An unintended extubation pre-identification and pull-off alarm method and system based on an identifier

Abstract

This invention provides a method for identifying unplanned extubation based on markers, comprising: inputting an image containing a marker into a marker recognition module for posture determination to obtain posture information, and cropping the marker to obtain a cropping frame containing the marker; inputting the cropping frame into a semantic segmentation module for supervision object segmentation to obtain the supervision object and the active body; determining the activity state of the active body on the supervision object based on the spatial position of the active body, the cropping frame, and / or the supervision object, and issuing an unplanned extubation supervision alarm. By classifying patient postures and actions, alarm and warning information are provided for different behavior models, thereby achieving the goal of accurately preventing unplanned extubation.

Description

Technical Field

[0001] This invention belongs to the field of catheter anti-dislodgement technology, specifically relating to a non-planned catheter removal identification method based on markers. Background Technology

[0002] To deliver necessary fluids and food to patients who cannot swallow under special circumstances, a nasogastric tube is inserted into the esophagus through the mouth or nostril. Unplanned removal of the trachea may damage the patient's mucous membranes and irritate the esophagus, causing severe physiological discomfort. Some patients with respiratory diseases cannot maintain adequate ventilation or oxygenation and cannot have sufficient gas exchange, so they usually use a ventilator to assist breathing. Unplanned removal of the trachea may cause respiratory arrest and death. Improper handling when changing the patient's position or clothes can cause excessive pulling and easily dislodge the tube. Patients may also manually remove the tube when they are uncomfortable or unconscious.

[0003] Such actions are often brief and difficult to prevent through real-time human monitoring. Chinese patent document, application number: 202410510955.5, discloses a method for monitoring and dynamically capturing patient extubation actions based on deep learning, but it does not solve the technical problem of how to prevent patients from extubating in advance. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention establishes different models to correspond to different patient postures and movements, thereby achieving the purpose of early warning and alarm for unplanned extubation.

[0005] The first aspect of the present invention provides a method for pre-identification and alarm of unplanned tube removal based on markers: including: inputting an image containing a passive marker into a passive marker recognition module for attitude determination to obtain attitude information, and performing passive marker cropping to obtain a cropping frame containing the passive marker;

[0006] The image containing the passive markers is input into the semantic segmentation module for supervised object segmentation to obtain supervised objects, active objects, and monitoring areas.

[0007] Based on the spatial location of the active body, the monitoring area, and / or the supervised object, determine the activity status of the active body on the supervised object, and issue an unplanned tube removal monitoring alarm.

[0008] A second aspect of the present invention provides: a marker-based unplanned tube removal pre-identification and removal alarm system, the system comprising at least one image acquisition terminal, at least one processor; and a memory storing instructions that, when executed by the at least one processor, implement the steps of the above-described method.

[0009] A third aspect of the present invention provides: a computer-readable storage medium for a method of pre-identification and removal alarm of unplanned tube removal based on identifiers, wherein a computer program / instructions are stored thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the above-described method.

[0010] The fourth aspect of the present invention provides a computer program product for a method of pre-identification and removal alarm of unplanned tube removal based on identifiers, comprising a computer program / instructions, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the above-described method.

[0011] The fifth aspect of the present invention provides: a network for unplanned tube removal pre-identification and removal alarm based on identifiers, characterized in that a first server and a second server are provided, wherein the first server or the second server is capable of executing the steps of the method described above.

[0012] The beneficial effects of this invention are: by classifying patient postures and behaviors, alarm and warning information are provided for different behavioral models, thereby achieving the goal of accurately preventing unplanned extubation. Attached Figure Description

[0013] Figure 1 Flowchart of alarm identification according to an embodiment of the present invention;

[0014] Figure 2 Flowchart of patient-restraining alarm in this invention embodiment;

[0015] Figure 3 Flowchart of the unrestrained patient early warning system according to an embodiment of the present invention;

[0016] Figure 4 Flowchart of the alarm process for identifying unrestrained and restrained patients according to an embodiment of the present invention;

[0017] Figure 5 Flowchart of constraint state information acquisition according to an embodiment of the present invention;

[0018] Figure 6 Flowchart of attitude information acquisition according to an embodiment of the present invention;

[0019] Figure 7 Flowchart of the frame acquisition process according to an embodiment of the present invention;

[0020] Figure 8 Flowchart of attitude information classification according to an embodiment of the present invention;

[0021] Figure 9 Flowchart of the supervision task issuance in this embodiment of the invention;

[0022] Figure 10 System structure diagram of an embodiment of the present invention;

[0023] Figure 11 Flowchart of posture information classification and recognition according to an embodiment of the present invention;

[0024] Figure 12 System structure diagram of an embodiment of the present invention;

[0025] Figure 13 System structure diagram of an embodiment of the present invention;

[0026] Figure 14 System structure diagram of an embodiment of the present invention;

[0027] Figure 15 System structure diagram of an embodiment of the present invention;

[0028] Figure 16 Schematic diagram of the algorithm structure in this embodiment of the invention;

[0029] Figure 17 An illustration of the early warning interface of the nurse station monitoring terminal in this embodiment of the invention;

[0030] Figure 18 An illustration of the alarm interface of the nursing station monitoring terminal during the extubation completion stage in an embodiment of the present invention;

[0031] Figure 19 Patient facial recognition image in an embodiment of the present invention;

[0032] Figure 20 Patient facial recognition image in an embodiment of the present invention;

[0033] Figure 21 A diagram of the bed interface of the nurse station monitoring terminal in this embodiment of the invention, showing how to disable the unplanned extubation recognition function with one click. Detailed Implementation

[0034] The following embodiments further illustrate the content of the present invention, but should not be construed as limiting the present invention. Any modifications or substitutions made to the methods, steps, or conditions of the present invention without departing from the spirit and essence of the invention are within the scope of the present invention.

[0035] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0036] In some embodiments of the present invention, such as Figure 1-21 As shown: A method for pre-identification and alarm of unplanned tube removal based on identifiers, characterized in that it includes:

[0037] An image containing a passive marker is input into a passive marker recognition module for pose determination to obtain pose information. The module then performs passive marker cropping to obtain a bounding box containing the passive marker, as shown in the image. Figure 19 , 20 The blue box in the middle;

[0038] The image containing passive markers is input into a semantic segmentation module for supervised object segmentation to obtain supervised objects, moving bodies, and monitoring areas. Supervised objects include, for example, identified tubing such as endotracheal tubes; moving bodies include, for example, hands, gripping devices, etc.; and monitoring areas include, for example,... Figure 19 , 20 The yellow box containing the face;

[0039] Based on the spatial location of the active body, the monitoring area, and / or the supervised object, determine the activity status of the active body on the supervised object, and issue an unplanned tube removal monitoring alarm.

[0040] These embodiments:

[0041] This solution primarily targets unplanned extubation procedures for patients in hospital ICUs. To address the need for pre-identification and alarm systems for unplanned extubation, the core identification targets are faces, tubes, and hands. Face recognition already boasts good accuracy within the existing technological framework; therefore, by further improving the accuracy of tracheal tube and hand recognition, accurate identification of unplanned extubation actions can be achieved. The monitoring objects can be endotracheal tubes, nasogastric tubes, etc.; passive markers can be special materials like adhesive tape, easily identifiable colors, etc.; and the frame of view includes a recognition frame containing the face, such as... Figure 19 The blue box indicates alarm information, including warning information and alarm information.

[0042] These embodiments:

[0043] Unplanned extubation, such as patients pulling out their own catheters when they are uncomfortable or unconscious, is a key issue in clinical risk management. In this embodiment, unplanned extubation is mainly divided into the following scenarios:

[0044] 1. Provide timely warnings regarding intubated patients' hands entering their face;

[0045] The patient does not have the right to be restrained:

[0046] By using facial recognition to detect faces and hands within the facial recognition area, if a hand is detected entering the patient's face and remains there for a period of time, it is considered a risk of extubation and an early warning is immediately generated.

[0047] Conditions involving patient constraints:

[0048] Hand intrusion warning: Timely alarms are issued for intubated patients' hands entering their faces, and the alarm information is promptly pushed to the nursing station monitoring terminal and nurses' mobile terminals.

[0049] Position recognition: supine position, left lateral position, right lateral position, prone position.

[0050] Identify tubing: endotracheal tube, high-flow oxygen catheter, gastric tube.

[0051] Both daytime and nighttime scenes can be recognized. The camera (image acquisition end) is installed directly above the foot of the bed, within 2.5 meters above the bed surface. By adjusting the focus, it can capture clear human faces.

[0052] In the recognition scenario, the optimal recognition conditions are: face is not severely obscured, pipelines are unobstructed, and the person is lying flat.

[0053] Medical staff issue a request for a testing bed and begin testing at the bed. If a hand is detected entering the facial area, an alert is issued.

[0054] 2. Issue an alarm when an intubated patient pulls the tube out of their nose or mouth;

[0055] Unplanned tube removal: An alarm is triggered for removal from the nose or mouth. If a hand is detected touching or near the tube in the face area and the tube is removed from the functional area, it is considered tube removal, and an alarm message is sent to the system.

[0056] Both daytime and nighttime scenes can be recognized. The camera (image acquisition unit) is installed directly above the foot of the bed, within 2.5 meters of the bed surface. By adjusting the focus, it can capture clear faces. The captured images can clearly identify faces, pipes, and tape colors.

[0057] In the recognition scenario, the optimal recognition conditions are as follows: the face is not severely blocked, the pipeline is not blocked, and the person is lying flat. When using the function of identifying and alarming for unplanned extubation, medical staff issue the detection bed, start the bed detection, and when it is recognized that the hand is close to the pipeline and the pipeline is out of the functional area, an alarm is given.

[0058] 3. Receive early warning information and turn off the monitoring of non-extubation beds;

[0059] 4. Push the alarm information to the nurse terminal in a timely manner;

[0060] The extubation status of patients is mainly recognized by video analysis. Different analysis rules are preset in the camera scenario. Once the target shows behaviors that violate the predefined analysis rules in the scenario, the system will trigger the preset linkage rules. The monitoring of unplanned extubation of patients is efficiently achieved.

[0061] In some embodiments of the present invention, as Figure 2 、 4 shown, the difference from the above embodiments is that the step of judging the activity state of the moving object with respect to the supervised object according to the moving object, the intercepting frame and / or the spatial position of the supervised object and sending out the supervision and alarm information for unplanned extubation includes:

[0062] The step of judging the activity state of the moving object with respect to the supervised object according to the moving object, the intercepting frame and / or the spatial position of the supervised object and sending out the supervision and alarm information for unplanned extubation includes:

[0063] Obtain the constraint status information of the supervised object;

[0064] Judge whether the supervised object is constrained. If the supervised object is constrained, then judge whether there is the moving object in the monitoring area. If it exists, send out the alarm information for unplanned extubation;

[0065] If the supervised object is not constrained, then judge whether there is the moving object in the monitoring area and whether the residence time of the moving object in the monitoring area exceeds the warning threshold. If it exceeds, send out the early warning information for unplanned extubation.

[0066] In some embodiments of the present invention, as Figure 3 、 4 shown, the difference from the above embodiments is that:

[0067] Obtain the constraint status information of the supervised object;

[0068] Judge whether the supervised object is constrained. If the supervised object is not constrained, then judge whether there is the moving object in the monitoring area and whether the residence time of the moving object in the monitoring area exceeds the warning threshold. If it exceeds, send out the early warning information for unplanned extubation.

[0069] In these embodiments: Passive marker recognition basis: To improve the accuracy of target object recognition, factors such as increasing color contrast and screen proportion are used to increase the distinction between target and non-target objects, thereby reducing the recognition difficulty. Conventional medical equipment and human body dimensions are relatively fixed, while in hospital wards, the colors of bed sheets, hospital gowns, and other environmental elements are mainly white and blue, resulting in a relatively uniform color scheme. Therefore, objects with clearly distinguishable colors can be selected as markers to assist in identification and judgment.

[0070] By changing the surface color of tubing such as endotracheal tubes, the color of the fixation straps and adhesive tape used for auxiliary fixation, etc., this embodiment selects colored adhesive tape with a significant color contrast to items related to the hospital bed, such as purple or orange, which makes image recognition easier to identify and analyze, and improves the accuracy of tubing identification.

[0071] The distinction between unrestrained and restrained patients is based on the following: the installation of tubes such as endotracheal tubes and gastric tubes, which play an important role in maintaining the patient's vital signs, requires strict monitoring. Different warning and alarm conditions are set. In order to prevent patients from extubating unconsciously, some patients will be specially monitored, and their behavior will also be restrained to a certain extent, such as restraining their hands with sleeves or straps to limit their mobility.

[0072] 1. For unrestrained patients, pre-identification is performed by determining whether the duration of the patient's arm marker within the facial recognition frame exceeds a threshold. In this embodiment, the threshold ranges from 0.5s to 5s, specifically 0.5s, 1s, 1.5s, 2s, 2.5s, 3s, 3.5s, 4.0s, 4.5s, and 5s; preferably 1s. The arm marker is a passive marker. A passive marker is colored adhesive tape with a clear color contrast to items related to the hospital bed, such as purple or orange, which facilitates image recognition and analysis.

[0073] 2. For restrained patients, pre-identification - an alarm will be triggered immediately if the patient enters the facial recognition system via an arm tag.

[0074] 3. Post-incident alarm - For patients with ordinary risk level intubation (generally unrestrained patients), the presence of a passive facial marker indicating whether the tube has fallen off can be used to determine if the tube has dislodged. For patients with severe risk level intubation (generally restrained patients), in addition to checking the presence of the passive facial marker, an active mother-daughter marker can also be installed to improve alarm accuracy.

[0075] Among them, the active marker is driven by a button battery and has a magnetic unit. It can be partially closed on the pipe and partially pasted and fixed to the patient's face. When the pipe is displaced, it can emit a separation alarm.

[0076] In some embodiments of the present invention, such as Figure 4, 19 As shown in Figure 20, the difference from the above embodiment is that:

[0077] The captured frame is input into the semantic segmentation module for supervised object segmentation, obtaining supervised objects, active bodies, and monitoring areas, including the following steps:

[0078] The captured bounding box is input into the semantic segmentation module for supervised object segmentation;

[0079] The system identifies the monitored object, the moving body, and the monitoring area. If the monitored object is not located within the monitoring area, a pipe-pulling alarm is issued.

[0080] In some embodiments of the present invention, such as Figure 5 , 17 As shown in Figure 18, the difference from the above embodiment is that the step of obtaining the constraint state information of the supervised object includes:

[0081] The information of the supervised object is associated with the information from the image acquisition terminal;

[0082] The constraint status information is obtained by associating the constraint status information with the information of the supervised object.

[0083] These embodiments:

[0084] In this embodiment, unrestrained patients have corresponding medical records and identifiers on the monitoring end. The medical records and identifiers can be used to determine whether a patient is restrained. The patient's corresponding bed number and the image acquisition device corresponding to the bed number, such as a high-definition camera, have corresponding device numbers. Based on this, a mapping or information management table is established for the patient, the patient's restraint status, the bed number, the camera, and the restraint status information of the camera's captured images, thus forming a judgment on the patient's restraint status.

[0085] Meanwhile, the identification can be generated at the monitoring terminal. For example, the restraint status can be selected on the operation interface, or restraint prompt buttons can be designed at the locations of patient-related equipment, such as ventilators and gastric tubes. When medical staff press the prompt button, the restraint patient identification is simultaneously generated at the monitoring terminal. This allows for quick acquisition of the patient's restraint status and facilitates further monitoring of the patient's extubation process.

[0086] In some embodiments of the present invention, such as Figure 1-4 As shown in Figures 6 and 16, the difference from the above embodiments is that the step of inputting the image containing the marker into the marker recognition module for posture determination to obtain posture information includes:

[0087] The image containing the marker is input into the marker recognition module, and several marker points are matched to obtain several marker points.

[0088] The posture of the supervised object is identified based on the distribution information of the marker points, and the posture information of the supervised object is obtained.

[0089] In some embodiments of the present invention, such as Figure 7 , 16 As shown, the difference from the above embodiment is that the step of capturing the marker to obtain a capture frame containing the passive marker includes:

[0090] Input the image containing the passive marker into the marker recognition module;

[0091] Perform semantic segmentation to obtain bounding boxes containing passive identifiers.

[0092] In some embodiments of the present invention, such as Figure 8 , 10 As shown, the difference from the above embodiments is that:

[0093] The step of identifying the posture of the supervised object based on the distribution information of the marker points and obtaining the posture information of the supervised object further includes:

[0094] Obtain the posture information of the supervised object, classify the posture information of the supervised object, obtain the posture information classification information, and send the posture information classification information to the monitoring terminal;

[0095] If the posture information classification information meets the recognition standard (lying upright), then passive marker interception will be performed.

[0096] These embodiments:

[0097] Taking YOLOv8 as an example: it uses the YOLOv8 human keypoint model pose and the YOLOv8 semantic segmentation model segment; the pose model can find 17 keypoints of the human body; the segmentation model is responsible for detecting tubes.

[0098] Among them, the keypoint model pose is a functional module in the YOLOv8 series specifically used for human pose estimation and keypoint detection. It can be used for real-time human pose estimation, action recognition, and behavior analysis. Correspondingly, it can also be combined and replaced with OpenPose multi-person keypoint detection system, MediaPipe Pose detection module, HRNet human pose estimation and keypoint detection network, and SimpleBaseline human keypoint detection model.

[0099] The semantic segmentation model is a functional module focused on image semantic segmentation. It can segment an image into multiple regions and assign semantic labels to each region, thereby achieving accurate identification and differentiation of different objects in the image. Correspondingly, U-Net image segmentation model, DeepLab series image segmentation model, Mask R-CNN image segmentation model, and PSPNet (Pyramid Scene Parsing Network) image segmentation model can be selected as combinations and replacements for the segment model.

[0100] The recognition result is shown in the figure. Figure 19 and 20 As shown: The pose model is used for detection, and the red box represents the pose model result; the lying posture is determined based on the facial key points of the pose; a yellow box is generated based on the facial key points to identify the person's face.

[0101] In some embodiments of the present invention, such as Figure 9 , 11 As shown in 12, 13, 14, 15, and 21, the difference from the above embodiments is that:

[0102] The identification method also includes a supervision task distribution step:

[0103] The monitoring terminal selects the monitored and unmonitored objects and issues monitoring tasks.

[0104] The corresponding image acquisition terminal for the monitored object will transmit the image containing the passive marker to the processor.

[0105] The alarm statistics are shown in Table 1.

[0106] Table 1

[0107]

[0108]

[0109] In some embodiments of the present invention, such as Figure 10 , 11 As shown in Figures 12, 13, 14, and 15, an unplanned tube removal pre-identification and removal alarm system based on identifiers is disclosed. The system includes at least one image acquisition terminal, one monitoring terminal, at least one processor, and a memory. The image acquisition terminal communicates with the processor, and the monitoring terminal communicates with the processor. The memory stores instructions that, when executed by at least one processor, implement the steps of the above-described method.

[0110] In some embodiments of the present invention, such as Figure 1As shown, the difference from the above embodiments is that: a computer-readable storage medium for an unplanned tube removal identification method based on identifiers stores a computer program / instructions thereon, characterized in that the computer program / instructions, when executed by a processor, implement the steps of the above method.

[0111] In some embodiments of the present invention, such as Figure 1 As shown, the difference from the above embodiments is that: a computer program product for a non-planned tube removal identification method based on identifiers includes a computer program / instructions, characterized in that the computer program / instructions implement the steps of the above method when executed by a processor.

[0112] In some embodiments of the present invention, such as Figure 1 As shown, the difference from the above embodiments is that: an unplanned tube removal identification network based on identifiers, characterized by a first server and a second server, wherein the first server or the second server is capable of performing the steps according to the method described above.

[0113] The image acquisition terminal captures images of the patient's face and adjacent areas, transmits these images to a server via a network, and uses AI algorithms to intelligently recognize the image content. The server then stores the recognition results in a database and transmits them to the monitoring terminal via the database, integration platform, and network. The image acquisition terminal is preferably a network camera, with each camera uniformly positioned above the bed. The camera's zoom function is used to adjust the image to capture the patient's upper body.

[0114] The processor can be a server, which can be an ECS server. It acquires and configures computing resources, such as CPU, memory, disk and network, according to the computing needs generated by the scene where the identification device is located.

[0115] The processor may include an intelligent recognition unit with a pre-installed image recognition program. It can determine the patient's posture, such as lying on their side or supine, based on the patient's face and surrounding area. It can also draw a facial recognition box in the image. Furthermore, it can identify the patient's tubes and unplanned extubation actions through semantic segmentation.

[0116] The memory can be a database. The database stores the images generated after the intelligent recognition unit recognizes them, and inserts, updates, deletes, and retrieves the generated image data based on the recognition information.

[0117] The network structure may include an integration platform, which is a software system or infrastructure used to enable interconnection, integration and collaboration between two or more different applications, business processes and data sources, such as intelligent identification units and monitoring terminals.

[0118] The network can be a wireless network or a combination of wired and wireless networks. Wireless networks include modules such as Bluetooth, Wi-Fi, GSM, GPRS, CDMA, CDMA2000, WCDMA, TD-SCDMA, Zigbee, or LoRa. By employing multiple wireless communication methods, the flexibility of wireless communication can be increased, and the needs of different users and occasions can be met. In particular, when using LoRa modules, their communication range is relatively long and their communication performance is relatively stable, making them suitable for occasions with high communication quality requirements.

[0119] The preferred monitoring terminal uses electronic screens to display and convey early warning and alarm information about unplanned extubation to medical staff.

[0120] The first monitoring terminal is a mobile terminal for nurses. It uses devices such as watches and mobile phones with display screens, sound prompts, and vibration alerts to issue warning and alarm signals. It can be carried by medical staff to monitor or receive warning and alarm information at any time.

[0121] The second monitoring terminal is the nurse station monitoring terminal, which uses desktop computers or other external display devices with screens, speakers, and interactive devices such as touch screens, keyboards, and mice for human-computer interaction. It allows monitoring, receiving and transmitting warnings and alarms to medical staff, or initiating and pausing video monitoring of a specific patient's bed.

[0122] The technical effects of this invention are multifaceted. By classifying patient postures and behaviors, alarm and warning information are provided for different behavioral models, thereby achieving the goal of accurately preventing unplanned extubation.

[0123] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0124] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0125] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0126] Although the present invention has been described in detail above with general descriptions, specific embodiments, and experiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for pre-identification and alarm of unplanned tube removal based on markers, characterized in that, include: An image containing passive markers is input into a passive marker recognition module for pose determination to obtain pose information, and passive markers are cropped to obtain a cropping box containing the passive markers. The image containing the passive markers is then input into a semantic segmentation module for supervised object segmentation to obtain supervised objects, moving bodies, and monitoring areas. The supervised objects include the identified pipes; the moving bodies include hands or gripping devices; and the monitoring areas include faces. Based on the spatial location of the active body, the monitoring area, and / or the monitored object, the activity status of the active body on the monitored object is determined, and an alarm message for unplanned tube removal is issued. This step specifically includes: Obtain the constraint status information of the supervised object; Determine whether the monitored object is constrained. If the monitored object is constrained, determine whether the active body exists in the monitoring area. If it exists, issue an alarm message for unplanned pipe removal. If the monitored object is not constrained, it is determined whether the active body exists in the monitoring area and whether the time the active body stays in the monitoring area exceeds the warning threshold. If it does, an unplanned tube removal warning message is issued. The method also includes a tube-pulling alarm step: The captured bounding box is input into the semantic segmentation module for supervised object segmentation; The system identifies the monitored object, the moving body, and the monitoring area. If the monitored object is not located within the monitoring area, a tube-pulling alarm is issued. The steps for obtaining attitude information by inputting an image containing markers into a marker recognition module for attitude determination include: The image containing the marker is input into the marker recognition module, and several marker points are matched to obtain several marker points. The posture of the supervised object is identified based on the distribution information of the marker points, and the posture information of the supervised object is obtained. The step of identifying the pose of the supervised object based on the distribution information of the marker points and obtaining the pose information of the supervised object further includes: Obtain the posture information of the supervised object, classify the posture information of the supervised object, obtain the posture information classification information, and send the posture information classification information to the monitoring terminal; If the posture information classification information meets the recognition criteria, then passive marker extraction will be performed.

2. The method for pre-identification and removal alarm of unplanned tube removal based on markers according to claim 1, characterized in that, The step of obtaining the constraint status information of the supervised object includes: The information of the supervised object is associated with the information from the image acquisition terminal; The constraint status information is obtained by associating the constraint status information with the information of the supervised object.

3. The method for pre-identification and removal alarm of unplanned tube removal based on markers according to claim 1, characterized in that, The steps for capturing markers and obtaining a capture frame containing passive markers include: Input the image containing the passive marker into the marker recognition module; Perform semantic segmentation to obtain bounding boxes containing passive identifiers.

4. The method for pre-identification and removal alarm of unplanned tube removal based on markers according to claim 1, characterized in that, The method also includes a step of supervising the task distribution: The monitoring terminal selects the monitored and non-monitored objects and issues monitoring tasks. The corresponding image acquisition terminal for the monitored object will transmit the image containing the passive marker to the processor.

5. A marker-based unplanned tube removal pre-identification and removal alarm system, the system comprising at least one image acquisition terminal, one monitoring terminal, at least one processor, and a memory; the image acquisition terminal communicates with the processor, and the monitoring terminal communicates with the processor; the memory stores instructions that, when executed by at least one processor, implement the steps of the method according to any one of claims 1-4.

Images