Personnel behavior monitoring method and device, electronic equipment and medium

Abstract

The invention provides a personnel behavior monitoring method and device, electronic equipment and a medium, and effectively solves the problems of incomplete supervision mechanism, fuzzy execution standard and risk early warning lag of an existing disinfection supervision system for a breeding house. The method comprises the steps that a configured image acquisition module triggers a target event and pushes the target event to a first message queue; the edge calculation module monitors a target event, acquires a target video clip acquired by the configured image acquisition module in real time, and extracts a plurality of key frame pictures from the target video clip; controlling a pre-trained behavior detection model to recognize the multiple key frame pictures from multiple recognition dimensions to obtain corresponding dimension results, and integrating the dimension results to obtain a recognition result; and the recognition result is pushed to a result display module, so that the result display module analyzes the recognition result and generates warning information, and behavior monitoring of the target person is ended based on the warning information.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and more specifically, to a method, device, electronic device, and medium for monitoring human behavior. Background Technology

[0002] In the biosecurity management system of large-scale livestock and poultry farming, the disinfection process for personnel entering and exiting the farm is a core control point for preventing the spread of pathogens. According to animal disease prevention technical specifications, personnel must strictly adhere to a three-tiered disinfection procedure before entering the farm pens: facial protection (wearing masks properly), hand disinfection (soaking hands in disinfectant for 10 seconds), and disinfection of contact surfaces (spraying door handles with alcohol). However, the current disinfection supervision system has the following technical deficiencies: 1. Incomplete regulatory mechanism: The existing disinfection process relies on subjective manual execution, lacks objective supervision methods, and cannot form a traceable disinfection record archive; 2. Vague Implementation Standards: Disinfection procedures lack quantifiable standards for assessment, resulting in deviations such as improper mask wearing, insufficient hand disinfection time, and low coverage of door handles. Field investigations revealed significant loopholes in the current disinfection process: approximately 31% of personnel did not wear masks correctly throughout the entire process, hand disinfection compliance was less than 35%, and the pass rate for door handles was only 62%. These deficiencies directly increase the risk of pathogen transmission by 27%-43%, posing a significant threat to the health of pig herds.

[0003] 3. Delayed risk warning: Traditional manual inspection methods cannot achieve real-time monitoring, and there is a time lag in the discovery of abnormal situations, leading to potential biosafety hazards. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, device, electronic device and medium for monitoring human behavior, which effectively solves the problems of incomplete regulatory mechanisms, vague implementation standards and delayed risk warnings in the existing disinfection supervision system for livestock pens.

[0005] In a first aspect, embodiments of this application provide a method for monitoring personnel behavior, applicable to a behavior monitoring system. The behavior monitoring system includes an image acquisition module, an edge computing module, and a result display module. The method includes: An image acquisition module is pre-configured. When a target person enters the target enclosure unit, the pre-configured image acquisition module triggers a target event and pushes the target event to the first message queue. From the first message queue, the edge computing module monitors the target event, obtains the target video segment collected in real time by the configured image acquisition module, and extracts multiple key frame images from the target video segment; The pre-trained behavior detection model is controlled to identify the multi-frame keyframe images from multiple recognition dimensions to obtain the corresponding dimensional results, and the dimensional results are integrated to obtain the recognition result; The identification results are pushed to the result display module, so that the result display module can parse the identification results and generate warning information, and terminate the behavior monitoring of the target personnel based on the warning information.

[0006] In conjunction with the first aspect, this application provides a first possible implementation of the first aspect, wherein the control of the pre-trained behavior detection model to identify the multi-frame keyframe images from multiple recognition dimensions to obtain corresponding dimensional results includes: The images of the target person are extracted from multiple keyframe images, and the target images are labeled with the target recognition dimensions based on the bounding boxes. Based on the target image and the corresponding standard image, the dimensional results corresponding to each recognition dimension are obtained.

[0007] In conjunction with the first aspect, this application provides a second possible implementation of the first aspect, wherein the identification dimensions include at least a handwashing dimension and a door handle dimension; The identification based on the target image and the corresponding standard image includes: Pre-configure corresponding standard actions for each recognition dimension, and decompose multiple key standard actions from the standard actions; Calculate the similarity between the key actions in the target image and the corresponding key standard actions, and obtain the dimension result based on the similarity.

[0008] In conjunction with the first aspect, this application provides a third possible implementation of the first aspect, wherein, in addition to calculating the similarity between the key action in the target image and the corresponding key standard action, it includes: Based on the pre-defined key standard actions of the handwashing dimension, multiple frames of video images of the key actions corresponding to the key standard actions are obtained. The duration of key disinfection actions is obtained based on the number of multiple video images, and the dimensional result is obtained by fusing the action duration with similarity.

[0009] In conjunction with the first aspect, this application provides a fourth possible implementation of the first aspect, wherein the identification dimension further includes the mask dimension; The identification based on the target image and the corresponding standard image includes: Extract the position of the mask in the target image and compare the position with a pre-set standard position to obtain the comparison result; Based on the comparison results, the dimension recognition results corresponding to the mask dimension are generated.

[0010] In conjunction with the first aspect, this application provides a fifth possible implementation of the first aspect, wherein pushing the recognition result to the result display module includes: The result display module and the second message queue are pre-configured to the target state so that the configured result display module can monitor the second message queue. The edge computing module is controlled to push the recognition result to the second message queue so that the result display module can obtain the recognition result.

[0011] In conjunction with the first aspect, this application provides a sixth possible implementation of the first aspect, wherein the pre-configured image acquisition module includes: Based on the environmental data of the target enclosure unit, a target configuration dimension is set for the image acquisition module; different configuration dimensions correspond to different configuration methods. The image acquisition module is configured by calling the configuration method corresponding to the target configuration dimension.

[0012] Secondly, embodiments of this application provide a personnel behavior monitoring device, applicable to a behavior monitoring system. The behavior monitoring system includes an image acquisition module, an edge computing module, and a result display module. The device includes: The entry module is used to pre-configure the image acquisition module. When the target person enters the target enclosure unit, the configured image acquisition module triggers a target event and pushes the target event to the first message queue. The monitoring module is used to monitor the target event from the first message queue, obtain the target video segment collected in real time by the configured image acquisition module, and extract multiple key frame images from the target video segment. An integration module is used to control a pre-trained behavior detection model to identify the multi-frame keyframe images from multiple recognition dimensions to obtain corresponding dimensional results, and to integrate the dimensional results to obtain the recognition result; The push module is used to push the recognition result to the result display module, so that the result display module can parse the recognition result and generate a warning message, and terminate the behavior monitoring of the target person based on the warning message.

[0013] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of any one of the human behavior monitoring methods described above.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of any one of the personnel behavior monitoring methods described above.

[0015] This application provides a method for monitoring personnel behavior, applicable to a behavior monitoring system. The behavior monitoring system includes an image acquisition module, an edge computing module, and a result display module. The method first pre-configures the image acquisition module. When a target person enters a target enclosure unit, the configured image acquisition module triggers a target event and pushes the target event to a first message queue. Next, from the first message queue, the edge computing module monitors the target event and acquires a target video segment captured in real-time by the configured image acquisition module, extracting multiple keyframe images from the target video segment. Then, a pre-trained behavior detection model is controlled to identify the multiple keyframe images from multiple recognition dimensions to obtain corresponding dimensional results, and these dimensional results are integrated to obtain a recognition result. Finally, the recognition result is pushed to the result display module, which then parses the recognition result and generates a warning message to terminate the behavior monitoring of the target person based on the warning message. Based on the above methods, this application realizes the detection of disinfection behaviors of personnel in the pen areas where livestock and poultry, such as pigs, live. This avoids the phenomenon of ambiguous implementation standards, improves the deficiencies in the existing disinfection supervision system, and also avoids the phenomenon of delayed risk warning based on the warning information. This ensures the safety and health of the pen areas where livestock and poultry, such as pigs, live and avoids the existence of biosecurity risks. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating a personnel behavior monitoring method provided in an embodiment of this application is shown; Figure 2 This illustration shows another flowchart of a personnel behavior monitoring method provided in an embodiment of this application; Figure 3 This illustration shows an adjustment diagram during the training process of the behavior detection model provided in an embodiment of this application; Figure 4 A schematic diagram illustrating the recognition results provided in an embodiment of this application is shown; Figure 5 A schematic diagram illustrating the key actions in handwashing provided in the embodiments of this application is shown; Figure 6 A schematic diagram illustrating key actions of the door handle dimension provided in an embodiment of this application is shown; Figure 7 This illustration shows a diagram illustrating the identification of mask dimensions according to an embodiment of this application; Figure 8 This diagram illustrates a structural block diagram of a personnel behavior monitoring device provided in an embodiment of this application. Figure 9 A structural block diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0019] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0020] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0021] The current disinfection supervision system suffers from technical defects such as incomplete supervision mechanisms, vague implementation standards, and delayed risk warnings, which lead to biosecurity risks in the areas where pigs and other livestock live, affecting the safety and health of pigs and other livestock.

[0022] Based on this, embodiments of this application provide a method, device, electronic device, and medium for monitoring human behavior, which are described below through embodiments.

[0023] Example 1 To facilitate understanding of this embodiment, a method for monitoring personnel behavior disclosed in this application will first be described in detail. For example... Figure 1 The diagram shown is a flowchart of a method for monitoring personnel behavior. Figure 2 The diagram shows another flowchart of a personnel behavior monitoring method. This application provides a personnel behavior monitoring method applicable to a behavior monitoring system, which includes an image acquisition module, an edge computing module, and a result display module. The method includes: S101. A pre-configured image acquisition module is used. When a target person enters the target enclosure unit, the pre-configured image acquisition module triggers a target event and pushes the target event to the first message queue. S102. From the first message queue, the edge computing module monitors the target event, obtains the target video segment collected in real time by the configured image acquisition module, and extracts multiple key frame images from the target video segment. S103. Control the pre-trained behavior detection model to identify the multi-frame key frame images from multiple recognition dimensions to obtain the corresponding dimension results, and integrate the dimension results to obtain the recognition result; S104. Push the recognition result to the result display module so that the result display module can parse the recognition result and generate a warning message, so as to end the behavior monitoring of the target person based on the warning message.

[0024] The behavior monitoring system described in this application includes an image acquisition module, an edge computing module, and a result display module. The image acquisition module includes a Hikvision camera, or other modules capable of image acquisition. The edge computing module can be a computer, a mainframe, or other intelligent device with data processing and analysis capabilities, capable of processing data from multiple enclosure units within the same enclosure area. The cloud management platform can receive data from multiple edge computing modules to form a three-tier architecture, enabling automated monitoring, intelligent judgment, and digital archiving of the disinfection process, effectively improving the level of biosafety control. The result display module is used to display warning information and the identification results.

[0025] In step S101, when this application is used, the image acquisition module is pre-configured so that it can clearly identify areas with poor lighting and other environmental conditions in the pens where pigs and other livestock live. After configuration, when a target person enters the target pen unit, the entry action must be continuous for more than 10 frames before it is considered as entering the target pen unit. This avoids the phenomenon of people passing by the target pen unit and being mistakenly identified as entering the target pen unit, improving fault tolerance and making the target event pushed to the first message queue more accurate. At this time, the configured image acquisition module generates a "person enters unit" tag. Based on this tag, The configured image acquisition module triggers a target event based on the image of a target person appearing in the acquired video stream. This target person is a staff member in the pen area, who can be an authenticated individual. The target event is an intrusion event, indicating that someone has entered the pen area. The target pen unit is one of the units in an area where a large number of pigs and other livestock live together. The module pushes the target event to a first message queue. The first message queue uses the Kafka message component, but other message queues can also be used. For example, when a target person enters a target pen unit, the Kafka message queue will contain the following messages representing the target event: [ {'datetime': '2025-09-11T15:23:48.000+08:00', 'regionName': 'XX Company Monitoring Center / XX Province / XX City / XX County / XX District-XXXX / XX14 Growth Farm / XX14 Growth Farm Zone 2 / Unit Entrance', 'srcName': 'Piglets Unit 15 Entrance XX14', 'cameraIndexCode': '49c2fd1ad1234450929dd445525c6cd1','img':' / pic?9d90011ff21do-6el*31-f8558bc1fa 27084b4*1719=6=8sp**214==t5*7715474255l3*4737=5o7*14-b=552f-4pi192o=14ea0=344'} . . . . ].

[0026] In a specific implementation of step S101, one embodiment is as follows: the pre-configured image acquisition module includes: S1011. Based on the environmental data of the target enclosure unit, set the target configuration dimension for the image acquisition module; different configuration dimensions correspond to different configuration methods. S1012. Invoke the configuration method corresponding to the target configuration dimension to configure the image acquisition module.

[0027] In steps S1011-S1012, before configuring the image acquisition module, this application pre-acquires environmental data of the target pen unit, including lighting data, pigsty structure data, etc. Based on the environmental data of the target pen unit, a target configuration dimension for the image acquisition module is set. The configuration dimensions pre-set in this application include hardware structure dimension and spatial state dimension. Different configuration dimensions correspond to different configuration methods. For example, the hardware structure dimension is used to install auxiliary equipment such as lighting for the camera, and while installing the auxiliary equipment, the camera's image configuration, such as dynamic balance and exposure value, is adjusted according to the camera's image exposure. The spatial state dimension is used to configure the camera's installation position. For example, if the camera is installed diagonally above the entrance of the target pen unit, within 5m of the unit door, the target configuration dimension is calculated based on the environmental data of the target pen unit, and the configuration method corresponding to the target configuration dimension is called to configure the image acquisition module to obtain a clear video stream.

[0028] In step S102, this application pre-configures the edge computing module to listen to the first message queue. After a message corresponding to the target event appears in the first message queue, the module captures the target event, starts streaming identification, and calls the SDK-python interface of the Hikvision camera (which has been configured with an image acquisition module) to control the camera to capture images in real time. This enables real-time acquisition of target video segments of the target person entering the target enclosure unit. The target video segment is the video segment in which the target person exists and is associated with the target event. Multiple key frame images are extracted from the target video segment. The key frame images are images of video frames in which the target person performs key actions. The extraction of which frames can be determined by analyzing the movement of the target person in the video segment using methods such as optical flow. Optical flow can calculate the pixel motion information between adjacent frames. When the average motion amplitude exceeds a set threshold, the frame is extracted. Other methods can also be used to achieve extraction.

[0029] In step S103, the edge computing module of this application sets up a pre-trained behavior detection model. The behavior detection model pre-sets multiple recognition dimensions, including handwashing, door handles, and masks. The door handle dimension refers to disinfecting door handles, and the mask dimension refers to people wearing masks. The most advanced and mature target recognition model, YOLOv8, is selected as the base model for the multiple recognition dimensions. Video footage of people wearing masks, washing hands, and disinfecting door handles is collected and labeled for model training, testing, and verification. After obtaining the multiple keyframe images, the pre-trained behavior detection model is controlled to recognize the multiple keyframe images from the multiple recognition dimensions to obtain the corresponding dimension results. The dimension results are then integrated to obtain the recognition result. That is, the recognition result is obtained by processing the corresponding dimension results based on the relationship between the multiple recognition dimensions. If the dimension result corresponding to one of the recognition dimensions is unqualified, the recognition result is unqualified. Only when the dimension results corresponding to all three recognition dimensions are qualified, the recognition result is qualified.

[0030] During training, the behavior detection model collects video footage of people entering the enclosure units. Frames are extracted from the videos, and those containing key behaviors are selected and labeled. The model is then trained using a YOLOv8 object detection model. The training process includes: feeding labeled data into the neural network to allow the model to learn the mapping between input and output; calculating the difference between the predicted result and the true label using a loss function; using an optimization algorithm to adjust the network parameters in reverse, continuously narrowing the gap until the model converges; and finally, parameter tuning and iteration are performed based on the training process. Figure 3 As shown, this leads to the trained behavior detection model.

[0031] In a specific implementation of step S103, one embodiment is as follows: the pre-trained behavior detection model is controlled to identify the multi-frame keyframe images from multiple recognition dimensions to obtain the corresponding dimensional results, including: S1031. Extract the images of the target person from multiple keyframe images, and annotate the target images corresponding to the target recognition dimensions based on the bounding boxes; S1032. Based on the target image and the corresponding standard image, perform identification to obtain the dimension result corresponding to each identification dimension.

[0032] In steps S1031-S1032, the behavior detection model extracts the image of the target person from multiple keyframe images. A deep learning-based target detection model, such as YOLOv8, can be used. A pre-trained YOLOv8 model can be loaded using OpenCV's DNN module to perform person detection on the keyframe images. This can quickly and accurately detect people in the image and provide the bounding box coordinates of the person. Based on the bounding box, the target image corresponding to the target recognition dimension is labeled. For example, in the mask dimension, the target image is the face of the target person, while for the handwashing and door handle dimensions, the target image is the hand of the target person. Recognition is performed based on the target image and the corresponding standard image. Each recognition dimension has a corresponding standard image, resulting in a dimension result for each recognition dimension. The dimension result includes "qualified" and "unqualified," with the "unqualified" result also including the reason for the unqualified result, such as... Figure 4 As shown, this is to create interpretability for the dimensional results.

[0033] In a specific implementation of step S1032, one embodiment is as follows: the identification dimensions include at least the handwashing dimension and the door handle dimension; The identification based on the target image and the corresponding standard image includes: A1. Pre-configure corresponding standard actions for each recognition dimension, and decompose multiple key standard actions from the standard actions; A2. Calculate the similarity between the key actions in the target image and the corresponding key standard actions, and obtain the dimension result based on the similarity.

[0034] In steps A1-A2, for the handwashing dimension and the door handle dimension, corresponding standard actions are pre-configured for each identification dimension. For example, the standard action set for the handwashing dimension is that the target person rubs their hands together inside the handwashing basin, and the standard action set for the door handle dimension is that the target person picks up the alcohol spray bottle to disinfect the door handle. Several key standard actions are extracted from these standard actions. Therefore, in the handwashing dimension, the key standard actions include rubbing both hands together, rubbing both hands together inside the handwashing basin, and rubbing hands together. Figure 5As shown; then, at the door handle level, key standard actions include the target person picking up the alcohol spray bottle, the alcohol spray bottle being moved, etc., such as... Figure 6 As shown, the similarity between key actions in the target image and corresponding key standard actions is calculated. The key standard actions are pre-set and stored in the behavior detection model in the form of action tags. When calculation is required, the key actions are matched against the database of key standard actions. If a match is found, the similarity is calculated. If no match is found, the next key action is called for matching. The dimension result is obtained based on the similarity. Thresholds are set for the two dimensions of handwashing and door handle disinfection. If the similarity is lower than that of the corresponding dimension, the result of the corresponding dimension is unqualified; otherwise, it is qualified. In a specific implementation of step A2, one embodiment includes: in addition to calculating the similarity between the key actions in the target image and the corresponding key standard actions, the following steps are also included: A21. Based on the pre-set key standard actions of the handwashing dimension, obtain multiple frames of video images of the key actions corresponding to the key standard actions. A22. The duration of key disinfection actions is obtained based on the number of multiple video images, and the dimensional result is obtained by fusing the duration of the actions with the similarity.

[0035] In steps A21-A22, this application, regarding the handwashing dimension, based on pre-defined key standard actions for the handwashing dimension, such as washing both hands, both hands in a handwashing box, and rubbing hands together, uses a behavior detection model to acquire multiple video images of the key actions corresponding to the key standard actions. Specifically, the behavior detection model matches each key action frame with the preset key standard actions, filtering out video frames belonging to the key actions. The action duration is directly calculated from the number of frames corresponding to the key action and the video frame rate. For example, if the original video frame rate (FPS) of the image acquisition module is determined, the action duration is calculated based on the number and the original video frame rate using the formula: Action Duration (seconds) = Number of Key Action Frames ÷ Video Frame Rate. This application also sets a duration threshold for the action duration. If the action duration is higher than the preset threshold, such as 10 seconds, and the similarity at this time is higher than the corresponding threshold, then the dimension result corresponding to the handwashing dimension is qualified. If the action duration is lower than the preset threshold or the similarity at this time is lower than the corresponding threshold, then the dimension result corresponding to the handwashing dimension is unqualified.

[0036] In a specific implementation of step S1032, one embodiment is as follows: the identification dimension also includes the mask dimension; The identification based on the target image and the corresponding standard image includes: B1. Extract the position of the mask in the target image and compare the position with a pre-set standard position to obtain the comparison result; B2. Based on the comparison results, generate the dimension recognition results corresponding to the dimensions of the mask.

[0037] In steps B1-B2, this application extracts the position of the mask in the target image from the "mask" category in the behavior detection model of the YoloV8 model, along with the mask dimension. It outputs the bounding box coordinates (mx1, my1, mx2, my2) of the mask in the image and also obtains the bounding box coordinates (x1, y1, x2, y2) of the face in the output target image, i.e., the pixel positions of the upper left and lower right corners of the face. To eliminate the influence of different image sizes, the mask coordinates are converted into proportional positions relative to the face. For example, the proportion of the distance between the upper edge of the mask and the forehead to the height of the face, and the proportion of the distance between the left and right edges of the mask and the cheeks to the width of the face are calculated. These positions are compared with pre-set standard positions to obtain the comparison results. The standard positions are defined according to the standard range defined by epidemic prevention regulations. Vertically, the upper edge of the mask must cover the bridge of the nose, and the lower edge must cover the chin, i.e., the vertical coverage of the mask accounts for 60%-80% of the vertical height of the face. Horizontally, the left and right edges of the mask must fit snugly against the cheeks without obvious gaps, i.e., the horizontal coverage of the mask accounts for 90% of the horizontal width of the face. The above compares the extracted actual mask position with the standard position, and calculates the deviation. For example, if the actual vertical coverage ratio is 40%, and the deviation from the standard lower limit of 60% is 20%, the comparison result includes two categories: qualified and unqualified. Figure 7 As shown, if there are facial obstructions such as glasses or hair, a model that can identify some mask features (such as ear loops and edge contours) can be selected first to reduce obstruction interference. Based on the comparison results, the dimension recognition results corresponding to the mask dimension are generated.

[0038] In step S104, when the target person cannot be captured from the acquisition range of the image acquisition module, the edge computing module pushes the recognition result to the result display module so that the result display module can parse the recognition result. The recognition result is organized into target enclosure unit data and a warning message is generated. Based on the warning message, the monitoring of the target person's behavior ends. The warning message is pushed to designated managers or management groups for risk warning. Based on the target enclosure unit data, a target enclosure unit behavior record report is generated. The unit behavior record report includes the following: personnel Entry details: Region - Area - Site - Equipment Alias ​​(Monitoring Name) - Camera ID, Batch Type - Unit Number - Personnel Entry Time, Entry Behavior Compliance Status (All Behaviors Combined), Single Behavior (Handwashing / Door Handle Spraying / Mask Wearing Recognition…) Recognition Result, Unit Manager; Personnel Entry Non-Compliance Data Statistics Bar Chart: Total Number of Entryes, Number of Non-Compliances, Non-Compliance Percentage; (Filtered by Region, Area, Batch Type, Unit Number, Combined Behavior, Single Behavior); Non-Compliance Business Push, Non-Compliance Correction; Daily Report Management, thereby solving the technical bottleneck of intelligent biosafety prevention and control for the smallest unit - personnel entering the enclosure. The target enclosure unit data includes the following basic data: unit entrance monitoring and maintenance, camera information monitoring (region, area, field, batch type, unit number, monitoring online status, camera ID, platform address, offline time, offline duration); monitoring online rate analysis bar chart, coverage bar chart; offline monitoring push, monitoring screen anomaly push.

[0039] Verified through field testing on four farms with a scale of over 100,000 pigs each, the behavioral monitoring system of this application achieved an accuracy rate of 98.7%, reduced the rate of unqualified personnel entering the pigpens from 80% to less than 2%, and increased the disinfection compliance rate to 98.2%, significantly outperforming traditional monitoring systems. The technical solution of this application is not only applicable to pig farming but can also be extended to biosafety-sensitive areas such as poultry farming, slaughtering and processing, and food production, providing key technical support for building a digital epidemic prevention system in modern agriculture.

[0040] In a specific implementation of step S104, one embodiment is as follows: pushing the recognition result to the result display module includes: S1041. Pre-configure the result display module and the second message queue to the target state so that the configured result display module can monitor the second message queue; S1042. Control the edge computing module to push the recognition result to the second message queue so that the result display module can obtain the recognition result.

[0041] In steps S1041-S1042, this application pre-configures the result display module and the second message queue to a target state, i.e., the target state is the state where the result display module monitors the second message queue. The second message queue can also be a Kafk message queue, which is not the same message queue as the first message queue. Connection parameters are configured, including the queue's IP address, port number, username, password, and the name of the target queue for push notifications, such as `mask_detection_result_queue`. Message format specifications are set, for example, uniformly using JSON format. Fields of the identification results, such as `image_id`, `detection_time`, `mask_result`, and `deviation_detail`, are clearly defined to ensure that the result display module can parse the results and monitor the second message queue. After the edge computing module obtains the identification result, it pushes the identification result to the second message queue. If the push fails, retry logic is triggered, such as retrying 3 times with an interval of 2 seconds between each retry. To prevent data loss, the result display module obtains the recognition result, parses and extracts the recognition result, and verifies the parsed data. For example, it checks whether "image_id" is unique and whether "mask_result" is a preset valid value (e.g., qualified or unqualified). If the verification fails, the message is marked as abnormal and temporarily stored to avoid affecting the display function and for front-end display.

[0042] Example 2 This application also provides a personnel behavior monitoring device, such as Figure 8 The diagram shows a block diagram of a personnel behavior monitoring device. The functions implemented by this device correspond to the steps of executing a personnel behavior monitoring method on a terminal device as described above. This device can be understood as a server component including a processor. The personnel behavior monitoring device described in this application is applicable to a behavior monitoring system, which includes an image acquisition module, an edge computing module, and a result display module. The device includes: The module 801 is used to pre-configure the image acquisition module. When the target person enters the target enclosure unit, the configured image acquisition module triggers a target event and pushes the target event to the first message queue. The monitoring module 802 is used to monitor the target event from the first message queue, obtain the target video segment collected in real time by the configured image acquisition module, and extract multiple key frame images from the target video segment. The integration module 803 is used to control the pre-trained behavior detection model to identify the multi-frame key frame images from multiple recognition dimensions to obtain the corresponding dimension results, and to integrate the dimension results to obtain the recognition result; The push module 804 is used to push the recognition result to the result display module, so that the result display module can parse the recognition result and generate a warning message, so as to end the behavior monitoring of the target person based on the warning message.

[0043] In one feasible implementation, the integration module includes: The annotation module is used to extract the images of the target person from multiple keyframe images and annotate the target images corresponding to the target recognition dimensions based on the bounding boxes. The recognition module is used to recognize the target image and the corresponding standard image to obtain the dimension result corresponding to each recognition dimension.

[0044] In one feasible implementation, the integration module further includes: The decomposition module is used to pre-configure corresponding standard actions for each recognition dimension and decompose multiple key standard actions from the standard actions; The calculation module is used to calculate the similarity between the key actions in the target image and the corresponding key standard actions, and to obtain the dimension result based on the similarity.

[0045] In one feasible implementation, the integration module also includes: The setting module is used to obtain multiple frames of video images of the key actions corresponding to the key standard actions based on the pre-set key standard actions of the handwashing dimension. The fusion module is used to obtain the action duration of key disinfection actions based on the number of multiple video images, and to fuse the action duration with similarity to obtain the dimensional result.

[0046] In one feasible implementation, the integration module further includes: The extraction module is used to extract the position of the mask in the target image and compare the position with a pre-set standard position to obtain a comparison result; The generation module is used to generate the dimension recognition result corresponding to the dimension of the mask based on the comparison result.

[0047] In one feasible implementation, the push module includes: The monitoring module is used to pre-configure the result display module and the second message queue to the target state, so that the configured result display module can monitor the second message queue; The acquisition module is used to control the edge computing module to push the recognition result to the second message queue so that the result display module can acquire the recognition result.

[0048] In one feasible implementation, the entry module includes: The configuration module is used to set target configuration dimensions for the image acquisition module based on the environmental data of the target enclosure unit; different configuration dimensions correspond to different configuration methods. The configuration module is used to call the configuration method corresponding to the target configuration dimension to configure the image acquisition module.

[0049] Example 3 This application also provides an electronic device, such as Figure 9 As shown, it includes: a processor 901, a memory 902, and a bus 903. The memory 902 stores machine-readable instructions that can be executed by the processor 901. When the electronic device is running, the processor 901 and the memory 902 communicate through the bus 903. When the machine-readable instructions are executed by the processor 901, the steps of any one of the personnel behavior monitoring methods described above are executed.

[0050] Example 4 This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of any one of the human behavior monitoring methods described herein.

[0051] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.

[0052] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0054] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a platform server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0055] The above are merely specific embodiments 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 method for monitoring personnel behavior, characterized in that, Applicable to a behavior monitoring system, the behavior monitoring system including an image acquisition module, an edge computing module, and a result display module, the method includes: An image acquisition module is pre-configured. When a target person enters the target enclosure unit, the pre-configured image acquisition module triggers a target event and pushes the target event to the first message queue. From the first message queue, the edge computing module monitors the target event, obtains the target video segment collected in real time by the configured image acquisition module, and extracts multiple key frame images from the target video segment; The pre-trained behavior detection model is controlled to identify the multi-frame keyframe images from multiple recognition dimensions to obtain the corresponding dimensional results, and the dimensional results are integrated to obtain the recognition result; The identification results are pushed to the result display module, so that the result display module can parse the identification results and generate warning information, and terminate the behavior monitoring of the target personnel based on the warning information.

2. The method according to claim 1, characterized in that, The pre-trained behavior detection model is controlled to identify the multi-frame keyframe images from multiple recognition dimensions to obtain corresponding dimensional results, including: The images of the target person are extracted from multiple keyframe images, and the target images are labeled with the target recognition dimensions based on the bounding boxes. Based on the target image and the corresponding standard image, the dimensional results corresponding to each recognition dimension are obtained.

3. The method according to claim 2, characterized in that, The identification dimensions include at least the handwashing dimension and the door handle dimension; The identification based on the target image and the corresponding standard image includes: Standard actions are pre-configured for each recognition dimension, and multiple key standard actions are decomposed from the standard actions; Calculate the similarity between the key actions in the target image and the corresponding key standard actions, and obtain the dimension result based on the similarity.

4. The method according to claim 3, characterized in that, In addition to calculating the similarity between the key actions in the target image and the corresponding key standard actions, the method includes: Based on the pre-defined key standard actions of the handwashing dimension, multiple frames of video images of the key actions corresponding to the key standard actions are obtained. The duration of key disinfection actions is obtained based on the number of multiple video images, and the dimensional result is obtained by fusing the action duration with similarity.

5. The method according to claim 2, characterized in that, The identification dimension also includes the mask dimension; The identification based on the target image and the corresponding standard image includes: Extract the position of the mask in the target image and compare the position with a pre-set standard position to obtain the comparison result; Based on the comparison results, the dimension recognition results corresponding to the mask dimension are generated.

6. The method according to claim 1, characterized in that, The step of pushing the recognition result to the result display module includes: The result display module and the second message queue are pre-configured to the target state so that the configured result display module can monitor the second message queue. The edge computing module is controlled to push the recognition result to the second message queue so that the result display module can obtain the recognition result.

7. The method according to claim 1, characterized in that, The pre-configured image acquisition module includes: Based on the environmental data of the target enclosure unit, target configuration dimensions are set for the image acquisition module; different configuration dimensions correspond to different configuration methods. The image acquisition module is configured by calling the configuration method corresponding to the target configuration dimension.

8. A personnel behavior monitoring device, characterized in that, Suitable for behavior monitoring systems, the behavior monitoring system includes an image acquisition module, an edge computing module, and a result display module, the device includes: The entry module is used to pre-configure the image acquisition module. When the target person enters the target enclosure unit, the configured image acquisition module triggers a target event and pushes the target event to the first message queue. The monitoring module is used to monitor the target event from the first message queue, obtain the target video segment collected in real time by the configured image acquisition module, and extract multiple key frame images from the target video segment. An integration module is used to control a pre-trained behavior detection model to identify the multi-frame keyframe images from multiple recognition dimensions to obtain corresponding dimensional results, and to integrate the dimensional results to obtain the recognition result; The push module is used to push the recognition result to the result display module, so that the result display module can parse the recognition result and generate a warning message, and terminate the behavior monitoring of the target person based on the warning message.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of a human behavior monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of a personnel behavior monitoring method as described in any one of claims 1 to 7.

Images