Intelligent key cabinet abnormal event real-time notification method and system

Abstract

The application relates to the technical field of identity recognition, and discloses a smart key cabinet abnormal event real-time notification method and system, which comprises the following steps: determining a key identifier set that can be operated by a user on a smart key cabinet according to identity authentication information; collecting radio frequency signals and use ranges of each key identifier in real time; generating inventory change time sequence data of the existence state of each cabinet door to a key and generating spatial trajectory jump data of each key; aligning the spatial trajectory jump data with the opening and closing state data of each cabinet door on a time axis, correlating and analyzing the key identifier set and the state alignment data, and obtaining abnormal event information; filtering the abnormal event information, selecting a transmission channel of the filtered abnormal event information, and obtaining a target transmission channel; triggering a notification mode corresponding to the target transmission channel, and notifying the abnormal event information to a remote management platform by using the notification mode. The application can improve the accuracy of smart key cabinet abnormal event real-time notification.

Description

Technical Field

[0001] This invention relates to the field of identity recognition technology, and in particular to a method and system for real-time notification of abnormal events in a smart key cabinet. Background Technology

[0002] Smart key cabinets, as important asset management devices, are widely used in properties, hospitals, parks, and other scenarios to achieve centralized storage and standardized management of keys. Existing smart key cabinets typically use RFID radio frequency identification technology to monitor the storage and retrieval status of keys, and verify user identity through facial recognition, IC cards, or QR codes, recording the time of key retrieval and return to generate an operation log.

[0003] The existing single identity authentication method is easily cracked and does not achieve fine-grained permission matching between users and keys through feature extraction and precise comparison with the cloud permission database. It cannot prevent illegal operations from the source. At the same time, the transmission of abnormal information mostly uses a single channel without hierarchical differentiation, which makes it easy for emergency abnormal information to be pushed out in a delayed manner. Therefore, the industry urgently needs a smart key cabinet management method that takes identity recognition and authentication as the core, and achieves accurate key permission matching, full life cycle spatiotemporal monitoring, and hierarchical and reliable transmission of abnormal information. Summary of the Invention

[0004] This invention provides a method and system for real-time notification of abnormal events in smart key cabinets, the main purpose of which is to solve the problem of low accuracy in real-time notification of abnormal events in smart key cabinets.

[0005] To achieve the above objectives, the present invention provides a real-time notification method for abnormal events in a smart key cabinet, comprising: Obtain the user's identity authentication information, and determine the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information; The smart key cabinet is controlled to open the cabinet door corresponding to each key identifier based on the set of key identifiers, and the radio frequency signal and usage range of each key identifier are collected in real time. Based on the radio frequency signal, inventory change time-series data for each cabinet door in relation to the key's presence status is generated; based on the usage scope and the inventory change time-series data, spatial trajectory jump data for each key is generated. Align the spatial trajectory jump data with the opening and closing status data of each cabinet door on the time axis to obtain status alignment data. Perform correlation analysis between the key identifier set and the status alignment data to obtain abnormal event information of the smart key cabinet. The abnormal event information is transmitted to a preset dual-layer detection unit. The outer layer of the dual-layer detection unit filters the abnormal event information, and the inner layer of the dual-layer detection unit filters the transmission channel of the filtered abnormal event information to obtain the target transmission channel. The notification method corresponding to the target transmission channel is triggered, and the abnormal event information is notified to the remote management platform using the notification method.

[0006] To address the aforementioned problems, the present invention also provides a real-time notification system for abnormal events in a smart key cabinet, the system comprising: The key identifier set determination module is used to obtain the user's identity authentication information and determine the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information. The data analysis module is used to control the smart key cabinet to open the cabinet door corresponding to each key identifier based on the set of key identifiers, and to collect the radio frequency signal and usage range of each key identifier in real time; The spatial trajectory jump data generation module is used to generate inventory change time-series data for each cabinet door based on the radio frequency signal, and to generate spatial trajectory jump data for each key based on the usage scope and the inventory change time-series data. The abnormal event information analysis module is used to align the spatial trajectory jump data with the opening and closing status data of each cabinet door on the time axis to obtain status alignment data, and to perform correlation analysis between the key identifier set and the status alignment data to obtain abnormal event information of the smart key cabinet. The transmission channel filtering module is used to transmit the abnormal event information to a preset dual-layer detection unit. The outer layer of the dual-layer detection unit filters the abnormal event information, and the inner layer of the dual-layer detection unit filters the filtered abnormal event information to obtain the target transmission channel. An abnormal event notification module is used to trigger the notification method corresponding to the target transmission channel and use the notification method to notify the remote management platform of the abnormal event information.

[0007] This invention achieves all-time and all-space trajectory tracking of keys through spatial positioning and dynamic electronic fences, expanding the anomaly judgment dimension from a single state of being in or out of the cabinet to spatial boundary crossing detection. Through multi-channel parallel filtering of the outer layer of a dual-layer detection unit and multi-dimensional hierarchical evaluation of the inner layer, intelligent noise reduction, dynamic classification, and differentiated channel selection for abnormal events are achieved. Through a multi-channel transmission resource pool, dynamic channel switching mechanism, and closed-loop management of notification confirmation, reliable delivery and traceability of critical abnormal events are ensured during network fluctuations or channel congestion, significantly improving the intelligence level and security capabilities of smart key cabinets. Therefore, the real-time notification method and system for abnormal events in smart key cabinets proposed in this invention can solve the problem of low accuracy in real-time notification of abnormal events in smart key cabinets. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating a real-time notification method for abnormal events in a smart key cabinet, as provided in an embodiment of the present invention. Figure 2 This is a functional module diagram of a real-time notification system for abnormal events in an intelligent key cabinet, provided as an embodiment of the present invention.

[0009] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0010] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0011] This application provides a method for real-time notification of abnormal events in a smart key cabinet. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0012] Reference Figure 1The diagram shown is a flowchart illustrating a real-time notification method for abnormal events in a smart key cabinet according to an embodiment of the present invention. In this embodiment, the real-time notification method for abnormal events in a smart key cabinet includes: S1. Obtain the user's identity authentication information, and determine the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information.

[0013] In this embodiment of the invention, the identity authentication information refers to the data submitted by the user before operating the key cabinet to prove their identity, including at least one of facial image, identification card information and identification code information.

[0014] In detail, the smart key cabinet is equipped with a facial recognition camera, an NFC card reader, and a QR code scanning module. When the user selects facial recognition, the camera captures the user's facial image and generates facial image data. When the user selects card swiping, the NFC card reader reads the card number and encrypted information stored in the user's IC card to generate identification card information. When the user selects QR code scanning, the scanning module reads the dynamic QR code displayed in the user's mobile app or APP and parses it to obtain the encrypted identification code information.

[0015] Furthermore, feature extraction is performed on the three types of information collected to generate standardized identity feature vectors, ensuring the uniqueness and comparability of the features.

[0016] In this embodiment of the invention, the key identifier set refers to the set of unique codes of all keys that the user is authorized to access, as determined by the system based on the user's permissions.

[0017] In this embodiment of the invention, determining the set of key identifiers operable by the user for the smart key cabinet based on the identity authentication information includes: Extract the facial image, identification card information, and identification code information from the identity authentication information; Feature extraction is performed on the face image, the identification card information, and the identification code information respectively to obtain a first identity feature vector, a second identity feature vector, and a third identity feature vector; The first identity feature vector, the second identity feature vector, or the third identity feature vector are compared with the pre-stored features in the preset cloud permission database; If the comparison matches, the system retrieves a list of key identifiers that the user can operate on the smart key cabinet from the cloud-based permission database, and generates a key identifier set based on the key identifier list.

[0018] In detail, for facial images, a pre-trained facial recognition neural network model (such as FaceNet) is invoked to map the facial image into a 128-dimensional or 512-dimensional floating-point vector, serving as the first identity feature vector. For identification card information, the encrypted data stored on the card is decrypted to extract the user ID, which is then converted into a fixed-length feature vector using a hash function or encoding algorithm, serving as the second identity feature vector. For identification code information, the QR code content is parsed to obtain the user's unique identifier, and a third identity feature vector is generated using a preset encoding rule. The obtained feature vectors are sent to a cloud-based permission database via a secure communication protocol (such as HTTPS). The cloud database pre-stores the identity features and corresponding permission information of all registered users. For facial feature vectors, the Euclidean distance or cosine similarity between the vector and the pre-stored facial features in the database is calculated; if the similarity exceeds a set threshold, a match is considered. For card and code features, the decrypted user ID is directly compared to the database, thus achieving online user identity verification and ensuring that only legitimate users can continue operating.

[0019] Specifically, after successful authentication, the cloud database returns unique identifiers (such as key IDs and RFID tag numbers) for all keys the user is authorized to operate, along with the corresponding cabinet door number, preset authorized usage duration, and authorized activity area. The local control unit of the smart key cabinet receives the data and organizes it into a set, namely the key identifier set. This key identifier set serves as the basis for authorization, controlling cabinet door opening and anomaly detection. If a mismatch is found, an unauthorized access attempt is recorded, and a local notification or remote alarm may be triggered. Through multimodal identity authentication and cloud-based permission verification, it is ensured that only authorized users can operate the corresponding keys.

[0020] Furthermore, after completing user identification and authentication and determining the set of operable key identifiers, only the authorization has been determined. Actual cabinet door control and key status monitoring have not yet been achieved. Only by opening the corresponding cabinet door according to the key identifier set and collecting the key's radio frequency signal and usage range in real time can the status perception of the key usage process be realized.

[0021] S2. Based on the set of key identifiers, control the smart key cabinet to open the cabinet door corresponding to each key identifier, and collect the radio frequency signal and usage range of each key identifier in real time.

[0022] In this embodiment of the invention, the main control unit of the smart key cabinet retrieves a set of key identifiers from the access data area. It then matches each key identifier in the set with the cabinet door identifier of the smart key cabinet to determine the corresponding cabinet door. Subsequently, it sends a door opening command to the door control module. The door control module only sends an unlock signal to the successfully matched doors, opening the corresponding doors. Unmatched doors remain locked. This achieves access-based door control for the smart key cabinet, ensuring the security of key usage.

[0023] Furthermore, after the cabinet door is opened, the key may be taken out or returned. It is necessary to determine the presence or disappearance of the key in real time by collecting the key's radio frequency signal, and to grasp the spatial activity boundary of the key by collecting the usage range, so as to realize the dynamic status monitoring of the key's usage process.

[0024] In this embodiment of the invention, the radio frequency signal refers to the wireless signal strength and status change information of the key tag detected by the RFID read / write unit. The scope of application refers to the continuous sequence of key coordinates in three-dimensional space obtained by the UWB positioning system after the key is retrieved.

[0025] In this embodiment of the invention, the real-time acquisition of the radio frequency signal and usage range of each key identifier includes: The key tag corresponding to each key identifier is scanned by a preset RFID reader / writer unit. When the signal strength of the key tag exceeds the preset first dynamic threshold, it is determined that the key corresponding to the key identifier is in existence, and the key identifier and the current timestamp are recorded. When the signal strength of the key tag is detected to be lower than the preset second dynamic threshold, it is determined that the key corresponding to the key tag is in a missing state, and the key tag and the current timestamp are recorded. The key identifier and the current timestamp are encapsulated into a radio frequency signal; The three-dimensional spatial coordinates of each key identifier corresponding to the key are collected in real time at a first sampling frequency by a preset ultra-wideband positioning base station, and the continuous position coordinates of the key in space are identified based on the three-dimensional spatial coordinates. The scope of use for each key identifier is determined based on the continuous position coordinates.

[0026] In detail, each key slot in the smart key cabinet is equipped with an RFID antenna connected to an RFID reader / writer module. All key tags are scanned cyclically at a fixed frequency (e.g., 10Hz). The first dynamic threshold is a value dynamically adjusted based on ambient noise and historical signal strength; for example, the default setting is -60dBm. When the signal strength of a key tag exceeds this threshold, it indicates that the key is in the cabinet. The system records the key ID and the current time, and updates its status to "present." The second dynamic threshold is typically lower than the first dynamic threshold, for example, -70dBm, and a hysteresis interval is introduced to avoid signal jitter. When the signal strength remains below the second threshold for more than a set time (e.g., 200ms), it is determined that the key has been removed, the key ID and the current time are recorded, and the status is updated to "disappeared." By using dual thresholds and duration judgment, false judgments caused by momentary signal interruptions are effectively filtered out. Each detected key status change event (including key ID, event type "removed or returned," and timestamp) is encapsulated into a radio frequency signal record according to a preset data format and stored in a local cache.

[0027] Specifically, at least three UWB positioning base stations are deployed around the smart key cabinet, forming a positioning network. Each key embeds a UWB tag. When a key is removed, the tag emits a pulse signal at a first sampling frequency (e.g., 5Hz). The base station calculates the key's three-dimensional coordinates (x, y, z) in real time using a Time Difference of Arrival (TDOA) algorithm. It calculates the maximum and minimum values ​​of the x, y, and z axes, using the extreme values ​​as boundaries to determine the key's horizontal and vertical movement limits in space, thus forming a standardized key usage range. The three-dimensional coordinates are continuously recorded to form a position sequence. The collected continuous position coordinates are organized chronologically to form the key's movement trajectory, i.e., usage range data. This data reflects the key's spatial position changes during removal, including the area it entered and the position it remained in.

[0028] Furthermore, the collected radio frequency signals and usage range data are raw and scattered data, which cannot directly reflect the changes in key inventory and spatial trajectory. Only by processing the raw data and generating standardized inventory change time-series data and spatial trajectory jump data can the spatiotemporal status of the keys be analyzed.

[0029] S3. Generate inventory change time-series data for each cabinet door based on the radio frequency signal, and generate spatial trajectory jump data for each key based on the usage range and the inventory change time-series data.

[0030] In this embodiment of the invention, the inventory change time series data refers to the sequence of events recording the change in the state of each key in the cabinet in chronological order, including retrieval events and return events and their occurrence times.

[0031] In this embodiment of the invention, generating time-series data on inventory changes for each cabinet door based on the radio frequency signal in relation to the key's presence status includes: The first change point is detected based on the radio frequency signal, where the signal strength of each key tag changes from above the first dynamic threshold to below the second dynamic threshold and the duration exceeds the first preset time. The key retrieval event is determined based on the first change point. The system detects the first change point where the signal strength of each key tag changes from below the second dynamic threshold to above the first dynamic threshold and the duration exceeds the second preset time, and determines the key return event based on the second first change point. Arrange the retrieval events and return events in chronological order, and mark the key as out of cabinet between adjacent retrieval events and return events, and mark the key as in cabinet between adjacent return events and retrieval events. Generate time-series data of inventory changes based on the arranged and marked events, including key identifiers, status markers, status start time, and status end time.

[0032] In detail, the RFID-collected radio frequency signals are analyzed in real time. When the signal strength of a key drops from above a first dynamic threshold (e.g., -60dBm) to below a second dynamic threshold (e.g., -70dBm), and this low signal state persists for more than a first preset duration (e.g., 200ms), the system confirms that the key has indeed been retrieved, and that it is not a false judgment caused by a momentary signal interruption. At this time, the key ID and retrieval timestamp (accurate to milliseconds) are recorded as a retrieval event. When the key is returned, its signal strength rises from below the second dynamic threshold to above the first dynamic threshold and persists for more than the second preset duration (e.g., 200ms), which is determined as a return event, and the key ID and return timestamp are recorded.

[0033] Specifically, all retrieval and return events for each key are sorted by time to form an event sequence. For the same key, the time interval between a retrieval event and a subsequent return event is marked as the "off-cabinet state"; the time interval between a return event and the next retrieval event is marked as the "in-cabinet state". The key identifier, state start time, and state end time corresponding to each state marker are extracted from the state change data. This information is then integrated into standardized inventory change time-series data according to a time-series data format. Each entry in the data contains a unique key identifier, the corresponding state marker, and the start and end times of that state. The generated inventory change time-series data is stored in the time-series data area of ​​the master control unit. For example, for key A, the following records are generated: "in-cabinet state" starts at t0 and ends at t1; "off-cabinet state" starts at t1 and ends at t2; "in-cabinet state" starts at t2. These records constitute the inventory change time-series data, accurately reflecting the real-time presence of the key in the cabinet.

[0034] Furthermore, inventory change time series data only reflects the time and inventory status changes of the keys, and cannot reflect the spatial movement trajectory of the keys after they leave the cabinet or whether there is any boundary violation. By combining the scope of use and inventory change time series data to generate spatial trajectory jump data, it is possible to achieve systematic analysis of the spatial trajectory of the keys after they leave the cabinet and accurately identify boundary violation.

[0035] In this embodiment of the invention, spatial trajectory jump data refers to key movement path data generated by combining the inventory status window and UWB trajectory, which includes compliant trajectory segments and out-of-bounds trajectory segments.

[0036] In this embodiment of the invention, generating spatial trajectory jump data for each key based on the scope of use and the inventory change time series data includes: Extract the time intervals marked as key-out-of-cabinet status from the inventory change time series data, and use the time intervals as the spatial trajectory acquisition window for each key; The spatial coordinates of each key are continuously acquired within the spatial trajectory acquisition window of each key to generate the original trajectory point sequence. Based on the scope of use, determine the boundary of the authorized activity area and the authorized activity height range of the key in space, and determine the spatial inclusion relationship between each trajectory point in the original trajectory point sequence and the boundary of the authorized activity area and the authorized activity height range; If the coordinates of a trajectory point are located outside the boundary of the authorized activity area or exceed the authorized activity height range, the trajectory point is marked as an out-of-bounds trajectory point, and out-of-bounds trajectory segments are extracted based on the spatiotemporal distribution of the out-of-bounds trajectory points. The out-of-bounds trajectory segments are spliced ​​together with compliant trajectory segments that are not marked as out of bounds in chronological order to generate spatial trajectory jump data. The spatial trajectory jump data includes the first out-of-bounds time, the farthest out-of-bounds distance, and the out-of-bounds duration for each out-of-bounds trajectory segment.

[0037] In detail, based on inventory change time-series data, it's known when each key is retrieved (start of departure) and when it's returned (end of departure). The departure time interval is the effective time period for collecting UWB trajectory data. Processing UWB data only within the departure window avoids invalid data collection while the key is in the locker, reducing the data processing burden. Within the departure window, the three-dimensional coordinates of the key are obtained from the UWB positioning system at a second sampling frequency (e.g., 5Hz). Each trajectory point includes a timestamp and (x,y,z) coordinates, forming an original trajectory point sequence. The second sampling frequency can be equal to or different from the first sampling frequency; setting it to 5Hz ensures trajectory accuracy while controlling data volume.

[0038] Specifically, from the key identifier set, each key is associated with a preset authorized activity area (e.g., a two-dimensional polygon describing the permitted activity range). The extreme values ​​of the horizontal coordinates of the usage range are determined as the boundary of the authorized activity area and the authorized activity height range (e.g., 0.5m to 2m above the ground), and the extreme values ​​of the vertical coordinates are determined as the authorized activity height range. For each point in the original trajectory point sequence, it is determined whether its coordinates are within the authorized polygon and whether its height is within the authorized height range. The three-dimensional spatial coordinates of each trajectory point are extracted, and it is determined whether its x and y axis coordinates are within the boundary of the authorized activity area and whether its z axis coordinate is within the authorized activity height range. If the coordinates are within the authorized polygon and the height is within the range, it is a compliant trajectory point; otherwise, it is an out-of-bounds trajectory point. Consecutive out-of-bounds trajectory points are combined into an out-of-bounds trajectory segment, and the start time, end time, and farthest out-of-bounds distance of the segment are recorded. The timestamp of the first trajectory point in the segment is taken as the first out-of-bounds time, the farthest distance between all trajectory points in the segment and the boundary of the authorized activity range is taken as the farthest out-of-bounds distance, and the time span of the segment is taken as the out-of-bounds duration.

[0039] Furthermore, the initial boundary crossing time, the furthest boundary crossing distance, and the duration of the boundary crossing are appended to the corresponding boundary crossing trajectory segment. Following a preset trajectory data format, the spliced ​​trajectory data and the appended boundary crossing information are integrated into standardized spatial trajectory jump data. For example, a key's trajectory consists of three segments: compliant trajectory segment A (0-5 seconds), boundary crossing trajectory segment B (5-10 seconds, boundary crossing distance 2.5 meters), and compliant trajectory segment C (10-15 seconds). The final generated spatial trajectory jump data includes these segments and their characteristic parameters, clearly reflecting the key's complete movement path and boundary crossing behavior during its time away from the cabinet.

[0040] Furthermore, the two types of data generated are still independent time-series data and trajectory data, which cannot be directly used for abnormal event identification. Only by aligning the spatial trajectory jump data with the cabinet door opening and closing status data on the time axis to generate status-aligned data, and then combining it with a determined set of key identifiers for correlation analysis, can accurate identification of various abnormal events be achieved.

[0041] S4. Align the spatial trajectory jump data with the opening and closing status data of each cabinet door on the time axis to obtain status alignment data. Perform correlation analysis between the key identifier set and the status alignment data to obtain abnormal event information of the smart key cabinet.

[0042] In this embodiment of the invention, the cabinet door opening and closing status data refers to the timestamp and cabinet door identifier of each cabinet door opening and closing event collected by the cabinet door sensor. The status alignment data refers to the comprehensive data obtained by matching the key trajectory with the cabinet door actions on the timeline, including the take-out action, the return action, and the trajectory points.

[0043] In this embodiment of the invention, aligning the spatial trajectory jump data with the opening and closing status data of each cabinet door along the time axis to obtain state alignment data includes: Extract trajectory segments for each key from the spatial trajectory jump data; Extend a first preset time window forward from the starting time timestamp of the trajectory segment of each key, and check within the first preset time window whether there is a cabinet door opening event corresponding to the trajectory segment; Extend a second preset time window forward from the end timestamp of the trajectory segment of each key as the center, and check whether there is a cabinet door closing event corresponding to the trajectory segment within the second preset time window; If a cabinet door opening event is found within the first preset time window and the cabinet door identifier carried by the cabinet door opening event is consistent with the authorized cabinet door corresponding to the key identifier, the timestamp of the trajectory starting point is bound to the timestamp of the cabinet door opening event to generate a retrieval action alignment record. If a cabinet door closing event is found within the second preset time window and the cabinet door identifier carried by the cabinet door closing event is consistent with the authorized cabinet door corresponding to the key identifier, then the trajectory endpoint timestamp is bound to the timestamp of the cabinet door closing event to generate a return action alignment record. The retrieved action alignment record, the returned action alignment record, and the trajectory point sequence in the trajectory segment are fused together to generate state alignment data.

[0044] In detail, the trajectory segment of each key consists of its spatial trajectory jump data, including a key identifier, a trajectory start timestamp, a trajectory end timestamp, and a sequence of trajectory points. A first preset time window is set to 5 seconds prior. For the trajectory start time t_start of a key, a cabinet door opening event is searched within the range [t_start-5s, t_start], meaning the cabinet door ID carried in the event should match the cabinet door ID where the key was authorized. If a match is found, it indicates that the key's retrieval was likely caused by this cabinet door opening. A second preset time window is set to 5 seconds backward. For the trajectory end time t_end, a cabinet door closing event is searched within the range [t_end, t_end+5s], again requiring a match in the cabinet door ID. If a match is found, it indicates that the key's return was likely accompanied by this cabinet door closing. If a cabinet door opening event is found within the first preset time window, and the cabinet door identifier carried by the opening event matches the authorized cabinet door corresponding to the key identifier, a retrieval action alignment record is generated. This retrieval action alignment record includes the key ID, retrieval time (retrieval trajectory start time or cabinet door opening time), and cabinet door ID. If a cabinet door closing event is found within the second preset time window, and the cabinet door identifier carried by the closing event matches the authorized cabinet door corresponding to the key identifier, a return action alignment record is generated. This return action alignment record includes the key ID, return time (retrieval trajectory end time or cabinet door closing time), and cabinet door ID. The retrieval action alignment record, the return action alignment record, and the trajectory point sequence in the trajectory segment are merged to generate state alignment data containing the key identifier, retrieval time, return time, retrieved cabinet door, returned cabinet door, trajectory point sequence, and the timestamp corresponding to each trajectory point.

[0045] Specifically, to ensure the reliability of cabinet door data, the cabinet door opening and closing status data is preprocessed. Cabinet door opening and closing events collected by the cabinet door sensors are acquired. These events include door opening and closing events, each carrying an event type, a timestamp, and a corresponding cabinet door identifier, generating raw cabinet door status data. The raw cabinet door status data undergoes a time-series integrity check. If a missing door closing event is detected between consecutive door opening events, or a missing door opening event is detected between consecutive door closing events, the missing events are interpolated based on historical operation patterns to generate repaired cabinet door status data, i.e., cabinet door opening and closing status data. For example, if cabinet door 5 is found to have opened at 10:05:30 and then opened again at 10:05:35 without a closing event in between, a closing event might be interpolated at 10:05:50 based on the average opening time (e.g., an average opening time of 20 seconds) to ensure data integrity.

[0046] Furthermore, the state alignment data only reflects the actual operation and trajectory of the key, without comparing it with the user's authorization permissions. By combining the key identifier set, i.e. the user's authorized operation range, for correlation analysis, various abnormal events such as unauthorized use and failure to return the key within the time limit can be accurately identified through permission verification, duration verification, trajectory verification, etc., realizing the core transformation from data fusion to anomaly identification.

[0047] In this embodiment of the invention, abnormal event information refers to events such as unauthorized access, failure to return within the time limit, unauthorized use, and lost keys identified through correlation analysis.

[0048] In this embodiment of the invention, the step of performing correlation analysis between the key identifier set and the state alignment data to obtain abnormal event information of the smart key cabinet includes: Extract the operation event records from the state alignment data, and compare the key identifiers in the operation event records with the authorized key identifiers in the key identifier set item by item; If a key identifier in an operation event log is detected to be outside the set of key identifiers, the operation event log will be marked as an illegal access event. If the key identifier in the operation event record is detected to belong to the set of key identifiers, the time difference between the retrieval time and the return time is taken as the actual usage time. If the actual usage time exceeds the preset authorized usage time and the excess part is greater than the tolerance threshold, the operation event record is marked as an overdue return event. The trajectory point sequence in the state alignment data is correlated with the spatial positions of the cabinet door being taken out and the cabinet door being returned. If there are off-cabinet trajectory points outside the spatial range of the cabinet door, a key activity heat map is generated based on the distribution of the off-cabinet trajectory points. The key activity heat map is then overlaid with the authorized activity area of ​​the key. If there is an out-of-bounds activity area outside the authorized area, it is marked as an out-of-bounds use event. The operation event record is checked for integrity. If there is a record of taking out the action but no corresponding record of returning the action and the current time exceeds the preset loss time, it is marked as a key loss event. The illegal access events, overdue return events, out-of-bounds use events, and lost key events are aggregated in chronological order to generate abnormal event information.

[0049] In detail, the process involves extracting retrieval and return action alignment records for each key from the status alignment data. An operation event record is generated based on these records, including the key ID, retrieval time, return time, cabinet door retrieval, and cabinet door return information. The key ID in this record is compared one by one with the current user's key identifier set. If a key identifier in the operation event record does not belong to the key identifier set, the record is marked as an illegal access event. For example, user A may only authorize access to keys 1 and 2, but the status alignment data shows that user A accessed key 3, triggering an illegal access event and recording the key ID, operation time, and cabinet door. If a key identifier in the operation event record belongs to the key identifier set, a preset authorized usage duration is obtained from the set (e.g., key 1 is authorized for 30 minutes). The tolerance threshold can be set to 10% (i.e., an alarm is only triggered if the usage exceeds 3 minutes). The actual usage duration is the return time minus the retrieval time. If it exceeds 30 minutes + 3 minutes, it is marked as overdue and the overdue duration is recorded.

[0050] Specifically, all points on the trajectory that leave the cabinet door (i.e., the key is not inside the cabinet) are considered as off-cabinet trajectory points. Kernel density estimation is performed on the off-cabinet trajectory points to generate a heat map, showing the densely populated areas of key activity during the time the key is away from the cabinet. Then, the heat map is overlaid with the authorized activity area of ​​the key (such as office area, laboratory, etc.). If there is an area outside the authorized area in the heat map, it is determined as out-of-bounds use, and the range and duration of the out-of-bounds area are recorded. The integrity of the operation event records is checked. If there is a record of a retrieval action but no corresponding record of a return action, and the current time exceeds the sum of the retrieval time plus the preset authorized use time plus the preset grace period, then the operation event is marked as a key loss event. All detected abnormal events are sorted by occurrence time to form an abnormal event information list. Each piece of information includes event type, event time, associated key ID, associated user ID, etc.

[0051] Furthermore, the generated abnormal event information may contain invalid data such as format errors, duplicates, and sensor malfunctions, and the corresponding transmission channel is not matched according to the severity of the abnormality. Direct transmission will lead to low information transmission efficiency and delay of important abnormal information. By using the outer layer filtering and inner layer channel screening of the dual-layer detection unit, invalid abnormal information can be filtered out, providing high-quality abnormal data and reliable transmission guarantee for remote notification.

[0052] S5. Transmit the abnormal event information to a preset dual-layer detection unit. The outer layer of the dual-layer detection unit filters the abnormal event information, and the inner layer of the dual-layer detection unit filters the transmission channel of the filtered abnormal event information to obtain the target transmission channel.

[0053] In this embodiment of the invention, the dual-layer detection unit refers to an intelligent processing module composed of an outer filter and an inner classifier. The outer filter cleans abnormal event information through multiple parallel channels, eliminating invalid, repetitive, and storm-like events. The inner classifier evaluates the severity of events and dynamically selects the most suitable transmission channel.

[0054] In this embodiment of the invention, the filtering of abnormal event information through the outer layer of the dual-layer detection unit includes: The abnormal event information is input into multiple parallel filtering channels of the outer filter of the dual-layer detection unit; The abnormal event information is time-appropriately verified by the first filtering channel in the parallel filtering channel. If the timestamp deviates from the current system time by more than the maximum allowable deviation, time synchronization correction is triggered, and the abnormal event information is re-input to the outer filter. The second filtering channel in the parallel filtering channel performs event repeatability detection on the time-verified abnormal event information. If the matching degree between the feature fingerprint of the abnormal event information and the feature fingerprint within the preset time window exceeds the similarity threshold, the abnormal event information is determined to be a duplicate event. The third filtering channel in the parallel filtering channel performs event storm suppression on the abnormal event information that passes the repeatability detection. If the frequency of abnormal events of the same key exceeds the storm suppression threshold within the statistical time window, the current event and historical events are identified as aggregated event information. Abnormal event information or aggregated event information that has been verified through all filtering channels is marked as valid abnormal event information, and the valid abnormal event information is passed to the inner layer of the dual-layer detection unit.

[0055] In detail, abnormal event information (which may contain multiple events) is sequentially sent to multiple parallel channels of the outer filter, with each channel responsible for a specific filtering rule. The first filtering channel verifies the time reasonableness of the abnormal event information, extracting the timestamp and comparing it with the current system time of the dual-layer detection unit to calculate the time deviation. If the time deviation exceeds the preset maximum allowable time deviation, the abnormal event information is determined to be a time anomaly, triggering a time synchronization correction command and temporarily storing the abnormal event information until it re-enters the outer filter after time synchronization correction. For example, if the event timestamp is 10 seconds later than the current system time (exceeding the allowable deviation ±2 seconds), it may be due to device time asynchrony. The system triggers NTP synchronization and temporarily stores the event, reprocessing it after synchronization. The second filtering channel performs event duplication detection on abnormal event information that has passed the time reasonableness verification, extracting the feature fingerprint of the abnormal event information. The feature fingerprint is generated by combining the event type, key identifier, and the event occurrence time at the minute level. The feature fingerprint is then fuzzily matched with the feature fingerprint database of processed events within a preset time window. If the matching degree exceeds a preset similarity threshold, the abnormal event information is determined to be a duplicate event, discarded, and the occurrence frequency count of the corresponding feature fingerprint is updated. For example, the feature fingerprint is "illegal access to key 1_10:05" (minute level). If an event with the same fingerprint has already occurred within the last 5 minutes, and the similarity exceeds 90%, it is considered a duplicate event and discarded directly to avoid duplicate notifications.

[0056] Specifically, the third filtering channel performs event storm suppression on abnormal event information that passes the event repetition detection. It obtains the frequency of abnormal events occurring within the most recent first statistical time window for the same key identifier. If the frequency exceeds the preset storm suppression threshold, the event aggregation mode is activated. The current abnormal event information is aggregated with the historical abnormal event information within the window to generate aggregated event information that includes event type distribution, occurrence time distribution, and frequency statistics. The aggregated event information replaces the original abnormal event information. For example, if the same key has 10 illegal attempts within 1 minute, instead of processing them one by one, a storm event is generated: aggregated information of 10 illegal access attempts by key 1 between 10:00 and 10:01, effectively reducing processing pressure.

[0057] In addition, abnormal event information undergoes data integrity verification. The event type, key identifier, timestamp, and user identifier fields are extracted from the abnormal event information, and each field is checked for null values ​​or format errors. If any are found, the abnormal event information is marked as a formatting error and stored in a pending review queue, not proceeding to subsequent processing. For abnormal event information that passes integrity verification, source credibility is assessed by obtaining real-time status data of the sensor module that generated the abnormal event information. This real-time status data includes the signal quality index of the RFID reader / writer module, the positioning accuracy index of the UWB positioning base station, and the operating status code of the cabinet door sensor. The signal quality index, positioning accuracy index, and operating status code are compared with their corresponding credibility thresholds. If any index is lower than the corresponding credibility threshold or the operating status code indicates a fault, the abnormal event information is marked as a sensor abnormal event and stored in a pending review queue, not proceeding to subsequent processing.

[0058] Furthermore, the outer layer of filtering only removes invalid abnormal information and does not match the corresponding transmission channel according to the severity of the abnormal event. Abnormal events of different severity require transmission channels with different priorities to ensure transmission efficiency and reliability. Through the inner layer of channel filtering, the optimal transmission channel can be accurately matched according to the abnormality level, ensuring that important abnormal information is transmitted with priority and reliability, and improving the timeliness of abnormal information notification.

[0059] In this embodiment of the invention, the target transmission channel refers to the communication channel selected based on the level of the abnormal event and used to transmit the corresponding abnormal information.

[0060] In this embodiment of the invention, the step of filtering the filtered abnormal event information through the inner layer of the dual-layer detection unit to obtain the target transmission channel includes: Extract the feature parameters of the filtered abnormal event information, and query the basic abnormality score of the abnormal event information based on the event type in the feature parameters; Obtain the associated key of the abnormal event information, determine the correction coefficient of the abnormal event information according to the number of associated keys, and correct the basic abnormal score according to the correction coefficient; The abnormality level of the abnormal event information is determined based on the corrected abnormality score. A transmission channel is selected from the transmission channel resource pool based on the abnormality level, and the real-time status of the selected transmission channel is queried. If the current queue length or transmission delay in the real-time state exceeds the threshold, the alternative transmission channels are switched sequentially from high to low according to the corrected anomaly score until an available channel that meets the transmission requirements is switched as the target transmission channel.

[0061] In detail, the system receives valid abnormal event information from the outer filter and extracts feature parameters such as event type, associated key identifier, out-of-bounds distance, or timeout duration. These feature parameters are then input into a multi-dimensional abnormality level assessment model. Based on the event type, the corresponding basic abnormality score is retrieved. For example, a predefined table of basic scores for event types is provided, such as 80 points for unauthorized access, 60 points for overdue return, 70 points for out-of-bounds use, 90 points for lost key, and 50 points for aggregated events. The basic score is then adjusted using spatiotemporal weighting based on the out-of-bounds distance, timeout duration, or user permission level. The greater the out-of-bounds distance and the longer the timeout duration, the higher the weighted score. Abnormal events involving high-privilege users are also weighted higher according to a preset coefficient. For example, for every 1 meter increase in out-of-bounds distance, the basic score is multiplied by (1 + 0.1); for every 5 minutes increase in timeout duration, the basic score is multiplied by (1 + 0.05); and the higher the user permission level (e.g., administrator), the smaller the weighting coefficient (e.g., 0.8). Retrieve the number of historical anomalies for the associated key or user within the most recent time window, and query the corresponding cumulative coefficient. The more historical anomalies, the larger the cumulative coefficient. For example, if there are 1 anomalies in the last hour, the coefficient is 1.0; 2 anomalies, 1.2; 3 anomalies, 1.5; and 4 anomalies, 1.9. Multiply the weighted adjusted score by the cumulative coefficient to generate the anomaly level score.

[0062] Specifically, the final anomaly score is compared with preset thresholds for emergency, important, and general events to determine the level of the anomaly. The thresholds are set as follows: emergency events ≥ 80, important events 60-80, general events 30-60, and alert events < 30. Based on the determined anomaly level, a corresponding target transmission channel is selected from the transmission channel resource pool. Emergency events correspond to high-priority channels with independent bandwidth, important events to priority sending queue channels, general events to best-effort sending channels, and alert events to idle network sending or local log channels. The level is determined based on the final score, and a default channel corresponding to that level is selected from the transmission channel resource pool. For example, the emergency channel is SMS + phone + local alarm, the important channel is APP push + local indicator light, the general channel is a public account, and the alert channel is log recording. The real-time load of each channel is monitored. If the current queue length exceeds the threshold or the estimated sending delay exceeds the threshold, a dynamic channel switching mechanism is activated. Alternative transmission channels are tried sequentially from highest to lowest anomaly score until a usable channel that meets the sending requirements is found as the final target transmission channel. If the current queue for the emergency channel exceeds 100 entries or the estimated delay exceeds 30 seconds, attempt to switch to the secondary emergency channel (such as the critical channel). If this is still not satisfactory, continue switching until an available channel is found.

[0063] Furthermore, the matching of transmission channels was completed. If abnormal event information is not pushed to the remote management platform through the target transmission channel, the management personnel cannot keep abreast of the abnormal situation of the smart key cabinet. By triggering the notification method corresponding to the target transmission channel, the abnormal event information can be notified to the remote management platform, which can realize the remote delivery of abnormal information and allow the management personnel to receive and handle abnormal events in a timely manner.

[0064] S6. Trigger the notification method corresponding to the target transmission channel, and use the notification method to notify the remote management platform of the abnormal event information.

[0065] In this embodiment of the invention, the notification method refers to a message push method pre-configured according to different transmission channels, including SMS, telephone voice, APP push, official account push, local sound and light alarm, etc.

[0066] In detail, the system obtains the target transmission channel information and retrieves the corresponding notification method for the target transmission channel based on the preset channel-notification method correspondence. Standardized abnormal event information is encapsulated according to the message format of the corresponding notification method, such as SMS format, APP push format, PC pop-up format, etc. The encapsulated abnormal information retains core content, namely event type, event time, associated key identifier, associated user identifier, and abnormal details. The encapsulated abnormal information is then sent to the communication module of the remote management platform through the target transmission channel, simultaneously triggering corresponding notification commands, such as triggering a pop-up command for a PC client, triggering a text message sending command for a mobile client, or triggering a push command for a mobile APP.

[0067] Specifically, after receiving abnormal event information and notification instructions, the remote management platform displays and alerts the information according to the corresponding notification method. If the target transmission channel is an emergency channel, an alarm SMS containing key information such as event type, key ID, time, and location is immediately sent to the preset administrator's mobile phone via the SMS gateway. At the same time, the automatic outbound telephone voice system is activated to broadcast the details of the abnormal event and sends an instruction to the local control unit of the smart key cabinet, triggering the buzzer to emit a continuous alarm sound and the red LED to flash, alerting on-site personnel. If the target transmission channel is an important channel, a notification is pushed through the mobile APP. The message content includes an event summary, key ID, time of occurrence, and on-site photos (if the cabinet has a camera that is linked to capture images). At the same time, the yellow indicator light of the corresponding key position on the cabinet is lit up to alert the administrator. If the target transmission channel is a general channel, a message is pushed through the official WeChat account. The message is added to the history list of the official WeChat account, and only a log is recorded locally without triggering an audible or visual alarm. If the target transmission channel is a notification channel, the event information is stored in the local database and uploaded to the cloud in batches when the network is idle (such as at night).

[0068] Furthermore, to ensure the reliability of notifications, a confirmation timer (e.g., 5 minutes) is started after the notification is sent, waiting for the management terminal to return a confirmation receipt (e.g., app click confirmation, SMS reply). If no confirmation is received within the preset time, the notification method is automatically escalated, for example, from a general channel to an important channel, or from app push to SMS + phone, until confirmation is received. All notification records and confirmation statuses are stored in a cloud database for easy traceability.

[0069] like Figure 2 The diagram shown is a functional block diagram of a real-time notification system for abnormal events in an intelligent key cabinet, provided in an embodiment of the present invention.

[0070] The intelligent key cabinet abnormal event real-time notification system 100 of the present invention can be installed in an electronic device. Depending on the functions implemented, the intelligent key cabinet abnormal event real-time notification system 100 may include a key identifier set determination module 101, a data analysis module 102, a spatial trajectory jump data generation module 103, an abnormal event information analysis module 104, a transmission channel filtering module 105, and an abnormal event notification module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0071] In this embodiment, the functions of each module / unit are as follows: The key identifier set determination module 101 is used to obtain the user's identity authentication information and determine the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information. The data analysis module 102 is used to control the smart key cabinet to open the cabinet door corresponding to each key identifier according to the key identifier set, and to collect the radio frequency signal and usage range of each key identifier in real time. The spatial trajectory jump data generation module 103 is used to generate inventory change time series data for each cabinet door based on the radio frequency signal for the key's presence status, and to generate spatial trajectory jump data for each key based on the usage scope and the inventory change time series data. The abnormal event information analysis module 104 is used to align the spatial trajectory jump data with the opening and closing status data of each cabinet door on the time axis to obtain status alignment data, and to perform correlation analysis between the key identifier set and the status alignment data to obtain abnormal event information of the smart key cabinet. The transmission channel filtering module 105 is used to transmit the abnormal event information to a preset dual-layer detection unit, filter the abnormal event information through the outer layer of the dual-layer detection unit, and filter the transmission channel of the filtered abnormal event information through the inner layer of the dual-layer detection unit to obtain the target transmission channel. The abnormal event notification module 106 is used to trigger the notification method corresponding to the target transmission channel and use the notification method to notify the abnormal event information to the remote management platform.

[0072] In detail, the modules in the real-time notification system for abnormal events of the smart key cabinet described in this embodiment of the invention employ the same methods as described above. Figure 1 This method uses the same technical means as the real-time notification method for abnormal events of a smart key cabinet described above, and can produce the same technical effect, so it will not be elaborated here.

[0073] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0074] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0075] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0076] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0077] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects. The scope of the invention is not limited to the foregoing description, and all variations within the meaning and scope of equivalents falling within the protection scope are intended to be included in the invention.

[0078] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0079] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems stated in a system claim may also be implemented by a single unit or system through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time notification of abnormal events in a smart key cabinet, characterized in that, The method includes: Obtain the user's identity authentication information, and determine the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information; The smart key cabinet is controlled to open the cabinet door corresponding to each key identifier based on the set of key identifiers, and the radio frequency signal and usage range of each key identifier are collected in real time. Based on the radio frequency signal, inventory change time-series data for each cabinet door in relation to the key's presence status is generated; based on the usage range and the inventory change time-series data, spatial trajectory jump data for each key is generated. Align the spatial trajectory jump data with the opening and closing status data of each cabinet door on the time axis to obtain status alignment data. Perform correlation analysis between the key identifier set and the status alignment data to obtain abnormal event information of the smart key cabinet. The abnormal event information is transmitted to a preset dual-layer detection unit. The outer layer of the dual-layer detection unit filters the abnormal event information, and the inner layer of the dual-layer detection unit filters the transmission channel of the filtered abnormal event information to obtain the target transmission channel. The notification method corresponding to the target transmission channel is triggered, and the abnormal event information is notified to the remote management platform using the notification method.

2. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The step of determining the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information includes: Extract the facial image, identification card information, and identification code information from the identity authentication information; Feature extraction is performed on the face image, the identification card information, and the identification code information respectively to obtain a first identity feature vector, a second identity feature vector, and a third identity feature vector; The first identity feature vector, the second identity feature vector, or the third identity feature vector are compared with the pre-stored features in the preset cloud permission database; If the comparison matches, the system retrieves a list of key identifiers that the user can operate on the smart key cabinet from the cloud-based permission database, and generates a key identifier set based on the key identifier list.

3. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The real-time acquisition of the radio frequency signal and usage range of each key identifier includes: The key tag corresponding to each key identifier is scanned by a preset RFID reader / writer unit. When the signal strength of the key tag exceeds the preset first dynamic threshold, it is determined that the key corresponding to the key identifier is in existence, and the key identifier and the current timestamp are recorded. When the signal strength of the key tag is detected to be lower than the preset second dynamic threshold, it is determined that the key corresponding to the key tag is in a missing state, and the key tag and the current timestamp are recorded. The key identifier and the current timestamp are encapsulated into a radio frequency signal; The three-dimensional spatial coordinates of each key identifier corresponding to the key are collected in real time, and the continuous position coordinates of the key in space are identified based on the three-dimensional spatial coordinates; The scope of use for each key identifier is determined based on the continuous position coordinates.

4. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The step of generating time-series data on inventory changes for each cabinet door based on the radio frequency signal for the presence of the key includes: The first change point is detected based on the radio frequency signal, where the signal strength of each key tag changes from above the first dynamic threshold to below the second dynamic threshold and the duration exceeds the first preset time. The key retrieval event is determined based on the first change point. The system detects the first change point where the signal strength of each key tag changes from below the second dynamic threshold to above the first dynamic threshold and the duration exceeds the second preset time, and determines the key return event based on the second first change point. Arrange the retrieval events and return events in chronological order, and mark the key as out of cabinet between adjacent retrieval events and return events, and mark the key as in cabinet between adjacent return events and retrieval events. Generate time-series data of inventory changes based on the arranged and marked events, including key identifiers, status markers, status start time, and status end time.

5. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The step of generating spatial trajectory jump data for each key based on the scope of use and the inventory change time series data includes: Extract the time intervals marked as key-out-of-cabinet status from the inventory change time series data, and use the time intervals as the spatial trajectory acquisition window for each key; The spatial coordinates of each key are continuously acquired within the spatial trajectory acquisition window of each key to generate the original trajectory point sequence. Based on the scope of use, determine the boundary of the authorized activity area and the authorized activity height range of the key in space, and determine the spatial inclusion relationship between each trajectory point in the original trajectory point sequence and the boundary of the authorized activity area and the authorized activity height range; If the coordinates of a trajectory point are located outside the boundary of the authorized activity area or exceed the authorized activity height range, the trajectory point is marked as an out-of-bounds trajectory point, and out-of-bounds trajectory segments are extracted based on the spatiotemporal distribution of the out-of-bounds trajectory points. The out-of-bounds trajectory segments are spliced ​​together with compliant trajectory segments that are not marked as out of bounds in chronological order to generate spatial trajectory jump data.

6. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The step of aligning the spatial trajectory jump data with the opening and closing status data of each cabinet door along the time axis to obtain status alignment data includes: Extract trajectory segments for each key from the spatial trajectory jump data; Extend a first preset time window forward from the starting time timestamp of the trajectory segment of each key, and check within the first preset time window whether there is a cabinet door opening event corresponding to the trajectory segment; Extend a second preset time window forward from the end timestamp of the trajectory segment of each key as the center, and check whether there is a cabinet door closing event corresponding to the trajectory segment within the second preset time window; If a cabinet door opening event is found within the first preset time window and the cabinet door identifier carried by the cabinet door opening event is consistent with the authorized cabinet door corresponding to the key identifier, the timestamp of the trajectory starting point is bound to the timestamp of the cabinet door opening event to generate a retrieval action alignment record. If a cabinet door closing event is found within the second preset time window and the cabinet door identifier carried by the cabinet door closing event is consistent with the authorized cabinet door corresponding to the key identifier, then the trajectory endpoint timestamp is bound to the timestamp of the cabinet door closing event to generate a return action alignment record. The retrieved action alignment record, the returned action alignment record, and the trajectory point sequence in the trajectory segment are fused together to generate state alignment data.

7. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The step of performing correlation analysis between the key identifier set and the state alignment data to obtain abnormal event information of the smart key cabinet includes: Extract the operation event records from the state alignment data, and compare the key identifiers in the operation event records with the authorized key identifiers in the key identifier set item by item; If a key identifier in an operation event log is detected to be outside the set of key identifiers, the operation event log will be marked as an illegal access event. If the key identifier in the operation event record is detected to belong to the set of key identifiers, and if the actual usage time corresponding to the key exceeds the preset authorized usage time and the excess part is greater than the tolerance threshold, it is marked as an overdue event. If there are off-cabinet trajectory points in the trajectory point sequence of the state alignment data that are outside the space range of the cabinet door, a key activity heat map is generated according to the distribution of the off-cabinet trajectory points, and the key activity heat map is overlaid with the authorized activity area of ​​the key for analysis. If there is an out-of-bounds activity area outside the authorized area, it is marked as an out-of-bounds use event. If the operation event log contains a record of taking out an action alignment but no corresponding record of returning an action alignment, and the current time exceeds the preset loss duration, it is marked as a key loss event. The illegal access events, overdue return events, out-of-bounds use events, and lost key events are aggregated in chronological order to generate abnormal event information.

8. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The filtering of abnormal event information through the outer layer of the dual-layer detection unit includes: The first filter channel in the parallel filter channel of the outer filter of the dual-layer detection unit is used to verify the time rationality of the abnormal event information. If the timestamp deviates from the current system time by more than the maximum allowable deviation, time synchronization correction is triggered, and the abnormal event information is re-inputted to the outer filter. The second filtering channel in the parallel filtering channel performs event repeatability detection on the time-verified abnormal event information. If the matching degree between the feature fingerprint of the abnormal event information and the feature fingerprint within the preset time window exceeds the similarity threshold, the abnormal event information is determined to be a duplicate event. The third filtering channel in the parallel filtering channel performs event storm suppression on the abnormal event information that passes the repeatability detection. If the frequency of abnormal events of the same key exceeds the storm suppression threshold within the statistical time window, the current event and historical events are identified as aggregated event information. Abnormal event information or aggregated event information that has been verified through all filtering channels is marked as valid abnormal event information, and the valid abnormal event information is passed to the inner layer of the dual-layer detection unit.

9. The method for real-time notification of abnormal events in a smart key cabinet as described in claim 1, characterized in that, The step of filtering the abnormal event information through the inner layer of the dual-layer detection unit to obtain the target transmission channel includes: Extract the feature parameters of the filtered abnormal event information, and query the basic abnormality score of the abnormal event information based on the event type in the feature parameters; Obtain the associated key of the abnormal event information, determine the correction coefficient of the abnormal event information according to the number of associated keys, and correct the basic abnormal score according to the correction coefficient; The abnormality level of the abnormal event information is determined based on the corrected abnormality score. A transmission channel is selected from the transmission channel resource pool based on the abnormality level, and the real-time status of the selected transmission channel is queried. If the current queue length or transmission delay in the real-time state exceeds the threshold, the alternative transmission channels are switched sequentially from high to low according to the corrected anomaly score until an available channel that meets the transmission requirements is switched as the target transmission channel.

10. A real-time notification system for abnormal events in a smart key cabinet, characterized in that, The system is used to execute a real-time notification method for abnormal events of a smart key cabinet as described in any one of claims 1-9, the system comprising: The key identifier set determination module is used to obtain the user's identity authentication information and determine the set of key identifiers that the user can operate on the smart key cabinet based on the identity authentication information. The data analysis module is used to control the smart key cabinet to open the cabinet door corresponding to each key identifier based on the set of key identifiers, and to collect the radio frequency signal and usage range of each key identifier in real time; The spatial trajectory jump data generation module is used to generate inventory change time-series data for each cabinet door based on the radio frequency signal, and to generate spatial trajectory jump data for each key based on the usage scope and the inventory change time-series data. The abnormal event information analysis module is used to align the spatial trajectory jump data with the opening and closing status data of each cabinet door on the time axis to obtain status alignment data, and to perform correlation analysis between the key identifier set and the status alignment data to obtain abnormal event information of the smart key cabinet. The transmission channel filtering module is used to transmit the abnormal event information to a preset dual-layer detection unit. The outer layer of the dual-layer detection unit filters the abnormal event information, and the inner layer of the dual-layer detection unit filters the filtered abnormal event information to obtain the target transmission channel. An abnormal event notification module is used to trigger the notification method corresponding to the target transmission channel and use the notification method to notify the remote management platform of the abnormal event information.

Images