Method, device and system for early warning of safe behavior of aerial refueling vehicle

Abstract

This application provides a method, device, and system for early warning of safety behaviors of aviation refueling trucks. The method utilizes an edge computing terminal. If the personnel in the video data acquired at the location in front of the truck are identified as staff, and after the staff have left the area captured by a designated camera once, the camera ranked first in the designated camera sequence is designated as the current camera. The current video data is then used to assess whether the behavior complies with the safety regulations for aviation refueling trucks. If the current camera is at the end of the designated camera sequence, a judgment result confirming compliance is sent to the cloud. If the current camera is not at the end of the designated camera sequence, the next camera is updated as the current camera, and video data captured by the current camera continues to be acquired. For behaviors that do not comply with the safety regulations for aviation refueling trucks, an early warning message is obtained and issued. Therefore, this application can effectively solve the problem of dormant monitoring terminal equipment and improve the efficiency of safety inspections of aviation refueling trucks.

Description

Technical Field

[0001] This application relates to the field of safety operation and maintenance technology, and in particular to a method, device and system for early warning of safety behavior of aviation refueling vehicles. Background Technology

[0002] Aviation refueling trucks are important facilities at airports and are fundamental to ensuring the smooth operation of airports. The safe operation and maintenance management of aviation refueling trucks is one of the key points of the daily operation and maintenance management work of aviation fuel companies.

[0003] In recent years, with the rapid development of video capture and mobile network transmission technologies, some companies have deployed a large number of cameras in their operation and maintenance areas and remotely transmitted the video data captured by these cameras to monitoring equipment deployed at the company's end. The monitoring equipment displays the monitoring images in order to supervise the monitored area. However, manually reviewing each image one by one is labor-intensive, has a large monitoring area, and results in low monitoring efficiency, which in turn leads to low safety inspection efficiency for aviation refueling trucks. Summary of the Invention

[0004] This application provides a method, device, and system for early warning of safety behaviors of aviation refueling trucks, which makes the safety inspection efficiency of aviation refueling trucks higher.

[0005] The technical solution provided in this application includes:

[0006] In a first aspect, embodiments of this application provide a method for early warning of safe behavior of an aviation refueling truck. This method is applied to an edge computing terminal of a detection system. The detection system further includes multiple cameras and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck, and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server. After determining that refueling has been completed, the method includes:

[0007] Acquire video data captured by a camera at the front of the vehicle; if the person in the video data is identified as a staff member, and after it is determined that the staff member has left the monitoring area captured by the camera at the front of the vehicle once, then the camera ranked first in the designated camera sequence is taken as the current camera, the video data captured by the current camera is acquired, and it is determined whether the behavior of the staff member in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling vehicles.

[0008] If the conditions are met, determine whether the current camera's order is at the end of the specified camera's order; if not, take the next camera in the specified camera's order as the current camera and return to the step of obtaining the video data captured by the current camera.

[0009] If the violation is not met, a warning message is obtained to indicate the violation, and the warning message is used to issue a warning.

[0010] Secondly, embodiments of this application provide a detection device for the safe behavior of an aviation refueling truck. This device is applied to an edge computing terminal of a detection system. The detection system further includes multiple cameras and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck, and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server. After determining that refueling has been completed, the device includes:

[0011] The video data acquisition unit is used to acquire video data captured by a camera at the front of the vehicle; if the person in the video data is identified as a staff member, and the staff member leaves the monitoring area captured by the camera at the front of the vehicle once, the behavior rule determination unit is triggered.

[0012] The behavior regulation determination unit is used to select the camera ranked first in the specified camera sorting as the current camera, acquire the video data captured by the current camera, and determine whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks; if it complies, the camera sorting determination unit is triggered; if it does not comply, the warning information acquisition unit is triggered.

[0013] The camera sorting determination unit is used to determine whether the current camera sorting is at the end of the specified camera sorting; if not, the data video reacquisition unit is triggered.

[0014] The data video acquisition unit is used to take the next camera in the current camera sorting in the specified camera sorting as the current camera, and return to execute the step of acquiring the video data captured by the current camera.

[0015] The warning information acquisition unit is used to acquire warning information indicating violations and to issue warnings using the warning information.

[0016] As can be seen from the above technical solution, the early warning method for safe behavior of aviation refueling trucks provided in this application is applied to the edge computing terminal of the detection system. Each camera in this detection system is installed at the front, rear, left, and right positions of the aviation refueling truck, and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to a cloud application server. After determining that refueling has been completed, if the personnel in the video data obtained at the front position are identified as staff, and after determining that the staff have left the area captured by the designated camera once, the camera ranked first in the designated camera sequence is designated as the current camera, provided that the current video data conforms to the safety behavior regulations for aviation refueling trucks. If the current camera is not at the end of the designated camera sequence, the next camera is updated as the current camera, and video data captured by the current camera continues to be acquired. For violations of the safety behavior regulations for aviation refueling trucks, early warning information for the violation is generated and used to issue a warning. As can be seen, the technical solution provided in this application no longer relies on cameras at designated locations to monitor a fixed area. Instead, it relies solely on cameras installed around the aviation refueling truck to monitor the area that the truck needs to monitor. Furthermore, it transforms traditional passive monitoring into active monitoring, automatically extracting and analyzing the camera footage to determine violations. This method requires minimal manpower and offers high detection efficiency. Therefore, the technical solution provided in this application demonstrates high efficiency in the safety inspection of aviation refueling trucks. Attached Figure Description

[0017] Figure 1(a) is a schematic diagram of the camera setup in the detection system for safe behavior of aviation refueling vehicles provided in this application;

[0018] Figure 1(b) is a schematic diagram of the structure of the detection system for the safety behavior of aviation refueling vehicles provided in this application;

[0019] Figure 2 A flowchart illustrating a method for early warning of safe behavior of aviation refueling vehicles provided in this application;

[0020] Figure 3 The network structure framework diagram of the YOLOv7 network model provided in this application;

[0021] Figure 4(a) is a schematic diagram of the ELAN module provided in this application;

[0022] Figure 4(b) is a schematic diagram of the MP-1 module provided in this application;

[0023] Figure 5 A schematic diagram of the ELAN-W module provided in this application;

[0024] Figure 6A schematic diagram of the framework of the SPPCS module provided in this application;

[0025] Figure 7(a) is a schematic diagram of the staff provided in this application in an actual work scenario;

[0026] Figure 7(b) is a schematic diagram of the key points of the extraction personnel provided in this application;

[0027] Figure 7(c) is a schematic diagram of the relative 3D coordinates of each key point belonging to the staff provided in this application;

[0028] Figure 8 A schematic diagram of the structure of a detection device for the safety behavior of an aviation refueling vehicle provided in this application;

[0029] Figure 9 This is a structural diagram of an electronic device provided in this application. Detailed Implementation

[0030] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0031] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0032] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0033] Figure 1(a) is a schematic diagram of the camera setup in the aviation refueling vehicle safety behavior detection system provided in this application; Figure 1(b) is a schematic diagram of the aviation refueling vehicle safety behavior detection system provided in this application.

[0034] As shown in Figures 1(a) and 1(b), the early warning method for the safety behavior of an aviation refueling truck provided in this application is applied to the edge computing terminal of a detection system. The detection system also includes multiple cameras and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck, as shown in Figures 1(a) and 1(b), and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server. In this embodiment, multiple cameras are installed at the front, rear, left, and right positions of the aviation refueling truck. This arrangement of cameras can capture all monitoring areas of the aviation refueling truck in all four directions without blind spots, thus covering the entire perimeter of the aviation refueling truck as much as possible. In the embodiment shown in Figure 1(b), the cameras are respectively installed at the front, rear, left, and right positions of the aviation refueling vehicle, and can be respectively referred to as the front camera (corresponding to the front camera in Figure 1(b), the right camera (corresponding to the rear camera in Figure 1(b), the left camera (corresponding to the left camera in Figure 1(b), and the right camera (corresponding to the right camera in Figure 1(b)). Multiple cameras can be connected to the aviation refueling vehicle's own onboard NVR (Network Video Recorder) system. The onboard NVR system is connected to an edge computing terminal, which in turn is connected to a cloud application server, allowing real-time monitoring of the videos captured by each camera through the onboard NVR system.

[0035] Compared to related technologies where manual monitoring is prone to missing footage and has lower accuracy, this embodiment utilizes multiple cameras to detect whether the video data captured by the current camera complies with the safety regulations for aviation fuel refueling trucks. This automatic, real-time detection allows for timely understanding of on-site information and automatic generation of judgment results. Multiple cameras can capture images within the visible range around the aviation refueling truck, providing comprehensive information and preventing omissions, thereby improving detection accuracy and ultimately enhancing the accuracy of safety inspections of aviation refueling trucks.

[0036] Compared to related technologies that deploy numerous cameras at airports, only the cameras near the vehicle are used for vehicle monitoring, leaving other unused cameras dormant, resulting in low utilization of the existing airport cameras. In this embodiment, multiple cameras are installed throughout the vehicle, utilizing only the onboard cameras, thus achieving a higher utilization rate of the vehicle's onboard cameras.

[0037] Compared to related technologies that involve remote transmission to monitoring equipment deployed at the enterprise level, which requires transmitting large amounts of video data collected by numerous cameras via 4G or 5G network cards to the enterprise's monitoring equipment for processing, the large data volume not only leads to slow network transmission speeds but may also result in transmission failures. This results in incomplete data on the enterprise's monitoring equipment, consequently lower accuracy in safety inspections of aviation refueling trucks. In this embodiment, the edge computing terminal obtains a determination result indicating whether the behavior of the personnel in the current video is compliant or non-compliant. Only the result or a small amount of data is uploaded to the cloud application server. This smaller data volume reduces the impact on network transmission, making the data from the monitoring equipment more complete and improving monitoring accuracy, thereby resulting in higher accuracy in safety inspections of aviation refueling trucks.

[0038] After confirming that the aviation refueling vehicle has completed refueling... Figure 2 A flowchart illustrating a method for early warning of safe behavior of aviation refueling vehicles provided in this application.

[0039] based on Figure 2 The process for implementing the early warning method for safe behavior of aviation refueling vehicles may include the following steps 101 to 106:

[0040] Step 101: Obtain video data captured by the camera at the front of the vehicle; if the person in the video data is identified as a staff member, and after determining that the staff member has left the monitoring area captured by the camera at the front of the vehicle once, proceed to step 102.

[0041] This step involves detecting personnel in the video stream captured by the camera at the front of the vehicle. It determines whether any personnel wearing safety helmets and work clothes have entered a designated area in front of the vehicle, which is within the area captured by the camera at the front of the vehicle. If the video data captured by the camera at the front of the vehicle shows personnel wearing safety helmets and work clothes, then the personnel are considered to be staff members.

[0042] It should be noted that the safety helmets and work clothes are specifically designated for use by personnel on aviation refueling vehicles.

[0043] Step 102: Select the camera ranked first in the specified camera sorting as the current camera, acquire the video data captured by the current camera, and determine whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks; if it complies, proceed to step 103; if it does not comply, proceed to step 106.

[0044] In this step, the camera order is specified by pre-sorting the cameras installed on the aviation refueling truck. As one embodiment, the specified camera order is: camera on the right side of the vehicle, camera at the rear of the vehicle, and camera on the left side of the vehicle. It should be noted that there can be multiple specified camera orders, and the above is merely one embodiment; this application does not limit the scope of the specification.

[0045] Step 103: Determine whether the current camera's order is at the end of the specified camera's order; if yes, proceed to step 104; if not, proceed to step 105.

[0046] In this step, it is determined whether all cameras in the specified camera sorting have been processed by judging whether the current camera sorting is at the end of the specified camera sorting. If the current camera sorting is not at the end of the specified camera sorting, it means that all cameras in the specified camera sorting have not been processed; if the current camera sorting is at the end of the specified camera sorting, it means that all cameras in the specified camera sorting have been processed.

[0047] Step 104: Send a determination result to the cloud application server to determine whether the behavior of the staff in the current video is compliant.

[0048] This embodiment only sends the judgment result to the cloud application server to avoid the phenomenon of unreliable remote transmission due to large video data traffic. In other words, it can improve the safety inspection efficiency of aviation refueling trucks.

[0049] Step 105: Select the next camera in the specified camera sorting as the current camera, and return to the step 102 to obtain the video data captured by the current camera.

[0050] In this application, "the next camera in the current camera sorting" can be understood as the camera in the specified camera sorting order, meaning the camera that is sorted next in the current camera sorting.

[0051] Step 106: Obtain warning information indicating violations and use the warning information to issue a warning. This allows for timely reminders of violations by staff, assisting them in correcting errors and ensuring safe conduct.

[0052] The warning information in this step can be generated by the edge computing terminal itself, or it can be generated by the cloud application server and sent to the edge computing terminal. Based on this, in some embodiments, implementing step 106 may include: generating warning information for a violation based on a determination that the behavior of a staff member in the current video constitutes a violation, issuing a warning using the warning information, and sending the warning information to the cloud application server. In other embodiments, implementing step 106 may include: sending a determination result to the cloud application server that the behavior of a staff member in the current video constitutes a violation, receiving warning information from the cloud application server indicating a violation, and issuing a warning using the warning information. In this embodiment, when the cloud application server receives the determination result, it can generate warning information based on the determination result and send the generated warning information to the edge computing terminal, thereby reducing the information processing burden on the edge computing terminal.

[0053] Warning information is at least used to alert users to violations. Warning information includes warning terms, such as "Please operate according to regulations," or "Please do not leave." Of course, the warnings using the aforementioned warning information can include, but are not limited to, displaying warning information, broadcasting warning information aloud, or using vibration alerts. Any method capable of issuing warnings to staff falls within the protection scope of this application's embodiments, and will not be listed individually here.

[0054] This embodiment uses a combination of cloud application servers and edge computing terminals for intelligent data analysis and processing. By combining the computing resources of cloud application servers and edge computing terminals, and based on business processes, it realizes intelligent safety operation and maintenance in scenarios such as aviation refueling trucks, effectively solving the problem of freeing up inefficient manpower and effectively improving the safety inspection efficiency of aviation refueling trucks.

[0055] This completes the description shown in Figure 1.

[0056] As can be seen from the above technical solutions, the technical solutions provided in this application, on the one hand, no longer send all video data captured by the camera to the cloud application server, but only send early warning information of violations to the cloud application server, greatly reducing the data transmitted over the network and thus improving network transmission speed; on the other hand, they no longer rely on cameras at designated locations to monitor the monitoring screen of a fixed area, but only rely on cameras installed around the aviation refueling truck to monitor the area that the aviation refueling truck needs to monitor. At the same time, they can transform traditional passive monitoring into active monitoring, automatically extracting and analyzing the monitoring screen captured by the cameras to determine violations. Monitoring requires no manpower and is highly efficient. Therefore, the technical solutions provided in this application can improve the safety inspection efficiency of aviation refueling trucks.

[0057] Compared to related technologies where monitoring results are obtained through manual monitoring, which suffers from low efficiency due to the large amount of monitoring data, the edge computing terminal of this application automatically generates judgment results, resulting in higher detection effectiveness and thus improving the efficiency of safety inspections of aviation refueling trucks. Simultaneously, the edge computing terminal obtains judgment results indicating whether the behavior of personnel in the current video is compliant or non-compliant; only the results or partial content can be viewed by personnel, resulting in a smaller amount of monitored data and higher monitoring efficiency.

[0058] This concludes the description of the above embodiments.

[0059] As an example, when the current camera is located on the right side of the vehicle, the specific implementation of step 102, which determines whether the behavior of the staff in the video data captured by the current camera conforms to the safety behavior regulations for aviation fuel refueling trucks, includes the following steps:

[0060] Step A1: Obtain video data captured by the camera at the right position of the vehicle; if the person in the video data is identified as a staff member, proceed to step A2; if the staff member in the video data is identified as leaving the monitoring area captured by the camera at the right position of the vehicle, proceed to step A4.

[0061] In some application scenarios where the aforementioned staff member leaves the monitoring area captured by the camera on the right side of the vehicle, such as when the staff member faints, a warning message is generated to call for help. In other application scenarios where the staff member leaves the monitoring area captured by the camera on the right side of the vehicle, such as when the staff member briefly leaves, a warning message is generated to indicate when the staff member should leave.

[0062] The video data for this step is the video data captured by the camera located on the right side of the vehicle.

[0063] Step A2: Detect designated key points of the staff in the video data. If the designated key point can form an acute angle, can form a straight line that is not perpendicular to the bottom surface, or can form a straight line that is within a set angle range with the bottom surface, then proceed to step A3. If the designated key point fails to form an acute angle, fails to form a straight line, or can form a straight line but the straight line does not form an angle within the set angle range with the bottom surface, then return to step A1.

[0064] The set angle range can be understood as an angle range of approximately 90°. As an example, the set angle range can be understood as [85° 90°].

[0065] In this step, if the specified key point can form an acute angle, then an action is considered to be completed. If the specified key point forms a straight line and the line is not perpendicular to the bottom surface, then an action is considered to be completed. If the specified key point can form a straight line and the line is within a set angle range with the bottom surface, then an action is considered to be completed. When all three actions are completed in sequence, step C will be executed to record a confirmation action completion mark.

[0066] If the specified key point cannot form an acute angle, it is considered that an action has not been completed. If the specified key point cannot form a straight line, it is considered that an action has not been completed. If the specified key point can form a straight line but the line is perpendicular to the bottom surface, it is considered that an action has not been completed. Or, if the specified key point can form a straight line but the line and the bottom surface are not within the set angle range, it is considered that an action has not been completed. If any of these actions is considered to have failed to be completed, it means that an action has not been completed and no record is made.

[0067] Only when all three actions mentioned above are confirmed to be completed is it considered that the staff has completed the three to four confirmation actions. The three to four confirmation actions are for final checks and confirmations after the aviation refueling truck has finished refueling. The checks begin at the front of the truck, proceeding to the right side, rear, and left side, completing a full circumference of the vehicle. The checks include: first, the staff checks the front of the aviation refueling truck to ensure there are no obstructions and that the lights are functioning properly; second, the staff checks the right side of the truck to ensure that the static electricity wires, work ladder, and work stool are in place, and re-checks the refueling connectors and hoses to ensure they are in place; third, the staff checks the rear of the truck to ensure that the fire extinguisher and warning flags are in place; and fifth, the staff checks the left side of the truck to ensure that the reel connector is in place, and re-checks the emergency pull rope and other equipment to confirm they are in place. Except for the front of the truck, each check requires a marked action: raising the arm at an acute angle to the elbow, then extending it straight out, and finally letting it hang naturally. Based on this, this embodiment determines whether the staff has completed the three to four confirmation actions by detecting the key points specified by the staff.

[0068] Step A3: Record a marker once to indicate that the set action has been confirmed as completed, and return to execute step 105.

[0069] Step A4: Check if the recorded marker is less than 2; if it is less than 2, proceed to step 106; if it is greater than or equal to 2, return to step 103.

[0070] During this detection process, the edge computing terminal acquires video data from the camera on the right side of the vehicle to determine whether the staff member has been in the area that the camera on the right side of the vehicle can capture. If the staff member leaves, it determines whether the confirmation action has been completed twice within the area monitored by the camera on the right side of the vehicle. If it is less than twice, step 103 is executed to issue a violation warning.

[0071] As another embodiment, when the current camera is a camera located at the rear of the vehicle, the specific implementation method of determining whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior specifications for aviation fuel refueling trucks in step 102 includes the following steps:

[0072] Step B1: Obtain video data captured by the camera at the rear of the vehicle; if the person in the video data is identified as a staff member, proceed to step B2; if the staff member in the video data is identified as leaving the monitoring area captured by the camera at the rear of the vehicle, proceed to step B4.

[0073] The video data in this step is the video data captured by the camera located at the rear of the vehicle.

[0074] Step B2: Detect designated key points of the staff in the video data. If the designated key point can form an acute angle, can form a straight line that is not perpendicular to the bottom surface, or can form a straight line that is within a set angle range with the bottom surface, then proceed to step B3. If the designated key point fails to form an acute angle, fails to form a straight line, or can form a straight line but the straight line does not form an angle within the set angle range with the bottom surface, then return to step B1.

[0075] Step B3: Record a marker indicating that the action has been confirmed as set, and return to step 105.

[0076] Step B4: Check if the recorded marker is less than 1; if it is less than 1, proceed to step 106; if it is greater than or equal to 1, return to step 103.

[0077] During this detection process, the edge computing terminal acquires video data from the camera at the rear of the vehicle to determine whether the staff member has been within the area that the camera at the rear of the vehicle can capture. If the staff member leaves the area, it determines whether the confirmation action has been completed once within the area monitored by the camera at the rear of the vehicle. If it is less than once, step 103 is executed to issue a violation warning.

[0078] As another embodiment, when the current camera is located on the left side of the vehicle, the specific implementation method of determining whether the behavior of the staff in the video data captured by the current camera conforms to the safety behavior regulations for aviation fuel refueling trucks in step 102 includes the following steps:

[0079] Step C1: Obtain video data captured by the camera at the left position of the vehicle; if the person in the video data is identified as a staff member, proceed to step C2; if the staff member in the video data is identified as leaving the monitoring area captured by the camera at the left position of the vehicle, proceed to step C4.

[0080] The video data for this step is the video data captured by the camera located on the left side of the vehicle.

[0081] Step C2: Detect the designated key points of the staff in the video data. If the designated key point can form an acute angle, can form a straight line and the straight line is not perpendicular to the bottom surface, or can form a straight line and the straight line is within a set angle range with the bottom surface, then step C3: If the designated key point fails to form an acute angle, fails to form a straight line, or can form a straight line but the straight line is not within a set angle range with the bottom surface, then return to step C1.

[0082] Step C3: Record a marker once to indicate that the set action has been confirmed as completed, and return to execute step 105.

[0083] Step C4: Check if the recorded marker is less than 2; if it is less than 2, proceed to step 106; if it is greater than or equal to 2, return to step 103.

[0084] During this detection process, the edge computing terminal acquires video data from the camera on the left side of the vehicle to determine whether the staff member has been in the area that the camera on the left side of the vehicle can capture. If the staff member leaves, it determines whether the confirmation action has been completed twice within the area monitored by the camera on the left side of the vehicle. If it is less than twice, step 103 is executed to issue a violation warning.

[0085] In another embodiment, identifying the personnel in the video data as staff members may include: inputting the video data into a pre-trained target detection algorithm YOLOv7 network model to identify whether the personnel in the video data are wearing designated safety helmets and designated work clothes, and identifying the personnel wearing designated safety helmets and designated work clothes as staff members.

[0086] In this embodiment, the target detection algorithm YOLOv7 network model is used to identify whether a person is wearing a specified work uniform, which refers to a uniform with a specified safety helmet and specified work clothes.

[0087] Taking the current video as an example of a camera positioned in front of the vehicle, the video data captured by the camera in front of the vehicle is input into a trained YOLOv7 network model. If the output indicates that the person is wearing a designated work uniform, then the person is a staff member. If the output indicates that the person is not wearing a designated work uniform, then the person is not a staff member.

[0088] In some embodiments, the process of training a YOLOv7 network model is as follows:

[0089] Step D1: Use the training set as training samples to iteratively train the YOLOv7 network model, and verify the output of the identification result used to identify whether the person in the training sample is the designated staff member with the labeled image. If the verification accuracy reaches the threshold, the training is considered complete, and the trained YOLOv7 network model is taken as the trained YOLOv7 network model. If the verification accuracy does not reach the threshold, continue to use the training samples to verify the YOLOv7 network model until the accuracy reaches the threshold.

[0090] In this step, image data previously captured by a pre-camera is used as training samples. Annotation tools are used to label images containing human figures, saving the annotations as XML files. These XML files contain category and location information. For each person's image data, annotations are performed specifically on the person wearing work clothes, covering three parts: the outer frame of the person, the frame of the safety helmet, and the frame of the work uniform top. The training sample dataset is then divided, with 80% of the training samples used as the training set and 20% used as the test set.

[0091] The trained YOLOv7 network model is tested using this as a test set to obtain a successfully trained YOLOv7 network model. The network structure of the YOLOv7 network model in this embodiment is as follows: Figure 3 As shown.

[0092] The ELAN and MP-1 modules of the YOLOv7 network model, which are implemented through, for example... Figure 3 The combination of various convolution operations with kernel sizes of 3×3 and 1×1, activation functions, max pooling operations, and normalization operations shown in the figure completes the extraction of image features from the training samples.

[0093] The ELAN module is a highly efficient feature extraction module, and its module structure is as follows: Figure 2 As shown, the system consists of two interconnected branches: a short branch and a long branch. The short branch extracts shallow features, while the long branch extracts deep features. Shallow features retain the texture and appearance information of the feature map, while deep features contain abstract semantic information. The ELAN module then concatenates the feature maps obtained from both branches to obtain a feature map that simultaneously contains both shallow and deep feature information. The CBS (Complete Basis Set) modules on each branch work together to extract image features. The short branch consists of one CBS module, and the long branch consists of five CBS modules, as shown in Figure 4(a). Here, the CBS module includes at least a convolutional layer (Conv), a batch normalization layer (Batch Normalization), and an activation function (Silu).

[0094] The MP-1 module also consists of two branches, but unlike the ELAN module, the two branches of MP-1 are of equal length, as shown in Figure 4(b). The upper branch of the MP-1 module consists of a max-pooling layer (MaxPool) and a CBS module, while the lower branch consists of two CBS modules. The purpose of using the MaxPool layer is to downsample and obtain key feature information from the feature map while reducing parameters. The role of the MP-1 module is to minimize feature information loss during the downsampling process of the feature map, while reducing parameters as much as possible.

[0095] The detection section includes at least the ELAN-W module, MP-2 module, and SPPCSPC module, and its function is to detect the signals transmitted through the network. Figure 3 The backbone network processes feature maps for prediction. The detection part has three detection heads, such as... Figure 3 The detection heads 1, 2, and 3 in the module are used to detect large, medium, and small targets in the image, respectively. This is because large feature maps retain more spatial information, including features of small targets. Therefore, large feature maps in large detection heads are typically used to detect small targets, while small feature maps in small detection heads, having undergone several convolutions, have more abstract information and are more suitable for detecting large targets. The ELAN-W module has a similar structure to the ELAN module, as shown below. Figure 5 As shown, the number of CBS modules used is the same as that used in the ELAN module structure, but the difference is that the ELAN-W module places greater emphasis on the fusion of shallow and deep features. The structure of the MP-2 module is completely identical to that of the MP-1 module, and the number of channels in its resulting feature map is twice that of the MP-1 module. For example... Figure 6 As shown, the SPPCSPC module is responsible for obtaining feature maps with different receptive fields by aggregating max pooling operations of different scales, such as... Figure 6The max pooling layers MaxPool13, MaxPool15, and MaxPool19 in the network allow the network model to adapt to images of different resolutions without increasing the network model's parameters, thus ensuring the network's lightweight nature.

[0096] Step D2: Validate the trained YOLOv7 network model using the test set. If the validation accuracy reaches the set threshold, it is considered that the loss of both the training set and the test set has decreased until it stabilizes, thus achieving the set iteration goal. If the validation accuracy does not reach the set threshold, the training set and the test set are reassigned, and the training in Step D1 and the validation in Step D2 are repeated until the loss of both the training set and the test set has decreased until it stabilizes.

[0097] Based on this, a trained YOLOv7 network model can be used as the training result to detect people and analyze whether they are staff members based on the topological relationship.

[0098] In this embodiment, the algorithm first detects the human body and the upper garment and hat. The algorithm first calls the YOLOv7 network model to detect the human body, work hat and work clothes, and determines whether the detection results of the image have a correlation, that is, whether the work hat and work clothes have an inclusion relationship with the human body detection box. If so, the person wearing the work clothes is a staff member.

[0099] As can be seen, the technical solution adopted in this embodiment can help determine whether the people around the aviation refueling vehicle are staff members, and whether the staff members have entered the camera's collection area.

[0100] As another embodiment, the designated key points include limb key points; wherein, the limb key points include hand points, elbow points, and shoulder points, and the detection of designated key points of the staff in the video data includes:

[0101] The hand, elbow, and shoulder points of the worker in the video data are input into a trained video 3D algorithm, the videoPose3D network model, to detect whether the hand, elbow, and shoulder points can form an acute angle, whether they can form a straight line that is not perpendicular to the bottom surface, and whether they can form a straight line with a set angle range to the bottom surface. If it is detected that the hand, elbow, and shoulder points can form an acute angle, can form a straight line that is not perpendicular to the bottom surface, and can form a straight line with a set angle range to the bottom surface, then a step is executed to record a marker indicating that the predetermined action has been confirmed, and the process returns to step 105. Otherwise, the step of acquiring the video data captured by the current camera continues.

[0102] In this embodiment, the current camera is a camera located on the right side of the vehicle, a camera located at the rear of the vehicle, or a camera located on the left side of the vehicle.

[0103] The videoPose3D network model in this step first uses an existing 2D pose detection algorithm to extract the 2D coordinates of key points on the human body in each frame of the video. Then, the videoPose3D network model further processes the extracted key points, finally outputting the relative 3D coordinates of each key point. With all data labeled, a temporal dilated convolutional model can be used; see [link to relevant documentation] for details. Figures 7(a) to 7(c) .

[0104] The coordinates (x, y) of each key point of the human body in each frame of the video are extracted using a 2D pose detection algorithm. These coordinates are then concatenated to form a three-dimensional array, which serves as the input to the videoPose3D network model.

[0105] Specifically, the number of channels in this array (the length of the third dimension) equals the number of joints J^2, representing the horizontal and vertical coordinates of each joint. The first dimension represents the size of the receptive field of the frame. Given a complete video, considering both historical and future information, the frame to be predicted should be located in the center of the current dimension. For real-time pose estimation, only historical information can be used (the convolution operation in this case is called causal convolution), and the frame to be predicted can be located at one end of the current dimension. For single-frame prediction, the length of the second dimension of this array is 1. For multi-frame prediction (assuming the number of frames is n, n≥2), the arrays from single-frame prediction are directly concatenated, and the length of the second dimension is n.

[0106] Let J = 17, the temporal receptive field be 243 frames, and the kernel length W = 3. Therefore, the input array of the videoPose3D network model has a shape of 243 1 34. First, it is passed sequentially through a convolutional layer with 1024 kernels of size 3:1, a batch normalization layer, a ReLU activation function, and a dropout layer. The resulting intermediate output has a shape of 241 1 1024. This convolution is actually a one-dimensional temporal convolution. Then, it is passed sequentially through B residual blocks, and finally through another convolutional layer to obtain a 1 1 51 output vector. The composition of the residual blocks is shown in the figure below. It is worth noting that the preceding convolution in the i-th residual block is a dilated convolution with a dilation factor of 1 ≤ i ≤ B, where B is the total number of residual blocks. The advantage of this is that the receptive field of the neurons can grow exponentially with the number of layers, allowing higher-level neurons to comprehensively consider information from all input frames, thus obtaining a more comprehensive and accurate judgment. Furthermore, when performing residual connections, since the input and output dimensions of the residual blocks are inconsistent, the input must be pruned before adding the input to the output.

[0107] See Figure 8 , Figure 8 This is a schematic diagram of a detection device 800 for the safe behavior of an aviation refueling truck provided in this embodiment. The device is applied to the edge computing terminal of the detection system, which also includes multiple cameras and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server. After determining that refueling has been completed, the device includes:

[0108] The video data acquisition unit 801 is used to acquire video data captured by a camera at the front of the vehicle; if the person in the video data is identified as a staff member, and the staff member leaves the monitoring area captured by the camera at the front of the vehicle once, the behavior rule determination unit 802 is triggered.

[0109] The behavior regulation determination unit 802 is used to select the camera ranked first in the specified camera sorting as the current camera, acquire the video data captured by the current camera, and determine whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks; if it complies, the camera sorting determination unit 803 is triggered; if it does not comply, the warning information acquisition unit 806 is triggered.

[0110] The camera sorting determination unit 803 is used to determine whether the current camera sorting is at the end of the specified camera sorting; if yes, the compliance result sending unit 804 is triggered; if no, the data video acquisition unit 805 is triggered.

[0111] The compliance result sending unit 804 is used to send a judgment result to the cloud application server, which determines that the behavior of the staff in the current video is compliant behavior;

[0112] The data video acquisition unit 805 is used to take the next camera in the current camera sorting in the specified camera sorting as the current camera, and return to execute the step of acquiring the video data captured by the current camera.

[0113] The warning information acquisition unit 806 is used to acquire warning information indicating violations and to issue warnings using the warning information.

[0114] In one embodiment of this application, the warning information acquisition unit 806 is specifically configured to: generate warning information for a violation based on a determination result indicating that the behavior of a staff member in the current video constitutes a violation, and use the warning information to issue a warning and send the warning information to the cloud application server. In another embodiment of this application, the warning information acquisition unit 806 is specifically configured to: send a determination result indicating that the behavior of a staff member in the current video constitutes a violation to the cloud application server, receive warning information indicating a violation sent by the cloud application server, and use the warning information to issue a warning.

[0115] In one embodiment of this application, the designated cameras are arranged in the following order: a camera at the right side of the vehicle, a camera at the rear of the vehicle, and a camera at the left side of the vehicle.

[0116] In one embodiment of this application, when the current camera is a camera located on the right side of the vehicle, the behavior regulation determination unit 802 includes a violation determination subunit for determining whether the behavior of the staff in the video data captured by the current camera conforms to the safety behavior regulations for aviation fuel refueling trucks. The violation determination subunit is specifically used for:

[0117] Acquire video data captured by a camera located on the right side of the vehicle. If the person in the video data is identified as a staff member, then a specified key point of the staff member in the video data is detected. When it is detected that the specified key point can form an acute angle, the specified key point can form a straight line that is not perpendicular to the bottom surface, and the specified key point can form a straight line that forms an angle within a set range with the bottom surface, then a marker is recorded to indicate that the set action has been confirmed as completed, and the process returns to the step of setting the next camera in the current camera sorting as the current camera. When it is detected that the specified key point cannot form an acute angle, the specified key point cannot form a straight line, or the specified key point can form a straight line but the formed straight line does not form an angle within the set range with the bottom surface, the process returns to the step of acquiring video data captured by a camera located on the right side of the vehicle.

[0118] If it is detected that a staff member in the video data leaves the monitoring area captured by the camera at the right position of the vehicle, then it is checked whether the recorded marker is less than 2; if it is less than 2, then the warning information acquisition unit 806 is triggered.

[0119] If the value is greater than or equal to 2, then return to the step of determining whether the current camera's sorting is at the end of the specified camera sorting.

[0120] In one embodiment of this application, when the current camera is a camera located behind the vehicle, the behavior regulation determination unit 802 includes a violation determination subunit for determining whether the behavior of the staff in the video data captured by the current camera conforms to the safety behavior regulations for aviation fuel refueling trucks. The violation determination subunit is specifically used for:

[0121] Acquire video data captured by a camera at the rear of the vehicle; if the person in the video data is identified as a staff member, then detect designated key points of the staff member in the video data. When it is detected that the designated key point can form an acute angle, the designated key point can form a straight line and the straight line is not perpendicular to the bottom surface, and the designated key point can form a straight line and the straight line is within a set angle range with the bottom surface, then record a mark to indicate that the set action has been confirmed and completed, and return to execute the step of taking the next camera in the current camera sorting as the current camera; when it is detected that the designated key point cannot form an acute angle, the designated key point cannot form a straight line, or the designated key point can form a straight line but the straight line is not within the set angle range with the bottom surface, then return to execute the step of acquiring video data captured by a camera at the rear of the vehicle.

[0122] If it is detected that a staff member in the video data has left the monitoring area captured by the camera at the rear of the vehicle, then it is checked whether the recorded marker is less than 1; if it is less than 1, then the warning information acquisition unit 806 is triggered.

[0123] If the value is greater than or equal to 1, then return to the step of determining whether the current camera's sorting is at the end of the specified camera sorting.

[0124] In one embodiment of this application, when the current camera is a camera located on the left side of the vehicle, the behavior regulation determination unit 802 includes a violation determination subunit for determining whether the behavior of the staff in the video data captured by the current camera conforms to the safety behavior regulations for aviation fuel refueling trucks. The violation determination subunit is specifically used for:

[0125] Acquire video data captured by a camera at the left position of the vehicle; if the person in the video data is identified as a staff member, then detect designated key points of the staff member in the video data. When it is detected that the designated key point can form an acute angle, the designated key point can form a straight line and the straight line is not perpendicular to the bottom surface, and the designated key point can form a straight line and the straight line is within a set angle range with the bottom surface, then record a mark to indicate that the set action has been confirmed and completed, and return to execute the step of taking the next camera in the current camera sorting as the current camera; when it is detected that the designated key point cannot form an acute angle, the designated key point cannot form a straight line, or the designated key point can form a straight line but the straight line is not within the set angle range with the bottom surface, then return to execute the step of acquiring video data captured by a camera at the left position of the vehicle.

[0126] If it is detected that a staff member in the video data leaves the monitoring area captured by the camera at the left position of the vehicle, then it is checked whether the recorded marker is less than 2; if it is less than 2, then the warning information acquisition unit is triggered.

[0127] If the value is greater than or equal to 2, then return to the step of determining whether the current camera's sorting is at the end of the specified camera sorting.

[0128] In one embodiment of this application, the video data acquisition unit 801 includes an identification subunit for identifying personnel in the video data as staff members. The identification subunit is specifically used for:

[0129] The video data is input into a pre-trained object detection algorithm YOLOv7 network model to identify whether the people in the video data are wearing designated safety helmets and designated work clothes, and to identify the people wearing the designated safety helmets and designated work clothes as staff members.

[0130] The specified key points mentioned above include limb key points. Of course, the limb key points mentioned above may include, but are not limited to, one or more of the following: limb points, head points, and torso points.

[0131] In one embodiment of this application, the limb key points include hand points, elbow points, and shoulder points. The detection of designated key points of the worker in the video data includes:

[0132] The hand, elbow, and shoulder points of the worker in the video data are input into the trained video 3D algorithm videoPose3D network model to detect whether the hand, elbow, and shoulder points can form an acute angle, whether the hand, elbow, and shoulder points can form a straight line and the straight line is not perpendicular to the bottom surface, and whether the hand, elbow, and shoulder points can form a straight line and the straight line is within a set angle range with the bottom surface;

[0133] If it is detected that the hand point, the elbow point, and the shoulder point can form an acute angle, the hand point, the elbow point, and the shoulder point can form a straight line that is not perpendicular to the bottom surface, and the hand point, the elbow point, and the shoulder point can form a straight line that is within a set angle range with the bottom surface, then the step of recording a marker indicating that the action has been confirmed as completed according to the set action is executed, and the step of setting the next camera in the current camera sorting in the specified camera sorting as the current camera is executed.

[0134] Otherwise, continue with the step of acquiring the video data captured by the current camera.

[0135] Therefore, the technical solution provided in this application, on the one hand, eliminates the need to send all video data captured by the camera to the cloud application server, but only sends warning information about violations to the cloud application server, greatly reducing the data transmitted over the network and thus improving network transmission speed; on the other hand, it eliminates the reliance on cameras at designated locations to monitor a fixed area, but relies solely on cameras installed around the aviation refueling truck to monitor the area that needs to be monitored. Simultaneously, it transforms traditional passive monitoring into active monitoring, automatically extracting and analyzing the monitoring images captured by the cameras to determine violations. Monitoring requires minimal manpower and is highly efficient. Thus, the technical solution provided in this application can improve the efficiency of safety inspections of aviation refueling trucks.

[0136] The specific implementation process of the functions and roles of each device in the above-mentioned apparatus can be found in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0137] Thirdly, this application embodiment also improves a detection system for the safety behavior of aviation refueling trucks. This detection system includes an edge computing terminal for the early warning method described in any of the above embodiments, multiple cameras, and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck, and is electrically connected to the edge computing terminal, which is also remotely connected to the cloud application server. On the one hand, it eliminates the need to send all video data captured by the cameras to the cloud application server, sending only early warning information of violations, significantly reducing data transmitted over the network and thus improving network transmission speed. On the other hand, it eliminates the reliance on cameras at specific locations to monitor a fixed area, instead relying solely on cameras installed around the aviation refueling truck to monitor the area that needs to be monitored. This transforms traditional passive monitoring into active monitoring, automatically extracting and analyzing the camera footage to determine violations, reducing manpower requirements and increasing monitoring efficiency. Therefore, the technical solution provided by this application embodiment can improve the efficiency of safety inspections of aviation refueling trucks.

[0138] Fourthly, this application also provides an electronic device. From a hardware perspective, a hardware architecture diagram can be found in [reference needed]. Figure 9 As shown, it includes: a machine-readable storage medium and a processor, wherein: the machine-readable storage medium stores machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the detection operation of the aviation refueling vehicle safety behavior disclosed in the above example.

[0139] The machine-readable storage medium provided in this application embodiment stores machine-executable instructions. When the machine-executable instructions are invoked and executed by a processor, the machine-executable instructions cause the processor to perform the detection operation of the aviation refueling vehicle safety behavior disclosed in the above example.

[0140] Here, a machine-readable storage medium can be any electronic, magnetic, optical, or other physical storage device that can contain or store information, such as executable instructions, data, etc. For example, a machine-readable storage medium can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0141] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which can take the form of a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email sending and receiving device, game console, tablet computer, wearable device, or any combination of these devices.

[0142] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0143] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, embodiments of this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0144] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0145] Furthermore, these computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0146] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0147] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units 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 application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0148] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for early warning of safety behaviors of aviation refueling vehicles, characterized in that, This method is applied to the edge computing terminal of a detection system, which also includes multiple cameras and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck, and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server. After determining that refueling has been completed, the method includes: Acquire video data captured by a camera at the front of the vehicle; if the person in the video data is identified as a staff member, and after it is determined that the staff member has left the monitoring area captured by the camera at the front of the vehicle once, then the camera ranked first in the designated camera sequence is taken as the current camera, the video data captured by the current camera is acquired, and it is determined whether the behavior of the staff member in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling vehicles. If the conditions are met, determine whether the current camera's order is at the end of the specified camera's order; if not, take the next camera in the specified camera's order as the current camera and return to the step of obtaining the video data captured by the current camera. If the violation is not met, a warning message is obtained to indicate the violation, and the warning message is used to issue a warning.

2. The early warning method for the safety behavior of aviation refueling vehicles according to claim 1, characterized in that, The step of obtaining early warning information to indicate violations and using the early warning information to issue an early warning includes: Based on the determination that the behavior of the staff in the current video constitutes a violation, a warning message for the violation is generated, and the warning message is used to issue a warning and sent to the cloud application server; or Send a determination result to the cloud application server to determine that the behavior of the staff in the current video is a violation, receive a warning message from the cloud application server indicating a violation, and use the warning message to issue a warning.

3. The early warning method for the safe behavior of aviation refueling vehicles according to claim 1, characterized in that, When the current camera is located on the right side of the vehicle, determining whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks includes: Acquire video data captured by a camera located on the right side of the vehicle. If the person in the video data is identified as a staff member, then a specified key point of the staff member in the video data is detected. When the specified key point forms an acute angle, forms a straight line that is not perpendicular to the bottom surface, or forms a straight line that is within a set angle range with the bottom surface, a marker is recorded to indicate that the set action has been confirmed and completed. Then, the process returns to the step of setting the next camera in the current camera sequence as the current camera. When the specified key point fails to form an acute angle, fails to form a straight line, or forms a straight line but the straight line does not form an angle within the set angle range with the bottom surface, the process returns to the step of acquiring video data captured by a camera located on the right side of the vehicle. If it is detected that a staff member in the video data leaves the monitoring area captured by the camera on the right side of the vehicle, then it is checked whether the recorded marker is less than 2; if it is less than 2, then the step of obtaining warning information to indicate the violation and using the warning information to issue a warning is executed. If the value is greater than or equal to 2, then return to the step of determining whether the current camera's sorting is at the end of the specified camera sorting.

4. The early warning method for the safety behavior of aviation refueling vehicles according to claim 3, characterized in that, When the current camera is located at the rear of the vehicle, determining whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks includes: Acquire video data captured by a camera at the rear of the vehicle; if the person in the video data is identified as a staff member, then detect designated key points of the staff member in the video data. When it is detected that the designated key point forms an acute angle, forms a straight line that is not perpendicular to the bottom surface, or forms a straight line that is within a set angle range with the bottom surface, then record a mark to indicate that the set action has been confirmed and complete, and return to execute the step of setting the next camera in the current camera sorting as the current camera; when it is detected that the designated key point fails to form an acute angle, fails to form a straight line, or forms a straight line but the straight line does not form an angle within the set angle range with the bottom surface, then return to execute the step of acquiring video data captured by a camera at the rear of the vehicle. If it is identified that a staff member in the video data has left the monitoring area captured by the camera at the rear of the vehicle, then it is checked whether the recorded marker is less than 1; if it is less than 1, then the step of obtaining warning information to indicate the violation and using the warning information to issue a warning is executed. If the value is greater than or equal to 1, then return to the step of determining whether the current camera's sorting is at the end of the specified camera sorting.

5. The early warning method for the safety behavior of aviation refueling vehicles according to claim 4, characterized in that, When the current camera is located on the left side of the vehicle, determining whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks includes: Acquire video data captured by a camera at the left position of the vehicle; if the person in the video data is identified as a staff member, then detect designated key points of the staff member in the video data. When it is detected that the designated key point forms an acute angle, forms a straight line that is not perpendicular to the bottom surface, or forms a straight line that is within a set angle range with the bottom surface, then record a marker indicating that the set action has been confirmed and complete, and return to execute the step of setting the next camera in the current camera sorting as the current camera; when it is detected that the designated key point fails to form an acute angle, fails to form a straight line, or forms a straight line but the straight line does not form an angle within the set angle range with the bottom surface, then return to execute the step of acquiring video data captured by a camera at the left position of the vehicle. If it is identified that the staff member in the video data has left the monitoring area captured by the camera at the left position of the vehicle, then it is checked whether the recorded marker is less than 2; If the value is less than 2, then the steps of obtaining warning information to indicate violations and using the warning information to issue a warning are performed. If the value is greater than or equal to 2, then return to the step of determining whether the current camera's sorting is at the end of the specified camera sorting.

6. The early warning method for the safe behavior of aviation refueling vehicles according to any one of claims 1 to 5, characterized in that, The individuals identified in the video data as staff members include: The video data is input into a pre-trained object detection algorithm YOLOv7 network model to identify whether the people in the video data are wearing designated safety helmets and designated work clothes, and to identify the people wearing the designated safety helmets and designated work clothes as staff members.

7. The early warning method for the safe behavior of aviation refueling vehicles according to any one of claims 3 to 5, characterized in that, The designated key points include limb key points; wherein, the limb key points include hand points, elbow points, and shoulder points, and the detection of designated key points of the staff in the video data includes: The hand, elbow, and shoulder points of the worker in the video data are input into the trained video 3D algorithm videoPose3D network model to detect whether the hand, elbow, and shoulder points form an acute angle, whether the hand, elbow, and shoulder points form a straight line and the straight line is not perpendicular to the bottom surface, and whether the hand, elbow, and shoulder points form a straight line and the straight line is within a set angle range with the bottom surface; If it is detected that the hand point, the elbow point, and the shoulder point form an acute angle, the hand point, the elbow point, and the shoulder point form a straight line that is not perpendicular to the bottom surface, and the hand point, the elbow point, and the shoulder point form a straight line that is within a set angle range with the bottom surface, then the step of recording a marker indicating that the action has been confirmed as completed according to the set action is executed, and the step of setting the next camera in the current camera sorting in the specified camera sorting as the current camera is executed. Otherwise, continue with the step of acquiring the video data captured by the current camera.

8. The early warning method for the safe behavior of aviation refueling vehicles according to any one of claims 1 to 5, characterized in that, The designated cameras are arranged in the following order: the camera on the right side of the vehicle, the camera at the rear of the vehicle, and the camera on the left side of the vehicle.

9. A detection device for the safety behavior of an aviation refueling vehicle, characterized in that, This device is applied to the edge computing terminal of a detection system, which also includes multiple cameras and a cloud application server. Each camera is installed at the front, rear, left, and right positions of the aviation refueling truck, and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server. After determining that refueling has been completed, the device includes: The video data acquisition unit is used to acquire video data captured by a camera at the front of the vehicle; if the person in the video data is identified as a staff member, and the staff member leaves the monitoring area captured by the camera at the front of the vehicle once, the behavior rule determination unit is triggered. The behavior regulation determination unit is used to select the camera ranked first in the specified camera sorting as the current camera, acquire the video data captured by the current camera, and determine whether the behavior of the staff in the video data captured by the current camera complies with the safety behavior regulations for aviation fuel refueling trucks; if it complies, the camera sorting determination unit is triggered; if it does not comply, the warning information acquisition unit is triggered. The camera sorting determination unit is used to determine whether the current camera's sorting is at the end of the specified camera sorting; if yes, it triggers the compliance result sending unit; if no, it triggers the data video reacquisition unit. The compliance result sending unit is used to send a judgment result to the cloud application server, which determines that the behavior of the staff in the current video is compliant behavior; The data video acquisition unit is used to take the next camera in the current camera sorting in the specified camera sorting as the current camera, and return to execute the step of acquiring the video data captured by the current camera. The warning information acquisition unit is used to acquire warning information indicating violations and to issue warnings using the warning information.

10. A detection system for the safety behavior of aviation refueling vehicles, characterized in that, The detection system includes an edge computing terminal that implements the early warning method for safe behavior of aviation refueling vehicles according to any one of claims 1 to 8, multiple cameras, and a cloud application server. Each of the cameras is installed at the front, rear, left, and right positions of the aviation refueling vehicle, and is electrically connected to the edge computing terminal. The edge computing terminal is also remotely connected to the cloud application server.

Images