Dynamic safety management and control system and method based on intelligent lock

Abstract

The invention discloses a dynamic safety management and control system and method based on an intelligent lock, and relates to the field of intelligent management and control, firstly, a background management platform generates an authorization data packet according to an operation task and issues the authorization data packet to a key management machine, and after an operator passes identity authentication, an intelligent key obtains authorization and interacts with a target lock; and after the verification is passed, the intelligent key drives the lock to open, reads the physical state of the lock after unlocking, sends state information to the key management machine through the real-time communication module, and finally uploads the state information to a background platform for analysis and database updating. According to the scheme, closed-loop management of authority control, operation verification and state feedback is achieved, safety and management efficiency are improved, real-time synchronization and traceability of the lock state are ensured, and remote monitoring and risk early warning are effectively supported.

Description

Technical Field

[0001] This application relates to the field of intelligent control, and more specifically, to a dynamic security control system and method based on smart locks. Background Technology

[0002] With the increasing demand for intelligent management, smart lock systems are being used more and more widely in critical infrastructure sectors such as power, energy, and rail transportation. Traditional mechanical locks, unable to achieve digital control, pose security risks such as easy key loss and untraceable operation records. In contrast, smart lock systems based on smart keys and back-end management platforms have achieved preliminary security control through functions such as authorization authentication and status recording.

[0003] However, existing technologies exhibit significant real-time limitations in practical applications: when the smart key is removed from the management unit for on-site operations, the inherent delay in its status feedback mechanism arises due to the passive design of most locks and the low-power requirements of the smart key. Specifically, after the operator completes the unlocking action, the physical state of the lock (such as open / closed status, abnormal prying, etc.) can only be temporarily stored in the smart key's local storage unit. Data can only be uploaded to the backend system in batches via wired connection after the operation is completed and the key is returned to the management unit. This delayed status synchronization mechanism prevents the management platform from promptly sensing real-time changes in the lock's status. In cases of unauthorized operation, forced entry, or accidental unlocking, there is a security response window of several hours or even days. Especially in complex scenarios involving multi-level interlocking and dynamic permission adjustments, asynchronous updates of status information can lead to security risks such as permission conflicts and operational misjudgments, severely limiting the application value of smart lock systems in real-time security management scenarios.

[0004] Therefore, a dynamic security management solution based on smart locks is desired. Summary of the Invention

[0005] This application is made in order to solve the above-mentioned technical problems.

[0006] According to one aspect of this application, a dynamic security management method based on a smart lock is provided, comprising: Based on the job task information, the back-end management platform generates an authorization data packet and sends it to the key management machine; Once the operator successfully authenticates their identity at the key management machine, the key management machine will send the corresponding authorization data packet to the smart key; The smart key interacts with the target smart lock. The smart key verifies the authorization information. If the verification is successful, the smart key drives the target smart lock to open. After the unlocking action is completed, the smart key reads the current physical state of the target smart lock. The smart key uses its built-in real-time communication module to send the current physical status of the target smart lock to the key management unit within the communication range; The key management device sends the current physical status of the target smart lock to the back-end management platform. The back-end management platform parses the current physical status of the target smart lock and updates the real-time status of the corresponding target smart lock in the database.

[0007] According to another aspect of this application, a dynamic security control system based on a smart lock is provided, comprising: The authorization data packet generation and sending module is used to generate authorization data packets based on job task information through the back-end management platform and send them to the key management machine; The identity authentication module is used so that when the operator successfully authenticates his identity at the key management machine, the key management machine will send the corresponding authorization data packet to the smart key. The current physical state reading module is used to interact with the target smart lock using a smart key. The smart key verifies the authorization information. If successful, it drives the target smart lock to open. After the unlocking action is completed, the smart key reads the current physical state of the target smart lock. The sending module is used to send the current physical status of the target smart lock to the key management machine within the communication range via the smart key's built-in real-time communication module. The parsing module is used to send the current physical status of the target smart lock to the back-end management platform through the key management machine. The back-end management platform parses the current physical status of the target smart lock and updates the real-time status of the corresponding target smart lock in the database.

[0008] According to another aspect of this application, an apparatus is provided, comprising: a processor and a memory, wherein a computer-executable program is stored on the memory, and when the computer-executable program is executed by the processor, it implements the above-described dynamic security control method based on a smart lock.

[0009] Compared with existing technologies, this application provides a dynamic security management system and method based on smart locks. First, the backend management platform generates an authorization data packet based on the task and sends it to the key management machine. After the operator is authenticated, the smart key obtains authorization and interacts with the target lock. Upon successful verification, the smart key drives the lock to open and reads its physical status after unlocking. The status information is then sent to the key management machine via a real-time communication module and finally uploaded to the backend platform for parsing and database updates. This solution achieves closed-loop management of access control, operation verification, and status feedback, improving security and management efficiency, ensuring real-time synchronization and traceability of lock status, and effectively supporting remote monitoring and risk warning. Attached Figure Description

[0010] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 This is a flowchart of a dynamic security management method based on a smart lock according to an embodiment of this application.

[0012] Figure 2 This is a flowchart of step S110 in the dynamic security management method based on smart locks according to an embodiment of this application.

[0013] Figure 3 This is a flowchart of step S111 in the dynamic security management method based on smart locks according to an embodiment of this application.

[0014] Figure 4 This is a block diagram of a dynamic security management system based on a smart lock according to an embodiment of this application. Detailed Implementation

[0015] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. It should be understood that the drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0016] To address the problems mentioned above in the background technology, this application proposes a dynamic security management method based on smart locks. Figure 1 This is a flowchart of a dynamic security management method based on a smart lock according to an embodiment of this application. Figure 1 As shown, the dynamic security management method based on smart locks according to an embodiment of this application includes: S110, the background management platform generates an authorization data packet based on the job task information and sends it to the key management machine; S120, after the operator successfully authenticates their identity at the key management machine, the key management machine sends the corresponding authorization data packet to the smart key; S130, the smart key interacts with the target smart lock, wherein the smart key verifies the authorization information, and if successful, drives the target smart lock to open, and after completing the unlocking action, the smart key reads the current physical state of the target smart lock; S140, the smart key uses its built-in real-time communication module to send the current physical state of the target smart lock to the key management machine within the communication range; S150, the key management machine sends the current physical state of the target smart lock to the background management platform, and the background management platform parses the current physical state of the target smart lock and updates the real-time state of the corresponding target smart lock in the database.

[0017] In other words, in traditional solutions, the reliance on contact-based data synchronization mechanisms during off-cabinet operations prevents real-time transmission of lock status information. This solution, however, integrates a low-power real-time communication module (such as Bluetooth, LoRa, or NB-IoT) into the smart key, establishing a wireless transmission link throughout the entire operation. Once the operator unlocks the lock, the smart key immediately reads the current physical state from the lock and transmits the status data in real-time to the communication coverage area of ​​the key management unit. The management unit then forwards this data to the backend management platform, achieving state synchronization with a latency of only seconds. This design breaks the limitation of traditional solutions that require waiting for the key to return before uploading data, enabling the backend platform to monitor key information such as lock opening / closing status and abnormal prying in real time, significantly shortening the security response time in case of unauthorized operations or equipment malfunctions. Furthermore, by dynamically converting task information into authorized data packets containing user permissions, lock identification, and timeframes, and combining this with a real-time status feedback mechanism, the backend platform can immediately revoke or adjust dynamic access permission rules upon detecting anomalies. This avoids conflicts caused by asynchronous permission updates in multi-level interlocking scenarios, thus achieving a closed-loop security management system in complex operating environments.

[0018] In step S110, the backend management platform generates an authorization data packet based on the job task information and sends it to the key management machine. It should be understood that in smart lock security management, the core function of generating the authorization data packet is to accurately transform the abstracted job requirements at the business layer into dynamic permission rules that can be executed at the physical layer. Job task information and access permissions are at different information abstraction levels; job tasks are business-level requirements, while access permissions are precise instructions at the physical layer. There is a need for information conversion and mapping between the two, requiring the backend management platform to generate relevant content that meets access permission requirements based on the job task information. Traditional job tasks (such as "Inspect rack 3 in room A" in a maintenance work order) often only describe business objectives, while lock management requires clearly defining the binding relationship between user permissions and physical locks (e.g., "Zhang San can operate the LOCK_A03_FRONT lock"). This cross-level permission conversion has long relied on manual parsing. When job information comes from heterogeneous data sources such as work order systems and equipment management systems, the mapping relationship between personnel information, lock location codes, and equipment logical names is prone to deviation.

[0019] Based on this, the backend management platform generates an authorization data packet based on the job task information and sends it to the key management machine. The technical concept of this application is to achieve seamless transformation from business requirements to execution instructions by constructing an automated mapping mechanism between job task information and physical lock permissions. When the backend management platform receives a job task, it first uses natural language processing technology to deeply analyze the work order description, accurately extracting core elements such as the person in charge, job time, and equipment location through semantic analysis and contextual reasoning. To address the lock identification deviation problem that is prone to occur in manual mapping in traditional solutions, the platform dynamically associates the "person in charge" with the real-time ID in the user database based on a preset mapping rule library, automatically matches the logical location in the "equipment description" (such as "front door of rack 3 in computer room A") to LOCK_A03_FRONT in the lock spatial coding system, and converts the planned time into an encrypted time window parameter. Through a dynamic permission rule engine, this data is encapsulated into an authorization data packet containing the user ID, lock ID, and valid time, and then encrypted and pushed to the designated key management machine. This process not only ensures that job information from heterogeneous data sources such as work order systems and equipment management systems can be accurately converted into executable permission instructions, but also lays the foundation for accurate interaction between smart keys and locks through machine-readable standardized authorization data formats, effectively avoiding security vulnerabilities caused by incorrect permission mapping.

[0020] Figure 2 This is a flowchart of step S110 in the dynamic security management method based on a smart lock according to an embodiment of this application. Specifically, in the embodiments of this application, as... Figure 2 As shown, in step S110, the background management platform generates an authorization data packet based on the job task information and sends it to the key management machine, including: S111, extracting key fields from the job task information, the key fields including the person in charge / member, the planned start and end time, and the description of the job location / equipment; S112, mapping the person in charge / member to a user ID; S113, mapping the description of the job location / equipment to the ID of the target smart lock; S114, converting the planned start and end time into a valid time window for authorization; S115, generating the authorization data packet based on the user ID, the ID of the target smart lock, and the valid time window for authorization.

[0021] Specifically, in step S111, key fields are extracted from the job task information. These key fields include the person in charge / members, the planned start and end times, and a description of the job location / equipment. It should be understood that job task information often describes business-level requirements in natural language, such as "Maintenance personnel Li Si will perform maintenance on server 5 in room B from 9:00 AM to 11:00 AM tomorrow." However, smart lock security control requires precise physical-level instructions, such as specifying who (user ID) can operate which lock (target smart lock ID) at what time (valid permission time window). Therefore, key information must be extracted from the job task information to translate business requirements into executable permission rules for the smart lock system.

[0022] Figure 3 This is a flowchart of step S111 in the dynamic security management method based on a smart lock according to an embodiment of this application. Figure 3 As shown, step S111 involves extracting key fields from the task information, including the person in charge / members, planned start and end times, and a description of the task location / equipment. This includes: S1111, performing word segmentation on the task information to obtain a sequence distribution of task description words; S1112, extracting a first task description word from the sequence distribution of the task description words, and using other parts of the sequence distribution of the task description words as context information for the first task description word; and S1113, performing word embedding encoding on the first task description word to obtain a semantic embedding encoding of the first task description word. S1114, Context semantic encoding is performed on the context information of the first task descriptor to obtain the target task descriptor context semantic information encoding vector; S1115, Based on the target task descriptor context semantic information encoding vector, granular interactive iterative semantic expression enhancement is performed on the semantic embedding encoding vector of the first task descriptor to obtain the first task descriptor semantic context enhancement encoding vector; S1116, The first task descriptor semantic context enhancement encoding vector is input into a classifier-based type determiner to determine the type label of the key field to which the first task descriptor belongs.

[0023] More specifically, in step S1111, the task information is segmented to obtain a sequence distribution of task description words. It should be understood that task information often exists in natural language (e.g., "Inspect the front door of rack 3 in room A" in a work order system), and its description includes multi-dimensional information such as business objectives, equipment location, and operational requirements. This textual information has unstructured characteristics, and when directly performing semantic parsing, the ambiguity of word boundaries and semantic adhesion issues can make it difficult to accurately identify key elements (such as specific lock locations and operational objects). For example, if "front door of rack 3" is not correctly segmented into independent semantic units such as "3," "rack," and "front door," the system may mistakenly identify "front door" as an independent device rather than a substructure of the rack, thus incorrectly mapping the lock ID. Based on this, in this application, the job task information is processed by word segmentation to decompose the work order text into discrete semantic units (such as decomposing "inspection of the front door of cabinet No. 3 in computer room A" into "inspection", "computer room A", "No. 3", "cabinet", and "front door"), forming a sequence distribution of job task description words with contextual relevance.

[0024] More specifically, in step S1112, a first task description word is extracted from the sequence distribution of the task description words, and the other parts of the sequence distribution of the task description words are used as the context information of the first task description word. Correspondingly, considering that the semantic understanding depth of the task description words directly affects the generation accuracy of the permission rules, strong dependencies exist between words in natural language text, and the independent semantics of a single word often have ambiguity and incompleteness. For example, the word "front door" in a work order, when taken out of context, may refer to various physical locks such as building doors, equipment compartment doors, or cabinet doors. If only isolated word parsing is relied upon, it is very easy to lead to incorrect lock ID mapping. Therefore, this application extracts the first task description word from the sequence distribution of the task description words and uses the other parts of the sequence distribution of the task description words as the context information of the first task description word. Specifically, by treating each word in the text sequence as the parsing focus in turn (e.g., processing "A computer room" first, then "number 3"), while retaining the remaining words as a dynamic context pool, the system can simulate the progressive cognitive process of human language understanding. For example, when parsing "inspecting the front door of rack 3 in computer room A", after the first descriptive word "inspection" is extracted, the remaining sequence "front door of rack 3 in computer room A" serves as its context, helping to determine that "inspection" belongs to the operation type field. When the focus shifts to "computer room A", the subsequent "front door of rack 3" provides spatial modification information, thereby accurately locating the topological position of the target lock. This step-by-step focusing mechanism allows the semantics of each word to be dynamically calibrated in both local and global contexts.

[0025] More specifically, in step S1113, word embedding encoding is performed on the first task description word to obtain a semantic embedding encoding vector for the first task description word. It should be understood that task description words often carry multi-dimensional semantic information, but their original text form is difficult for a computer to directly parse into operable permission parameters. For example, the term "front door" in a work order may correspond to the location attribute, structural attribute (such as an electronic lock or a mechanical lock), or functional attribute (such as an emergency exit sign) of a physical lock. Therefore, in order to understand and process the semantics of words based on vector calculation, and to provide a foundation for subsequent semantic analysis and processing, this application performs word embedding encoding on the first task description word to obtain a semantic embedding encoding vector for the first task description word. In particular, in one example of this application, text convolutional encoding can be used to perform word embedding encoding on the first task description word to more accurately represent the semantic information of the description word.

[0026] More specifically, in step S1114, the contextual semantic information of the first task description word is semantically encoded to obtain the contextual semantic information encoding vector of the target task description word. Correspondingly, considering the complex semantic relationships between words in a task description, isolated parsing of a single word often fails to reconstruct its true business intent. For example, the word "3" in a work order may refer to a rack number, equipment serial number, or room identifier, and its specific meaning is highly dependent on the context. If the word is parsed without context, the system may incorrectly map "3" to the server room area code instead of the target rack lock ID, leading to an offset in the scope of subsequent permission granting. Therefore, this application performs contextual semantic encoding on the contextual information of the first task description word to incorporate surrounding semantic information related to the first task description word into the consideration, thereby more comprehensively and accurately grasping the specific meaning of the description word in the entire task context and obtaining the contextual semantic information encoding vector of the target task description word. In particular, text convolutional encoding can also be used to perform contextual semantic encoding on the contextual information of the first task description word.

[0027] More specifically, in step S1115, based on the target task description word context semantic information encoding vector, the semantic embedding encoding vector of the first task description word is subjected to granular interactive iterative semantic expression enhancement to obtain the semantic context enhancement encoding vector of the first task description word. Furthermore, considering that the semantic ambiguity and context dependence of task description words constitute the main obstacles to key field extraction, due to the inherent polysemy and diversity of expressions in natural language, isolated analysis of individual words often makes it difficult to accurately determine their business attributes. For example, "No. 3" may refer to a cabinet number, equipment sequence, or work step, while "maintenance" may be associated with different lock operation permissions in different contexts. Traditional mapping methods based on static lexicons cannot capture such dynamic semantic relationships. When work order text contains cross-system terminology differences (such as recording "distribution box" as "power terminal") or omits key modifiers (such as only labeling "front door" without specifying the machine room area), it is easy to lead to lock ID matching errors. Therefore, it is necessary to establish a relational reasoning mechanism that can dynamically integrate the semantic ontology of words with the context environment to eliminate the permission mapping risks caused by semantic ambiguity. Based on this, in this application, the semantic embedding encoding vector of the first task description word is subjected to granular interactive iterative semantic expression enhancement based on the target task description word context semantic information encoding vector to obtain the semantic context enhancement encoding vector of the first task description word.

[0028] More specifically, in this embodiment, step S1115, based on the target task descriptor context semantic information encoding vector, performs granular interactive iterative semantic expression enhancement on the first task descriptor semantic embedding encoding vector to obtain the first task descriptor semantic context enhanced encoding vector, including: S1115-1, performing local latent semantic feature interaction on the target task descriptor context semantic information encoding vector and the first task descriptor semantic embedding encoding vector to obtain a set of target task guidance-first task descriptor local semantic feature interaction encoding vectors; S1115-2, based on the feature distribution characteristics of each target task guidance-first task descriptor local semantic feature interaction encoding vector in the set of target task guidance-first task descriptor local semantic feature interaction encoding vectors, S1115-3: Determine the local semantic attention weights of the local semantic feature interaction encoding vectors of the target task-guided first task descriptor to obtain a set of local semantic attention weights of the target task-guided first task descriptor; S1115-4: Based on the set of local semantic attention weights of the target task-guided first task descriptor, perform weighted strengthening on the set of local semantic feature interaction encoding vectors of the target task-guided first task descriptor to obtain a set of enhanced encoding vectors of local semantic feature interaction of the target task-guided first task descriptor; S1115-5: Input the set of enhanced encoding vectors of local semantic feature interaction of the target task-guided first task descriptor into a semantic aggregator based on a forward LSTM model to obtain the semantic context enhanced encoding vector of the first task descriptor.

[0029] Specifically, in this embodiment, step S1115-1, which involves performing local latent semantic feature interaction on the target task description context semantic information encoding vector and the first task description semantic embedding encoding vector to obtain a set of target task guidance-first task description local semantic feature interaction encoding vectors, includes: Local one-dimensional convolutional latent feature extraction is performed on the target task descriptor context semantic information encoding vector and the first task descriptor semantic embedding encoding vector to obtain a set of target task descriptor local latent context semantic information encoding vectors and a set of first task descriptor local latent semantic embedding encoding vectors. This process is expressed by the following formula:

[0030]

[0031] in, It is the encoding vector of the contextual semantic information of the target task descriptor. It is the semantic embedding encoding vector of the first task description word. For local one-dimensional convolutional latent feature extraction, The length of the one-dimensional convolution kernel. , , and These are the 1st, 2nd, and 3rd elements in the set of encoding vectors for the local latent context semantic information of the target task descriptor. The and the first Encoding vectors of local implicit contextual semantic information for each target task descriptor. , , and These are the 1st, 2nd, and 3rd local latent semantic embedding encoding vectors of the first task description words. The and the first The local latent semantic embedding encoding vector of the first task description words. yes and The number of vectors in the middle, and and Same length; The set of encoding vectors for the local implicit context semantic information of the target task descriptor and the set of encoding vectors for the local implicit semantic embedding of the first task descriptor are respectively subjected to semantic feature interaction to obtain the set of encoding vectors for the interaction of the target task guidance and the local semantic features of the first task descriptor. This process is expressed by the formula:

[0032] in, It is a dot product by position. It is added based on the position point. It is subtracted based on position. It is a cascading operation. The set of local semantic feature interaction weight matrices is the first... Local semantic feature interaction weight matrix, The set of local semantic feature interaction bias vectors is the first... Local semantic feature interaction bias vectors It is the first task-guided task description word local semantic feature interactive encoding vector set in the set of target task guidance - first task description word local semantic feature interactive encoding vector. The first task description word local semantic feature interaction encoding vector guided by the target task - the first task description word.

[0033] It is understandable that the semantic complexity of job task descriptions stems from the multidimensional correlation and contextual dependence of business terms. For example, the "3" in "Inspect rack 3 in computer room A" in a work order may exist simultaneously in the equipment numbering system of multiple computer rooms, and the "inspection" action may correspond to different levels of lock operation permissions. If the entire semantic encoding vector is directly processed globally, it is easy to cause deviations in the extraction of key elements due to feature confusion. Based on this, local one-dimensional convolutional latent feature extraction is performed on the context semantic information encoding vector of the target task description word and the semantic embedding encoding vector of the first job task description word, respectively. This is to use local one-dimensional convolution to perform hierarchical feature extraction on the context semantic vector and the core word embedding vector to obtain the set of local latent context semantic information encoding vectors of the target task description word and the set of local latent semantic embedding encoding vectors of the first job task description word.

[0034] Correspondingly, the semantic parsing of job task descriptions needs to address the dynamic relationship between core elements and their context. For example, when a work order contains "inspecting distribution cabinet No. 5 in area B," relying solely on the isolated semantics of "No. 5" may not accurately locate the lock ("No. 5" in different areas may point to different devices). If only the overall semantics are processed, the implicit relationship between "area B" as a spatial qualifier and "distribution cabinet" as a device type is easily overlooked. To address this, the local implicit contextual semantic information encoding vectors of each set of target task description words and the local implicit semantic embedding encoding vectors of the first job task description words are subjected to semantic feature interaction to construct a dynamic relationship model between core words and context at the level of local implicit features, resulting in a set of target task guidance-first job task description word local semantic feature interaction encoding vectors.

[0035] Specifically, in step S1115-2, based on the feature distribution characteristics of each target task guidance-first task description word local semantic feature interaction encoding vector in the set of target task guidance-first task description word local semantic feature interaction encoding vectors, the local semantic attention weights of each target task guidance-first task description word local semantic feature interaction encoding vector are determined to obtain the set of target task guidance-first task description word local semantic attention weights. This process is expressed by the formula:

[0036] in, yes The Middle 1 eigenvalue, To calculate the square of the Euclidean norm of a vector. yes The number of eigenvalues ​​in the middle. yes function, It is the first task-guided set of local semantic attention weights for the first task description words. Local semantic attention weights for the first task description word guided by the target task.

[0037] It is understandable that the semantic weight distribution of task descriptions exhibits significant imbalance. For example, the modifier "urgent" in "urgent repair of pump room 7 in area F" in a work order may imply a higher priority of timeliness for permissions, while the equipment type characteristic of "pump room" needs to be associated with the lock control strategy of a specific area. If the interaction encoding vector is equalized, it may weaken key semantic elements (such as the need for time window compression due to "urgent") while amplifying the impact of redundant information (such as the general verb "repair"), resulting in the generated permission rules failing to accurately reflect the urgency and specificity of business needs. To achieve adaptive importance assessment of interaction features, this application determines the local semantic attention weights of the interaction encoding vectors of the local semantic features of the first task description word guided by each target task based on the feature distribution characteristics of the interaction encoding vectors of the local semantic features of the first task description word guided by each target task, thereby obtaining a set of local semantic attention weights of the first task description word guided by the target task. By focusing and amplifying key interaction features through attention weights, the system can accurately capture potential risk elements in business requirements (such as special permission rules triggered by modifiers like "high pressure" and "hazardous chemicals"), while weakening the interference of non-core semantics.

[0038] Specifically, in this embodiment, step S1115-3, based on the set of local semantic attention weights of the target task-guided first task descriptor, weights and strengthens the set of local semantic feature interaction encoding vectors of the target task-guided first task descriptor to obtain the set of enhanced local semantic feature encoding vectors of the target task-guided first task descriptor, including: The set of local semantic attention weights for the first task descriptor guided by the target task is adjusted using a spatial constraint paradigm to obtain the set of adjusted local semantic attention weights for the first task descriptor guided by the target task. This process is expressed by the following formula: , ,

[0039]

[0040]

[0041]

[0042] in, It is the first task-guided set of translational covariance intensity coefficients in the first task description term. The first task description term, guided by the target task, is the translational covariance intensity coefficient. It is the first task-guided set of translational difference intensity coefficients for the first task description. The intensity coefficient of translational difference in the first task description, guided by the target task. It is the first task-guided set of frequency domain oscillation coupling coefficients in the first task description. The first task, guided by a specific objective, describes the word frequency domain oscillation coupling coefficient. Based on the natural constant The value of the logarithmic function with base 0. It is the first task-guided set of periodic feature compensation values ​​for the first task descriptor. Compensation value for periodic features of the first task description words guided by the target task. It is the first task-guided set of periodic phase auxiliary values ​​in the first task description. The periodic phase auxiliary value of the first task description, guided by the target task. and They are and The corresponding weighting coefficients, It is the first task-guided set of local semantic attention correction weights for the first task description words. Local semantic attention correction weights for the first task description words guided by the target task; Based on the set of local semantic attention correction weights for the first task description words guided by the target task, the set of interactive encoding vectors of local semantic features of the first task description words guided by the target task is weighted and strengthened to obtain the set of interactive strengthened encoding vectors of local semantic features of the first task description words guided by the target task. This process is expressed by the formula:

[0043]

[0044] in, , , and These are the 1st, 2nd, and 3rd elements in the set of local semantic feature interaction reinforcement encoding vectors of the first task-guided task description words. The and the first The first task-guided local semantic feature interaction reinforcement encoding vector of the target task description words. It is a set of target task-guided local semantic feature interaction reinforcement encoding vectors of the first task description words.

[0045] As mentioned above, when calculating the interaction encoding vectors of the local semantic features of each target task guide-first task descriptor, a spatial effect deconstruction mechanism of the interaction paradigm needs to be incorporated. Specifically, the interaction features between the encoding vectors of the local implicit contextual semantic information of the target task descriptor and the encoding vectors of the local implicit semantic embedding of the first task descriptor (such as...) should be incorporated. , , (etc.) are regarded as mathematical representations of different spatial constraint paradigms. These operations essentially define differentiated spatial collaborative constraint associations within the interaction space.

[0046] To enhance the optimization effect of paradigm adaptability aids on the local semantic attention weights of the first task description words guiding the target task, it is recommended to adjust the weights based on the spatial constraint paradigm of interaction features. Specifically, , This can be analyzed as a translation effect, where the direction of its characteristic gradient remains collinear with the direction of spatial interaction; and This corresponds to the oscillatory effect, where the direction of the characteristic gradient is orthogonal to the interaction direction. Based on this, a characteristic statistic is defined. , , The oscillation effect induces regional periodic coupling in the translation domain feature representation. This phenomenon can be quantified through a periodic feature compensation mechanism:

[0047] This formula shows that when the translation domain features are characterized When enhanced, the oscillation effect It exhibits nonlinear growth with increasing logarithmic dimension. Simultaneously, the oscillation effect also induces periodic phase-assisted evolution, the mathematical form of which is:

[0048] Finally, through collaborative optimization of spatial constraint paradigms, a target task-guided local semantic attention weight correction equation for the first task descriptor is constructed:

[0049] This dual correction mechanism effectively enhances the dynamic correlation between different spatial constraint paradigms, and strengthens the generalization ability and numerical stability of the local semantic attention weight calculation of the first task description word guided by the target task through paradigm adaptability.

[0050] Subsequently, based on the set of local semantic attention correction weights for the first task descriptor guided by the target task, the set of interactive encoding vectors for the local semantic features of the first task descriptor guided by the target task is weighted and strengthened. Through this attention-weighted strengthening, the system can adaptively increase the decision weights of key business elements, resulting in the set of enhanced encoding vectors for the interactive semantic features of the first task descriptor guided by the target task. For example, the risk identification features in "high-risk operations" can be dynamically bound to the enhanced audit strategy, or the member number feature in "multi-person collaboration" can be transformed into parallel authorization rules.

[0051] Specifically, in step S1115-4, the set of local semantic feature interaction-enhanced encoding vectors of the first task description word guided by the target task is input into a semantic aggregator based on a feedforward LSTM model to obtain the semantic context-enhanced encoding vector of the first task description word. This process is expressed by the formula:

[0052] in, It is forward LSTM encoding. It is the semantic context reinforcement encoding vector of the first task description word.

[0053] Finally, the set of locally semantic feature interaction-enhanced encoding vectors of the first task description word guided by the target task is input into a semantic aggregator based on a forward LSTM model to obtain the semantic context-enhanced encoding vector of the first task description word. Through the sequential reasoning capability of LSTM, the system can capture long-span semantic dependencies (such as the association between "quarterly maintenance" and the cross-month lock inspection plan), while ensuring policy compatibility between local features (e.g., dynamically binding the spatial features of "N-zone control room" with the permission rules of "two-person operation"). This deep aggregation mechanism ensures that the generated permission data package accurately reflects the details of business requirements while maintaining the global consistency of the security policy system, effectively avoiding permission conflicts or blind spots in control caused by isolated processing of semantic elements.

[0054] More specifically, in step S1116, the semantic context-enhanced encoding vector of the first task description word is input into a classifier-based type determiner to determine the type label of the key field to which the first task description word belongs. It should be understood that when processing task information, simply obtaining the semantic context-enhanced encoding vector is insufficient for the smart lock system to manage and utilize task information. Further structuring of this semantic information is needed to clarify the role and category of each description word in the task. The classifier has been pre-trained with a large number of samples, learning the unique semantic patterns and features of different types of key fields (responsible person / member, planned start and end time, task location / equipment description, etc.). When the semantic context-enhanced encoding vector of the first task description word is input into the classifier, the classifier first parses the vector, extracts the key features from the vector, and compares these features with the feature templates of various key fields it has already learned. For example, if the vector contains specific vocabulary features related to the names of people, and these features have a high degree of matching with the template features of the key field of the "person in charge / member" type, the classifier will tend to classify the descriptor as belonging to the "person in charge / member" category; if the vector presents time format-related features, such as vocabulary features representing year, month, day, hour, minute, etc., which match the feature template of "planned start and end time", it will be classified as the "planned start and end time" type.

[0055] Specifically, in step S112, the responsible person / member is mapped to a user ID. It should be understood that in a smart lock system, using natural language to describe the responsible person / member information (such as "Zhang San," "Li Si's team," etc.) is not conducive to accurate and efficient identification and management by the computer system. A user ID, however, is a unified and standardized identifier within the system, possessing uniqueness and determinism. Mapping the responsible person / member to a user ID enables the system to process personnel information in a standardized way, facilitating accurate identification and management of personnel in subsequent stages such as permission allocation and operation recording. Specifically, when the backend management platform receives the task information, it extracts the relevant information about the responsible person / member. This information is usually in natural language form, such as "Zhang San," "Li Si's team," etc. Then, the platform processes this naturally language-described responsible person / member information according to a preset mapping rule base. The mapping rule base pre-sets various rules for converting natural language into system-recognizable information. The platform parses the extracted information, identifying and matching key content such as names and titles, and using the rule base to find the corresponding mapping logic. For example, if the rule base specifies that "Zhang San" corresponds to user ID "001" in the system, then when "Zhang San" is extracted as the person in charge, a preliminary association can be made based on this rule. Then, the platform performs a real-time search in the user database. The user database stores detailed information about all users in the system, including unique user IDs. Based on the preliminary association determined by the mapping rule base, the platform accurately searches for the corresponding real-time ID in the user database. In this way, the person in charge / member information described in natural language is accurately mapped to a unified, standardized, unique, and deterministic user ID within the system.

[0056] Specifically, in step S113, the work location / equipment description is mapped to the ID of the target smart lock. Correspondingly, each smart lock has a unique ID as its identifier, which is the basis for the system to manage and control the locks. However, the work location / equipment description is usually in natural language (e.g., "Server rack 5 in computer room A," "backup equipment room in the power distribution room," etc.), which is difficult for computer systems to directly use for accurate identification and operation of smart locks. Therefore, it is necessary to convert the work location / equipment description into the ID of the target smart lock so that the system can accurately locate and manage the corresponding smart lock. Specifically, firstly, the work location / equipment description is extracted from the work task information. These descriptions are generally presented in natural language, such as "Server rack 5 in computer room A," "backup equipment room in the power distribution room," etc. Then, the mapping is completed using preset mapping rules and a lock space coding system. The extracted natural language description is parsed, and key information is decomposed. For example, for "front door of rack 3 in computer room A," key elements such as "computer room A," "3," "rack," and "front door" are identified. Next, based on pre-defined mapping rules, these key elements are matched with information in the lock space coding system. These mapping rules may be based on factors such as room number, device serial number, and specific location. For example, "Room A" might correspond to the "A" part of the code, "No. 3" to "03," and "Front door of the cabinet" to "FRONT." The system then searches the lock space coding system, which contains the unique ID of all smart locks and their corresponding location or device description information. By matching the information according to the mapping rules, the system can accurately locate the ID of the corresponding target smart lock, such as LOCK_A03_FRONT.

[0057] Specifically, in step S114, the planned start and end times are converted into an effective time window for permissions. It should be understood that the planned start and end times are the time range of the task expressed in natural language or a conventional time format, such as "October 15, 2024, 9:00 AM to 5:00 PM". For the permission management of a smart lock system, this expression method is not precise enough and is not convenient for the system to process. Specifically, when converting the planned start and end times into an effective time window for permissions, the planned start and end times are first extracted from the task information, which is usually presented in natural language or a conventional time format, such as "October 15, 2024, 9:00 AM to 5:00 PM". Next, the extracted time is parsed, separating the time elements such as year, month, day, hour, and minute. Then, these time elements are reorganized according to a format that the system can recognize and process, converting them into an effective time window for permissions. During the conversion process, the time is standardized to conform to the time parameter format of the system's access control. For example, it is converted into a range represented by seconds or a specific timestamp, so that the system can accurately control and manage permissions. This ensures that the smart key of the relevant operator can only open the target smart lock within the specified time window, and cannot be operated outside the time range.

[0058] Specifically, step S115 involves generating the authorization data packet based on the user ID, the target smart lock ID, and the valid time window of the permission. Specifically, in this embodiment, generating the authorization data packet based on the user ID, the target smart lock ID, and the valid time window of the permission includes: generating dynamic access permission rules based on the user ID, the target smart lock ID, and the valid time window of the permission; and packaging the dynamic access permission rules into the authorization data packet. Accordingly, the smart lock system needs to ensure that only authorized users can access specific locks within a specified time. The user ID uniquely identifies the operator, the target smart lock ID clarifies the operable object, and the valid time window of the permission limits the time range of the operation. By combining these three key elements to generate dynamic access permission rules, precise control over smart lock access can be achieved, preventing unauthorized personnel from accessing the lock at inappropriate times and improving system security. Specifically, firstly, the three types of key information—user ID, target smart lock ID, and valid time window of the permission—are integrated. Based on this information, dynamic access permission rules are generated. It associates and combines the user ID, the target smart lock ID, and the valid time window of the permission, clearly defining that within the valid time window, only personnel with a specific user ID can operate on the corresponding target smart lock. For example, the generated rule might be "User with user ID 123 has the right to access the smart lock with target smart lock ID 456 from 9:00 to 17:00 on May 1, 2025." This clearly defines the conditions for user access to smart locks, ensuring that only qualified personnel can perform the corresponding operations within the specified time, improving system security and management accuracy. After generating dynamic access permission rules, they are packaged into authorization data packets. Since dynamic access permission rules contain multiple parameters and conditions, direct transmission and storage would increase the complexity of data processing. Therefore, this relevant information is integrated and encapsulated according to a specific format to form a unified data unit, namely the authorization data packet. This facilitates efficient transmission and storage between various components of the smart lock system (such as the backend management platform, key management machine, and smart keys), ensuring the smooth operation of the entire smart lock system and achieving precise control over smart lock access.

[0059] In summary, step S110 clarifies that it achieves efficient transformation of business requirements into execution instructions by constructing an automated mapping mechanism between work tasks and physical lock permissions. The backend management platform uses natural language processing technology to parse work order information, accurately extracting key elements such as the person in charge, work time, and equipment location. Based on a preset rule base, it automatically matches these elements to the corresponding user ID, lock code, and time window parameters, generating an encrypted authorization data packet and pushing it to the key management machine. This process realizes the automatic conversion of heterogeneous data source information into standardized permission instructions, avoiding lock identification deviations caused by manual mapping, improving the accuracy and security of permission allocation, and ensuring accurate interaction between smart keys and locks.

[0060] In step S120, after the operator successfully authenticates their identity at the key management machine, the key management machine sends the corresponding authorization data packet to the smart key. That is, by authenticating on the key management machine, the operator's identity information can be verified, determining whether they are an authorized person. Only when identity authentication is successful does it mean that the operator has legitimate operating authority. Sending the authorization data packet at this point ensures that subsequent operations on the smart lock are performed within the authorized scope, effectively preventing unauthorized personnel from using the smart key to operate the smart lock and maintaining system security.

[0061] In step S130, the smart key interacts with the target smart lock. The smart key verifies the authorization information; if successful, it unlocks the target smart lock. After unlocking, the smart key reads the current physical state of the target smart lock. It should be understood that one of the main purposes of a smart lock is to prevent unauthorized access. The smart key's authorization verification ensures that only a smart key with valid permissions can unlock the smart lock. The authorization information includes key elements such as the user ID, the target smart lock ID, and the valid authorization time window. Verifying this information determines whether the unlocking request conforms to pre-set authorization rules, thus ensuring the security of the area or device protected by the smart lock. Specifically, when an operator arrives at the location of the target smart lock with the smart key, the interaction between the smart key and the target smart lock begins. The smart key first retrieves the authorization information from its own storage area, including key elements such as the user ID, the target smart lock ID, and the valid authorization time window. The smart key compares the target smart lock ID it carries with the actual lock ID to be opened to confirm a match. Simultaneously, it checks if the current time is within the valid access window and verifies the user ID's legitimacy. If all verifications pass, the smart key sends an unlocking command to the target smart lock, driving it to open. After unlocking, the smart key immediately begins reading the target smart lock's current physical state. The smart key integrates sensor modules capable of sensing the lock's status. These modules can detect whether the lock is open, closed, or has experienced abnormal tampering. These sensors collect the lock's physical state information and convert it into digital signals, storing it in the smart key's internal storage.

[0062] In step S140, the smart key uses its built-in real-time communication module to send the current physical state of the target smart lock to the key management unit within communication range. Specifically, the real-time communication module includes, but is not limited to, WAPI, Bluetooth, LoRa, and NB-IoT modules. That is, to achieve dynamic security control of the smart lock, the backend management platform needs to obtain the lock's physical state information in real time, such as whether it is open, closed, or subjected to abnormal prying. As a device that directly interacts with the lock, the smart key can obtain this state information after completing the unlocking action and send it out through its built-in real-time communication module, meeting the needs of real-time monitoring and enabling the backend management platform to promptly detect and handle potential security issues. Through the real-time communication module, the smart key can promptly send the current physical state of the target smart lock to the key management unit, which then forwards it to the backend management platform. This allows the backend management platform to grasp the lock's state changes in real time, achieving state synchronization with a second-level delay, shortening the security response window, and enabling rapid response in the event of unauthorized operation or forced entry.

[0063] Specifically, the real-time communication module built into the smart key plays a crucial role. This module supports multiple communication methods, such as WAPI, Bluetooth, LoRa, or NB-IoT, with the specific method chosen depending on the actual scenario and device configuration. The smart key first encodes the physical state data of the target smart lock, converting lock open / close and abnormal tampering status information into a digital signal format that the communication module can recognize and transmit, ensuring data accuracy and compatibility. Next, the real-time communication module begins operating, automatically searching for key management units within communication range. Taking Bluetooth communication as an example, the smart key's Bluetooth module emits a broadcast signal at a specific frequency. Upon receiving the signal, the key management unit's Bluetooth receiver negotiates a device connection. Once established, the smart key sends the encoded data frame by frame to the key management unit according to the established communication protocol. During transmission, a data verification and retransmission mechanism is employed to ensure data reliability. If the key management unit detects a verification error while receiving data, it sends a retransmission request to the smart key. Upon receiving the request, the smart key retransmits the corresponding data frame.

[0064] In step S150, the key management device sends the current physical status of the target smart lock to the backend management platform. The backend management platform parses the current physical status of the target smart lock and updates the real-time status of the corresponding target smart lock in the database. Specifically, the key management device, acting as an intermediary between the smart key and the backend management platform, collects the current physical status information of the target smart lock sent by the smart key. Sending this information to the backend management platform enables centralized data management. The backend management platform possesses powerful data storage and processing capabilities, allowing for unified management and analysis of large amounts of lock status data, avoiding the management inconvenience and security risks caused by data dispersion. The key management device sending lock status information to the backend management platform enables the platform to aggregate status data from various smart locks. Analysis of this data allows for the acquisition of an overall security posture, such as statistical analysis of the usage frequency of smart locks in a specific area and the probability of abnormal situations. These analytical results help managers understand the system's operational status and provide data support for optimizing system management strategies.

[0065] Specifically, after receiving the status data from the smart key, the key management unit first performs preliminary verification and organization. It checks the integrity and accuracy of the data to ensure no data loss or errors, guaranteeing the reliability of the data transmitted to the backend management platform. Subsequently, the key management unit sends the organized current physical status data of the target smart lock through the established communication connection with the backend management platform. This communication connection can be based on a wired network, such as Ethernet, or a wireless network, such as Wi-Fi, 4G, or 5G, depending on the specific application scenario and deployment environment. When sending data, the key management unit encapsulates it according to the communication protocol and data format pre-defined by the backend management platform, ensuring the platform can correctly identify and process the data. Upon receiving the current physical status data of the target smart lock from the key management unit, the backend management platform immediately initiates the data parsing process. Since this data is transmitted in a specific format, the backend management platform needs to unpack it according to the corresponding parsing rules. It separates and identifies the various fields in the data, extracting key information such as the lock's open or closed status and whether any abnormal prying has occurred. For complex data, further decoding and conversion may be required to transform it into a format that the system can understand and process. After data parsing, the backend management platform updates the real-time status of the corresponding target smart lock in the database based on the parsed current physical status information. The database stores detailed information about all smart locks, including their basic information and real-time status. The backend management platform locates the corresponding record in the database based on the unique identifier of the target smart lock. Then, it replaces the original status data with the newly parsed physical status information to ensure that the information in the database is consistent with the actual lock status. To ensure the accuracy and consistency of data updates, the backend management platform typically uses a transaction processing mechanism when updating the database. If any errors or anomalies occur during the update process, the system will automatically perform a rollback operation to ensure that there are no data inconsistencies in the database. After the update is complete, administrators can view the latest status of the target smart lock in real time through the backend management platform to promptly identify and address any potential security issues.

[0066] In summary, the dynamic security management method based on smart locks, as described in this application, is as follows: First, the backend management platform generates an authorization data packet based on the task and sends it to the key management machine. After the operator is authenticated, the smart key obtains authorization and interacts with the target lock. Upon successful verification, the smart key drives the lock to open and reads its physical status after unlocking. The status information is then sent to the key management machine via a real-time communication module and finally uploaded to the backend platform for parsing and database updates. This solution achieves closed-loop management of access control, operation verification, and status feedback, improving security and management efficiency, ensuring real-time synchronization and traceability of lock status, and effectively supporting remote monitoring and risk warning.

[0067] Figure 4 This is a block diagram of a dynamic security management system based on a smart lock according to an embodiment of this application. Figure 4 As shown, the dynamic security management system 100 based on smart locks according to an embodiment of this application includes: an authorization data packet generation and sending module 110, used to generate authorization data packets based on job task information through a background management platform and send them to a key management machine; an identity authentication module 120, used to send the corresponding authorization data packet to the smart key after the operator successfully authenticates their identity at the key management machine; a current physical state reading module 130, used to interact with the target smart lock using the smart key, wherein the smart key verifies the authorization information, and if successful, drives the target smart lock to open, and after completing the unlocking action, the smart key reads the current physical state of the target smart lock; a sending module 140, used to send the current physical state of the target smart lock to the key management machine within communication range through the smart key's built-in real-time communication module; and a parsing module 150, used to send the current physical state of the target smart lock to the background management platform through the key management machine, and the background management platform parses the current physical state of the target smart lock and updates the real-time state of the corresponding target smart lock in the database.

[0068] Those skilled in the art will understand that the specific operations of each step in the above-described dynamic security control system based on smart locks have been referenced above. Figures 1 to 3 The dynamic security management method based on smart locks has been described in detail, and therefore, its repeated description will be omitted.

[0069] In particular, this application also provides a device comprising: a processor and a memory, wherein the memory stores a computer-executable program, and when the computer-executable program is executed by the processor, it implements the above-described dynamic security management method based on a smart lock.

[0070] In summary, the above detailed description is intended to be illustrative rather than restrictive, and it should be understood that these embodiments are for illustrative purposes only and not for limiting the scope of protection of the invention. After reading the description of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent changes and modifications also fall within the scope defined by the claims of this invention.

Claims

1. A dynamic security management method based on smart locks, characterized in that, include: Based on the job task information, the back-end management platform generates an authorization data packet and sends it to the key management machine; Once the operator successfully authenticates their identity at the key management machine, the key management machine will send the corresponding authorization data packet to the smart key; The smart key interacts with the target smart lock. The smart key verifies the authorization information. If the verification is successful, the smart key drives the target smart lock to open. After the unlocking action is completed, the smart key reads the current physical state of the target smart lock. The smart key uses its built-in real-time communication module to send the current physical status of the target smart lock to the key management unit within the communication range; The key management device sends the current physical status of the target smart lock to the back-end management platform. The back-end management platform parses the current physical status of the target smart lock and updates the real-time status of the corresponding target smart lock in the database.

2. The dynamic security management method based on smart locks according to claim 1, characterized in that, The real-time communication module is a WAPI module, Bluetooth module, LoRa module, or NB-IoT module.

3. The dynamic security management method based on smart locks according to claim 1, characterized in that, Based on the job task information, the backend management platform generates an authorization data packet and sends it to the key management machine, including: Key fields are extracted from the task information, including the person in charge / members, planned start and end times, and description of the task location / equipment. Map the responsible person / member to a user ID; Map the work location / equipment description to the ID of the target smart lock; Convert the planned start and end times into an authorized valid time window; The authorization data packet is generated based on the user ID, the target smart lock ID, and the permission validity time window.

4. The dynamic security management method based on smart locks according to claim 3, characterized in that, Based on the user ID, the target smart lock ID, and the permission validity time window, the authorization data packet is generated, including: Based on the user ID, the target smart lock ID, and the permission validity time window, generate dynamic access permission rules; The dynamic access permission rules are packaged into the authorization data package.

5. The dynamic security management method based on smart locks according to claim 3, characterized in that, Key fields are extracted from the task information, including the person in charge / members, planned start and end times, and a description of the work location / equipment, including: The task information is segmented into words to obtain the sequence distribution of task description words; Extract the first task description word from the sequence distribution of the task description words, and use the other parts of the sequence distribution of the task description words as the context information of the first task description word; The first task description word is embedded and encoded to obtain the semantic embedding encoding vector of the first task description word; The contextual semantic information of the first task descriptor is encoded to obtain the contextual semantic information encoding vector of the target task descriptor; Based on the target task descriptor context semantic information encoding vector, the first task descriptor semantic embedding encoding vector is subjected to granular interactive iterative semantic expression enhancement to obtain the first task descriptor semantic context enhanced encoding vector. The semantic context-enhanced encoding vector of the first task descriptor is input into a classifier-based type determiner to determine the type label of the key field to which the first task descriptor belongs.

6. The dynamic security management method based on smart locks according to claim 5, characterized in that, Based on the target task descriptor context semantic information encoding vector, the first task descriptor semantic embedding encoding vector is subjected to granular interactive iterative semantic expression enhancement to obtain the first task descriptor semantic context enhanced encoding vector, including: The target task description context semantic information encoding vector and the first task description semantic embedding encoding vector are subjected to local implicit semantic feature interaction to obtain a set of target task guidance-first task description local semantic feature interaction encoding vectors; Based on the feature distribution characteristics of each target task guidance-first task description word local semantic feature interaction encoding vector in the set of target task guidance-first task description word local semantic feature interaction encoding vectors, the local semantic attention weight of each target task guidance-first task description word local semantic feature interaction encoding vector is determined to obtain the set of target task guidance-first task description word local semantic attention weights; Based on the set of local semantic attention weights of the first task description word guided by the target task, the set of local semantic feature interaction encoding vectors of the first task description word guided by the target task is weighted and strengthened to obtain the set of local semantic feature interaction strengthening encoding vectors of the first task description word guided by the target task. The set of local semantic feature interaction-enhanced encoding vectors of the first task description word guided by the target task is input into a semantic aggregator based on a forward LSTM model to obtain the semantic context-enhanced encoding vector of the first task description word.

7. The dynamic security management method based on smart locks according to claim 6, characterized in that, The target task description context semantic information encoding vector and the first task description semantic embedding encoding vector are subjected to local latent semantic feature interaction to obtain a set of target task guidance-first task description local semantic feature interaction encoding vectors, including: Local one-dimensional convolutional latent feature extraction is performed on the target task descriptor context semantic information encoding vector and the first task descriptor semantic embedding encoding vector to obtain the set of target task descriptor local latent context semantic information encoding vectors and the set of first task descriptor local latent semantic embedding encoding vectors. The set of encoding vectors for the local implicit context semantic information of the target task descriptor and the set of encoding vectors for the local implicit semantic embedding of the first task descriptor are respectively subjected to semantic feature interaction to obtain the set of encoding vectors for the interaction of the target task guidance and the local semantic feature of the first task descriptor.

8. The dynamic security management method based on smart locks according to claim 7, characterized in that, Based on the set of local semantic attention weights of the first task description words guided by the target task, the set of interactive encoding vectors of local semantic features of the first task description words guided by the target task is weighted and strengthened to obtain the set of interactive strengthened encoding vectors of local semantic features of the first task description words guided by the target task, including: The set of local semantic attention weights of the first task description words guided by the target task is modified by a spatial constraint paradigm to obtain the set of modified local semantic attention weights of the first task description words guided by the target task. Based on the set of local semantic attention correction weights of the first task description word guided by the target task, the set of local semantic feature interaction encoding vectors of the first task description word guided by the target task is weighted and strengthened to obtain the set of local semantic feature interaction strengthening encoding vectors of the first task description word guided by the target task.

9. A dynamic security control system based on smart locks, characterized in that, include: The authorization data packet generation and sending module is used to generate authorization data packets based on job task information through the back-end management platform and send them to the key management machine; The identity authentication module is used so that when the operator successfully authenticates his identity at the key management machine, the key management machine will send the corresponding authorization data packet to the smart key. The current physical state reading module is used to interact with the target smart lock using a smart key. The smart key verifies the authorization information. If successful, it drives the target smart lock to open. After the unlocking action is completed, the smart key reads the current physical state of the target smart lock. The sending module is used to send the current physical status of the target smart lock to the key management machine within the communication range via the smart key's built-in real-time communication module. The parsing module is used to send the current physical status of the target smart lock to the back-end management platform through the key management machine. The back-end management platform parses the current physical status of the target smart lock and updates the real-time status of the corresponding target smart lock in the database.

10. An apparatus comprising: A processor and a memory, wherein the memory stores a computer-executable program, which, when executed by the processor, implements the dynamic security management method based on a smart lock as described in any one of claims 1-8.

Images