Safety monitoring system applied to hospitalization area

Abstract

The invention provides a safety monitoring system applied to a hospitalization area, and relates to the technical field of computer processing, and the system comprises a patient feature registration module which is used for collecting and storing biological feature data and clinical parameters of a patient; the intelligent monitoring sub-module comprises a monitoring strategy engine and a behavior analysis unit and is used for generating a dynamic monitoring rule and executing multi-modal behavior analysis; the early warning response mechanism module is used for configuring a three-level early warning strategy and carrying out matching based on an analysis result so as to generate a multi-channel alarm; and the data communication module is used for feeding back to a nurse station based on the acquired multi-channel alarm. According to the safety monitoring system applied to the hospitalized area, all-weather and all-around monitoring and protection of hospitalized patients are realized, possible accidents such as falling and missing of the patients in the hospitalized period can be effectively prevented, timely and accurate early warning information can be provided for medical staff through real-time data analysis, and the safety of the hospitalized patients is improved. Therefore, the efficiency and quality of medical services are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of computer processing, and particularly to a security monitoring system applied to the inpatient area. Background Art

[0002] As an indispensable part of nursing work, the core purpose of the nurse's rounds is to accurately grasp the pathological conditions of inpatients through regular inspections, and to ensure that patients follow medical orders and do not leave the hospital without permission, so as to effectively prevent the occurrence of accidents. Especially at night, nurses are required to make a round of inspections every hour, and this requirement is even more stringent for elderly patients admitted to the psychiatric department. Because at night, elderly patients may increase the risk of falling due to getting up to use the toilet or moving in the corridor, and pathological emergencies may occur suddenly. These factors may lead to the unfortunate sudden death of elderly patients. Although high-frequency inspections can alleviate this risk to a certain extent, accidents often occur during inspections or during the intervals between two inspections, making it particularly difficult to completely avoid. In addition, the number of night shift nurses is limited, and the high-intensity inspection work undoubtedly brings a heavy physical burden to night shift nurses, thus affecting their response ability in case of emergencies.

[0003] The existing security system in the hospital (referring to the monitoring technology disclosed in the authorized publication number CN107925821B) aims to comprehensively monitor key areas such as hospital corridors and wards, and capture and record video images of the monitored areas in real time. Given the wide monitoring range, multi-screen observation has become an inevitable choice. And the current advanced security system already has the function of portrait extraction, and can accurately extract the head portrait information of the people in the video through photo frame-level analysis of the input video.

[0004] In view of this, if the existing security system in the hospital can be optimized and upgraded to automatically identify and recognize patient information during monitoring, and can immediately send an alarm to the nurse station when an abnormal situation is detected, this will undoubtedly greatly make up for the deficiencies of manual inspections. This measure can not only significantly reduce the inspection intensity of night shift nurses, but also achieve instant warning, rapid notification and efficient handling of patient accidents, and escort the safety of patients. Summary of the Invention

[0005] In view of the above technical problems, the technical solution adopted by the present invention is a security monitoring system applied to the inpatient area, including: A patient feature registration module for collecting and storing patient biometric data and clinical parameters; An intelligent monitoring sub-module, including a monitoring strategy engine and a behavior analysis unit, for generating dynamic monitoring rules and performing multi-modal behavior analysis; An early warning response mechanism module, configured with a three-level early warning strategy, and matching based on the analysis results to generate multi-channel alarms; The data communication module feeds back the obtained multi-channel alarms to the nurse station.

[0006] Preferably, the patient feature registration module includes: A feature vector database inputs a face image whose features are to be extracted from a surveillance video into a trained deep learning model. After being processed by a convolutional layer, a feature embedding layer, and an L2 normalization layer, a 128-dimensional feature vector is obtained; An electronic medical record interface unit automatically associates three clinical parameters: the patient's nursing level, the number of days after surgery, and the mobility assessment value.

[0007] Preferably, the monitoring strategy engine includes: A dynamic threshold calculation unit is used to automatically calculate and set a monitoring threshold according to the patient's clinical parameters. The monitoring threshold satisfies the formula: T = α×(1 + β×γ), where α is the base threshold, β is the nursing level coefficient set according to the patient's nursing level, and γ is the postoperative days correction factor set according to the number of days after surgery of the patient; A regional weight allocation model is used to divide the inpatient area into different monitoring levels, assign corresponding weights to each monitoring level, and increase the data processing priority of the system for the ward area, the bathroom, and the corridor based on the weights. Among them, for the ward area, the weight is 0.6; for the bathroom, the weight is 0.3; for the corridor, the weight is 0.1.

[0008] Preferably, the behavior analysis unit includes: An action capture layer extracts the input surveillance video, extracts each frame of picture from the video, annotates the collected pictures with a diverse image dataset containing human skeletal key points, converts the annotated data into the YOLO format, and inputs the pictures in the YOLO format into the YOLOv8-Pose model to obtain the skeletal key point coordinates and calculate the relative positions between the key points; A spatio-temporal feature extraction module obtains a data set about the skeletal key point coordinates within a predetermined time window; An action judgment module extracts data sets of the skeletal key point coordinates of multiple consecutive windows, calculates the average coordinates of the skeletal key points within each window, then connects the average coordinates of the skeletal key points of the consecutive windows to obtain a human body shape connection diagram, and loads a spatial gyroscope to calculate whether the human body shape connection diagram is in a balanced state.

[0009] Preferably, for the obtained human body shape connection diagram, loading a spatial gyroscope to calculate whether the human body shape connection diagram is in a balanced state, the processing steps include: S10. Process the human body image, identify and mark the key bone points of the human body, and generate a human body shape connection diagram based on these key bone points; S11. Connect the spatial gyroscope to the human body to be detected or its surrounding environment, and obtain the spatial attitude data output by the gyroscope in real time. The data includes angular velocity information in at least three axes; S12. According to the angular velocity data output by the spatial gyroscope, for each axis i (i = X, Y, Z), the angle change Δθi can be calculated by the following formula: Δθ i = ω i ×Δt, where ωi is the angular velocity of the i axis; Δθ i =(ω i,start + ω i,end ) / 2×Δt, where, ω i,star t and ω i,end are the angular velocities at the start and end of the time interval respectively. Obtain the tilt angles of the three axes of the spatial gyroscope respectively, and based on the angle synthesis method in three-dimensional space, obtain the tilt angle of the human body and its surrounding environment relative to the initial position; S13. Compare the calculated human body tilt angle with a preset balance threshold, and at the same time refer to the human body shape connection diagram to analyze the overall stability of the connection diagram or the relative position relationship between specific key points, and comprehensively determine whether the human body is in a balanced state; if the tilt angle is less than or equal to the balance threshold, and the connection diagram shows overall stability or the relative position relationship between key points is reasonable, it is determined to be in a balanced state; otherwise, it is determined to be in an unbalanced state.

[0010] Preferably, the three-level early warning strategy includes: Level 1 early warning: When a fall behavior is detected, trigger the audible and visual alarm device and pop up the patient's position and video playback link on the nurse station interface, with a response delay < 2 seconds; Level 2 early warning: When the residence time reaches 90% of the set threshold, start the voice prompt device; Level 3 early warning: When the patient's moving speed exceeds the safe range, generate a path prediction heat map and send it to the mobile nursing terminal.

[0011] Preferably, the intelligent monitoring sub-module includes a long short-term memory network (LSTM) path prediction model. Its input layer is designed to receive 10 consecutive groups of three-dimensional position coordinate data of the target object. Each group of data contains the three-dimensional coordinates of the target object at a certain moment, and the output layer is designed to output the prediction result of the moving trajectory of the target object within the next 30 seconds. The result can be a series of continuous three-dimensional coordinate points or a vector representation of the moving path.

[0012] Preferably, the early warning response mechanism module further includes a time monitoring mechanism: when the monitored target stays in a non-bed area for more than a set duration, a stay warning is initiated, and a dual verification mechanism is adopted through stay time calculation: the combination of the first frame detection timestamp and the confirmation of the positions of 5 consecutive frames.

[0013] The present invention has at least the following beneficial effects: 1. By collecting and storing the biometric data (such as face images) and clinical parameters (such as nursing level, number of days after surgery, mobility assessment value) of patients, it is ensured that the system can accurately identify patients and perform personalized monitoring settings according to their clinical conditions.

[0014] 2. Automatically adjust the monitoring threshold according to the clinical parameters of patients and set the monitoring weights for different inpatient areas, so as to achieve accurate monitoring of patients. And by extracting the coordinates of the skeletal key points in the monitoring video, action capture and spatio-temporal feature extraction are carried out to judge whether the behavior of patients is abnormal. And when abnormal behavior or the patient's state exceeds the safe range is detected, corresponding warning measures will be triggered, such as sound and light alarms, voice prompts, path prediction heat maps, etc., to ensure that medical staff can respond quickly and take necessary intervention measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a module diagram of a safety monitoring system applied to an inpatient area provided in Embodiment 1 of the present invention; Figure 2 It is a unit diagram of a patient feature registration module provided in Embodiment 1 of the present invention; Figure 3 It is a unit diagram of an intelligent monitoring sub-module provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present invention.

[0018] It should be noted that the terms "first", "second", etc. in the specification, claims and the above-mentioned drawings of the present invention are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or server comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0019] Embodiment 1 This embodiment provides a safety monitoring system applied to the inpatient area. The system includes the following architecture, as Figures 1-3 shown: A patient feature registration module for collecting and storing patient biometric data and clinical parameters; Specifically, before the system executes, it is necessary to input the facial features of the inpatients and the data set of the bone key point coordinates in the standing state, so as to obtain the patient biometric data required, that is, to obtain the feature vector database. It should be noted that the feature vector database contains the face images to be extracted with features obtained from the monitoring video input into the trained deep learning model, and after being processed by the convolutional layer, the feature embedding layer and the L2 normalization layer, a 128-dimensional feature vector is obtained. The clinical parameters refer to the electronic medical record interface unit, which is used to automatically associate three clinical parameters: the patient's nursing level, the number of days after surgery, and the action ability evaluation value. In order to facilitate the later warning of the system, not only the location information of the patient will be reported to the nurse station, but also the three clinical parameters of the patient's nursing level, the number of days after surgery, and the action ability evaluation value will be reported, so that the nurses going to the rescue can perform first aid faster. Through the cooperation of the feature vector database and the electronic medical record interface unit, the system can achieve accurate identification and personalized monitoring of patients, improving the accuracy and effectiveness of monitoring.

[0020] An intelligent monitoring sub-module, including a monitoring strategy engine and a behavior analysis unit, for generating dynamic monitoring rules and performing multi-modal behavior analysis; Specifically, the monitoring strategy engine in this embodiment includes: A dynamic threshold calculation unit for automatically calculating and setting a monitoring threshold according to the patient's clinical parameters. The monitoring threshold satisfies the formula: T = α × (1 + β × γ), where α is the basic threshold, β is the nursing level coefficient set according to the patient's nursing level, and γ is the postoperative days correction factor set according to the patient's number of days after surgery; The regional weight allocation model is used to divide the inpatient area into different monitoring levels, assign corresponding weights to each monitoring level, and increase the data processing priority for the ward area, bathroom, and corridor based on the weights. Among them, the weight of the ward area is 0.6; the weight of the bathroom is 0.3; the weight of the corridor is 0.1.

[0021] Further, the behavior analysis unit in the above embodiment includes: The action capture layer extracts the input monitoring video, extracts each frame of picture from the video, annotates the collected pictures to form a diverse image dataset containing human skeletal key points, converts the annotated data into the YOLO format, and inputs the YOLO-format pictures into the YOLOv8-Pose model to obtain the skeletal key point coordinates and calculate the relative positions between the key points; The spatio-temporal feature extraction module obtains a data set about the skeletal key point coordinates within a predetermined time window; The action judgment module extracts the data sets of the skeletal key point coordinates of multiple consecutive windows, calculates the average coordinates of the skeletal key points within each window, then connects the average coordinates of the skeletal key points of the consecutive windows to obtain a human body shape connection diagram, and loads a spatial gyroscope to calculate whether the human body shape connection diagram is in a balanced state.

[0022] In the above technology, by extracting the input monitoring video, extracting each frame of picture from the video, annotating the collected pictures, forming a diverse image dataset containing human skeletal key points, extracting images frame by frame from the monitoring video, preprocessing each frame of image, such as denoising, enhancement, etc., using a deep learning model (such as YOLOv8-Pose) to detect and annotate the human skeletal key points, and converting the annotated data into the YOLO format for subsequent processing.

[0023] Further, obtaining a data set about the skeletal key point coordinates within a predetermined time window for subsequent action judgment. The detailed steps include: Set a predetermined time window, such as 5 seconds or 10 seconds; Continuously obtain the data of the skeletal key point coordinates within the time window; Store and organize these data to form a spatio-temporal feature data set.

[0024] And the action judgment module, the execution steps include: Extract the skeletal key point coordinate data within consecutive time windows; Calculate the average of the skeletal key point coordinates within each time window to obtain the average coordinates; Connect the average coordinates of the consecutive time windows to form a human body shape connection diagram; Load the data of the space gyroscope, calculate the body tilt angle, and compare it with the preset balance threshold.

[0025] At the same time, refer to the overall stability of the human body form connection diagram or the relative position relationship between specific key points, and comprehensively determine whether the human body is in a balanced state.

[0026] Suppose a patient is walking in the ward and the monitoring video captures this process. Through the motion capture layer, the system extracts the key body skeleton points in each frame of the image and marks them. Then, the spatio-temporal feature extraction module obtains the coordinate data of these key points within a predetermined time window. Finally, the motion judgment module analyzes this data and finds that the patient's body remains balanced during walking and no abnormal behaviors such as falling occur. Therefore, the system does not trigger an early warning response.

[0027] It should be noted that to determine whether the obtained human body form connection diagram is in a balanced state by loading the space gyroscope, the processing steps include: S10. Process the human body image, identify and mark the key skeleton points of the human body, and generate a human body form connection diagram based on these key skeleton points; S11. Connect the space gyroscope to the human body to be detected or its surrounding environment, and obtain the spatial attitude data output by the gyroscope in real time. The data includes at least the angular velocity information of three axes; S12. According to the angular velocity data output by the space gyroscope, for each axis i (i = X, Y, Z), the angle change Δθi can be calculated by the following formula: Δθ i = ω i ×Δt, where ωi is the angular velocity of the i axis; Δθ i =(ω i,start + ω i,end ) / 2×Δt, where, ω i,star t and ω i,end are the angular velocities at the start and end of the time interval respectively. Obtain the tilt angles of the three axes of the space gyroscope respectively, and based on the angle synthesis method in three-dimensional space, obtain the tilt angle of the human body and its surrounding environment relative to the initial position; S13. Compare the calculated human body tilt angle with the preset balance threshold. At the same time, refer to the human body form connection diagram, analyze the overall stability of the connection diagram or the relative position relationship between specific key points, and comprehensively determine whether the human body is in a balanced state; if the tilt angle is less than or equal to the balance threshold, and the connection diagram shows overall stability or the relative position relationship between key points is reasonable, then it is determined to be in a balanced state; otherwise, it is determined to be in an unbalanced state.

[0028] Suppose a patient is standing in the bathroom but is a bit shaky. The surveillance video captures this process and extracts the key skeletal points through the motion capture layer. Then, the system obtains the patient's tilt angle data through a spatial gyroscope. After calculation, it is found that the patient's tilt angle exceeds the preset balance threshold, and the human body form connection diagram also shows an unstable state. Therefore, the system triggers an early warning response to prompt the medical staff to pay attention to the patient's safety status.

[0029] Through the above dynamic adjustment of the monitoring threshold and setting of regional weights, the system can achieve precise monitoring of patients, reduce the possibility of false alarms and missed alarms, and improve the accuracy and reliability of monitoring. Secondly, the skeletal key point coordinates in the surveillance video are extracted through the motion capture layer, and spatio-temporal feature extraction and motion judgment are carried out. Among them, the spatio-temporal feature extraction module obtains a data set of the skeletal key point coordinates for subsequent motion judgment. The motion judgment module then loads a spatial gyroscope to assist in judging the patient's balance state, so as to achieve timely early warning of abnormal behaviors such as falls.

[0030] The early warning response mechanism module configures a three-level early warning strategy and makes a match based on the analysis results to generate multi-channel alarms; Specifically, the three-level early warning strategy includes: Level 1 early warning: When a fall behavior is detected, trigger the audible and visual alarm device and pop up the patient's location and video playback link on the nurse station interface, with a response delay < 2 seconds; Level 2 early warning: When the detention time reaches 90% of the set threshold, start the voice prompt device; Level 3 early warning: When the patient's moving speed exceeds the safe range, generate a path prediction heat map and send it to the mobile nursing terminal.

[0031] By configuring the three-level early warning strategy and taking different early warning measures, the system can achieve timely early warning and effective intervention for abnormal behaviors, ensuring the safety of patients.

[0032] Secondly, the intelligent monitoring sub-module includes a long short-term memory network (LSTM) path prediction model. Its input layer is designed to receive 10 consecutive groups of three-dimensional position coordinate data of the target object. Each group of data contains the three-dimensional coordinates of the target object at a certain moment, and the output layer is designed to output the prediction result of the moving trajectory of the target object within the next 30 seconds. This result can be a series of continuous three-dimensional coordinate points or a vector representation of the moving path.

[0033] The Long Short-Term Memory (LSTM) path prediction model can predict the future movement trajectories of patients. This model receives continuous three-dimensional position coordinate data as input and outputs the predicted results of the movement trajectories within a certain period in the future. This function provides healthcare staff with predictive information about the possible movement paths of patients, helping to take preventive measures in advance. By introducing the LSTM path prediction model, the system can accurately predict the future movement paths of patients, providing valuable information support for healthcare staff, helping to take preventive measures in advance and avoid potential safety risks.

[0034] In addition, the warning response mechanism module in this embodiment further includes a time monitoring mechanism: when the monitored target stays in a non-bed area for more than the set duration, a stay warning is triggered, and a dual verification mechanism is adopted through the calculation of the stay time: the combination of the first-frame detection timestamp and the confirmation of the positions of 5 consecutive frames.

[0035] Its purpose is to prevent mistakes and maintain. When the patient maintains one posture in the ward area, the toilet, and the corridor, it indicates that there must be an emergency with the patient, or the monitoring device is damaged, or there are problems such as bugs in the system. At this time, the nurse in the nurse station needs to go to check. If there is no danger to the patient, a fault report is made through the nurse station system for timely repair.

[0036] The data communication module feeds back multi-channel alarms obtained to the nurse station.

[0037] Specifically, a local area network is used to establish a data transmission system, and the present invention can be a computer terminal or a mobile phone terminal. When the nurse is at work, they can directly access the system by running the designed app, so that they can still obtain the situation of the ward during the inspection or activity process.

[0038] In summary, by collecting and storing the biometric data (such as face images) and clinical parameters (such as nursing level, number of days after surgery, and action ability assessment value) of patients, it is ensured that the system can accurately identify patients and perform personalized monitoring settings according to their clinical conditions.

[0039] Automatically adjust the monitoring threshold according to the clinical parameters of the patient, and set the monitoring weights for different inpatient areas to achieve precise monitoring of the patient. And by extracting the bone key point coordinates in the monitoring video, action capture and spatio-temporal feature extraction are performed to determine whether the patient's behavior is abnormal. And when abnormal behavior or the patient's state exceeds the safe range is detected, corresponding warning measures will be triggered, such as sound and light alarms, voice prompts, path prediction heat maps, etc., to ensure that healthcare staff can respond quickly and take necessary intervention measures.

[0040] Embodiment 2 An embodiment of the present invention provides a non-transitory computer-readable storage medium, in which at least one instruction or at least one segment of program is stored, and the at least one instruction or the at least one segment of program is loaded and executed by a processor to implement the steps: Collect and store the biometric data and clinical parameters of the patient, and use a monitoring policy engine and a behavior analysis unit to generate dynamic monitoring rules and perform multi-modal behavior analysis. Based on the analysis results, match the pre-configured three-level warning policy and implement multi-channel alarms for abnormal events to transmit and interact with the data of the nurse station.

[0041] Those of ordinary skill in the art can understand that all or part of the processes in the above-described embodiment methods can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-described method embodiments. Among them, any reference to a memory, storage, database, or other medium used in the various embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0042] Those skilled in the art can clearly understand that for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is used as an example. In practical applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0043] Embodiment III An embodiment of the present invention provides an electronic device, including a processor and a memory, in which at least one instruction or at least one segment of program is stored, and the at least one instruction or the at least one segment of program is loaded and executed by the processor to implement the steps: Collect and store the patient's biometric data and clinical parameters, and use the monitoring strategy engine and behavior analysis unit to generate dynamic monitoring rules and perform multi-modal behavior analysis. Based on the analysis results, match the pre-configured three-level warning strategy and implement multi-channel alarms for abnormal events to transmit and interact with the data at the nurse station.

[0044] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the above-disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A safety monitoring system applied to inpatient areas, characterized in that: include: Patient characteristic registration module, used to collect and store patient biometric data and clinical parameters; The intelligent monitoring submodule includes a monitoring strategy engine and a behavior analysis unit, which is used to generate dynamic monitoring rules and perform multimodal behavior analysis; The early warning response mechanism module configures three-level early warning strategies and matches them based on the analysis results to generate multi-channel alarms; The data communication module feeds back the acquired multi-channel alarm to the nurse station.

2. A safety monitoring system for inpatient areas according to claim 1, characterized in that: The patient characteristics registration module comprises: Feature vector database, obtains the face image to be extracted from the surveillance video and inputs it into the trained deep learning model. After being processed by the convolution layer, feature embedding layer and L2 normalization layer, a 128-dimensional feature vector is obtained; The electronic medical record interface unit automatically associates three clinical parameters: patient care level, number of days after surgery, and mobility assessment value.

3. A safety monitoring system for inpatient areas according to claim 1, characterized in that: The monitoring strategy engine includes: A dynamic threshold calculation unit, used to automatically calculate and set a monitoring threshold according to the patient's clinical parameters, wherein the monitoring threshold satisfies the formula: T=α×(1+β×γ), wherein α is a basic threshold, β is a nursing level coefficient set according to the patient's nursing level, and γ is a postoperative days correction factor set according to the patient's postoperative days; The regional weight allocation model is used to divide the inpatient area into different monitoring levels, and assign corresponding weights to each monitoring level. Based on the weights, the system increases the data processing priority for the bed area, bathroom and corridor, where the bed area has a weight of 0.6; the bathroom has a weight of 0.3; and the corridor has a weight of 0.

1.

4. A safety monitoring system for inpatient areas according to claim 1, characterized in that: The behavior analysis unit comprises: The motion capture layer extracts the recorded surveillance video and extracts each frame of the video, annotates the collected images with a diverse image dataset containing human skeleton key points, converts the annotated data into YOLO format, and inputs the YOLO format images into the YOLOv8-Pose model to obtain the coordinates of the skeleton key points and calculate the relative positions between the key points; A spatiotemporal feature extraction module, which acquires a data set about the coordinates of the skeleton key points in a predetermined time window; The action judgment module extracts a data set of the coordinates of multiple skeleton key points in consecutive windows, obtains the average coordinates of the skeleton key points in each window, and then connects the average coordinates of the skeleton key points in consecutive windows to obtain a human body morphology connection diagram, and loads it into a spatial gyroscope to calculate whether the human body morphology connection diagram is in a balanced state.

5. A safety monitoring system for inpatient areas according to claim 4, characterized in that: The obtained human body shape connection diagram is loaded into a space gyroscope to calculate whether the human body shape connection diagram is in a balanced state, and the processing steps include: S10, processing the human body image, identifying and marking the key bone points of the human body, and generating a human body morphology connection diagram according to these key bone points; S11, connecting the spatial gyroscope to the human body to be detected or the environment in which it is located, and acquiring the spatial posture data output by the gyroscope in real time, wherein the data includes angular velocity information of at least three axes; S12. Based on the angular velocity data output by the space gyroscope, for each axis i (i=X, Y, Z), the angle change Δθi can be calculated by the following formula: Δθ i =ω i ×Δt, Where ωi is the angular velocity of the i-axis; Dth i =(ω i,start +oh i,end ) / 2×Δt, Among them, ω i,star t and ω i,end The angular velocities at the beginning and end of the time interval are respectively used to obtain the tilt angles of the three axes of the space gyroscope, and based on the angle synthesis method in three-dimensional space, the tilt angles of the human body and its environment relative to the initial position are obtained; S13. Compare the calculated human body inclination angle with the preset balance threshold, and refer to the human body morphology connection diagram at the same time, analyze the overall stability of the connection diagram or the relative position relationship between specific key points, and comprehensively determine whether the human body is in a balanced state; if the inclination angle is less than or equal to the balance threshold, and the connection diagram shows overall stability or the relative position relationship between key points is reasonable, it is determined to be a balanced state; otherwise, it is determined to be an unbalanced state.

6. A safety monitoring system for inpatient areas according to claim 1, characterized in that: The three-level early warning strategy includes: Level 1 warning: When a fall is detected, the sound and light alarm device is triggered and a pop-up window is displayed on the nurse station interface to display the patient's location and video playback link, with a response delay of less than 2 seconds; Level 2 warning: When the detention time reaches 90% of the set threshold, the voice prompt device is activated; Level 3 warning: When the patient's movement speed exceeds the safe range, a path prediction heat map is generated and sent to the mobile nursing terminal.

7. A safety monitoring system for inpatient areas according to claim 1, characterized in that: The intelligent monitoring submodule includes a long short-term memory network (LSTM) path prediction model, whose input layer is designed to receive 10 consecutive sets of three-dimensional position coordinate data of the target object, each set of data contains the three-dimensional coordinates of the target object at a certain moment, and the output layer is designed to output the predicted movement trajectory of the target object within the next 30 seconds, which can be a series of continuous three-dimensional coordinate points or a vector representation of the movement path.

8. The safety monitoring system for inpatient areas according to claim 1, characterized in that: The early warning response mechanism module also includes a time monitoring mechanism: when the monitored target stays in the non-bed area for more than a set time, a detention warning is activated, and a double verification mechanism is adopted through the detention time calculation: the first frame detection timestamp is combined with 5 consecutive frame position confirmation.

9. A non-transitory computer-readable storage medium, wherein at least one instruction or at least one program is stored in the non-transitory computer-readable storage medium, characterized in that: The at least one instruction or the at least one program is loaded and executed by the processor to implement the method for a safety monitoring system applied to an inpatient area as described in any one of claims 1 to 8.

10. An electronic device, characterized in that: It includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the method of the safety monitoring system applied to the inpatient area as described in any one of claims 1-8.

Images