A system and method for standardized detection and autonomous alerting of personnel testing

Abstract

The present disclosure relates to a personnel experiment standardized detection and autonomous alarm system and method, belonging to the field of intelligent laboratory management. The system comprises: an acquisition module for acquiring laboratory video stream information; a visual perception module for identifying irregular behavior in the laboratory video stream information using an improved YOLOv11 model, wherein the model is a YOLOv11 model embedded with a global grouping coordinate attention module in a backbone network; a UWB positioning module for analyzing the position coordinates of personnel in the laboratory and outputting personnel position coordinates with ID information and time stamps; a data fusion and decision module for associating the irregular behavior data and the personnel position coordinates, binding the irregular behavior and the irregular operation personnel through a multi-modal data fusion algorithm to obtain a binding result; and an alarm module for performing an alarm operation according to the binding result. The present disclosure can automatically and accurately identify irregularities and specific irregular personnel and automatically alarm.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent laboratory management technology, and in particular to a standardized detection and autonomous alarm system and method for personnel testing. Background Technology

[0002] The detection and alarm of the operational standardization of personnel at the test site are mainly carried out by manual inspection and video playback spot checks, which has the following defects: the proportion of human factors is large, and the test results are easily affected by the personal experience and work status of the inspectors; the real-time performance of the detection is poor, and problems cannot be detected in time, and the alarm messages issued are prone to delays; the detection blind area is large, and it is not possible to achieve all-weather, all-round detection.

[0003] Currently, mature general-purpose target detection models (such as the YOLO series algorithms) can identify and issue alarms for personnel violations. However, existing models lack sufficient accuracy in identifying complex violations, especially in complex and dynamic laboratory environments where they struggle to extract discriminative detailed features. Therefore, in scenarios such as "not wearing a safety belt while working at height" or "not wearing a safety helmet during experiments," the system is prone to missed or false detections. Secondly, regarding positioning, most systems currently use Wi-Fi or Bluetooth positioning based on signal strength. This results in low positioning accuracy and is easily affected by the environment. They cannot obtain the centimeter-level positioning data needed to determine behaviors such as "personnel entering restricted areas" or "distance between personnel and hazardous equipment." Therefore, they remain at the stage of post-event recording and delayed alarms, unable to achieve autonomous alarm functionality.

[0004] Therefore, it is necessary to design a system and method that can automatically and accurately identify violations and specific individuals involved in the violations. Summary of the Invention

[0005] Based on this, it is necessary to address the aforementioned issues by providing a standardized detection and autonomous alarm system and method for personnel testing.

[0006] To solve the above problems, the present disclosure adopts the following technical solution:

[0007] Firstly, this disclosure provides a standardized detection and autonomous alarm system for personnel testing, including:

[0008] The acquisition module is used to acquire laboratory video stream information;

[0009] The visual perception module is used to identify violations in laboratory video stream information using an improved YOLOv11 model and output violation data with timestamps. The violation data includes images or videos of violations. The input of the improved YOLOv11 model is laboratory video stream information. The improved YOLOv11 model is a YOLOv11 model with a global grouped coordinate attention module embedded in the backbone network.

[0010] The UWB positioning module is used to analyze the location coordinates of personnel in the laboratory and output the location coordinates of personnel with ID information and timestamps.

[0011] The data fusion and decision-making module is used to associate the violation data with the personnel location coordinates, and to bind the violation with the violator through a multimodal data fusion algorithm to obtain the binding result;

[0012] The alarm module is used to perform alarm operations based on the binding result.

[0013] In a preferred embodiment, the global grouped coordinate attention module includes:

[0014] The grouping unit is used to group the input feature map along the channel dimension to obtain the grouped feature map;

[0015] The first pooling unit performs a global average pooling operation along the height direction on each grouped feature map to obtain... ;

[0016] The second pooling unit performs max pooling on each grouped feature map along the height direction to obtain... ;

[0017] The third pooling unit performs global average pooling on each grouped feature map along the width direction to obtain... ;

[0018] The fourth pooling unit is used to perform max pooling on each grouped feature map along the width direction to obtain... ;

[0019] Convolutional units are used for... , , and Perform convolution processing;

[0020] The weight generation unit is used to... The results of convolution processing and The convolution results are summed to obtain the first result, which is used to... The results of convolution processing and The convolution results are added together to obtain the second result. The first and second results are then processed by the Sigmoid function to generate spatial attention weights.

[0021] The weighting module is used to weight the input feature map based on spatial attention weights to obtain the output feature map.

[0022] In a preferred embodiment, the data fusion and decision-making module is specifically used for:

[0023] Convert the pixel coordinates of the image or video corresponding to the violation to world coordinates;

[0024] Establish the mapping relationship between the UWB positioning coordinate system and the world coordinate system.

[0025] Convert the world coordinates to UWB coordinates;

[0026] The violation data output by the visual perception module and the personnel location coordinates are matched in time, and the matching condition is that the time difference is less than a preset time threshold.

[0027] Calculate the Euclidean distance between the UWB coordinates corresponding to the violation data and the location coordinates of the personnel matching the time.

[0028] Based on the number of people whose Euclidean distance is less than the spatial distance threshold;

[0029] Determine the category of the violation;

[0030] The individuals responsible for the violations will be identified based on the number of personnel and the type of violation.

[0031] Link violations to the individuals who violated the rules.

[0032] In a preferred embodiment, the data fusion and decision module is further specifically used to obtain a list of equipment operators, and the step of determining the violator based on the number of personnel and the type of violation specifically involves determining the violator based on the number of personnel, the type of violation, the list of equipment operators, and Euclidean distance.

[0033] In a preferred embodiment, determining the violator based on the number of personnel and the type of violation specifically involves: when the number of personnel is 1, identifying the personnel whose Euclidean distance is less than the spatial distance threshold as violators; when the number of personnel is greater than 1 and the type of violation falls within a first category, selecting the personnel with the smallest Euclidean distance as violators; when the number of personnel is greater than 1 and the type of violation falls within a second category, retrieving the list of equipment operators for the relevant equipment during the time period of the violation, filtering out the total number of personnel whose Euclidean distance is less than the spatial distance threshold and who are located in the list of equipment operators; when the total number of personnel is 1, identifying the personnel whose Euclidean distance is less than the spatial distance threshold and who are located in the list of equipment operators as violators; and when the total number of personnel is greater than 1, identifying the personnel with the smallest Euclidean distance and who are located in the list of equipment operators as violators.

[0034] In a preferred embodiment, the UWB positioning module is used to analyze the location coordinates of personnel in the laboratory based on the transmission time of radio waves from the tag to the base station. The personnel in the laboratory wear the tag, which is capable of emitting radio waves and has a unique ID.

[0035] In a preferred embodiment, the alarm module is specifically used to generate an alarm command based on the output of the data fusion and decision module.

[0036] In a preferred embodiment, the alarm module is specifically used to issue an alarm using an audible and visual alarm and to send alarm information to the terminal of the safety management personnel.

[0037] In a preferred embodiment, the alarm module is further used to record videos or pictures of violations, and to record the ID of the person who violated the rules, the location coordinates of the person who violated the rules, and the time of the violation.

[0038] Secondly, this disclosure provides a standardized detection and autonomous alarm method for personnel testing, including:

[0039] Obtain laboratory video stream information;

[0040] The improved YOLOv11 model is used to identify violations in laboratory video stream information to obtain violation data, which includes images or videos of violations. The input of the improved YOLOv11 model is the laboratory video stream information, and the improved YOLOv11 model is a YOLOv11 model with a global grouped coordinate attention module embedded in the backbone network.

[0041] The location coordinates of personnel in the laboratory are analyzed to obtain personnel location coordinates with ID information and timestamps, and the personnel location coordinates are coordinates in the UWB coordinate system;

[0042] The violation data and the personnel location coordinates are associated, and the violation is bound to the violator through a multimodal data fusion algorithm to obtain the binding result;

[0043] An alarm operation will be performed based on the binding result.

[0044] The aforementioned standardized detection and autonomous alarm system and method for personnel testing utilizes an improved YOLOv11 model with a global grouping attention mechanism for violation detection and identification. This significantly enhances the ability to extract features of violations from small targets and complex backgrounds, improving the accuracy of automatic violation identification in real-world laboratory scenarios. By employing a data fusion and decision-making module, it analyzes the results of visual perception and UWB fusion to identify the violating personnel, effectively overcoming the bottleneck of pure vision systems' inability to accurately distinguish responsible parties in occluded, backlit, or multi-person gathering scenarios. This achieves precise and automatic binding of violations with the violating personnel, and automatic alarm operation is implemented based on the accurate binding results. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the system structure in one embodiment of the present disclosure;

[0046] Figure 2 This is a structural diagram of the GGCA module in one embodiment of the present disclosure;

[0047] Figure 3 This is a flowchart illustrating a method in one embodiment of the present disclosure. Detailed Implementation

[0048] The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0049] A standardized detection and autonomous alarm system for personnel testing includes: an acquisition module, a visual perception module, a UWB positioning module, a data fusion and decision-making module, and an alarm module;

[0050] The acquisition module is used to acquire laboratory video stream information;

[0051] The visual perception module is used to identify violations in laboratory video stream information using an improved YOLOv11 model; the visual perception module outputs violation data with timestamps, the violation data including images or videos of violations, the input of the improved YOLOv11 model is laboratory video stream information, and the improved YOLOv11 model is a YOLOv11 model with a global grouped coordinate attention module embedded in the backbone network.

[0052] The UWB positioning module is used to analyze the location coordinates of personnel in the laboratory and output the location coordinates of personnel with ID information and timestamps.

[0053] The data fusion and decision-making module is used to associate the violation data with the personnel location coordinates, and to bind the violation with the violator through a multimodal data fusion algorithm to obtain the binding result;

[0054] The alarm module is used to perform alarm operations based on the binding result.

[0055] Understandably, the violation data includes not only time information (time stamps) but also images or videos of the violation.

[0056] Understandably, the personnel position coordinates output by the UWB positioning module are coordinates in the UWB coordinate system.

[0057] Understandably, linking violations to the person committing the violation is equivalent to linking the violation event to the ID information of the person committing the violation.

[0058] It is understood that the alarm operation can be issuing an alarm command, that is, it is not limited whether the alarm module includes a device for displaying the alarm (such as a screen or light) and an alarm device that emits sound.

[0059] Understandably, the system automatically performs standardized testing and autonomous alarms for personnel testing.

[0060] The system will be further described in the following embodiments.

[0061] The acquisition module is used to acquire laboratory video stream information; the laboratory video stream information is based on data captured by cameras. The cameras are located in the laboratory.

[0062] The visual perception module, based on laboratory video stream information captured by a camera, uses an improved YOLOv11 model to achieve real-time detection and identification of violations. The Global Grouped Coordinate Attention (GGCA) mechanism is embedded into the YOLOv11 model backbone, enabling the model to effectively extract features of small targets and complex background features from the image.

[0063] In one embodiment, the visual perception module includes a plurality of cameras (i.e., cameras).

[0064] The structure diagram of the GGCA module is as follows: Figure 2As shown, the main idea is to calculate attention weights based on the global information of the feature map in the height and width dimensions, and then perform weighted processing on the original features to enhance the feature representation ability.

[0065] The GGCA module includes a grouping unit, a first pooling unit, a second pooling unit, a third pooling unit, a fourth pooling unit, a convolution unit, a weight generation unit, and a weighting module; the grouping unit is used to process the input feature map. Grouping along the channel dimension yields grouped feature maps; the first pooling unit performs global average pooling on each grouped feature map along the height direction. The second pooling unit is used to perform max pooling on each grouped feature map along the height direction to obtain... The third pooling unit is used to perform global average pooling along the width direction on each grouped feature map to obtain... The fourth pooling unit is used to perform max pooling on each grouped feature map along the width direction to obtain... Convolutional units are used for... , , and Perform convolution processing; the weight generation unit is used to... The results of convolution processing and The convolution results are summed to obtain the first result, which is used to... The results of convolution processing and The convolutional results are added together to obtain the second result. The first and second results are then processed by the Sigmoid function to generate spatial attention weights. The weighting module is used to weight the input feature map based on the spatial attention weights to obtain the output feature map.

[0066] Specifically, the GGCA module first processes the input feature map Grouped along the channel dimension, into Group, For integers greater than or equal to 2, form the grouped feature maps. Input feature map The scale is , Indicates altitude, Indicates width, Represents the number of channels, and the feature map after grouping. The scale is Then, global average pooling and max pooling operations are performed along the height and width directions within each group to obtain global contextual features (pooled feature maps). The dimensions of the results of the global average pooling and max pooling operations along the height direction are... The dimension of the result of global average pooling and max pooling operations along the width direction is... The pooled feature maps are then fed into the same convolutional unit to obtain the global average pooling and max pooling results in the height direction, respectively. And the results of global average pooling and max pooling in the width direction. This convolutional unit consists of two 1×1 convolutional layers and a batch normalization layer (…). Figure 2 (normalization layer) and ReLU activation function ( Figure 2 The input feature map is composed of a ReLU activation function, followed by a 1×1 convolutional layer, a batch normalization layer, another ReLU activation function, and yet another 1×1 convolutional layer. This allows for dimensionality reduction (first 1×1 convolutional layer) and restoration (second 1×1 convolutional layer) of the feature map. During the convolutional unit processing, only the channel dimension is changed, not the height and width. This effectively combines inter-channel feature information while compressing computational and parameter counts and preserving key contextual features. Then, the outputs are summed, and spatial attention weights are generated using the sigmoid function. Finally, these weights are used to weight the input feature map to obtain the output feature map. The formula is shown below.

[0067] (1)

[0068] in, The final output feature map has a dimension of . , Indicates the input feature map, For the Sigmoid function, This represents the result of global average pooling in the height direction. This represents the result of max pooling in the height direction. This represents the result of global average pooling in the width direction. This represents the result of max pooling in the width direction. This indicates that the spatial attention weights generated by processing the first result using the Sigmoid function are... This indicates that the second result is processed by the Sigmoid function to generate spatial attention weights.

[0069] The GGCA module employs a channel grouping strategy, which reduces computational complexity while capturing multi-scale contextual information by fusing different pooling operations, thereby effectively improving the model's ability to extract key features in complex scenarios.

[0070] UWB positioning module, where UWB stands for Ultra Wide Band.

[0071] The UWB positioning module is specifically used to analyze the location coordinates of personnel in the laboratory based on the transmission time of radio waves from the tag to the base station. Personnel in the laboratory wear the tag, which can emit radio waves and has a unique ID, which serves as the personnel's ID information.

[0072] The UWB positioning module is a key component of the entire system. By deploying fixed base stations in the laboratory and having personnel wear portable tags, the transmission time ("time of flight") of radio waves from the tag to the base station is obtained using nanosecond-level narrow pulse signals, thus providing the centimeter-level position coordinates of the tag (personnel). Compared to the limitations of vision systems in low-light or occluded environments, this method ensures more accurate spatial location information, and the unique tag ID is a crucial identifier for distinguishing different individuals from those violating regulations.

[0073] In one specific embodiment, the UWB positioning module includes a positioning coordinate determination unit, a base station located in the laboratory, and a tag worn by each person in the laboratory. The tag can transmit radio waves to the base station, which receives the emitted radio waves. The positioning coordinate determination unit determines the tag's location coordinates based on the time it takes for the radio waves to travel from the tag to the base station, thus determining the personnel's location coordinates. Each tag has a unique ID. In one embodiment, the positioning coordinate determination unit may be located within the base station.

[0074] The visual perception module outputs violations, such as "at a certain time t, at the image pixel coordinates..." "Violation detected," while the UWB positioning module outputs "At a certain time t, the tag ID is..." Personnel in laboratory space coordinates The data fusion and decision-making module is used to obtain the output results of the visual perception module and the UWB positioning module, and to associate the data output by the visual perception module and the UWB positioning module. Through multimodal data fusion algorithms (including spatiotemporal correlation), it binds the violation behavior with the violator.

[0075] In actual use, the alarm decision engine usually determines whether to trigger an alarm.

[0076] The data fusion and decision-making module is specifically used for:

[0077] Used to convert the pixel coordinates of images or videos corresponding to violations into world coordinates;

[0078] Used to establish the mapping relationship between the UWB positioning coordinate system and the world coordinate system.

[0079] Used to convert the world coordinates to UWB coordinates;

[0080] This is used to perform time matching between the violation data output by the visual perception module and the personnel location coordinates, with the matching condition being that the time difference is less than a preset time threshold.

[0081] The Euclidean distance used to calculate the UWB coordinates corresponding to the violation data and the personnel location coordinates that match the time.

[0082] Used to determine the number of people whose Euclidean distance is less than a spatial distance threshold;

[0083] Used to determine the category of violation;

[0084] Used to identify those who violate regulations based on the number of personnel and the type of violation;

[0085] This is used to link violations with the individuals who violated the rules.

[0086] Furthermore, the data fusion and decision-making module is also specifically used to obtain a list of equipment operators. The step of determining the violator based on the number of personnel and the type of violation is specifically: determining the violator based on the number of personnel, the type of violation, the list of equipment operators, and Euclidean distance.

[0087] Specifically: When the number of personnel is 1, the personnel whose Euclidean distance is less than the spatial distance threshold are identified as violators; when the number of personnel is greater than 1 and the type of violation falls within the first type range, the personnel with the smallest Euclidean distance are identified as violators; when the number of personnel is greater than 1 and the type of violation falls within the second type range, the personnel retrieve the list of equipment operators related to the violation during the time period of the violation, filter out the total number of personnel whose Euclidean distance is less than the spatial distance threshold and who are located in the list of equipment operators, when the total number of personnel is 1, identify the personnel whose Euclidean distance is less than the spatial distance threshold and who are located in the list of equipment operators as violators; and when the total number of personnel is greater than 1, identify the personnel with the smallest Euclidean distance and who are located in the list of equipment operators as violators.

[0088] Understandably, the aforementioned behaviors of the data fusion and decision-making module are not performed manually, but are implemented automatically.

[0089] Preferably, when the total number of personnel is 0 and the number of personnel is 0, the person committing the violation is manually identified. This may involve prompting a binding failure and then proceeding with the manual assessment. At least one of the data fusion and decision module, the acquisition module, or the alarm module is further used to acquire the result of the manually identified person committing the violation or to directly acquire the binding result between the manually performed violation and the person committing the violation.

[0090] As an example and not a limitation, the multimodal data fusion algorithm process consists of the following steps:

[0091] Step 1: The data fusion and decision-making module performs coordinate system unification and alignment.

[0092] Camera calibration techniques are used to determine the camera's internal parameters (focal length, principal point, etc.) and external parameters (position and orientation relative to the world coordinate system), and the pixel coordinates in the image / time-frequency data are converted to world coordinates, as shown in the following formula:

[0093] (2)

[0094] in, The pixel coordinates of the point The x-coordinate of the point in pixels. The pixel ordinate of the point Let the world coordinates of the point be... These are the X-axis, Y-axis, and Z-axis coordinates of the point's world coordinates, respectively. Scale factor: Represents the Z-axis coordinate of the pixel in the camera coordinate system. Let be the projection matrix. The formula is shown below.

[0095] (3)

[0096] in, For the camera intrinsic parameter matrix, For the camera extrinsic matrix: For rotation matrix, It is a translation vector.

[0097] At the same time, a mapping relationship between the UWB positioning coordinate system and the world coordinate system is established, and the world coordinates are converted into UWB coordinates, as shown in the following formula.

[0098] (4)

[0099] in, World coordinates; Locate the coordinates for UWB; It is a rotation matrix, specifically a 3×3 orthogonal rotation matrix.

[0100] Step 2: The data fusion and decision-making module performs spatiotemporal correlation matching.

[0101] Time synchronization: Both the visual perception module and the UWB positioning module add timestamps to their output data. The timestamp representing the violation data is used This represents the timestamp indicating the location coordinates of personnel in the laboratory, output by the UWB positioning module. When performing correlation, only the time difference is considered. Less than the preset time threshold The data is matched to ensure time consistency.

[0102] Spatial association: For a detected violation, its pixel coordinates projected onto UWB coordinates are... The data fusion and decision-making module searches for the same time point (time difference less than) in the data output by the UWB positioning module. The location of personnel in the laboratory near the UWB coordinate system coordinate point of the violation. Calculate the UWB coordinates of the violation. UWB coordinates of each person who meets the time requirement Euclidean distance between As shown in the following formula.

[0103] (5)

[0104] in, This represents the X-axis coordinate of personnel in the laboratory within the UWB positioning coordinate system. This represents the Y-axis coordinate of personnel in the laboratory within the UWB positioning coordinate system. This represents the Z-axis coordinate of personnel in the laboratory within the UWB positioning coordinate system; This indicates the X-axis coordinate of the violation in the UWB positioning coordinate system. This indicates the Y-axis coordinate of the violation in the UWB positioning coordinate system. This indicates the Z-axis coordinate of the violation in the UWB positioning coordinate system.

[0105] Step 3: The data fusion and decision-making module performs collaborative verification and identity binding.

[0106] Set a spatial distance threshold ,satisfy The number of personnel is .like If the binding fails (the system will display a binding failure message), instructions and images / video clips of the violation will be pushed to the administrator (via the alarm module) for manual intervention to determine the person responsible for the violation. The system directly associates the violation with the individual (tag ID), completing the identity verification. If If multiple individuals simultaneously meet the distance requirement (i.e., multiple people gathered), the conflict resolution mechanism will be activated based on the type of violation: When the violation is "personnel entering a prohibited area" or "testing without wearing a safety helmet," the person with the closest Euclidean distance will be selected as the violator. When the violation is "working at height without a safety belt" or "insufficient safe distance between personnel and hazardous equipment," the system will perform the following steps: retrieve the list of operators of the relevant equipment during the corresponding time period; and, if the requirement is met... From the personnel, select those who are also on the operation list, and record the number as follows: .like Then, manual intervention will be used to determine the person who violated the regulations; if The system will then bind that person as a violator of the rules; if Then, the person with the closest Euclidean distance is selected as the person who violated the rules.

[0107] Furthermore, in step 4, alarm decision generation occurs. That is, once successfully bound, the alarm decision engine determines that a violation has occurred. The data fusion and decision module immediately sends a command to the alarm module, triggering real-time alarm commands (on-site audible and visual alarms, sending SMS reminders to the administrator, etc.).

[0108] In one embodiment, the alarm module is specifically used to activate an audible and visual alarm and to send alarm information to the terminal of the safety management personnel. The alarm module responds in real time based on the output of the data fusion and decision-making module. If the data fusion and decision-making module determines that a violation has occurred and has matched and bound the violation to the violator, it will simultaneously activate two levels of alarms: first, it will activate an audible and visual alarm (such as a buzzer or warning light) in the laboratory to serve as a deterrent and warning; second, it will send alarm information to the terminal of the safety management personnel (such as a PC or mobile device), including the type of violation, the violator, the location of the violation, and on-site images or video clips of the violation, which can serve as a means of remote monitoring and handling.

[0109] In one embodiment, the alarm module generates / triggers an alarm command based on the results of the data fusion and decision module. In another embodiment, after the data fusion and decision module completes the binding of the violation and the violator, it generates an alarm command and sends it to the alarm module.

[0110] In one preferred implementation, the alarm and intervention module is also used to record the entire process of the violation, that is, to completely record the entire process of the event identification and alarm in the log database, forming a complete chain of evidence for event tracing. Specifically, the alarm and intervention module is used to record video or images of the violation, and to record the ID of the violator, the location coordinates of the violator, and the time of the violation. It may also include information such as the type of violation.

[0111] Preferably, the alarm intervention module adopts differentiated alarm and intervention measures for different types of violations. The alarm operation is based on the binding result, specifically on the category of the violation. This design enhances the timely alerting and rapid intervention capabilities for high-risk violations, improving the level of intelligence in security management.

[0112] See Figure 3 This embodiment also provides a standardized detection and autonomous alarm method for personnel testing, including:

[0113] Step 1: Obtain laboratory video stream information;

[0114] Step 2: Use the improved YOLOv11 model to identify violations in the laboratory video stream information to obtain violation data. The violation data includes images or videos of violations. The input of the improved YOLOv11 model is the laboratory video stream information. The improved YOLOv11 model is a YOLOv11 model with a global grouped coordinate attention module embedded in the backbone network.

[0115] Step 3: Analyze the location coordinates of personnel in the laboratory to obtain personnel location coordinates with ID information and timestamps. The personnel location coordinates are coordinates in the UWB coordinate system.

[0116] Step four: Associate the violation data with the personnel location coordinates, and use a multimodal data fusion algorithm to bind the violation with the violator to obtain the binding result;

[0117] Step 5: Perform an alarm operation based on the binding result.

[0118] In this embodiment, step four specifically includes:

[0119] Convert the pixel coordinates of the image or video corresponding to the violation to world coordinates;

[0120] Establish the mapping relationship between the UWB positioning coordinate system and the world coordinate system.

[0121] Convert the world coordinates to UWB coordinates;

[0122] The violation data output by the visual perception module and the personnel location coordinates are matched in time, and the matching condition is that the time difference is less than a preset time threshold.

[0123] Calculate the Euclidean distance between the UWB coordinates corresponding to the violation data and the location coordinates of the personnel matching the time.

[0124] Based on the number of people whose Euclidean distance is less than the spatial distance threshold;

[0125] Determine the category of the violation;

[0126] The individuals responsible for the violations will be identified based on the number of personnel and the type of violation.

[0127] Link violations to the individuals who violated the rules.

[0128] Specifically, step four further includes: obtaining a list of equipment operators; the step of determining the violator based on the number of personnel and the type of violation is specifically: determining the violator based on the number of personnel, the type of violation, the list of equipment operators, and Euclidean distance.

[0129] The specific steps for determining the person violating the rules based on the number of personnel and the type of violation are as follows:

[0130] If the number of personnel is equal to 1, then personnel whose Euclidean distance is less than the spatial distance threshold are identified as violators. If the number of personnel is not equal to 1, then the violation is identified as belonging to the first type (e.g., whether it belongs to "personnel entering a prohibited area" or "testing without wearing a safety helmet"). If so, then the personnel with the smallest Euclidean distance are selected as violators. If it does not belong to the first type (specifically, it belongs to the second type, for example, "not wearing a safety belt while working at height" or "insufficient safety distance between personnel and dangerous equipment"), then the list of equipment operators related to the violation is retrieved within the time period of the violation. The total number of personnel whose Euclidean distance is less than the spatial distance threshold and who are in the list of equipment operators is selected. If the total number of personnel is equal to 1, then the personnel whose Euclidean distance is less than the spatial distance threshold and who are in the list of equipment operators are identified as violators. If the total number of personnel is not equal to 1, then the total number of personnel is determined as greater than 1. If so, then the personnel with the smallest Euclidean distance and who are in the list of equipment operators are identified as violators.

[0131] When the total number of personnel is neither equal to nor greater than 1 (i.e., equal to 0), the person who violated the rules is identified manually.

[0132] When the number of personnel is neither equal to nor greater than 1 (i.e., equal to 0), the person violating the rules is identified manually.

[0133] In this embodiment, step three specifically involves: analyzing the location coordinates of personnel in the laboratory based on the transmission time of radio waves from the tag to the base station. The personnel in the laboratory wear the tag, the tag is capable of emitting radio waves, and the tag has a unique ID.

[0134] In this embodiment, step five specifically includes: generating an alarm command based on the output of the data fusion and decision-making module. Furthermore, step five also includes using an audible and visual alarm to send alarm information to the terminal of the safety management personnel.

[0135] In this embodiment, step 5 further includes recording the video or image of the violation, the ID of the person violating the rules, the location coordinates of the person violating the rules, and the time of the violation.

[0136] In specific implementation, the personnel test standardization detection and autonomous alarm method can refer to the personnel test standardization detection and autonomous alarm system in any of the above embodiments to achieve personnel test standardization detection and autonomous alarm. The specific implementation steps will not be repeated.

[0137] An electronic device can be implemented according to the method of this disclosure, the electronic device comprising: a memory; one or more processors; one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for executing the personnel test standardization detection and autonomous alarm method according to any of the above embodiments.

[0138] This disclosure also provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the steps of the personnel testing standardization detection and autonomous alarm method described in any of the above embodiments.

[0139] This disclosure discloses a standardized detection and autonomous alarm system and method for personnel testing. It utilizes an improved YOLOv11 model with a global grouping attention mechanism for violation detection and identification, significantly enhancing the ability to extract features of violations from small targets and complex backgrounds, and improving the accuracy of automatic violation identification in real-world laboratory scenarios. By employing a data fusion and decision-making module, it analyzes the results of visual perception and UWB fusion to identify the violating personnel, effectively overcoming the bottleneck of pure vision systems' inability to accurately distinguish responsible persons in occluded, backlit, or multi-person gathering scenarios, achieving precise and automatic binding of violations with violating personnel. Based on the precise binding results, this disclosure implements automatic alarm operation.

[0140] Specifically, this disclosure has the following effects:

[0141] 1. Improved recognition accuracy: By introducing a global group attention mechanism to improve the YOLOv11 model, the ability to extract features of violations in small targets (such as not wearing a safety helmet) and complex backgrounds is significantly enhanced, and the recognition accuracy and robustness in real laboratory scenarios are greatly improved.

[0142] 2. Precise Identity Binding: Utilizing visual-UWB fusion technology and a conflict determination mechanism based on violation type, precise and automatic binding of violations with responsible persons is achieved with high accuracy. Specifically, firstly, the identified violation events are associated with personnel location information obtained through UWB positioning using a unified spatiotemporal reference; secondly, a secondary identity verification is performed using the equipment operator list; thus achieving precise and automated association between violations and violators in both single-person and multi-person gathering scenarios.

[0143] 3. Improve regulatory efficiency: By automating the entire process from identification and binding to alarm, the time and economic costs previously spent on manual inspections or post-event verification have been significantly reduced, thus improving work efficiency.

[0144] 4. Complete and reliable chain of evidence: The system can automatically link and retain violations, responsible person IDs, time, location and on-site images to form an unalterable and complete chain of evidence, providing a strong basis for tracing events and defining responsibilities.

[0145] 5. Intelligent decision-making and intervention: Differentiated alarm and intervention measures are adopted for different types of violations to enhance the ability to promptly remind and quickly intervene in high-risk violations and improve the level of intelligence in safety management.

[0146] This disclosure automates the entire process from identification and binding to alarm, significantly reducing the time and economic costs previously spent on manual inspections or post-event verification, and improving the efficiency of standardized detection and alarm work.

[0147] This disclosure enhances the YOLOv11 model's ability to capture key behavioral features under complex experimental conditions by embedding a global grouped coordinate attention module. This significantly improves recognition accuracy and stability while further enhancing detection sensitivity. Simultaneously, by incorporating UWB (centimeter-level) positioning information, it can accurately locate personnel even in blind spots, ensuring the correctness of behavior-identity association verification. Furthermore, by combining multimodal data for fusion analysis, it achieves a shift from "passive recording" to "active alarm," constructing an intelligent monitoring system with comprehensive perception, high-precision judgment, and rapid response, thereby significantly improving the efficiency of laboratory safety management.

[0148] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The embodiments described above are merely illustrative of several implementations of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these all fall within the protection scope of this disclosure. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A standardized detection and autonomous alarm system for personnel testing, characterized in that, include: The acquisition module is used to acquire laboratory video stream information; The visual perception module is used to identify violations in laboratory video stream information using an improved YOLOv11 model and output violation data with timestamps. The violation data includes images or videos of violations. The input of the improved YOLOv11 model is laboratory video stream information. The improved YOLOv11 model is a YOLOv11 model with a global grouped coordinate attention module embedded in the backbone network. The UWB positioning module is used to analyze the location coordinates of personnel in the laboratory and output the location coordinates of personnel with ID information and timestamps. The data fusion and decision-making module is used to associate the violation data with the personnel location coordinates, and to bind the violation with the violator through a multimodal data fusion algorithm to obtain the binding result; The alarm module is used to perform alarm operations based on the binding result.

2. The standardized personnel testing and autonomous alarm system according to claim 1, characterized in that, The global grouped coordinate attention module includes: The grouping unit is used to group the input feature map along the channel dimension to obtain the grouped feature map; The first pooling unit performs a global average pooling operation along the height direction on each grouped feature map to obtain... ; The second pooling unit performs max pooling on each grouped feature map along the height direction to obtain... ; The third pooling unit performs a global average pooling operation along the width direction on each grouped feature map to obtain... ; The fourth pooling unit is used to perform max pooling on each grouped feature map along the width direction to obtain... ; Convolutional units are used for... , , and Perform convolution processing; The weight generation unit is used to... The results of convolution processing and The convolution results are summed to obtain the first result, which is used to... The results of convolution processing and The convolution results are added together to obtain the second result. The first and second results are then processed by the Sigmoid function to generate spatial attention weights. The weighting module is used to weight the input feature map based on spatial attention weights to obtain the output feature map.

3. The standardized personnel testing and autonomous alarm system according to claim 1, characterized in that, The data fusion and decision-making module is specifically used for: Convert the pixel coordinates of the image or video corresponding to the violation to world coordinates; Establish the mapping relationship between the UWB positioning coordinate system and the world coordinate system. Convert the world coordinates to UWB coordinates; The violation data output by the visual perception module and the personnel location coordinates are matched in time, and the matching condition is that the time difference is less than a preset time threshold. Calculate the Euclidean distance between the UWB coordinates corresponding to the violation data and the location coordinates of the personnel matching the time. Based on the number of people whose Euclidean distance is less than the spatial distance threshold; Determine the category of the violation; The individuals responsible for the violations will be identified based on the number of personnel and the type of violation. Link violations to the individuals who violated the rules.

4. The standardized personnel testing and autonomous alarm system according to claim 3, characterized in that, The data fusion and decision module is also specifically used to obtain a list of equipment operators. The step of determining the violator based on the number of personnel and the type of violation is specifically: determining the violator based on the number of personnel, the type of violation, the list of equipment operators, and Euclidean distance.

5. The standardized personnel testing and autonomous alarm system according to claim 4, characterized in that, The specific steps for determining the violator based on the number of personnel and the type of violation are as follows: when the number of personnel is 1, the personnel whose Euclidean distance is less than the spatial distance threshold are identified as violators; when the number of personnel is greater than 1 and the type of violation belongs to the first type, the personnel with the smallest Euclidean distance are selected as violators. When the number of personnel is greater than 1 and the type of violation falls within the second category, the list of equipment operators for the relevant equipment during the time period of the violation is retrieved. The total number of personnel whose Euclidean distance is less than the spatial distance threshold and who are located in the list of equipment operators is selected. When the total number of personnel is 1, the personnel who meet the requirement of having an Euclidean distance less than the spatial distance threshold and who are located in the list of equipment operators are identified as the violators. When the total number of personnel is greater than 1, the personnel with the smallest Euclidean distance and who are located in the list of equipment operators are identified as the violators.

6. The standardized personnel testing and autonomous alarm system according to claim 1, characterized in that, The UWB positioning module is used to analyze the location coordinates of personnel in the laboratory based on the transmission time of radio waves from the tag to the base station. The personnel in the laboratory wear the tag, which is capable of emitting radio waves and has a unique ID.

7. The standardized personnel testing and autonomous alarm system according to claim 1, characterized in that, The alarm module is specifically used to generate alarm commands based on the output of the data fusion and decision-making module.

8. The standardized personnel testing and autonomous alarm system according to claim 7, characterized in that, The alarm module is specifically used to issue an alarm using an audible and visual alarm and to send alarm information to the terminal of the safety management personnel.

9. The standardized personnel testing and autonomous alarm system according to claim 1, characterized in that, The alarm module is also used to record videos or pictures of violations, including the ID of the person committing the violation, the location coordinates of the person committing the violation, and the time of the violation.

10. A standardized detection and autonomous alarm method for personnel testing, characterized in that, include: Obtain laboratory video stream information; The improved YOLOv11 model is used to identify violations in laboratory video stream information to obtain violation data, which includes images or videos of violations. The input of the improved YOLOv11 model is the laboratory video stream information, and the improved YOLOv11 model is a YOLOv11 model with a global grouped coordinate attention module embedded in the backbone network. The location coordinates of personnel in the laboratory are analyzed to obtain personnel location coordinates with ID information and timestamps, and the personnel location coordinates are coordinates in the UWB coordinate system; The violation data and the personnel location coordinates are associated, and the violation is bound to the violator through a multimodal data fusion algorithm to obtain the binding result; An alarm operation will be performed based on the binding result.

Images