in the adapted device list in the grid model;

[0030] Calculate the optimal action path from the current coordinates of each device to the target protection area based on the trajectory deduction algorithm, and sort the effectiveness of the list of compatible devices according to the response level;

[0031] The optimal action path and the efficiency ranking are integrated into the grid model to generate a visual security execution sequence.

[0032] Preferably, when the central control unit performs logical verification on the detection indicators and execution strategies, a dual audit mode of feature integrity verification mechanism and biological association verification mechanism is adopted. The feature integrity verification mechanism is used to confirm the validity of data collection, and the biological association verification mechanism is used to resolve logical contradictions between features.

[0033] Preferably, the system further comprises:

[0034] an instruction encoding module connected to the central control unit, the instruction encoding module being configured to convert the final security instruction into a device control code, and to send the device control code to a designated security device via the biometric interface to initiate a security program;

[0035] A data archiving module connected to the central control unit is used to store the biometric data stream, the first feature topology map, the second feature behavior chain, the security level and the final security instruction, and generate a complete detection record chain according to the timeline.

[0036] Preferably, the present invention further includes a multi-dimensional compliance detection method based on face recognition security management, which is applied to the multi-dimensional compliance detection system based on face recognition security management as described above, and the method includes:

[0037] Step 1: Using an image acquisition module to acquire a biometric data stream of a target object, and dividing the biometric data stream into a static feature data stream and a dynamic feature data stream based on a preset feature classification strategy;

[0038] Step 2: Performing three-dimensional deformation modeling on the static feature data stream through a feature reconstruction module to generate a first feature topology map, and simultaneously performing micro-expression tracking on the dynamic feature data stream to generate a second feature behavior chain;

[0039] Step 3: Using the compliance analysis module, according to a preset compliance matrix, the first feature topology map and the second feature behavior chain are fused and mapped to a corresponding security level, and the security level is used as a detection indicator for the current object;

[0040] Step 4: Using the security decision module to call the target security plan in the preset strategy database based on the security level, and using the target security plan as the execution strategy for the current object;

[0041] Step 5: Use a central control unit to perform logic verification on the detection indicators and the execution strategy to generate a final security instruction.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] In terms of feature acquisition and processing, the system's image acquisition module and feature reconstruction module work together to achieve comprehensive biometric feature acquisition and in-depth processing. The image acquisition module, using a pre-set feature classification strategy, meticulously divides the biometric data stream into static and dynamic feature streams. These streams cover a wide range of features, including skeletal structure, texture, geometric proportions, capillary distribution, and thermal imaging, providing a rich data foundation for subsequent, precise analysis. The feature reconstruction module performs three-dimensional deformation modeling on static features to generate a primary feature topology map, accurately depicting the static structural characteristics of the face. It also tracks dynamic features through micro-expression tracking, generating secondary feature behavior chains based on dimensions such as muscle movement and pupil changes, capturing changes in a person's emotions and behavior. This multi-dimensional feature acquisition and reconstruction enables more comprehensive and accurate identification and analysis of a person's identity and status than traditional methods that rely solely on static image recognition, significantly improving recognition accuracy and reliability. For example, in airport security scenarios, this system not only enables rapid and accurate passenger identification but also allows for the detection of nervousness or unusual emotions through micro-expression tracking, promptly identifying potential security risks.

[0044] At the compliance analysis and security level assessment level, the compliance analysis module performs data fusion and security level mapping based on a preset compliance matrix. The preset compliance matrix has two modes: a static feature model and a dynamic behavior model, which can flexibly adapt to the security needs of different scenarios. In the static feature model, the security level is determined by weighted fusion of the first feature topology map and the second feature behavior chain, and the fusion value interval mapping is used to determine the security level. The dynamic behavior model uses an incremental learning algorithm to dynamically adjust the feature association analysis results to generate a continuous security set. This multi-mode, dynamic security level assessment mechanism can accurately measure personnel compliance and potential risks, providing a scientific basis for subsequent security decisions. In the access control management of confidential areas of the enterprise, security levels can be accurately divided according to employee identity characteristics and behavioral performance, and access rights can be reasonably restricted to effectively protect enterprise information security.

[0045] In terms of security decision-making and execution, the security decision-making module invokes targeted security solutions from a pre-set strategy database based on security levels, ensuring targeted and effective decisions. The central control unit performs logical verification of detection indicators and execution strategies, employing a dual audit model of feature integrity verification and biometric correlation verification to ensure the validity of data collection and the rationality of the logic between features, generating reliable final security instructions. The device linkage module selects a list of compatible devices based on the instructions and generates a security execution sequence. By loading a spatial grid model, locating device coordinates, calculating optimal action paths, and ranking them for effectiveness, it achieves efficient linkage of security devices. This series of operations makes the security management process more intelligent and efficient, significantly shortening the response time from detection to decision-making and execution, effectively improving the timeliness and effectiveness of security management. For example, in the event of an emergency in a large shopping mall, the system can quickly coordinate and coordinate surrounding surveillance equipment, access control systems, alarm systems, and other devices to rapidly respond to and handle the crisis.

[0046] In terms of data management, the instruction encoding module converts final security instructions into device control codes, enabling accurate instruction transmission and precise device control. The data archiving module stores key data such as biometric data streams, feature topology maps, behavioral chains, security levels, and final security instructions, and generates a complete detection record chain based on a timeline. This not only facilitates full traceability and auditing of the security management process, providing rich data support for subsequent security strategy optimization, but also complies with relevant regulatory requirements for data management and security auditing. For example, in judicial investigations, a complete detection record chain can serve as strong evidence to help restore the truth of an incident. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a working principle diagram of the multi-dimensional compliance detection system based on face recognition security management according to the present invention;

[0048] Figure 2 This is a diagram showing the working principle of the feature reconstruction module for tracking micro-expressions on dynamic feature data streams;

[0049] Figure 3 A diagram showing the working principle of the image acquisition module that divides the biometric data stream based on a preset feature classification strategy;

[0050] Figure 4 Generate a working principle diagram of the security execution sequence for the device linkage module;

[0051] Figure 5 This is the working principle diagram of the central control unit logic verification. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] See also Figure 1-Figure 5 The present invention provides a multi-dimensional compliance detection system based on face recognition security management, and the specific implementation steps are as follows:

[0054] Image Acquisition Module: This module is responsible for acquiring the target subject's biometric data stream. In practical applications, the image acquisition module can use a variety of devices, such as high-definition cameras and infrared cameras, to ensure comprehensive and accurate biometric data. After data acquisition, the biometric data stream is divided into static and dynamic feature data streams based on a pre-set feature classification strategy.

[0055] Feature Reconstruction Module: Based on the static feature data stream transmitted by the image acquisition module, the feature reconstruction module performs 3D deformable modeling. Using advanced modeling techniques, it simulates the 3D structural changes of the face and generates a first feature topology map that clearly displays the spatial relationships between static facial features. For the dynamic feature data stream, the feature reconstruction module performs micro-expression tracking. By analyzing and processing the continuous frame sequence in the dynamic feature data stream, it further generates a second feature behavior chain that reflects the dynamic behavioral characteristics of the target object.

[0056] Compliance Analysis Module: The compliance analysis module fuses the first feature topology graph and the second feature behavior chain based on a pre-set compliance matrix. The pre-set compliance matrix is a set of pre-defined rules based on extensive security data and actual application scenarios. By mapping the fused data to a corresponding security level and using that level as a detection indicator for the current object, the security compliance of the target object is determined.

[0057] Security Decision-Making Module: Based on the security level determined by the Compliance Analysis Module, the Security Decision-Making Module calls a target security plan from the preset policy database. The preset policy database stores a variety of security plans for different security levels. The Security Decision-Making Module uses the called target security plan as the execution policy for the current object, providing specific operational guidance for subsequent security protection work.

[0058] Central Control Unit: This unit performs logical verification of detection indicators and execution strategies. It employs a dual audit model, combining feature integrity verification and biometric correlation verification, to ensure data accuracy and logical rationality. After verification, it generates final security instructions, which are used to control relevant security equipment to execute specific security operations and achieve secure management of the target object.

[0059] The technical solution of the present invention is further described in detail below through five embodiments:

[0060] Example 1:

[0061] In this embodiment, the specific process of micro-expression tracking in a dynamic feature data stream by the feature reconstruction module is described in detail. First, the continuous frame sequence in the dynamic feature data stream is divided into a baseline expression group and an abnormal expression group. The baseline expression group serves as a reference for expressions in a normal state, while the abnormal expression group contains expression data that may indicate abnormal behavior. A preset optical flow parsing model is used to calculate muscle movement in the abnormal expression group. By analyzing the movement changes of facial muscles between different frames, a dynamic feature set is generated. This dynamic feature set accurately reflects the characteristic information of facial muscle movement. The pupil change data in the dynamic feature data stream is segmented along the time axis, and the iris vibration characteristics of each sequence are extracted to construct a behavior map. The dynamic feature set is then temporally and spatially correlated with the behavior map, and a second feature behavior chain is generated by comprehensively considering the feature relationships in the temporal and spatial dimensions. This generated second feature behavior chain more comprehensively and accurately reflects the micro-expression changes of the target object, providing a strong basis for subsequent security level determination.

[0062] In a high-security facility with strict access control requirements, such as the core office area of a large financial institution, a multi-dimensional compliance detection system based on facial recognition security management was deployed. Multiple high-definition cameras were installed as image acquisition modules within the institution. These cameras clearly capture the faces of people entering the area, continuously generating biometric data streams. These data streams are transmitted in real time to a feature reconstruction module, where the dynamic feature data stream undergoes micro-expression tracking.

[0063] When a staff member approaches the access control system, the system automatically divides a continuous sequence of frames from the dynamic feature data stream captured by the camera into different groups. Assuming this sequence contains 100 frames, the system will classify the first 30 frames as the baseline expression group, representing the person's normal expressions, based on pre-defined rules such as the amplitude and frequency of expression changes. The remaining 70 frames are classified as the abnormal expression group, which is used for subsequent analysis of possible abnormal expressions.

[0064] A preset optical flow parsing model is used to calculate muscle movement for abnormal expression groups. This model analyzes subtle changes in facial muscles frame by frame, such as the upward or downward movement of the corners of the mouth, or the raised or furrowed eyebrows. Taking the corners of the mouth as an example, the model tracks their displacement and angle changes between frames. Combining the attachment points and movement patterns of the facial muscles, it calculates information such as the movement amplitude and speed of the relevant muscles. After analyzing 70 frames of abnormal expression groups, a dynamic feature set was generated, encompassing movement features of multiple areas, including the corners of the mouth, eyebrows, and muscles surrounding the eyes. This set details the movement of facial muscles during this period.

[0065] The system processes pupil change data in chronological segments, assuming each segment lasts 0.5 seconds. Pupil change data for each segment is extracted from the dynamic feature data stream, including pupil dilation, iris vibration frequency and amplitude, and other information. This feature data is organized and analyzed to construct a behavioral map reflecting the patterns of pupil and iris changes. For example, the map might show a sudden increase in iris vibration frequency and a slight pupil contraction during a certain period.

[0066] The system comprehensively analyzes the dynamic feature set and the behavioral graph, considering their correlation in time and space. For example, if the dynamic feature set shows a sudden downward movement of the corners of the mouth at a certain moment, and the behavioral graph shows abnormal iris vibration around the same time, the system will correlate and match these two features in the spatiotemporal dimensions. Through this matching, facial muscle movement and pupil and iris change characteristics are organically combined to generate a second characteristic behavior chain. This behavior chain fully presents the dynamic process of the staff member's facial expression and eye changes when passing through the access control, providing key evidence for subsequent determination of abnormal behavior and security level. If the characteristics in the behavior chain differ significantly from normal behavior patterns, it may indicate that the person is experiencing abnormal emotions such as tension and anxiety, and their safety requires further assessment.

[0067] Example 2:

[0068] This embodiment focuses on the specific implementation of a preset feature classification strategy. The preset feature classification strategy includes a basic feature set and an auxiliary feature set. The basic feature set includes skeletal structure identifiers, texture identifiers, and geometric proportion identifiers, while the auxiliary feature set includes capillary distribution identifiers and thermal imaging identifiers. Each identifier corresponds to an independent data analysis channel. The system also provides a biometric interface for communication between the image acquisition module, feature reconstruction module, compliance analysis module, and security decision-making module and the biosensor network. During operation, the image acquisition module receives composite feature packets from the biosensor network in real time via the biometric interface. The protocol tags of the composite feature packets are matched against the identifiers in the basic feature set to isolate the basic feature segments. The extended tags of the composite feature packets are traversed according to the identifiers in the auxiliary feature set to extract the auxiliary feature segments. The basic and auxiliary feature segments are aligned according to the acquisition time sequence and then written to the static feature storage area and dynamic feature buffer, respectively. This method enables efficient and accurate classification and storage of biometric data streams, laying a solid foundation for subsequent feature reconstruction and analysis.

[0069] Take the security management system of a high-end scientific research park as an example. There are many important laboratories in the park, which store a large amount of confidential scientific research data. The access control of personnel is extremely strict, and a multi-dimensional compliance detection system based on facial recognition security management is adopted.

[0070] A variety of biometric sensors have been deployed at key locations, such as entrances and exits to the research park and passageways within each laboratory building. These devices form a biometric sensor network that continuously collects biometric data from individuals and packages it into composite feature packets for transmission. The system's biometric recognition interface receives these composite feature packets. After acquiring the data through the biometric recognition interface, the image acquisition module begins processing based on pre-set feature classification strategies.

[0071] The basic feature set in the preset feature classification strategy includes skeletal structure identifiers, texture feature identifiers, and geometric proportion identifiers. When a researcher enters the main entrance of the campus, the image acquisition module receives a composite feature packet from the biosensor network. The module first matches the protocol tag of the composite feature packet with the skeletal structure identifier in the basic feature set. For example, the skeletal structure identifier appears as a specific code in the protocol tag, such as "0x0101" for the zygomatic structure identifier and "0x0102" for the mandibular structure identifier. Through this matching, the data segments related to skeletal structure in the composite feature packet are accurately separated, forming the skeletal structure portion of the basic feature segment. Next, the protocol tags are matched again based on the texture feature identifiers, such as "0x0201" for the skin texture fineness identifier and "0x0202" for the facial wrinkle identifier, to separate the data segments related to the texture features. Similarly, the corresponding geometric proportion feature data segments are separated based on geometric proportion identifiers, such as those for eye spacing and the relative position and proportion of facial features. These data segments together constitute the basic feature segment.

[0072] The auxiliary feature set includes a capillary distribution identifier and a thermal imaging feature identifier. Based on the capillary distribution identifier, the image acquisition module traverses the extended tags of the composite feature package. Assuming the capillary distribution identifier is coded "0x0301" in the extended tag, the module searches the composite feature package for data containing this identifier and extracts relevant data on capillary distribution, such as capillary density and blood flow velocity in different facial regions. For the thermal imaging feature identifier, assuming its code is "0x0302" in the extended tag, the module extracts thermal imaging feature data, such as facial temperature distribution and hotspot locations. This data constitutes the auxiliary feature segment.

[0073] After collecting the basic and auxiliary feature segments, the image acquisition module aligns them according to the acquisition time sequence. Because different devices in the biosensor network may collect data at slightly different times, the times of the basic and auxiliary feature segments must be aligned. For example, if the skeletal structure data is collected at 10:00:00.001 and the capillary distribution data is collected at 10:00:00.003, their times must be aligned to ensure data consistency. After alignment, the basic feature segments are written to the static feature storage area. These data reflect relatively stable features of the researcher's face, such as bone structure and texture, and do not change rapidly over time. The auxiliary feature segments are written to the dynamic feature buffer, because capillary distribution and thermal imaging features can change due to changes in a person's mood or physical condition and are therefore dynamic feature data. This series of operations achieves precise classification and storage of the biometric data stream, providing an organized and accurate data foundation for subsequent feature reconstruction and analysis, enabling the system to more efficiently determine identity and security compliance.

[0074] Example 3:

[0075] This embodiment details the method for determining the security level under different models of the preset compliance matrix. When the preset compliance matrix adopts a static feature model, the security level is the interval mapping result of the weighted fusion value of the first feature topology map and the second feature behavior chain. In the actual calculation process, the first feature topology map and the second feature behavior chain are assigned corresponding weights according to the importance of different features, and the two are fused and calculated. Then, according to the pre-set interval mapping rules, the fusion value is mapped to the corresponding security level interval to determine the security level of the current object. When the preset compliance matrix adopts a dynamic behavior model, the security level is a continuous security set that dynamically adjusts the correlation analysis results of the first feature topology map and the second feature behavior chain through an incremental learning algorithm. The incremental learning algorithm can continuously update the analysis results of the correlation relationship between features as new data is continuously input, and then dynamically adjust the security level, so that the judgment of the security level is more in line with changes in actual conditions.

[0076] For example, consider a highly classified military base's personnel entry and exit management scenario. Each entrance and exit is equipped with high-precision facial recognition equipment to collect biometric data streams from entrants. Suppose soldier A is about to enter the base's core area, and the system begins its workflow.

[0077] When the preset compliance matrix uses a static feature model:

[0078] The system's feature reconstruction module first processes the collected biometric data stream to generate a first feature topology map and a second feature behavior chain. The first feature topology map reflects the static structural features of Soldier A's face, such as the facial bone contours and the relative positions of the facial features. The second feature behavior chain records dynamic features such as micro-expression changes during the recognition process.

[0079] The compliance analysis module will perform weighted fusion calculation on the first feature topology map and the second feature behavior chain according to the pre-set weight. Assume that the weight corresponding to the first feature topology map is , the weight corresponding to the second characteristic behavior chain is ,and ( 、 is a weight coefficient set based on actual security requirements and experience, used to indicate the importance of different features in security level judgment). The calculation formula of the weighted fusion value is: ,in represents the weighted fusion value, Represents the quantized value of the first characteristic topology graph, Indicates the quantized value of the second characteristic behavior chain. For example, after the system quantization process, , ,set up , ,but .

[0080] Afterwards, the compliance analysis module will convert the weighted fusion value into Mapped to the corresponding security level interval. Assuming that the security level is divided into three intervals: "low risk (0-50)", "medium risk (51-80)" and "high risk (81-100)", since the calculated Therefore, soldier A is mapped to the "medium risk" security level, and the system will take corresponding security measures accordingly, such as strengthening the monitoring of his subsequent actions.

[0081] When the default compliance matrix uses a dynamic behavior model:

[0082] As Soldier A moves through the base, the system continuously collects his biometric data. The feature reconstruction module continuously updates the first feature topology map and the second feature behavior chain. The compliance analysis module uses an incremental learning algorithm to perform correlation analysis on this new data. The incremental learning algorithm continuously adjusts its understanding of the relationship between the first feature topology map and the second feature behavior chain based on the new input data. For example, after entering the core area, Soldier A heads to a specific meeting room. During his journey, the system detects unusual pauses in his walking path and facial micro-expressions that indicate nervousness. These new dynamic features are incorporated into the analysis.

[0083] Through dynamic adjustments in the incremental learning algorithm, the system generates a continuous safety set to represent Soldier A's safety status. This continuous safety set is not a simple discrete level, but rather a numerical range that more accurately reflects the changing safety level. Suppose Soldier A's safety set initially ranges from [60, 70], indicating a relatively safe but potentially risky state. As abnormal behavior emerges, the incremental learning algorithm adjusts the association analysis results, changing the safety set range to [50, 60], indicating an increased risk. Based on this change, the system promptly notifies security personnel to monitor Soldier A's actions and implement further security measures, such as dispatching personnel to inquire about the situation, to ensure base safety.

[0084] Example 4:

[0085] This embodiment focuses on describing the workflow of the device linkage module. The system is also provided with a device linkage module connected to the central control unit, and the device linkage module is connected to the security device database through a biometric interface. The device linkage module filters the list of compatible devices from the security device database according to the device control parameters in the final security instruction. During the screening process, a comprehensive judgment is made based on factors such as the function, performance and adaptability of the device to the current scene. A security execution sequence is generated to optimize the security management process. The specific steps are: loading the spatial grid model of the monitoring area, and locating the spatial coordinate node of each device in the list of compatible devices in the grid model. Based on the trajectory deduction algorithm, the optimal action path from the current coordinates of each device to the target protection area is calculated, and the list of compatible devices is sorted by efficiency according to the response level. The optimal action path and efficiency ranking are integrated into the grid model to generate a visual security execution sequence. In this way, through the work of the device linkage module, efficient collaborative work of security equipment can be achieved, and the effect of security management can be improved.

[0086] Take, for example, the terminal security management scenario at a large international airport. This vast terminal, with its frequent flow of people and luggage, places extremely high demands on security. Therefore, a multi-dimensional compliance monitoring system based on facial recognition security management has been deployed. Within the terminal, a large number of security devices are installed at key locations such as entrances and exits, waiting areas, and baggage handling areas. These devices form a security device database. Furthermore, a high-precision spatial grid model is constructed for the entire terminal's surveillance area. The size of each grid is customized based on actual monitoring needs. For example, in the crowded waiting area, the grid is more finely divided, perhaps with a side length of 1 meter; while in the relatively open baggage handling corridor, the grid side length may be 2 meters.

[0087] After the system determines the security level of a target object through the compliance analysis module and the security decision module invokes the corresponding target security plan, the central control unit generates a final security command and sends it to the device linkage module. Suppose the device control parameters in the final security command indicate that a person with a low security level exhibiting abnormal behavior in the terminal waiting area needs to be monitored and controlled.

[0088] The device linkage module first selects a list of compatible devices from the security device database based on these device control parameters. During this process, the module considers a variety of factors. For example, for surveillance needs, it prioritizes cameras with a wide field of view, high resolution, and intelligent tracking capabilities, such as the high-definition panoramic cameras installed above the terminal area. For personnel control needs, it selects devices with alarm functions and personnel interception capabilities, such as smart gates and security robots installed at key entrances.

[0089] After determining the list of compatible devices, the device linkage module loads the spatial grid model of the monitored area and locates the spatial coordinate nodes of each device within the model. For example, a high-definition panoramic camera, installed at a specific height on a pillar in the waiting area, has coordinates (X1, Y1, Z1) in the grid model using a pre-defined coordinate system. Meanwhile, a smart gate, installed at the entrance, has coordinates (X2, Y2, Z2).

[0090] The trajectory deduction algorithm calculates the optimal path from each device's current coordinates to the target protection zone (i.e., the specific location in the waiting area where the abnormal person is located). The trajectory deduction algorithm takes into account various factors, such as the flow of people within the terminal, the width of the passageways, and the degree of congestion. For the security robot, which is originally located at a charging point on the edge of the waiting area, the algorithm uses real-time personnel flow data to plan a path that avoids crowded areas, passes through relatively unobstructed passages, and quickly reaches the target area. For smart gates, the algorithm calculates the optimal strategy for adjusting the gate status, guiding surrounding personnel away from the area where the abnormal person is located, and preventing the abnormal person from escaping.

[0091] The list of compatible devices is sorted by their response level. This level is determined based on factors such as the device's functional characteristics and the time it takes to reach the target area. For example, a security robot with fast movement and real-time communication capabilities has a higher response level, followed by a high-definition panoramic camera, and then a smart gate.

[0092] The optimal action path and efficiency ranking are integrated into the grid model to generate a visual security execution sequence. On the large screen in the airport security monitoring center, staff can visually view the action path of each security device, as well as their execution order and priority. The security robot will quickly drive to the location of the abnormal person according to the planned path. The high-definition panoramic camera will automatically adjust the shooting angle to focus on the area, and the smart gate will also operate accordingly according to the instructions. This device linkage and visual execution sequence achieves rapid response and efficient handling of abnormal situations, significantly improving the security management level of the airport terminal.

[0093] Example 5:

[0094] This embodiment introduces the working method of the instruction encoding module and data archiving module in the system. The system also includes an instruction encoding module and a data archiving module connected to the central control unit. The instruction encoding module converts the final security instruction into a device control code, and sends the device control code to the designated security device through the biometric interface to start the security program. During the conversion process, the final security instruction is converted into a corresponding code format according to the communication protocol and control requirements of different security devices to ensure that the security device can accurately receive and execute the instruction. The data archiving module is used to store the biometric data stream, the first feature topology map, the second feature behavior chain, the security level and the final security instruction. A complete detection record chain is generated according to the timeline. The data archiving module uses reliable data storage technology, such as database storage or cloud storage, to ensure the security and integrity of the data. The generation of a complete detection record chain facilitates subsequent retrospective analysis of the security management process, providing data support for optimizing system performance and improving security management strategies.

[0095] Take the headquarters building of a large enterprise, for example. This building houses several key departments and stores a large amount of commercial secrets and important information, making security management crucial. The building employs a multi-dimensional compliance monitoring system based on facial recognition security management. Image acquisition devices are installed at each entrance, key area corridors, and the doors of important offices. These devices continuously collect biometric data streams from people entering the building. When an employee prepares to enter a confidential area storing core information, the system initiates a series of workflows, in which the instruction encoding module and the data archiving module play a key role.

[0096] After the various modules of the system work together, the central control unit generates the final security command. For example, if the employee's identity is determined to be legitimate and their behavior is normal, the final security command is to allow them to enter the protected area and notify the relevant access control device to open. At this point, the command encoding module begins to work. It converts the final security command into a device control code based on the pre-defined communication protocol with the access control device. Assuming that the access control device uses a specific RS-485 communication protocol, the command encoding module converts the "allow entry" command into a hexadecimal code that complies with the protocol, such as "0x010x020x03..." After encoding, the command encoding module transmits this device control code to the designated access control device via the biometric interface. Upon receiving the code, the access control device decodes it and recognizes it as a command to allow entry. It then initiates the security program, driving the motor to open the access control device, allowing the employee to enter the protected area.

[0097] Throughout the entire process, the data archiving module simultaneously performs data storage. It stores the employee's biometric data stream obtained from the image acquisition module. This biometric data stream contains static and dynamic facial feature information, such as facial bone structure, texture features, and micro-expression changes. Simultaneously, it stores the first feature topology map and the second feature behavior chain generated by the feature reconstruction module. The first feature topology map displays the three-dimensional structural relationship of the employee's static facial features, while the second feature behavior chain records the micro-expression changes during the employee's entry into the building. The data archiving module also stores the security level determined by the compliance analysis module (assuming the employee's security level is "safe"), as well as the final security instructions generated by the central control unit.

[0098] The data archiving module generates a complete chain of inspection records along a timeline. Each inspection process for an employee entering or exiting the building generates a separate record, containing detailed information such as the inspection time, biometric data flow, feature topology, feature behavior chain, security level, and final security instructions. For example, at 9:10 AM on October 15th, the data archiving module compiles the relevant data into a single record, timestamped with 9:10 AM, and stores the remaining data in the database according to the corresponding fields. In this way, a complete chain of inspection records is established.

[0099] When an enterprise needs to conduct retrospective analysis of security management, it can retrieve relevant records from the data archiving module at any time. For example, if an anomaly is detected in a confidential area, security management personnel can filter out the detection records of all personnel entering the area within a specific time period based on the time range. By reviewing these records and analyzing changes in the biometric characteristics of entrants, the basis for determining security levels, and the execution of final security instructions, it is possible to quickly locate potential security risks, thereby optimizing system performance and improving security management strategies. For example, adjusting the parameters of the preset compliance matrix and strengthening the monitoring of specific behavioral characteristics can enhance the overall security protection capabilities of the building.

[0100] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0101] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional compliance detection system based on face recognition security management, characterized by: include: An image acquisition module is used to obtain a biometric data stream of a target object and divide the biometric data stream into a static feature data stream and a dynamic feature data stream based on a preset feature classification strategy; a feature reconstruction module, configured to perform three-dimensional deformation modeling on the static feature data stream to generate a first feature topology map, and to perform micro-expression tracking on the dynamic feature data stream to generate a second feature behavior chain; A compliance analysis module, configured to fuse the first feature topology map and the second feature behavior chain according to a preset compliance matrix, and map the data to a corresponding security level, and use the security level as a detection indicator for the current object; A security decision module is used to call a target security solution in a preset strategy database based on the security level and use the target security solution as an execution strategy for the current object; The central control unit is used to perform logic verification on the detection indicators and the execution strategy to generate a final security instruction.

2. The multi-dimensional compliance detection system based on face recognition security management according to claim 1 is characterized in that: The feature reconstruction module performs micro-expression tracking on the dynamic feature data stream, including: Dividing a continuous frame sequence in the dynamic feature data stream into a baseline expression group and an abnormal expression group, and performing muscle movement calculation on the abnormal expression group based on a preset optical flow parsing model to generate a dynamic feature set; Performing time-axis segmentation processing on pupil change data in the dynamic feature data stream, extracting iris vibration features of each sequence and constructing a behavior map; The dynamic feature set is matched with the behavior graph in a temporal and spatial manner to generate the second feature behavior chain.

3. The multi-dimensional compliance detection system based on face recognition security management according to claim 1 is characterized in that: The preset feature classification strategy includes a basic feature set and an auxiliary feature set; the basic feature set includes a bone structure identifier, a texture feature identifier, and a geometric proportion identifier; the auxiliary feature set includes a capillary distribution identifier and a thermal imaging feature identifier, and each identifier corresponds to an independent data analysis channel.

4. The multi-dimensional compliance detection system based on face recognition security management according to claim 3 is characterized in that: The system further includes a biometric interface, which is used to realize communication docking between the image acquisition module, the feature reconstruction module, the compliance analysis module, and the security decision module and the biosensor network respectively; The image acquisition module divides the biometric data stream based on the preset feature classification strategy, including: receiving a composite feature packet from a biosensor network in real time via the biometric recognition interface, and matching a protocol tag of the composite feature packet according to an identifier in the basic feature set to separate a basic feature segment; traversing the extended tags of the composite feature package according to the identifier in the auxiliary feature set to extract the auxiliary feature segment; The basic feature segments and the auxiliary feature segments are aligned according to the acquisition timing and then written into the static feature storage area and the dynamic feature buffer area respectively.

5. The multi-dimensional compliance detection system based on face recognition security management according to claim 1 is characterized in that: When the preset compliance matrix adopts a static feature model, the security level is an interval mapping result of the weighted fusion value of the first feature topology map and the second feature behavior chain; When the preset compliance matrix adopts a dynamic behavior model, the security level is a continuous security set that is dynamically adjusted by an incremental learning algorithm based on the correlation analysis results of the first characteristic topology map and the second characteristic behavior chain.

6. The multi-dimensional compliance detection system based on face recognition security management according to claim 1 is characterized in that: It also includes a device linkage module connected to the central control unit, and the device linkage module is connected to the security device database through the biometric interface; The device linkage module is used to filter the list of compatible devices from the security device database according to the device control parameters in the final security instruction, and generate a security execution sequence to optimize the security management process.

7. The multi-dimensional compliance detection system based on face recognition security management according to claim 6 is characterized in that: The device linkage module generates a security execution sequence including: Loading a monitoring area spatial grid model, and locating the spatial coordinate node of each device in the adapted device list in the grid model; Calculate the optimal action path from the current coordinates of each device to the target protection area based on the trajectory deduction algorithm, and sort the effectiveness of the list of compatible devices according to the response level; The optimal action path and the efficiency ranking are integrated into the grid model to generate a visual security execution sequence.

8. The multi-dimensional compliance detection system based on face recognition security management according to claim 1 is characterized in that: When the central control unit performs logical verification on the detection indicators and execution strategies, it adopts a dual audit mode of feature integrity verification mechanism and biological correlation verification mechanism. The feature integrity verification mechanism is used to confirm the validity of data collection, and the biological correlation verification mechanism is used to resolve logical contradictions between features.

9. The multi-dimensional compliance detection system based on face recognition security management according to claim 1 is characterized in that: Also includes: an instruction encoding module connected to the central control unit, the instruction encoding module being configured to convert the final security instruction into a device control code, and to send the device control code to a designated security device via the biometric interface to initiate a security program; A data archiving module connected to the central control unit is used to store the biometric data stream, the first feature topology map, the second feature behavior chain, the security level and the final security instruction, and generate a complete detection record chain according to the timeline.

10. A multi-dimensional compliance detection method based on face recognition security management, applied to the multi-dimensional compliance detection system based on face recognition security management according to any one of claims 1 to 9, characterized in that: include: Step 1: Using an image acquisition module to acquire a biometric data stream of a target object, and dividing the biometric data stream into a static feature data stream and a dynamic feature data stream based on a preset feature classification strategy; Step 2: Performing three-dimensional deformation modeling on the static feature data stream through a feature reconstruction module to generate a first feature topology map, and simultaneously performing micro-expression tracking on the dynamic feature data stream to generate a second feature behavior chain; Step 3: Using the compliance analysis module, according to a preset compliance matrix, the first feature topology map and the second feature behavior chain are fused and mapped to a corresponding security level, and the security level is used as a detection indicator for the current object; Step 4: Using the security decision module to call the target security plan in the preset strategy database based on the security level, and using the target security plan as the execution strategy for the current object; Step 5: Use a central control unit to perform logic verification on the detection indicators and the execution strategy to generate a final security instruction.

Images