An AI recognition-based home care management and control method and system

Abstract

This application proposes an AI-based home care management method and system. Belonging to the field of home care, the method includes: collecting the facial image of the visiting nurse and comparing it with a nurse qualification database for verification; upon successful verification, the nursing service is initiated; collecting real-time video streams including the nurse's operating area and the patient's activity area; performing local real-time analysis of the video stream on an edge computing terminal, including determining whether there are any operational violations; determining whether there are any safety hazards or high-risk patients; when operational violations or safety hazards are determined, the edge computing terminal immediately triggers an alert; after the service is completed, the entire video data is encrypted and uploaded to cloud storage, and the entire data undergoes batch quality control verification to generate a safety verification report; based on high-frequency violation types, training courses from the nurse training system are automatically matched and pushed to the corresponding nurse's portal. This enables intelligent, real-time, and traceable safety verification throughout the entire home care process.

Description

Technical Field

[0001] This application relates to the field of home care, and more specifically, to a home care management method and system based on AI recognition. Background Technology

[0002] With the widespread adoption of "Internet Plus" nursing services, home-based nursing care has become an important supplement to primary healthcare services. However, existing home-based nursing safety management has the following main shortcomings: Blind spots and inefficiency in quality control supervision: Traditional safety inspections mainly rely on manual post-event spot checks, which have limited coverage and poor timeliness, and cannot achieve real-time closed-loop supervision of the entire nursing process, resulting in non-standard operations and difficulty in timely detection of risks and hidden dangers.

[0003] Delayed risk warnings and difficulties in providing evidence in disputes: Existing technologies lack real-time intelligent identification methods, and the verification of nurses' qualifications, operating procedures, and environmental risks relies mainly on manual judgment, resulting in serious delays in risk warnings; at the same time, the nursing process lacks objective, complete, and traceable video records, making it difficult to provide effective evidence in the event of adverse events or medical disputes.

[0004] Therefore, there is an urgent need for a security verification method that can realize intelligent, real-time, and traceable security checks throughout the entire home care process. Summary of the Invention

[0005] The purpose of this application is to provide a home care management method and system based on AI recognition, which can realize intelligent, real-time, and traceable security verification of the entire home care process.

[0006] This application is implemented as follows: Firstly, this application provides a home care management method based on AI recognition, comprising the following steps: S1: Detects a nursing service request, obtains the patient's signed informed consent authorization information, and collects the facial image of the visiting nurse for comparison and verification with the nurse's professional qualification database. If the verification is successful, the nursing service is initiated. S2: Acquire real-time video streams including nurse operating areas and patient activity areas; S3: The edge computing terminal performs local real-time analysis of the real-time video stream, including: Based on the target detection model, key targets and their behavioral sequences in nursing operations are identified, and matched with a pre-set nursing operation standard rule base to determine whether there are any operational violations. Based on scene semantic segmentation, identify environmental risk sources and abnormal patient postures, match them with preset risk level rules, and determine whether there are safety hazards or high-risk patients. S4: When an operational violation or security risk is detected, the edge computing terminal will immediately trigger an alert and upload the violation / risk information and corresponding video clips to the cloud. S5: After the service is completed, the entire video data will be encrypted and uploaded to the cloud storage. The cloud AI platform will conduct batch quality control and verification of the entire data and generate a security verification report. S6: Based on the high-frequency violation types in the security audit report, automatically match the training courses in the nurse training system and push them to the corresponding nurse's port.

[0007] Based on the first aspect, the step S1 of collecting the facial image of the nurse who makes the home visit and comparing it with the nurse's professional qualification database includes: obtaining the order service items, automatically matching the nurse's special training qualification certificate according to the order service items, and failing the verification if there is no corresponding qualification or the certificate has expired.

[0008] Based on the first aspect, step S2, which involves acquiring a real-time video stream including the nurse's operating area and the patient's activity area, includes: Real-time video streams, including the nurse's work area and the patient's activity area, are captured using wearable badge cameras worn by nurses and / or portable mobile cameras fixed in the work area.

[0009] Based on the first aspect, the target detection model in step S3 is a lightweight YOLOv8-Nano model, and the nursing operation standard rule base includes hand hygiene compliance rules, aseptic operation rules, three checks and seven rights execution rules, medical waste classification rules, and special operation procedure rules; behavioral temporal analysis includes tracking the continuous frame motion trajectory of the nurse's hands, calculating the motion duration and motion integrity.

[0010] Based on the first aspect, the environmental risk sources in step S3 include slippery ground, debris accumulation, insufficient light, and flammable and explosive materials; abnormal patient postures include high-risk actions such as falls, bed falls, and tube slippage.

[0011] Based on the first aspect, the warning in step S4 is divided into three levels: Level 1 warning is given to nurses by voice reminder at the edge terminal without involving back-end intervention; Level 2 warning sends a pop-up notification to the quality control terminal, which is followed up by quality control personnel; Level 3 warning triggers the emergency linkage process and automatically contacts the nearest medical staff for support.

[0012] Based on the first aspect, in step S5, the cloud-based AI platform performs batch quality control verification on all data, including: automatically calculating quality control scores by order dimension, filtering low-scoring orders and violations with confidence levels below the threshold for manual review, and generating quality control statistical reports by nurse dimension and region dimension.

[0013] Secondly, this application provides a home care management system based on AI recognition, including: The verification module is used to detect nursing service requests, obtain the informed consent authorization information signed by the patient, and collect the facial image of the nurse making the home visit and compare it with the nurse's professional qualification database for verification. If the verification is successful, the nursing service is initiated. The acquisition module is used to acquire real-time video streams including the nurse's operating area and the patient's activity area; The analysis module is used by the edge computing end to perform local real-time analysis of the real-time video stream. The analysis includes: Based on the target detection model, key targets and their behavioral sequences in nursing operations are identified, and matched with a pre-set nursing operation standard rule base to determine whether there are any operational violations. Based on scene semantic segmentation, identify environmental risk sources and abnormal patient postures, match them with preset risk level rules, and determine whether there are safety hazards or high-risk patients. The early warning module is used to trigger an early warning immediately when an operational violation or security risk is detected, and at the same time upload the violation / risk information and corresponding video clips to the cloud. The verification module is used to encrypt and upload all video data to cloud storage after the service is completed. The cloud AI platform performs batch quality control verification on all data and generates a security verification report. The training module is used to automatically match training courses from the nurse training system based on the high-frequency violation types in the safety audit report and push them to the corresponding nurses' ports.

[0014] Thirdly, this application provides an electronic device, characterized in that it comprises: Memory, used to store one or more programs; processor; The above method is implemented when one or more programs are executed by the processor.

[0015] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0016] Compared with the prior art, this application has at least the following advantages or beneficial effects: This invention provides a home care management method and system based on AI recognition. It verifies the facial image of the visiting nurse against a nurse qualification database; upon successful verification, the nursing service is initiated. A camera captures real-time video streams including the nurse's operating area and the patient's activity area. An edge computing terminal performs local real-time analysis of the video stream, including determining whether there are operational violations, safety hazards, or high-risk patients. When an operational violation or safety hazard is detected, the edge computing terminal immediately triggers an alert and uploads the violation / risk information and corresponding video clips to the cloud. All video data is encrypted and uploaded to cloud storage. The cloud AI platform performs batch quality control verification on all data, generating a safety verification report. Based on the high-frequency violation types in the safety verification report, training courses from the nurse training system are automatically matched and pushed to the corresponding nurses. This allows for rapid identification of violations, immediate intervention, and prevention of escalation. AI-automated scoring and report generation provide data support for nurse performance. AI-driven training content improves weak areas. It reduces adverse events and disputes, optimizes manpower allocation, and saves quality control costs. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a home care management method based on AI recognition, as described in this application. Figure 2 This is a schematic diagram of the structure of a home care management system based on AI recognition according to this application; Figure 3 This is a schematic diagram of the structure of an electronic device according to this application; icon: 1. Verification module; 2. Data acquisition module; 3. Analysis module; 4. Early warning module; 5. Inspection module; 6. Training module; 7. Processor; 8. Memory; 9. Communication interface. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the various embodiments and features described below can be combined with each other. Example

[0021] This application provides a home care management and control method and system based on AI recognition, which can realize intelligent, real-time, and traceable security verification of the entire home care process.

[0022] Please refer to Figure 1 This AI-based home care management method includes the following steps: S1: Detects a nursing service request, obtains the patient's signed informed consent authorization information, and collects the facial image of the visiting nurse for comparison and verification with the nurse's professional qualification database. If the verification is successful, the nursing service is initiated. The process of collecting facial images of nurses making home visits and comparing them with the nurses' professional qualification database includes: obtaining the order service items, automatically matching the nurse's special training qualification certificate according to the order service items, and failing the verification if there is no corresponding qualification or the certificate has expired.

[0023] Specifically, after a patient places an order via a mobile app or other means, a nursing service request is generated. There are no specific limitations on how nursing service requests are generated. Before accepting an order, the nurse must capture their facial image using the nurse's app or a wearable smart badge camera. This image is then compared 1:N with the National Health Commission's registered nurse database and the hospital's nursing staff database. The validity of the nurse's license and on-duty status are also verified. After successful verification, the order service items (such as "PICC maintenance" or "catheterization") are matched, and the system automatically checks whether the nurse possesses the corresponding specialized training certification (such as a PICC maintenance training certificate or a catheterization operation certificate). If the qualifications are inconsistent or the certificate is expired, the nurse is prohibited from accepting the order, a supplementary training notification is sent, and the interception record is reported to the quality control center.

[0024] This setup eliminates unqualified or out-of-scope services at the source, proactively preventing risks and operational accidents caused by nurses' lack of qualifications. AI facial recognition and qualification matching automates manual verification, increasing efficiency by over 90%, and the entire verification process is recorded for easy traceability.

[0025] S2: Acquire real-time video streams including nurse operating areas and patient activity areas; The step of acquiring real-time video streams containing the nurse's operating area and the patient's activity area includes: acquiring real-time video streams containing the nurse's operating area and the patient's activity area through a wearable badge camera worn by the nurse and / or a portable mobile camera fixed in the operating area.

[0026] Specifically, after a nurse arrives at the patient's home, they wear a wearable smart badge with a camera (first-person view) and, with the patient's authorization, fix a portable mobile camera (fixed view) in the operating area. Both cameras have one-click blocking and one-click power-off functions, allowing the patient to pause recording at any time. During camera recording, the system automatically identifies private areas (such as bedrooms and bathrooms) and performs real-time AI blurring, retaining only the operating area and the area where the patient is active. The camera captures video at 25 frames per second and transmits the video stream in real-time to an edge computing gateway (or an edge computing module integrated into the badge camera) via 5G / WiFi. When network instability is detected, the video data is first cached locally and automatically uploaded synchronously after the network recovers. If the data is identified as critical, it is encrypted during transmission using SSL encryption and data anonymization to eliminate the risk of data leakage and ensure secure transmission.

[0027] This dual-camera setup takes into account both the details of the nurse's operation and the overall environment, avoiding blind spots. One-click masking and privacy blurring functions respect patient privacy and increase acceptance. Local caching ensures no data loss even when the network is offline, guaranteeing complete video recording throughout the service process.

[0028] S3: The edge computing terminal performs local real-time analysis of the real-time video stream, including: Based on the target detection model, key targets and their behavioral sequences in nursing operations are identified, and matched with a pre-set nursing operation standard rule base to determine whether there are any operational violations. Based on scene semantic segmentation, identify environmental risk sources and abnormal patient postures, match them with preset risk level rules, and determine whether there are safety hazards or high-risk patients. The target detection model is a lightweight YOLOv8-Nano model, and the nursing operation standard rule base includes hand hygiene compliance rules, aseptic operation rules, three checks and seven rights execution rules, medical waste classification rules, and special operation procedure rules; behavioral temporal analysis includes tracking the continuous frame motion trajectory of nurses' hands, calculating the motion duration and motion integrity.

[0029] Environmental risk sources include slippery ground, debris accumulation, insufficient lighting, and flammable and explosive materials; abnormal patient postures include high-risk actions such as falls, bed falls, and tube slippage.

[0030] Specifically, the edge computing unit (deployed at the name tag camera or local gateway) performs real-time analysis of the video stream. This includes: capturing a 25 frames / second video stream from the camera in real time, and then extracting frames at the edge at 10 frames / second to balance computing power and real-time performance. The extracted frames undergo adaptive brightness enhancement (to suit low-light indoor environments), Gaussian filtering for noise reduction, and the region of interest (ROI) containing the nurse's upper body and operating area is cropped to remove background interference. The output is a standardized RGB image frame (640×480 resolution).

[0031] Load a pre-trained lightweight YOLOv8-Nano model (INT8 quantization, 5MB size) to detect the following targets: nurse's hands, sterile packs, disinfectant supplies, medical waste bins, sharps containers, specialized operating instruments (urinary catheters, dressing packs, oxygen tubing, etc.), and patient operating sites. Output the target category, bounding box coordinates, and confidence score, filtering out invalid targets with a confidence score <0.8. Use the MediaPipePose algorithm to extract the coordinates of 21 key points on the nurse's hands (wrist, palm, and finger joints). Track the hand key point trajectory for 3-5 consecutive frames and analyze the action sequence: determine if there is a continuous action chain of "contact with disinfectant supplies → rubbing hands → drying"; calculate the number of frames for the action duration and convert it to action duration (e.g., rubbing hands for ≥20 seconds) using the frame rate (10 frames / second); determine whether the hand has entered a sterile area and whether sterile gloves are worn.

[0032] The rule engine's matching and violation determination include: Retrieve the digital nursing operation rule base (JSON format). Rule example: Poor hand hygiene: Rubbing hands for less than 20 seconds after contact with disinfectant or lacking the action of rubbing / drying hands is considered a Level 1 warning; local voice reminders will be issued.

[0033] Aseptic operation violation: Touching the sterile area without wearing sterile gloves for ≥2 seconds is judged as a level 2 warning, and a pop-up window and voice prompt will be issued on the quality control terminal.

[0034] Compare each behavioral feature with the rule base: If you only touch the disinfectant but do not rub your hands together, it is considered "inadequate hand hygiene (missing action)" and a Level 1 warning is issued. If the hand-rubbing motion lasts only 10 seconds, it is judged as "inadequate hand hygiene (insufficient duration)" and a Level 1 warning is issued. If sterile gloves are not worn when directly handling sterile packages, it is considered a "violation of sterile operation" and a level two warning is issued.

[0035] Environmental risk and abnormal patient posture recognition include: Scene semantic segmentation: A lightweight semantic segmentation model is used to divide the operation area, patient activity area, and privacy area, while privacy area analysis is shielded.

[0036] Hazard and posture detection: Detect slippery ground (texture and reflection), piled debris, flammable and explosive materials, key points of the patient's limbs, tubing (drainage tubes, PICC lines), and family members.

[0037] Risk level assessment: General hazards (Level 1 warning): dim lighting, a few pieces of clutter, the patient sits up slightly but shows no tendency to fall; High-risk hazards (Level 2 warning): The ground is obviously slippery, the patient gets up to get out of bed (high risk of fall), and the tubing is obviously twisted; Major risks (Level 3 warning): Sudden convulsions / loss of consciousness in the patient, forced interference by family members during the procedure, or tube slippage.

[0038] Tiered early warning system: Level 1 warning only prompts nurses to make corrections via local voice; Level 2 warning sends a pop-up window to the quality control system and assigns a dedicated person to follow up; Level 3 warning automatically dials the quality control center and notifies nearby medical staff for support.

[0039] A violation is only confirmed if the recognition results are consistent across three consecutive frames, thus avoiding misjudgment based on a single frame.

[0040] If the results are contradictory, increase the frame extraction frequency to 15 frames / second and re-identify.

[0041] With this setup, edge computing enables local analysis with a response time of ≤1 second, unaffected by network conditions, ensuring real-time intervention. The lightweight model and attitude estimation algorithm are adapted to the edge computing power at the grassroots level, achieving an accuracy rate of ≥95%. The digital rule engine transforms paper-based nursing guidelines into automatic judgments, solving the pain point of traditional manual sampling's inability to provide real-time and comprehensive coverage.

[0042] S4: When an operational violation or security risk is detected, the edge computing terminal will immediately trigger an alert and upload the violation / risk information and corresponding video clips to the cloud. Furthermore, the early warning system is divided into three levels: Level 1 early warning is a voice reminder to nurses from the edge terminal without involving back-end intervention; Level 2 early warning sends a pop-up notification to the quality control terminal, which is followed up by quality control personnel; Level 3 early warning triggers the emergency response process and automatically contacts the nearest medical staff for support.

[0043] Specifically, when the edge terminal determines that there is an operational violation or security risk, an alert is issued based on the risk level: Level 1 alert: The badge camera will broadcast prompts such as "Please complete the hand hygiene procedure correctly," and the information will be recorded locally without interrupting service.

[0044] Level 2 warning: In addition to local voice messages, a pop-up window on the cloud-based quality control terminal displays the violation footage and nurse information, allowing quality control personnel to remotely supervise rectification.

[0045] Level 3 warning: Automatically dials the quality control center and pushes emergency support instructions to nearby nurses' mobile apps, linking with 120 (emergency services).

[0046] Meanwhile, violation / risk information and corresponding video clips (5 seconds before and after) are encrypted (AES-256) and uploaded to the cloud. If the upload fails, the clips are cached locally and automatically synchronized after the network is restored.

[0047] This tiered early warning system avoids undue disruption to routine care while ensuring that major risks are addressed promptly. Encrypted data uploads protect privacy and security, while local caching prevents data loss.

[0048] S5: After the service is completed, the entire video data will be encrypted and uploaded to the cloud storage. The cloud AI platform will conduct batch quality control and verification of the entire data and generate a security verification report. Furthermore, the cloud-based AI platform performs batch quality control checks on all data, including: automatically calculating quality control scores by order dimension, filtering low-scoring orders and violations with confidence levels below the threshold for manual review, and generating quality control statistical reports by nurse dimension and region dimension.

[0049] Specifically, after the service is completed, all video data is encrypted and uploaded to cloud-based distributed storage (HDFS, stored for ≥6 months). The cloud-based AI platform performs batch checks daily at midnight. Basic order information verification: Verify nurse qualifications, order matching accuracy, and completeness of informed consent form.

[0050] Batch video content analysis: Utilizes a high-precision YOLOv8-L model for frame-by-frame analysis to generate a list of violations (violation type, timestamp, and confidence level).

[0051] Quality control score calculation: The maximum score is 100 points. Level 1 violations deduct 5 points, Level 2 violations deduct 10 points, and Level 3 violations deduct 20 points. A score of ≥90 is excellent, 80-89 is satisfactory, and <80 is unsatisfactory.

[0052] Manual review: Orders with scores <80, violations with confidence levels <0.9, and 5% of randomly selected qualified orders are then confirmed a second time by quality control personnel.

[0053] Report generation: Outputs "Single Order Safety Verification Report", "Nurse Monthly Quality Control Statistics Report", and "Regional Quality Control Analysis Report".

[0054] With this setup, AI-powered full-scale verification replaces manual order-by-order review, improving quality control efficiency by over 80%. Quantitative scoring and multi-dimensional reports provide objective data support for performance evaluation and management decisions.

[0055] S6: Based on the high-frequency violation types in the security audit report, automatically match the training courses in the nurse training system and push them to the corresponding nurse's port.

[0056] Specifically, the system compiles weekly statistics on high-frequency violation types (e.g., "30% of violations were related to aseptic technique") and automatically matches them with corresponding courses in the ADDIE training system (e.g., "Home Aseptic Dressing Change Practice Course"). Training tasks are then pushed to nurses via their mobile apps, requiring them to complete the learning and assessment within a specified timeframe. After training, the system compares the nurse's violation rate before and after training, generates a training effectiveness evaluation report, and pushes for retraining if no improvement is seen.

[0057] This setup breaks down data silos between AI verification and training, creating a closed loop of verification, analysis, training, and improvement. It enables precise delivery of personalized training, enhancing the relevance and effectiveness of training and continuously optimizing nursing quality.

[0058] Please refer to Figure 2 Based on the same inventive concept, this embodiment also provides an AI-based home care management system, including: Verification module 1 is used to detect nursing service requests, obtain the informed consent authorization information signed by the patient, and collect the facial image of the nurse making the home visit and compare it with the nurse's professional qualification database for verification. If the verification is successful, the nursing service is initiated. Acquisition module 2 is used to acquire real-time video streams including the nurse's operating area and the patient's activity area; Analysis module 3 is used for local real-time analysis of the real-time video stream at the edge computing end. The analysis includes: Based on the target detection model, key targets and their behavioral sequences in nursing operations are identified, and matched with a pre-set nursing operation standard rule base to determine whether there are any operational violations. Based on scene semantic segmentation, identify environmental risk sources and abnormal patient postures, match them with preset risk level rules, and determine whether there are safety hazards or high-risk patients. Early warning module 4 is used to trigger an early warning immediately when an operational violation or security risk is detected, and simultaneously upload the violation / risk information and corresponding video clips to the cloud. Module 5 is used to encrypt and upload the entire video data to the cloud storage after the service is completed. The cloud AI platform performs batch quality control verification on the full data and generates a security verification report. Training Module 6 is used to automatically match training courses from the nurse training system based on the high-frequency violation types in the safety audit report and push them to the corresponding nurses' ports.

[0059] For a detailed implementation of the AI-based home care management system, please refer to the above-mentioned implementation of the AI-based home care management method. Further details will not be provided here.

[0060] Please refer to Figure 3This embodiment also provides an electronic device, characterized in that it includes: Memory 8 is used to store one or more programs; Processor 7; Processor 7 is connected to memory 8 via communication interface 9; When one or more programs are executed by processor 7, all or some of the above methods are implemented.

[0061] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by processor 7, implements all or part of the methods described above.

[0062] The memory 8 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0063] Processor 7 can be an integrated circuit chip with signal processing capabilities. This processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0064] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A home care management method based on AI recognition, characterized in that, Includes the following steps: S1: Detects a nursing service request, obtains the patient's signed informed consent authorization information, and collects the facial image of the visiting nurse for comparison and verification with the nurse's professional qualification database. If the verification is successful, the nursing service is initiated. S2: Acquire real-time video streams including nurse operating areas and patient activity areas; S3: The edge computing terminal performs local real-time analysis on the real-time video stream, the analysis including: Based on the target detection model, key targets and their behavioral sequences in nursing operations are identified, and matched with a pre-set nursing operation standard rule base to determine whether there are any operational violations. Based on scene semantic segmentation, identify environmental risk sources and abnormal patient postures, match them with preset risk level rules, and determine whether there are safety hazards or high-risk patients. S4: When an operational violation or security risk is detected, the edge computing terminal will immediately trigger an alert and upload the violation / risk information and corresponding video clips to the cloud. S5: After the service is completed, the entire video data will be encrypted and uploaded to the cloud storage. The cloud AI platform will conduct batch quality control and verification of the entire data and generate a security verification report. S6: Based on the high-frequency violation types in the security check report, automatically match the training courses in the nurse training system and push them to the corresponding nurse's port.

2. The home care management method based on AI recognition according to claim 1, characterized in that, Step S1, which involves collecting the facial image of the nurse making the home visit and comparing it with the nurse's professional qualification database, includes: obtaining the order service items, automatically matching the nurse's special training qualification certificate according to the order service items, and failing the verification if there is no corresponding qualification or the certificate has expired.

3. The home care management method based on AI recognition according to claim 1, characterized in that, Step S2, which involves acquiring a real-time video stream including the nurse's operating area and the patient's activity area, includes: Real-time video streams, including the nurse's work area and the patient's activity area, are captured using a wearable badge camera worn by the nurse and / or a portable mobile camera fixed in the work area.

4. The home care management method based on AI recognition according to claim 1, characterized in that, The target detection model in step S3 is a lightweight YOLOv8-Nano model. The nursing operation standard rule base includes hand hygiene compliance rules, aseptic operation rules, three checks and seven rights execution rules, medical waste classification rules, and special operation procedure rules. The behavior time sequence analysis includes tracking the continuous frame motion trajectory of the nurse's hands and calculating the motion duration and motion integrity.

5. The home care management method based on AI recognition according to claim 1, characterized in that, The environmental risk sources mentioned in step S3 include slippery ground, piles of debris, insufficient light, and flammable and explosive materials; the abnormal patient postures include high-risk actions such as falls, bed falls, and tube slippage.

6. The home care management method based on AI recognition according to claim 1, characterized in that, The warning mentioned in step S4 is divided into three levels: Level 1 warning is given to nurses by voice reminder from the edge terminal, without involving back-end intervention; Level 2 warnings send pop-up notifications to the quality control system, which are then followed up by quality control personnel; Level 3 warnings trigger emergency response procedures, automatically contacting nearby medical personnel for support.

7. The home care management method based on AI recognition according to claim 1, characterized in that, The cloud-based AI platform described in step S5 performs batch quality control checks on all data, including: automatically calculating quality control scores by order dimension, filtering low-scoring orders and violations with confidence levels below the threshold for manual review, and generating quality control statistical reports by nurse dimension and region dimension.

8. A home care management system based on AI recognition, characterized in that, include: The verification module is used to detect nursing service requests, obtain the informed consent authorization information signed by the patient, and collect the facial image of the nurse making the home visit and compare it with the nurse's professional qualification database for verification. If the verification is successful, the nursing service is initiated. The acquisition module is used to acquire real-time video streams including the nurse's operating area and the patient's activity area; The analysis module is used by the edge computing end to perform local real-time analysis on the real-time video stream, and the analysis includes: Based on the target detection model, key targets and their behavioral sequences in nursing operations are identified, and matched with a pre-set nursing operation standard rule base to determine whether there are any operational violations. Based on scene semantic segmentation, identify environmental risk sources and abnormal patient postures, match them with preset risk level rules, and determine whether there are safety hazards or high-risk patients. The early warning module is used to trigger an early warning immediately when an operational violation or security risk is detected, and at the same time upload the violation / risk information and corresponding video clips to the cloud. The verification module is used to encrypt and upload all video data to cloud storage after the service is completed. The cloud AI platform performs batch quality control verification on all data and generates a security verification report. The training module is used to automatically match training courses from the nurse training system based on the high-frequency violation types in the security audit report and push them to the corresponding nurses' ports.

9. An electronic device, characterized in that, include: Memory, used to store one or more programs; processor; When the one or more programs are executed by the processor, the method as described in any one of claims 1-7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images