A configuration implementation method of an ICU bed intelligent video monitoring system

Abstract

The application discloses an intelligent video monitoring system configuration implementation method for an ICU bed, which is characterized in that the video acquisition, behavior analysis and system management are decoupled through the configuration function division of the monitoring system, and a decentralized communication mechanism is constructed based on a data distribution service, so that the system can be deployed and run in a distributed manner on heterogeneous computing devices. Through frame-by-frame analysis of bedside camera video data on the edge side, continuous monitoring of the patient's local behavior is realized, and alarm information is generated when it is determined that there is a risk of extubation. The method supports dynamically configuring the monitoring scale and algorithm instances according to the number of ICU beds and monitoring requirements, thereby improving the scalability, stability and practical application adaptability of the system.

Description

Technical Field

[0001] This invention relates to the field of medical information technology and intelligent monitoring technology, specifically to a configuration implementation method for an intelligent video monitoring system for ICU beds. Background Technology

[0002] The Intensive Care Unit (ICU) is a core medical facility for providing continuous life support and close monitoring of critically ill patients. ICU patients typically require multiple life-sustaining or therapeutic tubes, such as endotracheal tubes, nasogastric tubes, central venous catheters, and urinary catheters. Because patients may experience confusion, agitation, pain, or delirium, they may exhibit abnormal or dangerous movements, making unplanned extubation highly likely. This not only interrupts treatment and increases the risk of complications, but in severe cases, can even endanger the patient's life, while also significantly increasing the workload of medical staff.

[0003] Currently, clinical measures to prevent and control the risk of abnormal or dangerous actions (such as unplanned extubation) in ICU patients mainly rely on manual rounds, nursing experience and judgment, and physical or pharmacological interventions. For example, increasing bedside rounds, using limb restraint devices, or adjusting sedation and analgesia protocols can reduce the probability of extubation. However, these methods are highly dependent on the continuous attention and experience of medical staff, and in situations such as nighttime, shift changes, or staff shortages, monitoring blind spots and response delays are still unavoidable. In addition, over-reliance on physical restraints or sedative drugs may lead to patient discomfort, increased complications, and ethical controversies, making it difficult to meet the development needs of modern ICU's refined management and humanistic care.

[0004] With the development of computer vision and intelligent video analytics technologies, video surveillance-based patient behavior monitoring solutions are gradually being introduced into medical settings. Existing technologies typically deploy cameras in wards or beside beds to monitor patient activity, combining human posture recognition and action classification algorithms to identify macroscopic behaviors such as turning over, sitting up, and getting out of bed, assisting in fall warnings or rehabilitation assessments. However, most of these intelligent video monitoring solutions are designed for general wards or open spaces, focusing on large-scale, overall movements. They lack the ability to specifically identify subtle, continuous, and highly dangerous behavioral intentions in ICU environments, such as "gradually approaching, touching, and pulling on tubing."

[0005] On the other hand, existing video surveillance systems typically employ a centralized or semi-centralized deployment model at the system architecture level. This means that a server uniformly accesses multiple camera video streams and centrally runs video analysis algorithms. This approach is easily limited by computing power bottlenecks, concurrent processing capabilities, and system stability when the monitoring scale expands, making it difficult to flexibly adjust to changes in the number of ICU beds or the different monitoring needs of different beds. Furthermore, existing systems often tightly couple video acquisition, algorithm analysis, and management services, lacking a good decoupling mechanism between the algorithm module and the hardware platform. This results in high costs for system expansion, algorithm upgrades, or heterogeneous hardware integration, which is detrimental to long-term operation in scenarios like ICUs where the number of beds dynamically changes and monitoring needs are highly differentiated.

[0006] In view of this, the present invention proposes a configuration implementation method for an intelligent video monitoring system for ICU beds. Summary of the Invention

[0007] The purpose of this invention is to provide a configurable implementation method for an intelligent video monitoring system for ICU beds. This method aims to solve the technical problems in existing ICU bed video monitoring systems, such as the tight coupling between intelligent video analysis algorithms and hardware platforms, centralized system deployment, insufficient scalability, and difficulty in flexibly configuring and heterogeneously expanding the system according to changes in the number of ICU beds and the monitoring needs of different beds. These problems result in limited monitoring scale and high system upgrade and maintenance costs.

[0008] In a first aspect, the present invention provides a configurable implementation method for an intelligent video monitoring system for ICU beds, used for video monitoring and early warning of abnormal and dangerous actions of patients (such as unplanned extubation) in an intensive care unit environment, comprising the following steps:

[0009] S1, the monitoring system is divided into multiple configuration modules according to the configuration function. The configuration module includes at least a data acquisition module for video stream access and frame-by-frame decoding, a video analysis algorithm module for analyzing patient behavior and outputting alarm results, a Web interface service module for local resource scheduling and interface communication, and a management client module for centralized configuration of monitoring areas and aggregation and display of alarm results.

[0010] S2, build a decentralized communication mechanism based on the Data Distribution Service (DDS). Each configuration module interacts with data by publishing and subscribing to data topics, replacing the direct calling method between modules and realizing the logical separation between modules;

[0011] S3 enables distributed configuration deployment in a heterogeneous hardware environment, allowing the monitoring system to run in a distributed architecture consisting of a single computing device or multiple computing devices. Specifically, the data acquisition module accesses the video stream from the bedside camera and publishes frame-by-frame video data, the video analysis algorithm module subscribes to the video data and generates alarm results when it determines that there is a risk of unplugging the tube, and the Web interface service module registers the configuration of the cameras under the jurisdiction of this device and provides parameter configuration and data access services.

[0012] S4, the system monitors alarm results in real time and drives alarm devices to perform prompting operations;

[0013] S5, the management client module runs on the management terminal, and uses the web interface of each computing device to send parameters and aggregate results, but does not participate in the decoding and analysis of video data.

[0014] As a preferred technical solution of the present invention, each configuration module is encapsulated as an independent computing node; each computing node is developed based on a distributed system framework that supports DDS communication, and through the dynamic discovery and quality of service management mechanism of the underlying DDS middleware, automatic networking and transparent data distribution across processes or physical devices are realized.

[0015] As a preferred technical solution of the present invention, the system is built on a hardware environment that supports heterogeneous computing, the hardware environment including CPUs, GPUs or edge computing devices based on different instruction set architectures; each computing device is connected to a unified video data broadcast pool through a local area network to support the distributed operation of the configuration module on different computing resources.

[0016] As a preferred technical solution of the present invention, the video analysis algorithm module extracts temporal image features from the spatial topological relationship between the patient's hand and each indwelling tube to identify the continuous behavioral evolution process from the approach state, contact state to the force application state, thereby distinguishing between normal physiological activities and extubation risk behaviors.

[0017] As a preferred embodiment of the present invention, the video analysis algorithm module stores the generated monitoring data in real time into a local database on the edge device side; the local database is an embedded lightweight database, which is connected to and establishes a data read / write connection with the video analysis algorithm module, and completes data storage and query through the local database interface; the monitoring data is output to the management client module after receiving a data query request from the management client module.

[0018] As a preferred technical solution of the present invention, the Web interface service module supports the dynamic distribution of ROI area parameters, which allows medical staff to remotely adjust the monitoring range of the algorithm module for the indwelling area of ​​the tubing in a specific bed through the visual interface of the management client module, and to use the HTTP interface to realize the key frame retrospective of historical alarm screens.

[0019] As a preferred technical solution of the present invention, the same video data topic is subscribed to in parallel by multiple video analysis algorithm instances with different functions, so as to simultaneously perform the tasks of extubation intention recognition and fall risk assessment for the same ICU bed.

[0020] As a preferred technical solution of the present invention, in step S4, the alarm device adopts a portable wireless vibration reminder terminal. After receiving the alarm result, it performs vibration reminder and simultaneously parses the alarm information to display the corresponding bed number and risk level on the terminal screen.

[0021] As a preferred technical solution of the present invention, the system has the ability to be flexibly reconfigured. When the number of ICU monitoring beds changes, the management client module completes the system topology update by scanning the IP address and interface status of the Web service.

[0022] As a preferred technical solution of the present invention, a single computing device can simultaneously launch multiple data acquisition module instances based on the concurrent processing capability of its hardware decoder. Each instance corresponds to an RTSP video stream from a camera and publishes it to an independent DDS topic.

[0023] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0024] This invention designs an ICU bed video monitoring system as a hardware-software decoupled configuration method. Video acquisition, behavior analysis, and system management are divided into independent configuration modules, and communication between modules is achieved based on a decentralized data distribution mechanism. Therefore, when the monitoring scale changes or the algorithm is upgraded, there is no need to adjust the overall system structure, significantly reducing the complexity of system expansion and maintenance. Furthermore, because the monitoring system can be deployed in a distributed manner on heterogeneous computing devices, video analysis tasks are completed at the edge, avoiding the computing power bottleneck and single point of failure risk of centralized servers, thus improving the stability and reliability of the system in high-concurrency ICU monitoring scenarios. Further, by configuring the video analysis algorithm module to run independently at the bed level and combining it with the dynamic distribution of monitoring area parameters, continuous analysis of local patient behavior is achieved. This allows the system to generate early warning information in the early stages of unplanned extubation, transforming traditional post-event alarms into pre-event interventions, reducing the risk of unplanned extubation events, and improving the overall safety and adaptability of ICU bed monitoring. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0026] Figure 1 This is a diagram of the internal system configuration architecture of a single monitoring device according to the present invention;

[0027] Figure 2 This is a diagram of the distributed architecture of the present invention;

[0028] Figure 3 This is a flowchart of the configuration implementation method of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings.

[0030] Throughout the accompanying drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. The described embodiments are only a part of the embodiments of this application, not all of them. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0031] Example 1

[0032] In high-risk medical environments such as the Intensive Care Unit (ICU), unplanned extubation due to patient confusion, pain, or agitation often necessitates the placement of multiple life-sustaining tubes, including nasogastric tubes, endotracheal tubes, and urinary catheters. Unplanned extubation not only interrupts treatment and worsens the patient's condition but can also be life-threatening, significantly increasing the workload of medical staff. This embodiment utilizes specialized bedside video monitoring equipment combined with advanced intelligent video analytics algorithms to monitor and identify patient behavior in real time. The system accurately captures the characteristics of a patient's hand approaching or grasping the tubes and distinguishes between normal activity and extubation tendencies based on time-series data analysis. The intelligent extubation monitoring triggers a multi-level early warning mechanism upon detecting high-risk behavior, such as local audio-visual alerts to warn the patient or direct alarm messages to the nursing station. This allows medical staff to intervene before extubation occurs, transforming traditional "post-event handling" into "pre-event warning," significantly reducing the incidence of unplanned extubation and improving medical safety and nursing efficiency.

[0033] like Figure 3 As shown, this embodiment provides a configurable implementation method for an intelligent video monitoring system for ICU beds. It enables video monitoring and intelligent analysis of one or more ICU beds using a single monitoring device. The hardware deployment is flexible and not limited to specific hardware; the number of monitoring videos and algorithms can be flexibly configured according to the computing resources of the specific hardware. The specific steps include:

[0034] Step S1: Configuration design of the monitoring system.

[0035] The monitoring system is divided into multiple configuration modules based on its configuration functions, such as Figure 1 As shown, the configuration module includes at least a data acquisition module for video stream access and frame-by-frame decoding, a video analysis algorithm module for analyzing patient behavior and outputting alarm results, a Web interface service module for local resource scheduling and interface communication, and a management client module for centralized configuration of monitoring areas and aggregation and display of alarm results; the modules communicate with each other through a data distribution mechanism to form a loosely coupled software architecture.

[0036] Step S2, based on the DDS data communication mechanism.

[0037] The configuration module implements a general communication design through DDS. Each configuration module interacts with data by subscribing to and broadcasting data topics, rather than using direct calls between modules, thereby achieving logical decoupling between the data acquisition module, the video analysis algorithm module, and the Web interface service module.

[0038] Specifically, each configuration module is encapsulated as an independent computing node; each computing node is developed based on a distributed system framework that supports DDS communication, and through the dynamic discovery and quality of service management mechanism of the underlying DDS middleware, automatic networking and transparent data distribution across processes or physical devices are achieved.

[0039] When any configuration module is replaced or upgraded, it will not affect the normal operation of other modules.

[0040] Step S3: Distributed deployment of intelligent video surveillance system in heterogeneous environment.

[0041] Based on the above configuration design, the intelligent video surveillance system is deployed in a heterogeneous hardware environment, which includes CPUs, GPUs or edge computing devices based on different instruction set architectures; each computing device is connected to a unified video data broadcast pool through a local area network to support the distributed operation of the configuration module on different computing resources.

[0042] like Figure 2 As shown, the video surveillance system is deployed in a distributed manner, with multiple monitoring devices interconnected via a network. Each monitoring device independently runs its own configured software system. There are no uniform restrictions on the hardware model, architecture, or operating system type of different monitoring devices. Each monitoring device is connected to a camera on a different ICU bed. The camera video data is locally accessed and analyzed by the corresponding monitoring device. Different monitoring devices do not interfere with each other. Video data from one camera can also be simultaneously accessed by multiple different monitoring devices as needed to achieve parallel processing of different analysis tasks. The system includes the following sub-steps.

[0043] Step S3-1: Frame-by-frame broadcasting by the data acquisition module. The data acquisition module receives video data from a single camera, decodes the video stream output by the camera, and broadcasts the frame-by-frame video data via DDS.

[0044] Specifically, the data acquisition module is used to establish a connection with the bedside camera, receive video stream data from the camera, and decode the video stream to generate frame-by-frame image data. Each camera corresponds to one data acquisition module instance, and the data acquisition module broadcasts the frame-by-frame image data to the video data channel.

[0045] This can be understood as a single monitoring device being able to run multiple data acquisition module instances simultaneously, depending on network bandwidth and video decoding capabilities, to access video data from multiple cameras.

[0046] Step S3-2: The video analysis algorithm module subscribes to the frame-by-frame video data broadcast by the data acquisition module, analyzes and processes the video data, and outputs the risk assessment result.

[0047] Specifically, the specific analytical characteristics of the video analytics algorithm module are determined by the algorithm designer. When an algorithm instance only supports analysis of one video stream, a separate algorithm module instance is run for each video stream accessed by the device. When an algorithm instance supports analysis of multiple video streams, the number of running algorithm module instances is configured according to actual needs. Furthermore, different ICU beds can be configured with different video analytics algorithms, and different algorithm module instances can run simultaneously on the same monitoring device. After completing the analysis, the video analytics algorithm module performs necessary business processing on the analysis results, including alarm decisions and data storage, and broadcasts the processing results via DDS.

[0048] This can be understood as follows: The video analysis algorithm module extracts temporal image features from the spatial topological relationship between the patient's hand and each indwelling tube to identify the continuous intentional actions from "approaching" to "touching" and then to "grasping and pulling," thereby distinguishing between normal physiological activities and tube removal tendencies.

[0049] The video analysis algorithm module stores the generated monitoring data into the local database on the edge device in real time; it allows for flexible selection of different data persistence schemes, from embedded lightweight databases to large server databases, based on the heterogeneous hardware capabilities of the devices, so as to achieve full and rational utilization of hardware and software resources; this module usually only transmits data when responding to data query requests from the management client module, thereby effectively reducing the bandwidth load of the local area network.

[0050] Step S3-3: Registration and service of the Web interface service module. Only one Web interface service module is deployed on each monitoring device performing video analysis tasks. The Web interface service module supports dynamic distribution of ROI area parameters, allowing medical staff to remotely adjust the monitoring range of the algorithm module for specific bed tubing areas through the visual interface of the management client module, and to use the HTTP interface to review key frames of historical alarm images.

[0051] Specifically, the Web interface service module is used to register the configuration information of the cameras connected to the device and to provide an HTTP interface service. By calling this interface, functions such as setting monitoring areas, querying monitoring data, and reviewing video footage can be achieved. This Web interface service module only provides services to cameras connected to the device, i.e., to the ICU beds currently managed by the device.

[0052] It should be noted that in the heterogeneous deployment method described in step S3, the system of the present invention can be deployed on a single server with strong computing power or on multiple devices with weak computing power to form a distributed system; it supports multiple video analysis algorithm instances with different functions to subscribe to the same frame-by-frame video data topic in parallel, and the business processing content can be configured according to specific application requirements, such as writing the analysis results to the database or sending an alarm signal when abnormal behavior is detected. The above business processing is completed in the video analysis algorithm module to simultaneously perform the tasks of extubation intention recognition and fall risk assessment for the same ICU bed.

[0053] Step S4: Alarm device docking. Based on the real-time analysis results output in step S3-2, the alarm device is docked. When the analysis results meet the preset alarm conditions, the alarm device is driven to perform an alarm operation. The alarm device is a portable wireless vibration alert terminal. After receiving the alarm result, it performs a vibration alert and simultaneously parses the alarm information to display the corresponding bed number and risk level on the terminal screen.

[0054] The system has the ability to be reconfigured flexibly. When the number of ICU monitoring beds increases or decreases, the management client module automatically completes the system topology update by scanning the IP address and interface status of the Web service, realizing the expansion and contraction of nodes without being noticed.

[0055] Step S5: Deployment and centralized management of the management client module. Deploy the management client module on any computing device within the local area network, and configure the video surveillance equipment through the corresponding visual client. Through the management client module, realize unified configuration, video analysis task management and monitoring data acquisition for each ICU bed monitoring camera.

[0056] To further explain, a single computing device can support the simultaneous launch of multiple data acquisition module instances based on the concurrent processing capability of its hardware decoder. Each instance corresponds to one RTSP video stream from a camera and is published to an independent DDS topic.

[0057] This can be understood as follows: The management client module runs as a client on the management terminal device in the network. By accessing the Web interface service modules provided by each monitoring device, it enables unified configuration, status query, and data acquisition of distributed monitoring devices. The management client module only undertakes configuration and management functions and does not participate in the actual analysis and calculation of video data, thereby reducing the risk of centralized dependence on the system.

[0058] Example 2

[0059] Building upon Example 1, this example utilizes five NVIDIA Jetson Orin Nano 8G edge computing devices, each connected to two RTSP video streams. A Windows 10 PC is used with the client program, i.e., the management client module, installed. The alarm device is a wireless vibration alert device carried by medical staff; when an alarm is triggered, the device vibrates and displays the bed number on the screen.

[0060] Step S1: Configuration design of the monitoring system. The management client module is a web application; the data acquisition module performs RTSP stream decoding; the video analysis algorithm module analyzes only one video stream; and the web interface service module registers the two video streams currently connected to the device.

[0061] Step S2: The distributed system framework based on DDS communication develops the data acquisition module and video analysis algorithm into nodes. First, a web interface service module is written using Flask technology, and then a node program is developed to subscribe to real-time data through DDS and run the Flask program, providing a real-time interface.

[0062] Step S3: Heterogeneous deployment of the intelligent video surveillance system.

[0063] Step S3-1: On each edge computing device, start two data acquisition node instances, connect to two different video streams, and broadcast frame-by-frame data respectively;

[0064] Step S3-2: On the same edge computing device, start two video analysis algorithm node instances and connect to the frame-by-frame data of the two videos; use a lightweight embedded database for local persistence, with each monitoring device in its local database, and real-time result data is entered into the database, without using a public database within the ring network; send real-time alarm signals based on the result data.

[0065] Step 3-3: Start a web interface service node and register the parameters of the two cameras with the service before starting. The web interface service node provides HTTP interface services for local settings, data services, etc.

[0066] Step S4: Connect the alarm device. Match the alarm device with the signal transmitter to ensure proper operation.

[0067] Step S5: Deploy the management client module. On a Windows 10 PC, deploy and run the main control client. The interface can then be accessed via a browser on the local machine or other PCs within the network segment. After configuring the monitoring device information, you can configure each camera, such as setting ROI, querying monitoring data, and viewing historical screenshots.

[0068] Example 3

[0069] Monitoring cameras are installed next to the ICU beds that require video monitoring. These cameras are used to capture video images of the patients in the corresponding beds and are connected to the ICU's internal local area network via a network. Edge computing devices are connected to the local area network via a wired network, enabling them to access the video data output by the cameras. Following the configuration deployment method described in Embodiments 1 and 2, the data acquisition module, video analysis algorithm module, and Web interface service module are deployed on the edge computing device, and the alarm device is connected to the system. At the same time, the main control client is deployed and run on the management terminal.

[0070] During operation, the video monitoring system continuously collects and analyzes video data from ICU beds. When the video analysis algorithm module identifies a patient's risk of unplanned extubation based on real-time video data, the system generates corresponding alarm information and sends a notification to medical staff through the alarm device. The alarm device uses vibration to alert the staff, and when the vibration device generates an alarm, the corresponding ICU bed number indicating the risk of extubation is displayed simultaneously.

[0071] After receiving an alarm notification, medical staff can go to the corresponding bed to check the patient's condition and take necessary manual intervention measures as needed to prevent unplanned extubation. They can also configure and adjust the monitoring system through the main control client, including modifying the parameters of the camera monitoring area, querying historical monitoring data, and viewing historical alarm screenshots.

[0072] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A configurable implementation method for an intelligent video monitoring system for ICU beds, used for video monitoring and early warning of abnormal and dangerous actions of patients in an intensive care unit environment, characterized in that, Includes the following steps: S1, the monitoring system is divided into multiple configuration modules according to the configuration function. The configuration module includes at least a data acquisition module for video stream access and frame-by-frame decoding, a video analysis algorithm module for analyzing patient behavior and outputting alarm results, a Web interface service module for local resource scheduling and interface communication, and a management client module for centralized configuration of monitoring areas and aggregation and display of alarm results. S2, build a decentralized communication mechanism based on the Data Distribution Service (DDS). Each configuration module interacts with data by publishing and subscribing to data topics, replacing the direct calling method between modules and realizing the logical separation between modules; S3 enables distributed configuration deployment in a heterogeneous hardware environment, allowing the monitoring system to run in a distributed architecture consisting of a single computing device or multiple computing devices. Specifically, the data acquisition module accesses the video stream from the bedside camera and publishes frame-by-frame video data, the video analysis algorithm module subscribes to the video data and generates alarm results when it determines that there is a risk of unplugging the tube, and the Web interface service module registers the configuration of the cameras under the jurisdiction of this device and provides parameter configuration and data access services. S4, the system monitors alarm results in real time and drives alarm devices to perform prompting operations; S5, the management client module runs on the management terminal, and uses the web interface of each computing device to send parameters and aggregate results, but does not participate in the decoding and analysis of video data.

2. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 1, characterized in that, Each configuration module is encapsulated as an independent computing node; each computing node is developed based on a distributed system framework that supports DDS communication, and through the dynamic discovery and quality of service management mechanism of the underlying DDS middleware, it realizes automatic networking and transparent data distribution across processes or physical devices.

3. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 2, characterized in that, The system is built on a hardware environment that supports heterogeneous computing, which includes CPUs, GPUs or edge computing devices based on different instruction set architectures; each computing device is connected to a unified video data broadcast pool through a local area network to support the distributed operation of the configuration module on different computing resources.

4. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 1, characterized in that, The video analysis algorithm module extracts temporal image features from the spatial topological relationship between the patient's hand and each indwelling tube to identify the continuous behavioral evolution process from the approach state, contact state to the force application state, thereby distinguishing between normal physiological activities and extubation risk behaviors.

5. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 4, characterized in that, The video analysis algorithm module stores the generated monitoring data in real time into a local database on the edge device side. The local database is an embedded lightweight database. The embedded lightweight database is connected to and establishes a data read / write connection with the video analysis algorithm module, and completes data storage and query through the local database interface. The monitoring data is output to the management client module after receiving a data query request from the management client module.

6. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 5, characterized in that, The Web interface service module supports dynamic distribution of ROI area parameters, allowing medical staff to remotely adjust the monitoring range of the algorithm module for specific bed tubing areas through the visual interface of the management client module, and to use the HTTP interface to recall key frames of historical alarm images.

7. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 6, characterized in that, The same frame-by-frame video data topic is subscribed to in parallel by multiple video analysis algorithm instances with different functions, so as to simultaneously perform extubation intention recognition and fall risk assessment tasks for the same ICU bed.

8. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 1, characterized in that, In step S4, the alarm device uses a portable wireless vibration alert terminal. After receiving the alarm result, it performs a vibration alert and simultaneously parses the alarm information to display the corresponding bed number and risk level on the terminal screen.

9. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 1, characterized in that, The system has the ability to be flexibly reconfigured. When the number of ICU monitoring beds changes, the management client module completes the system topology update by scanning the IP address and interface status of the Web service.

10. The configuration implementation method of the intelligent video monitoring system for ICU beds according to claim 1, characterized in that, Based on the concurrent processing capability of its hardware decoder, a single computing device can simultaneously launch multiple data acquisition module instances, each instance corresponding to one RTSP video stream from a camera and publishing it to an independent DDS topic.

Images