A lock opening service scene real-time monitoring method, device and storage medium
Patent Information
- Application Number
- CN202611178485.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-05
- Publication Date
- 2026-11-03
AI Technical Summary
The current regulation of locksmith services suffers from incomplete regulatory coverage, insufficient privacy protection, lack of risk warning mechanisms, and inadequate real-time data processing, making it difficult to achieve fully automated regulation and real-time privacy protection.
It employs audio and video acquisition modules, scene detection modules, privacy desensitization modules, behavior analysis modules, audio semantic analysis modules, and positioning modules. By combining real-time positioning, multi-dimensional analysis, and real-time privacy desensitization, it automatically identifies the start and end points of services, provides real-time warnings of abnormal behaviors and semantics, and achieves full-process supervision and privacy protection.
It achieves automated monitoring of the lock-unlocking service process, accurately identifies service nodes, prevents abnormal behavior, protects user privacy, supports data retrieval and integration with monitoring platforms, and meets the standardization and security requirements of special security industries.
Smart Images
Figure CN123033191A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent monitoring and privacy protection technology, specifically relating to a method, device and storage medium for real-time monitoring of lock-unlocking service scenarios. Background Technology
[0002] Locksmith services, as a special security industry directly related to the safety of people's property and residences, are required by relevant regulations to implement full-process supervision, accurately record the service process, and retain records for future reference. Currently, the supervision of locksmith services largely relies on manual recording or simple audio and video recording methods, which have many shortcomings in practical application: The regulatory coverage is incomplete. Existing methods cannot achieve automated real-time monitoring of the service process, making it difficult to accurately identify the start and end points of the service. This can easily lead to problems such as excessive recording of irrelevant content or omission of key service links, failing to meet the compliance requirements of full-process supervision.
[0003] Insufficient privacy protection. Locksmith services are mostly conducted within users' residences, and the audio and video recording process inevitably collects private information such as the user's home environment, facial features, and voice information. If the raw audio and video data is directly stored, it is very easy to cause user privacy leaks, posing significant security risks.
[0004] The risk warning mechanism is lacking. During the recording process, it is impossible to conduct real-time analysis and warnings for abnormal behaviors such as unauthorized lock picking and malicious damage to locks, as well as abnormal semantics such as maliciously demanding high prices and threatening users, making it difficult to prevent violations and security risks during the service process in a timely manner.
[0005] Insufficient real-time data processing. Existing applications typically perform privacy desensitization in batches after recording, failing to achieve real-time desensitization and making it difficult to simultaneously ensure regulatory effectiveness and privacy protection.
[0006] Therefore, there is an urgent need for a technical solution that can achieve automated supervision of the entire lock-unlocking service process, real-time privacy desensitization, accurate identification of service start and end points, and abnormal behavior and semantic warning functions, so as to solve the pain point of the difficulty in balancing supervision and privacy protection in existing technologies and meet the regulatory standards for special security services. Summary of the Invention
[0007] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method, device, and storage medium for real-time monitoring of lock-unlocking service scenarios, achieving the dual goals of full-process supervision of lock-unlocking services and protection of user privacy. It also has real-time anomaly warning capabilities, improving the standardization and security of lock-unlocking services.
[0008] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a real-time monitoring method for lock-unlocking service scenarios, used for full-process supervision of the lock-unlocking service process, comprising the following steps: Step 1: Initialization Deployment and Parameter Configuration: Deploy the audio and video acquisition module, scene detection module, privacy desensitization module, behavior analysis module, audio semantic analysis module, storage module, and positioning module on the lock-picking service device; configure the lock-picking tool feature library, abnormal semantic keyword library, and positioning matching threshold. The lock-picking tool feature library contains image features and morphological features of common lock-picking tools, the abnormal semantic keyword library contains illegal semantic keywords, and the positioning matching threshold is used to limit the allowable deviation range between the on-site positioning address and the task target address. Step 2: Automatic identification and recording of service start point: The location module collects on-site geographical location information in real time and compares it with the target location address issued by the dispatch task; when the location deviation is within the preset location matching threshold, the location matching is determined to be successful, and the device starts the audio and video monitoring mode; the audio and video acquisition module collects on-site audio and video streams in real time, and the scene detection module identifies the service start time based on the preset unlocking service start characteristics. When the start characteristics are detected, audio and video recording is automatically started, and the behavior analysis module, audio semantic analysis module and privacy desensitization module are started simultaneously, entering the real-time monitoring mode; Step 3: Real-time Scene Detection and Multi-dimensional Analysis: During the recording process, three types of real-time analysis are performed simultaneously on the audio and video streams: the scene detection module identifies scene elements in the audio and video streams, filters scenes unrelated to the unlocking service, and monitors scene stability; the behavior analysis module identifies personnel actions and determines whether there is any abnormal behavior; and the audio semantic analysis module parses the semantic and emotional features of the audio to identify abnormal semantics and abnormal emotions. Step 4: Real-time privacy desensitization: The audio and video streams are simultaneously desensitized in real time through the privacy desensitization module. The processing includes desensitization of personnel privacy and desensitization of sound privacy. Step 5: Real-time alerts and handling: When abnormal behavior, abnormal semantics, or abnormal scene are detected, the corresponding type of alert is triggered, the alert information is pushed, and an alert log is generated and stored. Step 6: Automatic identification of service end point and termination of recording: Based on the preset lock unlocking service end characteristics, the service end point is automatically identified, audio and video recording is terminated, and the operation of each analysis module and desensitization module is stopped; Step 7: Data storage and playback after anonymization: Encrypt and store the anonymized audio and video files, behavior analysis reports, audio semantic analysis reports, early warning logs, and service start and end time information, supporting playback of anonymized audio and video and integration with the public security supervision platform.
[0009] Furthermore, in step 2, the judgment priority of location address matching is higher than that of the initial feature recognition. The device will only start the audio and video monitoring mode after the location matching is successful. The unlocking service start feature in step 2 and the unlocking service end feature in step 6 are both judged by fusing at least two features to reduce the probability of false recording and omission.
[0010] Furthermore, in step 3, the behavior analysis module obtains the hand position and the object the hand is in contact with through an end-to-end hand target detection algorithm, obtains accurate hand segmentation results through a pixel-level hand segmentation model, and then inputs the multi-frame segmentation result sequence into a multimodal large model to infer and judge whether there are any abnormalities in the hand movements; the behavior analysis module also analyzes the standardization of the unlocking operation and records non-standard operations based on the usage method and operation rhythm of the unlocking tool.
[0011] Furthermore, in step 3, the audio semantic analysis module converts the audio signal into a spectrogram, extracts audio emotion features and semantic features, compares them with the abnormal semantic keyword library, and performs semantic understanding through a large language model to comprehensively identify abnormal semantics and abnormal emotions.
[0012] Furthermore, in step 4, the personnel privacy desensitization specifically involves: using a deep learning human body segmentation model to accurately segment all personnel regions in the audio and video stream, assigning differentiated fill colors to different personnel regions to completely cover the personnel's facial and body features, and smoothing the segmentation edges; the sound privacy desensitization specifically involves: removing the speaker's voiceprint features through audio frequency shifting and timbre reconstruction processing, while retaining the effective sounds related to the unlocking operation.
[0013] Furthermore, step 5 specifically includes: Step 51: When abnormal behavior, abnormal semantics, or scene abnormality is detected, trigger the corresponding behavior abnormality warning, semantic abnormality warning, or scene abnormality warning. Step 52: Issue an audible and visual warning signal on-site, and simultaneously push the warning information, including the warning type, warning time, and corresponding audio and video clips, to the monitoring platform and service personnel terminals; if the abnormal state continues for more than the preset time, the monitoring platform will automatically trigger manual intervention prompts. Step 53: Synchronously store all warning information and corresponding processing procedures to generate a warning log.
[0014] Furthermore, in step 7, the data is stored in an encrypted manner, and data retrieval is supported by service order number, service time, and service personnel number; the storage module presets a data retention period, and audio and video files that exceed the retention period are automatically deleted.
[0015] The present invention also provides a real-time monitoring device for unlocking service scenarios, including an embedded AI processor, a memory, and a computer program and AI model stored in the memory and executable on the embedded AI processor; when the embedded AI processor executes the computer program and AI model, it implements the steps of the real-time monitoring method for unlocking service scenarios as described in any one of claims 1 to 7.
[0016] Furthermore, it also includes audio and video acquisition equipment, speakers, communication modules, display modules, and positioning modules; the audio and video acquisition equipment is used to acquire on-site audio and video streams; the speakers are used to output early warning signals; the communication module is used to realize data communication with the monitoring platform and service personnel terminals; the display module is used to display service status and early warning information; the positioning module is used to collect on-site geographical location information in real time and compare and match it with the target location address of the dispatched task.
[0017] The present invention also provides a computer-readable storage medium storing a computer program and an AI model, wherein when the computer program and the AI model are executed by a processor, the steps of the real-time monitoring method for unlocking service scenarios described in the first aspect are implemented.
[0018] (III) Beneficial Effects Compared with the prior art, the present invention provides a method, device and storage medium for real-time monitoring of lock-unlocking service scenarios, which has the following beneficial effects: This lock-picking service scenario real-time monitoring method, equipment, and storage medium automatically identifies the service start and end points through a scenario detection module, accurately filters irrelevant content, and records only lock-picking service-related segments to avoid omissions, misrecordings, and redundant recordings. At the same time, it combines behavioral analysis and audio semantic analysis to achieve multi-dimensional real-time supervision of the service process, meeting the full-process supervision requirements of special security services, effectively standardizing lock-picking service behavior, and preventing industry chaos such as exorbitant lock-picking fees and illegal operations.
[0019] Employing real-time privacy desensitization technology, this system uses human segmentation algorithms to accurately segment individuals and fill them with differentiated colors. Combined with audio voiceprint processing technology, it removes the speaker's identity features, achieving complete indistinguishability between the person's image and voice. At the same time, it fully preserves the audio and video information related to the unlocking operation, balancing the effectiveness of supervision and privacy protection. This addresses the pain point of privacy leakage in existing technologies and aligns with the development trend of audio and video privacy protection.
[0020] By using behavioral analysis and audio semantic analysis, abnormal behaviors and semantics can be identified in real time, triggering timely warnings and pushing them to the monitoring platform and service personnel terminals. This enables rapid response and handling of abnormal situations, effectively preventing security risks such as forced lock picking and malicious threats, and protecting users' property and personal safety.
[0021] It adopts a real-time de-identification and real-time analysis processing method, supports data retrieval, playback and connection with the regulatory platform, which facilitates regulatory verification and service traceability; at the same time, it sets up an automatic data expiration and deletion mechanism to reduce storage costs and meet data compliance management requirements. Attached Figure Description
[0022] Figure 1 This is a flowchart illustrating a real-time monitoring and early warning method for a lock-unlocking service process according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the hand movement behavior recognition process of the behavior analysis module in an embodiment of the present invention; Figure 3 This is a block diagram of the hardware device structure for the unlocking service process in an embodiment of the present invention; Figure 4 Schematic diagram of service start-up process with dual verification and recording; Figure 5 A schematic diagram of a two-dimensional real-time privacy de-identification process; Figure 6 Schematic diagram of audio semantic anomaly recognition process; Figure 7 Schematic diagram of multi-source anomaly early warning processing flow; Figure 8 A schematic diagram of the service termination determination and data archiving process. Detailed Implementation
[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1:
[0024] like Figure 1 As shown in the figure, this embodiment provides a real-time monitoring method for lock-unlocking service scenarios. The specific implementation steps are as follows: Step 1: Initial Deployment and Parameter Configuration. Deploy a high-definition camera and dual microphones as audio and video acquisition modules on the portable lock-unlocking monitoring terminal device. Simultaneously deploy scene detection modules, privacy desensitization modules, behavior analysis modules, audio semantic analysis modules, early warning modules, encrypted storage modules, and GPS / BeiDou dual-mode positioning modules; and configure the relevant model libraries and parameters. Lock picking tool feature library: Input the image features and morphological features of common lock picking tools such as single-sided keys, cross keys, plum blossom keys, lock picking guns, crowbars, etc., and train to build a tool recognition model; Service behavior feature library: Input hand action features of normal unlocking behavior (inserting tools into the keyhole, turning tools, adjusting locks) and action features of abnormal behavior (violent lock picking, malicious lock smashing, unauthorized key duplication), and train a hand behavior recognition model based on deep learning algorithm; Abnormal semantic keyword library: Input semantic keywords such as exorbitant prices, mandatory payment, no opening without payment, etc., which are illegal demands, threats, and insults, and establish a semantic comparison model; Human body segmentation model: Set the segmentation accuracy to no less than 90%, and configure contour dilation and burr removal preprocessing logic to ensure the accuracy of human body segmentation; Location matching threshold: The allowable range of location deviation is set to 30 meters. The location module collects the latitude and longitude information of the site in real time and compares it with the latitude and longitude of the target location address issued by the dispatch task. If the deviation is ≤30 meters, the location matching is considered successful.
[0025] Step 2: Automatic Identification and Recording of Service Starting Point. The positioning module collects the latitude and longitude information of the on-site geographical location in real time and compares it with the latitude and longitude of the target location address issued by the dispatch task. When the difference between the two is ≤30 meters, the positioning is considered to be successful, and the device automatically starts to enter the audio and video monitoring mode. Subsequently, the high-definition camera and dual microphones with noise reduction collect the on-site audio and video stream in real time. The scene detection module analyzes the audio and video stream in real time. When both the appearance of a lock-picking tool in the field of view and the service personnel issuing a voice related to starting to unlock the door are detected simultaneously, it is determined to be the service starting point, and audio and video recording is automatically started. At the same time, the behavior analysis module, audio semantic analysis module, and privacy desensitization module are activated. If the positioning matching fails, the device maintains a low-power standby state and does not start the monitoring and recording functions. If the positioning matching is successful but only a single starting feature is detected, the audio and video monitoring mode is maintained, and recording is not started to avoid false triggering.
[0026] Step 3: Real-time scene detection and multi-dimensional analysis: Step 31. Scene Detection: The scene detection module identifies scene elements in the audio and video stream in real time. If it detects scenes unrelated to the lock-unlocking service, such as service personnel resting or the acquisition device being blocked, it automatically filters the corresponding audio and video segments, retaining only the images and sounds related to lock operation. At the same time, it monitors the stability of the service scene in real time. If a sudden change in the scene occurs, such as the acquisition device being blocked, it triggers a scene anomaly prompt.
[0027] Step 32. Behavior Analysis: The behavior analysis module performs target detection and tracking of hands, analyzing the hand movements of people in the video stream in real time. The specific recognition process is as follows: Figure 2As shown: First, an end-to-end hand target detection algorithm is used to obtain the hand position and the object the hand is in contact with. Then, a pixel-level hand segmentation model is used to obtain accurate hand segmentation results. Finally, the multi-frame segmentation result image sequence is input into a multimodal large model to infer and determine whether there are any abnormalities in hand movements. If it is detected that a person is using a crowbar to forcibly pry open the lock body, it is determined as abnormal behavior, and a signal is sent to the early warning module simultaneously. At the same time, the standardization of the unlocking operation is analyzed. If improper use of tools is detected, an operation non-standard warning is triggered.
[0028] Step 33. Audio Semantic Analysis: The audio semantic analysis module converts the acquired audio stream into a spectrogram, extracts short-term detail features (such as the sound of tools turning) and long-term semantic features (such as the dialogue of service personnel), and compares them with an abnormal semantic keyword database. Simultaneously, it uses a large language model to perform deep language understanding on the extracted semantic information to further confirm scene and language anomalies. If the module detects service personnel stating "I won't unlock the door without 500 yuan," it is determined to be abnormal semantics, and a signal is simultaneously sent to the warning module.
[0029] Step 4. Real-time privacy desensitization: Step 41. Personnel Privacy Desensitization: The privacy desensitization module uses a deep learning human body segmentation model to accurately segment all personnel areas, such as service personnel and users, in the audio and video stream. Different personnel areas are filled with high-contrast differentiated colors such as blue, green, and yellow. After filling, the facial and body features of the personnel are completely indistinguishable. At the same time, the segmentation edges are smoothed to avoid obvious segmentation marks and abrupt filling.
[0030] Step 42. Voice privacy desensitization: The audio stream is processed in real time. Through audio frequency shift and timbre reconstruction technology, the frequency characteristics and timbre of the audio are adjusted to remove the speaker's voiceprint characteristics. The processed sound cannot identify the speaker's identity, but can clearly restore the effective sounds related to unlocking operations such as the sound of tools turning and locks opening.
[0031] Step 5. Real-time early warning and handling: Step 51. Warning Trigger: When the warning module receives abnormal behavior signals from the behavior analysis module, abnormal semantic signals from the audio semantic analysis module, or abnormal scene signals from the scene detection module, it immediately triggers the corresponding type of warning.
[0032] Step 52. Warning Execution: The warning module controls the speaker and indicator lights on the device to emit audible and visual warning signals; at the same time, it pushes the warning information, including the warning type, warning time, and corresponding audio and video clips, to the monitoring platform and the mobile terminals of service personnel; the monitoring platform simultaneously pops up a prompt for manual intervention, and staff can view the desensitized audio and video in real time and remotely remind service personnel to correct their behavior; if the abnormal state continues for more than the preset time, the monitoring prompt level will be automatically upgraded.
[0033] Step 53. Warning Record: Record all warning information, processing time, and processing results to the warning log and store them synchronously and encrypted in the storage module.
[0034] Step 6: Automatic service end point identification and recording termination scene detection module continuously monitors the audio and video stream. When both the lock is successfully opened (lock cylinder rotation signal) and the service personnel put away the unlocking tools are detected simultaneously, it is determined to be the service end point. The audio and video recording is automatically terminated, the work of each analysis module and the desensitization module is stopped, and the device enters standby mode. If the service personnel are detected to have paused operation for more than 3 minutes without reasonable semantic explanation, the recording is automatically paused and will continue to be recorded after the operation is detected to be resumed.
[0035] Step 7: Data Storage and Playback after De-identification. De-identified audio and video files, behavior analysis reports, audio semantic analysis reports, warning logs, and service start and end time information are stored in the device's local storage module using AES encryption. The stored content is automatically uploaded to the monitoring platform after the device connects to a designated Wi-Fi network in a specified location. The system supports data retrieval by keywords such as service order number, service time, and service personnel number. Supervisory personnel can play back the de-identified audio and video through the monitoring platform to clearly view the unlocking process, but cannot identify the personnel's appearance or voice. The data retention period is set to 6 months; audio and video files exceeding this period are automatically deleted. Example 2:
[0036] like Figure 3 As shown, this embodiment provides a real-time monitoring device for lock-unlocking service scenarios, used to implement the monitoring method described in Embodiment 1. The device includes an embedded AI processor, a memory, an audio / video acquisition module, an early warning module, a communication module, a display module, and a positioning module.
[0037] The audio and video acquisition module includes a high-definition camera and dual microphones for acquiring on-site audio and video streams; The warning module includes LED indicator lights and a horn, used to emit audible and visual warning signals; The communication module integrates 4G / 5G and WIFI communication units to realize data communication with the monitoring platform and service personnel terminals, and to complete the push of early warning information and the uploading of stored data; The display module is a small LCD screen used to display service status (recording, standby, monitoring) and warning information; The memory stores computer programs and various AI algorithm models; When the embedded AI processor executes the computer program and AI model, it implements all the steps of the real-time monitoring method for the unlocking service scenario described in Embodiment 1. The positioning module uses a GPS / BeiDou dual-mode positioning unit to collect on-site geographical location information in real time, compare and match it with the target positioning address issued by the dispatch task, trigger the audio and video monitoring mode to start, and the positioning accuracy is ≤30 meters, which meets the on-site positioning matching requirements. Example 3:
[0038] This embodiment provides a computer-readable storage medium storing a computer program and an AI model. When the computer program and AI model are executed by a processor, they implement the steps of the real-time monitoring method for unlocking service scenarios described in Embodiment 1. The computer-readable storage medium can be a solid-state drive, a portable hard drive, or other physical media capable of storing programs and models.
Claims
1. A real-time monitoring method for lock-unlocking service scenarios, used for full-process supervision of the lock-unlocking service process, characterized in that, Includes the following steps: Step 1: Initialization Deployment and Parameter Configuration: Deploy the audio and video acquisition module, scene detection module, privacy desensitization module, behavior analysis module, audio semantic analysis module, storage module, and positioning module on the lock-picking service device; configure the lock-picking tool feature library, abnormal semantic keyword library, and positioning matching threshold. The lock-picking tool feature library contains image features and morphological features of common lock-picking tools, the abnormal semantic keyword library contains illegal semantic keywords, and the positioning matching threshold is used to limit the allowable deviation range between the on-site positioning address and the task target address. Step 2: Automatic identification and recording of service start point: The location module collects on-site geographical location information in real time and compares it with the target location address issued by the dispatch task; when the location deviation is within the preset location matching threshold, the location matching is determined to be successful, and the device starts the audio and video monitoring mode; the audio and video acquisition module collects on-site audio and video streams in real time, and the scene detection module identifies the service start time based on the preset unlocking service start characteristics. When the start characteristics are detected, audio and video recording is automatically started, and the behavior analysis module, audio semantic analysis module and privacy desensitization module are started simultaneously, entering the real-time monitoring mode; Step 3: Real-time Scene Detection and Multi-dimensional Analysis: During the recording process, three types of real-time analysis are performed simultaneously on the audio and video streams: the scene detection module identifies scene elements in the audio and video streams, filters scenes unrelated to the unlocking service, and monitors scene stability; the behavior analysis module identifies personnel actions and determines whether there are any abnormal behaviors. The audio semantic analysis module parses the semantic and emotional features of the audio to identify abnormal semantics and abnormal emotions; Step 4: Real-time privacy desensitization: The audio and video streams are simultaneously desensitized in real time through the privacy desensitization module. The processing includes desensitization of personnel privacy and desensitization of sound privacy. Step 5: Real-time alerts and handling: When abnormal behavior, abnormal semantics, or abnormal scene are detected, the corresponding type of alert is triggered, the alert information is pushed, and an alert log is generated and stored. Step 6: Automatic identification of service end point and termination of recording: Based on the preset lock unlocking service end characteristics, the service end point is automatically identified, audio and video recording is terminated, and the operation of each analysis module and desensitization module is stopped; Step 7: Data storage and playback after anonymization: Encrypt and store the anonymized audio and video files, behavior analysis reports, audio semantic analysis reports, early warning logs, and service start and end time information, supporting playback of anonymized audio and video and integration with the public security supervision platform.
2. The real-time monitoring method for a lock-unlocking service scenario according to claim 1, characterized in that, In step 2, the location address matching judgment has a higher priority than the initial feature recognition. The device will only start the audio and video monitoring mode after the location matching is successful. The unlocking service start feature in step 2 and the unlocking service end feature in step 6 are both judged by using at least two feature fusion methods to reduce the probability of false recording and omission.
3. The real-time monitoring method for a lock-unlocking service scenario according to claim 1, characterized in that, In step 3, the behavior analysis module obtains the hand position and the object the hand is in contact with through an end-to-end hand target detection algorithm, obtains accurate hand segmentation results through a pixel-level hand segmentation model, and then inputs the multi-frame segmentation result sequence into a multimodal large model to infer and judge whether there are any abnormalities in the hand movements. The behavior analysis module also analyzes the standardization of the unlocking operation and records non-standard operations based on the usage and operation rhythm of the unlocking tool.
4. The real-time monitoring method for a lock-unlocking service scenario according to claim 1, characterized in that, In step 3, the audio semantic analysis module converts the audio signal into a spectrogram, extracts audio emotion features and semantic features, compares them with the abnormal semantic keyword library, and performs semantic understanding through a large language model to comprehensively identify abnormal semantics and abnormal emotions.
5. The real-time monitoring method for a lock-unlocking service scenario according to claim 1, characterized in that, In step 4, the personnel privacy desensitization is specifically as follows: a deep learning human body segmentation model is used to accurately segment all personnel regions in the audio and video stream, and different fill colors are assigned to different personnel regions to completely cover the facial and body features of the personnel, and the segmentation edges are smoothed; the sound privacy desensitization is specifically as follows: the speaker's voiceprint features are removed through audio frequency shift and timbre reconstruction processing, while retaining the effective sounds related to the unlocking operation.
6. The real-time monitoring method for a lock-unlocking service scenario according to claim 1, characterized in that, Step 5 specifically includes: Step 51: When abnormal behavior, abnormal semantics, or scene abnormality is detected, trigger the corresponding behavior abnormality warning, semantic abnormality warning, or scene abnormality warning. Step 52: Issue an audible and visual warning signal on-site, and simultaneously push the warning information, including the warning type, warning time, and corresponding audio and video clips, to the monitoring platform and service personnel terminals; if the abnormal state continues for more than the preset time, the monitoring platform will automatically trigger manual intervention prompts. Step 53: Synchronously store all warning information and corresponding processing procedures to generate a warning log.
7. The real-time monitoring method for a lock-unlocking service scenario according to claim 1, characterized in that, In step 7, the data is stored in an encrypted manner and can be retrieved by service order number, service time, and service personnel number. The storage module has a preset data retention period, and audio and video files that exceed the retention period are automatically deleted.
8. A real-time monitoring device for lock-unlocking service scenarios, characterized in that: It includes an embedded AI processor, a memory, and a computer program and AI model stored in the memory and executable on the embedded AI processor; when the embedded AI processor executes the computer program and AI model, it implements the steps of the real-time monitoring method for unlocking service scenarios as described in any one of claims 1 to 7.
9. A real-time monitoring device for a lock-unlocking service scenario according to claim 8, characterized in that, It also includes audio and video acquisition equipment, speakers, communication modules, display modules, and positioning modules; the audio and video acquisition equipment is used to acquire on-site audio and video streams; the speakers are used to output early warning signals; the communication modules are used to realize data communication with the monitoring platform and service personnel terminals; the display modules are used to display service status and early warning information; and the positioning modules are used to collect on-site geographical location information in real time and compare and match it with the target location address of the dispatched task.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program and an AI model. When the computer program and the AI model are executed by a processor, they implement the steps of the real-time monitoring method for unlocking service scenarios as described in any one of claims 1 to 7.