Dynamic scheduling elevator intelligent rescue system based on AI camera visual recognition

Abstract

The application discloses a dynamic scheduling elevator intelligent rescue system based on an AI camera visual recognition, and belongs to the technical field of elevator safety monitoring. The system comprises an AI camera visual recognition module, an elevator state acquisition module, a fault determination module, a nearby dynamic scheduling module, a rescue linkage execution module, a privacy security management module and a cloud platform management center. The AI camera visual recognition module, the elevator state acquisition module, the fault determination module, the nearby dynamic scheduling module, the rescue linkage execution module, the privacy security management module and the cloud platform management center are arranged, so that intelligent identification, rapid determination, accurate scheduling and standardized rescue of an elevator trapping event are realized. Through automatic and accurate identification of the elevator trapping event, through AI visual recognition and fusion judgment of elevator operation state data, the passenger does not need to actively report an alarm, and the reliability and response speed of trapping detection are greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of elevator safety monitoring technology, specifically referring to a dynamic scheduling intelligent rescue system for elevators based on AI camera visual recognition. Background Technology

[0002] Elevators are indispensable vertical transportation equipment in modern buildings. However, elevator entrapment is a frequent occurrence in daily life. When people are trapped, they either need to ring an alarm to alert others or call for help, but this often results in problems such as delayed alarms, false alarms, missed alarms, inaccurate dispatch of rescue personnel, and slow response. Existing traditional monitoring systems only allow video viewing and cannot automatically identify the entrapment status, leaving the elevator's location unknown. Some intelligent systems simply push information without accurately locating the elevator, personnel, their qualifications, or their availability for dynamic dispatch, leading to low rescue efficiency. Furthermore, elevator car monitoring generally suffers from insufficient privacy protection and non-standard data retention. To address these issues, a dynamic dispatch intelligent elevator rescue system based on AI camera visual recognition is proposed. Summary of the Invention

[0003] The purpose of this invention is to provide a dynamic scheduling intelligent rescue system for elevators based on AI camera visual recognition, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a dynamic scheduling elevator intelligent rescue system based on AI camera visual recognition, comprising an AI camera visual recognition module, an elevator status acquisition module, a fault determination module, a nearby dynamic scheduling module, a rescue linkage execution module, a privacy and security management module, and a cloud platform management center.

[0005] The camera visual recognition module is configured in the elevator car to collect the car video stream in real time. AI edge computing is used to identify elevator stoppage, excessive time spent by people, number of trapped people, dangerous behaviors such as prying open doors / climbing / falling, abnormal shaking of the car, and camera obstruction status.

[0006] The elevator status acquisition module communicates with the elevator controller to collect real-time data on operating status, door lock status, leveling signal, floor position, speed, fault code, and maintenance mode signal.

[0007] The fault determination module is connected to the AI ​​visual recognition module and the elevator status acquisition module. It integrates visual features and elevator electrical signals to determine that the elevator stops operating and people are trapped in the car when it is not under maintenance, and outputs a entrapment alarm signal.

[0008] The nearby dynamic dispatch module has a built-in database of rescue personnel (qualifications, positions, real-time location GPS / Bluetooth). After receiving a distress alarm, it sorts the personnel by distance, qualifications, and busy / idle status and automatically generates the optimal rescue dispatch plan.

[0009] The rescue linkage execution module includes car voice broadcast, two-way intercom, real-time background video monitoring, multi-level alarm push (App / SMS / telephone), electronic rescue work order, and rescue progress tracking.

[0010] The privacy and security management module is used to perform AI analysis without storing videos under normal circumstances; after a person entrapment incident is triggered, the video is automatically encrypted and can only be viewed by authorized personnel; the video records are automatically and periodically deleted after the rescue is completed.

[0011] The cloud platform management center communicates with the above modules to achieve global real-time monitoring, alarm management, personnel dispatch dashboard, rescue ledger archiving, statistical analysis, and access control.

[0012] Preferably, the fault determination module adopts a dual-condition fusion determination: if the elevator stops operating at a non-level floor, the door lock is disconnected, and the AI ​​identifies that someone has been lingering for a period of time greater than or equal to a preset time, such as ≥90 seconds, and the fault is not in maintenance mode, it is determined to be a valid entrapment.

[0013] Preferably, the nearby dynamic scheduling module supports three levels of scheduling: priority to property management personnel with certificates, nearby maintenance personnel, and security personnel. If no order is accepted within a preset time, the order will be automatically upgraded and reassigned, such as after 3 minutes.

[0014] Preferably, the AI ​​camera visual recognition module supports human posture recognition, door-opening action detection, people counting, fall detection, and occlusion detection.

[0015] Preferably, the privacy and security management module employs video desensitization, face blurring, tiered authorization, and automatic deletion mechanisms with preset time intervals, such as 7 days.

[0016] Preferably, the rescue linkage execution module supports remote assistance and guidance for elevator power outage, brake release, and manual operation, as well as audio and video recording of the entire rescue process.

[0017] A preferred method for dynamic scheduling of intelligent elevator rescue based on AI camera visual recognition includes the following steps:

[0018] S1. AI vision real-time recognition of car status and passenger behavior;

[0019] S2. Collect elevator operation and fault signals;

[0020] S3. Multi-source data fusion to determine trapped individuals;

[0021] S4. Intelligent dispatching and order assignment to nearby personnel;

[0022] S5. Remote reassurance and risk intervention;

[0023] S6. Rescue Execution and Closed-Loop Archiving;

[0024] S7. Privacy and security management.

[0025] The AI ​​camera visual recognition module, with cameras positioned inside the elevator car using latitude and longitude positioning (employing edge AI cameras), enables real-time identification of the number of people inside the car, their status, analysis of people's requests for help and calls for help, elevator door closing status, designated target detection alarms, AI intelligent two-way voice intercom, and voice reassurance functions. It also tracks whether people are stuck, requesting help, calling for help, door closing status, inactivity exceeding a set time, or trapped behavior via intelligent voice.

[0026] The elevator status acquisition module collects elevator operation signals in real time, including: elevator prohibition signal, leveling signal, door operator signal, running direction signal, stop signal, overspeed signal, and uncertain running direction signal. It also collects signals for non-leveling, stopped status, elevator prohibition signal, overspeed signal, and power outage signal.

[0027] The fault determination module collects fault code signals, including: no elevator fault, safety circuit disconnection fault, door operator fault, car in unlocked area fault, car unexpected movement fault, speed governor activation, wire rope broken strand fault, elevator running stop fault, brake fault, and overshoot / undershoot fault. The system combines an AI camera visual recognition module, an elevator status acquisition module, and a fault determination module; if all three conditions are met, a person is identified as trapped, eliminating false alarms.

[0028] The nearby dynamic scheduling module is activated immediately after a person is trapped, and through multi-dimensional intelligent calculation, it optimizes the rescue personnel.

[0029] The system automatically matches and notifies nearby maintenance and rescue personnel and property management staff based on the geographical location of cameras deployed inside the elevator car. Upon receiving a rescue request, the system automatically dispatches the task to the first maintenance and rescue personnel; if there is no response, it dispatches to the second maintenance and rescue personnel, eliminating the need for manual assignment. This provides the fastest dispatch capability, covering multi-level rescue force deployment, and is suitable for scenarios with diverse rescue response needs.

[0030] The system automatically dispatches and notifies the best rescue personnel based on factors such as the current online / busy status of nearby project maintenance personnel, the elevator brand they mainly repair, their historical rescue response speed, years of service, rescue success rate, and other factors.

[0031] Government regulatory authorities can assign and dispatch all rescue personnel and level-two rescue personnel based on the elevator's location using latitude and longitude coordinates.

[0032] The rescue linkage execution module automatically pushes the following information to the responding rescue personnel: elevator location, floor location, number of trapped people, on-site video, one-click voice reassurance, navigation rescue progress, and real-time monitoring and transmission of navigation rescue route.

[0033] Once rescuers respond to a work order, this module immediately initiates full-process linkage support, allowing rescuers to obtain on-site information in advance and carry out rescues accurately and efficiently. It is applicable to all on-site implementation stages of rescue operations for trapped individuals.

[0034] The privacy and security management module states that the AI ​​camera only performs behavior recognition and does not collect or upload facial information.

[0035] Event information will not be pushed out or live video will not be broadcast if the person is not trapped or if there is no elevator malfunction.

[0036] The video data will be automatically deleted one month after the incident ends.

[0037] Meets personal information protection requirements.

[0038] This module adheres to data compliance requirements throughout the process, while also considering rescue monitoring and privacy protection. It is suitable for all locations with strict requirements for personal information security, and completely solves the pain point of privacy leakage in traditional elevator monitoring.

[0039] The cloud platform management center describes the daily management scenarios for property / maintenance units.

[0040] The cloud platform enables unified real-time monitoring of all elevators within the jurisdiction, full recording of entrapment incidents, statistical analysis of rescue data, automatic generation of reports such as rescue duration, response rate, and fault type, optimization of rescue personnel dispatching plans, and improvement of elevator operation and maintenance efficiency.

[0041] In a scenario where government regulatory departments oversee the entire region, they can use the cloud platform's data visualization function to view the real-time elevator operation status within the area, monitor the use of cameras and voice prompts to reassure trapped individuals, track the progress of entrapment incidents, and coordinate rescue personnel dispatch. This allows them to coordinate rescue resources across the entire region, assess the response time and rescue quality of maintenance units, and achieve standardized and intelligent supervision of elevator safety. This approach is applicable to the city / district special equipment regulatory departments' comprehensive elevator safety management.

[0042] In the data traceability and review scenario, for each trapped person rescue incident, the cloud platform retains complete event judgment, dispatch, and rescue records, which facilitates subsequent fault review, responsibility determination, and process optimization. If a rescue dispute occurs, it can provide objective data evidence and provide data support for elevator renovation, maintenance, and upgrades.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] 1. This invention achieves intelligent identification, rapid judgment, precise dispatching, and standardized rescue of elevator entrapment incidents by setting up an AI camera visual recognition module, an elevator status acquisition module, a fault judgment module, a nearby dynamic dispatching module, a rescue linkage execution module, a privacy and security management module, and a cloud platform management center.

[0045] 2. This invention enables automatic and accurate identification of elevator entrapment incidents by integrating AI visual recognition with elevator operation status data. It can quickly and accurately determine entrapment incidents without relying on passengers to actively report them, avoiding problems such as missed reports, false reports, and untimely or inaccurate information from manual alarms, thus greatly improving the reliability and response speed of entrapment detection.

[0046] 3. This invention improves the efficiency of rescue dispatch and shortens the time to arrive at the scene. The nearby dynamic dispatch module can automatically match the optimal rescue force according to the real-time location, qualifications and busy status of rescue personnel, realize the dispatch of nearby orders and multi-level linkage, significantly reduce the time spent on rescue response and arrival, and improve the efficiency of emergency response.

[0047] 4. This invention strengthens safety management during the rescue process, reduces secondary risks, and the rescue linkage execution module can conduct real-time voice reassurance, two-way communication, and risk behavior intervention to effectively dissuade dangerous actions such as prying open doors and climbing. Combined with remote monitoring and on-site guidance, it improves the safety of the rescue process and avoids accidents caused by passengers trying to save themselves.

[0048] 5. By balancing monitoring needs with passenger privacy protection, the privacy and security management module achieves AI analysis without storage under normal circumstances, encrypted recording after a person is trapped, and automatic clearing after the rescue is completed. While meeting the needs of rescue supervision and evidence collection, it strictly protects the privacy of people in the car, which meets data security and compliance requirements.

[0049] 6. This invention achieves full-process digital and visual management. The cloud platform management center uniformly supervises and records the entrapment alarm, dispatch process, rescue execution, and result archiving, forming a complete closed-loop ledger, which facilitates traceability, statistical analysis, and operation and maintenance optimization, thereby improving the level of intelligent elevator safety management.

[0050] 7. This invention features strong system versatility and wide applicability. The entire system can be adapted to both newly built elevators and the renovation of old elevators. It is compatible with elevator control systems of different brands, has flexible deployment and strong scalability, and can operate stably in various scenarios such as residential communities, commercial complexes, hospitals, and office buildings. Attached Figure Description

[0051] Figure 1This is a schematic diagram of the structure of the intelligent elevator rescue system based on AI camera visual recognition according to the present invention. Detailed Implementation

[0052] 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.

[0053] Example

[0054] Please see Figure 1 As shown, the present invention provides a technical solution including an AI camera visual recognition module, an elevator status acquisition module, a fault determination module, a nearby dynamic dispatch module, a rescue linkage execution module, a privacy and security management module, and a cloud platform management center.

[0055] The camera visual recognition module is configured in the elevator car to collect the car video stream in real time. AI edge computing is used to identify elevator stoppage, excessive time spent by people, number of trapped people, dangerous behaviors such as prying open doors / climbing / falling, abnormal shaking of the car, and camera obstruction status.

[0056] The elevator status acquisition module communicates with the elevator controller to collect real-time data on operating status, door lock status, leveling signal, floor position, speed, fault code, and maintenance mode signal.

[0057] The fault determination module is connected to the AI ​​visual recognition module and the elevator status acquisition module. It integrates visual features and elevator electrical signals to determine that the elevator stops operating and people are trapped in the car when it is not under maintenance, and outputs a entrapment alarm signal.

[0058] The nearby dynamic dispatch module has a built-in database of rescue personnel (qualifications, positions, real-time location GPS / Bluetooth). After receiving a distress alarm, it sorts the personnel by distance, qualifications, and busy / idle status and automatically generates the optimal rescue dispatch plan.

[0059] The rescue linkage execution module includes car voice broadcast, two-way intercom, real-time background video monitoring, multi-level alarm push (App / SMS / telephone), electronic rescue work order, and rescue progress tracking.

[0060] The privacy and security management module is used to perform AI analysis without storing videos under normal circumstances; after a person entrapment incident is triggered, the video is automatically encrypted and can only be viewed by authorized personnel; the video records are automatically and periodically deleted after the rescue is completed.

[0061] The cloud platform management center communicates with the above modules to achieve global real-time monitoring, alarm management, personnel dispatch dashboard, rescue ledger archiving, statistical analysis, and access control.

[0062] Preferably, the fault determination module adopts a dual-condition fusion determination: if the elevator stops operating at a non-level floor, the door lock is disconnected, and the AI ​​identifies that someone has been lingering for a period of time greater than or equal to a preset time, such as ≥90 seconds, and the fault is not in maintenance mode, it is determined to be a valid entrapment.

[0063] Preferably, the nearby dynamic scheduling module supports three levels of scheduling: priority to property management personnel with certificates, nearby maintenance personnel, and security personnel. If no order is accepted within a preset time, the order will be automatically upgraded and reassigned, such as after 3 minutes.

[0064] Preferably, the AI ​​camera visual recognition module supports human posture recognition, door-opening action detection, people counting, fall detection, and occlusion detection.

[0065] Preferably, the privacy and security management module employs video desensitization, face blurring, tiered authorization, and automatic deletion mechanisms with preset time intervals, such as 7 days.

[0066] Preferably, the rescue linkage execution module supports remote assistance and guidance for elevator power outage, brake release, and manual operation, as well as audio and video recording of the entire rescue process.

[0067] A preferred method for dynamic scheduling of intelligent elevator rescue based on AI camera visual recognition includes the following steps:

[0068] S1. AI vision real-time recognition of car status and passenger behavior;

[0069] S2. Collect elevator operation and fault signals;

[0070] S3. Multi-source data fusion to determine trapped individuals;

[0071] S4. Intelligent dispatching and order assignment to nearby personnel;

[0072] S5. Remote reassurance and risk intervention;

[0073] S6. Rescue Execution and Closed-Loop Archiving;

[0074] S7. Privacy and security management.

[0075] The AI ​​camera visual recognition module, with cameras positioned inside the elevator car using latitude and longitude positioning (employing edge AI cameras), enables real-time identification of the number of people inside the car, their status, analysis of people's requests for help and calls for help, elevator door closing status, designated target detection alarms, AI intelligent two-way voice intercom, and voice reassurance functions. It also tracks whether people are stuck, requesting help, calling for help, door closing status, inactivity exceeding a set time, or trapped behavior via intelligent voice.

[0076] The elevator status acquisition module collects elevator operation signals in real time, including: elevator prohibition signal, leveling signal, door operator signal, running direction signal, stop signal, overspeed signal, and uncertain running direction signal. It also collects signals for non-leveling, stopped status, elevator prohibition signal, overspeed signal, and power outage signal.

[0077] The fault determination module collects fault code signals, including: no elevator fault, safety circuit disconnection fault, door operator fault, car in unlocked area fault, car unexpected movement fault, speed governor activation, wire rope broken strand fault, elevator running stop fault, brake fault, and overshoot / undershoot fault. The system combines an AI camera visual recognition module, an elevator status acquisition module, and a fault determination module; if all three conditions are met, a trapped person event is determined, eliminating false alarms.

[0078] The nearby dynamic scheduling module is activated immediately after a person is trapped, and through multi-dimensional intelligent calculation, it optimizes the rescue personnel.

[0079] The system automatically matches and notifies nearby maintenance and rescue personnel and property management staff based on the geographical location of cameras deployed inside the elevator car. Upon receiving a rescue request, the system automatically dispatches the task to the first maintenance and rescue personnel; if there is no response, it dispatches to the second maintenance and rescue personnel, eliminating the need for manual assignment. This provides the fastest dispatch capability, covering multi-level rescue force deployment, and is suitable for scenarios with diverse rescue response needs.

[0080] The system automatically dispatches and notifies the best rescue personnel based on factors such as the current online / busy status of nearby project maintenance personnel, the elevator brand they mainly repair, their historical rescue response speed, years of service, rescue success rate, and other factors.

[0081] Government regulatory authorities can assign and dispatch all rescue personnel and level-two rescue personnel based on the elevator's location using latitude and longitude coordinates.

[0082] The rescue linkage execution module automatically pushes the following information to the responding rescue personnel: elevator location, floor location, number of trapped people, on-site video, one-click voice reassurance, navigation rescue progress, and real-time monitoring and transmission of navigation rescue route.

[0083] Once rescuers respond to a work order, this module immediately initiates full-process linkage support, allowing rescuers to obtain on-site information in advance and carry out rescues accurately and efficiently. It is applicable to all on-site implementation stages of rescue operations for trapped individuals.

[0084] The privacy and security management module states that the AI ​​camera only performs behavior recognition and does not collect or upload facial information.

[0085] Event information will not be pushed out or live video will not be broadcast if the person is not trapped or if there is no elevator malfunction.

[0086] The video data will be automatically deleted one month after the incident ends.

[0087] Meets personal information protection requirements.

[0088] This module adheres to data compliance requirements throughout the process, while also considering rescue monitoring and privacy protection. It is suitable for all locations with strict requirements for personal information security, and completely solves the pain point of privacy leakage in traditional elevator monitoring.

[0089] The cloud platform management center describes the daily management scenarios for property / maintenance units.

[0090] The cloud platform enables unified real-time monitoring of all elevators within the jurisdiction, full recording of entrapment incidents, statistical analysis of rescue data, automatic generation of reports such as rescue duration, response rate, and fault type, optimization of rescue personnel dispatching plans, and improvement of elevator operation and maintenance efficiency.

[0091] In a scenario where government regulatory departments oversee the entire region, they can use the cloud platform's data visualization function to view the real-time elevator operation status within the area, monitor the use of cameras and voice prompts to reassure trapped individuals, track the progress of entrapment incidents, and coordinate rescue personnel dispatch. This allows them to coordinate rescue resources across the entire region, assess the response time and rescue quality of maintenance units, and achieve standardized and intelligent supervision of elevator safety. This approach is applicable to the city / district special equipment regulatory departments' comprehensive elevator safety management.

[0092] In the data traceability and review scenario, for each trapped person rescue incident, the cloud platform retains complete event judgment, dispatch, and rescue records, which facilitates subsequent fault review, responsibility determination, and process optimization. If a rescue dispute occurs, it can provide objective data evidence and provide data support for elevator renovation, maintenance, and upgrades.

[0093] In this embodiment, based on visual perception and state fusion, the entire process from identification to rescue of trapped persons is automated and efficient. The core indicators are as follows:

[0094] Anomaly detection response time ≤ 2 seconds, and entrapment event detection accuracy ≥ 95%;

[0095] The response time for rescue forces is ≤30 seconds, and the average rescue arrival time is ≤10 minutes.

[0096] The entire process of data is traceable and monitorable, meeting compliance and operation and maintenance requirements.

[0097] Core process phased implementation details:

[0098] First step: Deployment and debugging of the AI ​​camera visual recognition module;

[0099] 1. Equipment selection and deployment:

[0100] Technology Selection: Select AI cameras that support edge computing, with core parameters meeting the following requirements: resolution ≥ 1080P, frame rate 25fps, low light ≤ 0.01Lux; built-in core algorithm models for personnel stagnation, door prying, falls, and car shaking, with latency ≤ 500ms; support PoE power supply to adapt to the complex environment of elevator cars.

[0101] Installation and implementation:

[0102] Installation location: Centered on the top of the elevator car or above the side wall (2.5-3 meters from the ground), avoiding blind spots such as corners of the car and door tracks, to ensure coverage of the entire car space.

[0103] Wiring specifications: Concealed wiring is used, laid along the cable trays on the top of the car; power lines and signal lines are separated to avoid electromagnetic interference; cameras are physically isolated from the elevator control system so as not to affect the original operation of the elevator.

[0104] Permission configuration: Enable only necessary capture functions and disable irrelevant recording modes; capture data is only transmitted to the specified module and sensitive videos are not stored locally.

[0105] Algorithm debugging and optimization:

[0106] In collaboration with algorithm providers, we fine-tuned the model specifically for elevator scenarios.

[0107] Personnel lingering determination: Set a configurable time threshold (30 seconds - 5 minutes) to exclude false identification of passengers going up and down the elevator normally;

[0108] Abnormal behavior recognition: Optimize the scene adaptability of algorithms for door-opening and falling, and reduce false alarms caused by elevator vibration and changes in lighting.

[0109] On-site testing: Simulate 10+ abnormal scenarios (such as deliberately prying open doors, prolonged lingering, and simulated falls) to debug algorithm parameters and ensure recognition accuracy ≥95% and false alarm rate ≤3%.

[0110] The second step: Connecting and integrating the elevator status acquisition module with data;

[0111] Interface adaptation and data acquisition:

[0112] Integration Solution:

[0113] Interface type: Select RS485 / CAN bus interface or Ethernet interface according to the elevator main controller model; when the old elevator does not have a standard interface, install a signal acquisition converter.

[0114] Data Protocol: Compatible with elevator manufacturers' proprietary protocols (such as Mitsubishi and Otis proprietary protocols), or data can be converted via the OPCUA protocol to ensure data interoperability.

[0115] Content collection landing:

[0116] Real-time data collection: elevator operating status (going up / going down / stopped), door lock status (locked / unlocked), leveling signal (whether leveled), floor position (accurate to floor + centimeter offset);

[0117] Fault-related information: Fault codes (real-time synchronized elevator main controller fault codes), maintenance mode signals (distinguishing between normal operation and maintenance status). Data transmission: Employing encrypted transmission protocols such as HTTPS, with a data update frequency of ≥1 time / second to ensure real-time status.

[0118] Compatibility testing:

[0119] For elevators of different brands and models such as Mitsubishi, Hitachi, and Kone, interface compatibility tests were conducted to ensure a data acquisition success rate of ≥99%; elevator fault scenarios such as door lock failure and leveling failure were simulated to verify the synchronization accuracy of fault codes and status data, with no data loss or delay.

[0120] The third stage: Deployment and implementation of the fault diagnosis module algorithm;

[0121] Implementation of multi-source data fusion algorithm:

[0122] Algorithm deployment: An edge computing + cloud backup architecture is adopted, with the core fusion algorithm deployed on the local edge gateway to ensure normal judgment even when the network is disconnected.

[0123] Implementation of the judgment logic:

[0124] Input data: AI visual recognition results (abnormal behavior / person presence) + elevator status data (operating status, fault codes, maintenance mode);

[0125] Judgment rules: Implemented through logical AND operation, it must simultaneously satisfy the following four conditions:

[0126] Maintenance mode = No;

[0127] Elevator status = out of service;

[0128] Door lock status = locked;

[0129] AI visual recognition = There is someone remaining in the elevator car;

[0130] Alarm output: Generate standardized entrapment alarm signals (including elevator ID, floor location, fault code, duration of entrapment, and video clip link) and push them to the nearest dynamic dispatch module in real time.

[0131] Algorithm verification and calibration:

[0132] Simulated scenario testing: Covering various scenarios such as normal shutdown, maintenance, actual entrapment, and accidental triggering, to verify the accuracy of the judgment logic:

[0133] Eliminating false alarms: When the elevator is out of service during maintenance, no alarm will be triggered even if maintenance personnel are present; Ensuring no missed alarms: When the elevator is out of service during non-maintenance periods and personnel are present, an alarm will be triggered 100% of the time. Continuous calibration: After deployment, the algorithm weights are optimized based on actual operating data to keep the false alarm rate below 3%.

[0134] The fourth step: Building the nearest dynamic scheduling module and implementing the scheduling logic;

[0135] Rescue worker database construction:

[0136] Database Design: Includes core fields—Personnel ID, Qualification Level (Basic / Intermediate / Advanced Maintenance), Skills / Specialties (e.g., elevator repair for a specific brand), Real-time Location (collected via mobile app / location terminal, with personnel authorization), Busy / Idle Status (Idle / In Rescue / On Leave), Contact Information, and Rescue Radius (default 5 km, configurable). Data Maintenance: Integrates with the maintenance unit's management system, automatically synchronizing personnel qualification expiration reminders and personnel change information; personnel location is updated every 30 seconds to ensure dispatch accuracy.

[0137] Implementation of scheduling algorithm and order dispatch mechanism:

[0138] Core algorithm: A multi-dimensional weighted algorithm based on distance priority, qualification matching, and busy / idle status, with configurable weights (default: distance 40%, qualification 30%, busy / idle status 30%).

[0139] Order dispatch process:

[0140] Upon receiving a trapped person alarm signal, the system extracts the elevator location and fault type; it filters rescue personnel within the rescue radius, whose qualifications match, and who are available from the database, and generates a candidate list by sorting them according to an algorithm; it automatically sends a dispatch instruction to the top-ranked personnel (triple reach via APP push + SMS reminder + telephone notification); if the order is not accepted within 1 minute, it is automatically dispatched to the second-ranked candidate, and a timeout reminder is sent to the maintenance administrator. Manual intervention: Administrators can view the dispatch process on the cloud platform and manually adjust the dispatch results (e.g., assigning senior maintenance personnel for special faults).

[0141] Fifth stage: Implementation and coordination of the rescue coordination execution module;

[0142] Core functionality deployment:

[0143] Voice reassurance and two-way communication:

[0144] It has a built-in standardized reassurance script library (such as "Hello, the elevator is temporarily malfunctioning. We have arranged for rescue personnel to arrive within X minutes. Please remain calm and do not pry open the doors"), and supports custom modifications. It also features microphones and speakers adapted to the acoustic environment of the elevator car (volume adjustable), enabling two-way communication with rescue personnel via the network and their mobile app / cloud platform, with clear and smooth audio quality.

[0145] Multi-level alarm push:

[0146] Target audience: rescue personnel, maintenance managers, property managers, regulatory authorities (configured as needed);

[0147] Push notification content: Elevator ID, location, fault type, number of trapped people (estimated by AI recognition), and dispatch status of rescue personnel;

[0148] Push notification channels: App notifications, SMS, phone calls, and cloud platform alert pop-ups.

[0149] Electronic work orders and progress tracking:

[0150] Electronic work orders are automatically generated (including alarm time, elevator information, fault details, and dispatcher). Rescue personnel can update the progress via the APP (order accepted → departure → arrival → rescue in progress → rescue completed). The progress is synchronized to the cloud platform in real time, allowing administrators to view the entire process time nodes (such as order acceptance time and arrival time).

[0151] Data synchronization and collaboration:

[0152] During the rescue: Data is synchronized to the cloud platform in real time (full process log) and the privacy and security module (video clips, intercom recordings);

[0153] After the rescue is completed: the work order is automatically archived, and the rescue results (such as the number of trapped people, rescue duration, and cause of failure) are synchronized to the cloud platform to form a closed loop.

[0154] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0155] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A dynamic scheduling intelligent elevator rescue system based on AI camera visual recognition, characterized in that: It includes an AI camera visual recognition module, an elevator status acquisition module, a fault determination module, a nearby dynamic dispatch module, a rescue linkage execution module, a privacy and security management module, and a cloud platform management center; The camera visual recognition module is configured in the elevator car to collect the car video stream in real time, and AI edge computing is used to identify dangerous behaviors, abnormal shaking of the car, and camera obstruction status. The elevator status acquisition module communicates with the elevator controller to collect real-time data on operating status, door lock status, leveling signal, floor position, speed, fault code, and maintenance mode signal. The fault determination module is connected to the AI ​​visual recognition module and the elevator status acquisition module. It integrates visual features and elevator electrical signals to determine that the elevator stops operating and people are trapped in the car when it is not under maintenance, and outputs a entrapment alarm signal. The nearby dynamic dispatch module has a built-in database of rescue personnel. After receiving a distress alarm, it sorts the personnel by distance, qualifications, and busy / idle status and automatically generates the optimal rescue dispatch plan. The rescue linkage execution module includes car voice broadcast, two-way intercom, real-time background video monitoring, multi-level alarm push, electronic rescue work order, and rescue progress tracking. The privacy and security management module is used to perform AI analysis without storing videos under normal circumstances; after a person entrapment incident is triggered, the video is automatically encrypted and can only be viewed by authorized personnel; the video records are automatically and periodically deleted after the rescue is completed. The cloud platform management center communicates with the above modules to achieve global real-time monitoring, alarm management, personnel dispatch dashboard, rescue ledger archiving, statistical analysis, and access control.

2. The intelligent elevator rescue system based on AI camera visual recognition according to claim 1, characterized in that: The fault determination module adopts a dual-condition fusion determination: if the elevator stops operating at a non-level floor, the door lock is disconnected, and AI identifies that someone has been lingering for a period of time greater than or equal to a preset time, and the elevator is not in maintenance mode, it is determined to be a valid trapped person.

3. The intelligent elevator rescue system based on AI camera visual recognition according to claim 1, characterized in that: The nearby dynamic scheduling module supports three levels of scheduling: priority to property management personnel with certificates, nearby maintenance personnel, and security personnel. If no order is accepted within a preset time, the system will automatically upgrade and dispatch orders.

4. The intelligent elevator rescue system based on AI camera visual recognition according to claim 1, characterized in that: The AI ​​camera visual recognition module supports human posture recognition, door-opening action detection, people counting, fall detection, and occlusion detection.

5. The intelligent elevator rescue system based on AI camera visual recognition according to claim 1, characterized in that: The privacy and security management module employs video desensitization, face blurring, tiered authorization, and automatic deletion mechanisms with preset time limits.

6. The intelligent elevator rescue system based on AI camera visual recognition according to claim 1, characterized in that: The rescue linkage execution module supports remote assistance and guidance for elevator power outage, brake release, and manual operation, as well as audio and video recording of the entire rescue process.

7. The method for intelligent elevator rescue based on AI camera visual recognition, implemented in the AI ​​camera visual recognition-based dynamic scheduling elevator intelligent rescue system according to claim 1, is characterized in that... Includes the following steps: S1. AI vision real-time recognition of car status and passenger behavior; S2. Collect elevator operation and fault signals; S3. Multi-source data fusion to determine trapped individuals; S4. Intelligent dispatching and order assignment to nearby personnel; S5. Remote reassurance and risk intervention; S6. Rescue Execution and Closed-Loop Archiving; S7. Privacy and security management.

Images