A safety monitoring method for electric power construction operations, an intelligent safety helmet and a monitoring device

Abstract

The present invention provides a safety monitoring method for electric power construction operations, an intelligent safety helmet and a monitoring device. A safety monitoring method for electric power construction operations mainly includes a first part and a second part. The first part is to determine the types of high-value target images to be captured by the intelligent safety helmet; the second part is to adjust the image acquisition direction of the intelligent safety helmet and adjust the high-value target images in the captured electric power construction operation images; identify the construction behaviors of the adjusted high-value target images, and upload the identification results and the adjusted high-value target images. By adjusting the high-value target images in the electric power construction operation images, the present invention can reduce the utilization of communication resources and improve the monitoring efficiency.

Description

Technical Field

[0001] The present invention relates to the field of distribution network operations, and particularly to a safety monitoring method for electric power construction operations, an intelligent safety helmet, and a monitoring device. Background Art

[0002] In modern distribution network operations, safety monitoring in the scenarios of electric poles and iron towers is of crucial importance. Firstly, when operating personnel conduct inspections and maintenance at high altitudes, there are various safety risks such as falling and equipment failures. In addition, the loosening of the anchor bolts at the base of the iron tower may lead to unstable structures, thereby threatening the safety of operating personnel. Therefore, these all pose higher requirements for safety monitoring means. However, the current monitoring means mostly rely on manual inspections or uploading and identifying a large amount of image data, making it difficult to achieve real-time and flexible monitoring. Moreover, when uploading images or identifying violations in the distribution network operation scenario, a large amount of communication resources are used, resulting in a waste of resources and low accuracy in identifying construction behaviors. Therefore, there is an urgent need for a safety monitoring method for electric power construction operations, etc., to improve the safety and efficiency in distribution network operations. Summary of the Invention

[0003] Aiming at the deficiencies in the prior art, the present invention provides a safety monitoring method for electric power construction operations, an intelligent safety helmet, and a monitoring device, which mainly solves the problems of low recognition efficiency and low accuracy of the safety monitoring method for electric power construction operations in the prior art.

[0004] The purpose of the present invention is achieved through the following solutions:

[0005] According to the first aspect of the present invention, a safety monitoring method for electric power construction operations is provided. The monitoring method includes the following steps: S1. Obtain the operation positioning information of the intelligent safety helmet, and based on the obtained operation positioning information and the pre-configured power grid single map data, determine the operation line where the intelligent safety helmet is located; and determine the high-value target image type to be captured by the intelligent safety helmet based on the operation line where the intelligent safety helmet is located; S2. Collect the electric power construction operation images corresponding to the operation line determined in S1, and use the pre-trained target detection model to identify the high-value target image type in the electric power construction operation images. When the high-value target image type in the electric power construction operation images is inconsistent with the high-value image type to be captured by the intelligent safety helmet, send a prompt message to guide the adjustment of the image acquisition direction of the intelligent safety helmet; and adjust the high-value target images in the captured electric power construction operation images; perform construction behavior recognition on the adjusted high-value target images, and upload the recognition results and the adjusted high-value target images for re-construction behavior recognition.

[0006] In some embodiments, in the step S1, the operation line where the intelligent safety helmet is located is determined through the following steps: based on the operation positioning information of the intelligent safety helmet, the pole tower positions within a preset range where the operation positioning information is located are obtained from the power grid single map data, and the distances from each pole tower position within the preset range to the operation positioning information of the intelligent safety helmet are calculated; the operation line where the pole tower position with the minimum distance to the operation positioning information of the intelligent safety helmet among the obtained pole tower positions within the preset range is located is used as the operation line where the intelligent safety helmet is located.

[0007] In some embodiments, in the step S1, the high-value target image type to be captured by the intelligent safety helmet is determined through the following steps: the operation plan system is queried according to the identification information of the operation line to obtain the operation plan of the operation line; the high-value target image type to be captured by the intelligent safety helmet is obtained by matching the keywords in the operation plan of the operation line in the pre-configured high-value target relationship table.

[0008] In some embodiments, in the step S2, the high-value target images in the electric power construction operation images include a moving target area and a static target area, and both the moving target area and the static target area include multiple sub-target areas; the adjusted high-value target image is obtained through the following method: for the position of each sub-target area, the adjusted sub-target area image is obtained by adjusting the viewing angle and focal length of the camera, and all the adjusted sub-target area images are used as the adjusted high-value target image; among them, the sub-target area that is at the center of the current screen and the size of the sub-target area reaches a preset ratio relative to the size of the current screen is used as the adjusted sub-target area image.

[0009] In some embodiments, for the position of each sub-target area, the viewing angle and focal length of the camera are adjusted through the following steps: the deviation value is calculated based on the center position of the target box of the sub-target area and the center position of the current screen; the camera angle is moved based on the deviation value so that the two centers coincide; the ratio of the target box size to the current screen size at this time is calculated and compared with the preset ratio range, and when both the horizontal ratio and the vertical ratio in the ratio are within the preset ratio range, the adjustment of the camera focal length is stopped.

[0010] Preferably, in the step S2, when the result of the construction behavior recognition of the adjusted high-value target image is a violation, a voice prompt is issued.

[0011] Preferably, in the step S2, the construction behavior is recognized again through the following method:

[0012] Using a pre-configured target recognition large model, identify the construction behaviors in the uploaded high-value target images. When the construction behaviors in the uploaded high-value target images are illegal, send the illegal operation warning information to the intelligent safety helmet that uploaded the high-value target image.

[0013] According to the second aspect of the present invention, there is provided an intelligent safety helmet, which includes a controller and a camera group, a positioning module, a communication module, and an alarm connected to the controller; the camera group is used to collect power construction operation images; the controller is used to control the communication module to send the operation positioning information of the intelligent safety helmet determined by the positioning module, and control the communication module to receive the high-value image types to be captured by the intelligent safety helmet and the operation line where the intelligent safety helmet is located determined by the management server based on the operation positioning information of the intelligent safety helmet; the controller also collects power construction operation images corresponding to the operation line where the safety helmet is located through the camera group; and uses a pre-trained target detection model to identify the high-value target image types in the power construction operation images. When the high-value target image types in the power construction operation images are inconsistent with the high-value image types to be captured by the intelligent safety helmet, send a prompt message to guide the adjustment of the image acquisition direction of the intelligent safety helmet; and adjust the high-value target images in the captured power construction operation images; identify the construction behaviors of the adjusted high-value target images, and upload the identification results and the adjusted high-value target images for re-construction behavior identification.

[0014] According to the third aspect of the present invention, there is provided a power construction operation safety monitoring device, which includes a management server and an intelligent safety helmet provided according to the second aspect of the present invention;

[0015] The management server is used to obtain the operation positioning information of the intelligent safety helmet, and determine the operation line where the intelligent safety helmet is located based on the obtained operation positioning information and the pre-configured power grid one-map data; and determine the high-value image types to be captured by the intelligent safety helmet according to the operation line where the intelligent safety helmet is located and send the high-value image types to be captured by the intelligent safety helmet to the intelligent safety helmet; the management server is also used to perform re-construction behavior identification based on the high-value target images uploaded by the intelligent safety helmet.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Triggering the determination of the high-value target image types to be captured by the intelligent safety helmet through the operation positioning information of the intelligent safety helmet can improve the accuracy of the monitoring method by increasing the effective data information in the power construction operation images. Moreover, adjusting the high-value target images in the captured power construction operation images for construction behavior identification can improve the accuracy of monitoring and the accuracy of secondary monitoring (secondary monitoring can be understood as re-construction behavior identification). Brief Description of the Drawings

[0017] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0018] Figure 1 It is a schematic flowchart of the safety monitoring method for distribution network operations in the embodiments of the present invention.

[0019] Figure 2 It is a schematic diagram of the scene image in the embodiments of the present invention.

[0020] Figure 3 It is a schematic diagram of image matching in the embodiments of the present invention.

[0021] Figure 4 It is a schematic diagram of the CA attention module in the embodiments of the present invention.

[0022] Figure 5 It is a schematic structural diagram of the reparameterized convolution module in the embodiments of the present invention.

[0023] Figure 6 It is a schematic structural diagram of the SF module in the embodiments of the present invention. Detailed Embodiments

[0024] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below through specific embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0025] As recorded in the background art, "when uploading or identifying violations of images in the distribution network operation scenario, a large amount of communication resources are used, resulting in waste of resources and low accuracy of construction behavior identification." In response to these problems, the embodiments of the present invention can reduce the amount of invalid data in the power construction operation images by determining the types of high-value target images to be captured by the intelligent safety helmet and adjusting the high-value target images in the captured power construction operation images, thereby reducing the use of communication resources and improving the monitoring accuracy by increasing the amount of valid data in the images. According to an embodiment of the present invention, a safety monitoring method for power construction operations is provided as Figure 1 shown. A safety monitoring method for power construction operations mainly includes two parts. The first part is: determining the types of high-value target images to be captured by the intelligent safety helmet; the second part is: adjusting the image acquisition direction of the intelligent safety helmet and adjusting the high-value target images in the captured power construction operation images; identifying the construction behavior of the adjusted high-value target images and uploading the identification results and the adjusted high-value target images.

[0026] Specifically, a safety monitoring method for electric power construction operations includes the following steps: S1. Obtain the operation positioning information of the intelligent safety helmet, and based on the obtained operation positioning information and the pre-configured power grid map data, determine the operation line where the intelligent safety helmet is located; and determine the type of high-value target image to be captured by the intelligent safety helmet based on the operation line where the intelligent safety helmet is located; S2. Collect the electric power construction operation images corresponding to the operation line determined in S1, and use the pre-trained target detection model to identify the type of high-value target image in the electric power construction operation images. When the type of high-value target image in the electric power construction operation images is inconsistent with the type of high-value image to be captured by the intelligent safety helmet, send a prompt message to guide the adjustment of the image acquisition direction of the intelligent safety helmet; and adjust the high-value target image in the captured electric power construction operation images; perform construction behavior recognition on the adjusted high-value target image, and upload the recognition result and the adjusted high-value target image for re-construction behavior recognition.

[0027] In the embodiment of the present invention, the type of high-value target image to be captured by the intelligent safety helmet is triggered and determined through the operation positioning information of the intelligent safety helmet, which can improve the accuracy of the monitoring method by increasing the effective data information in the electric power construction operation images. Moreover, adjusting the high-value target image in the captured electric power construction operation images for construction behavior recognition can improve the accuracy of monitoring and the accuracy of secondary monitoring. Among them, when the high-value target image label in the electric power construction operation images is inconsistent with the type of high-value image to be captured by the intelligent safety helmet, a prompt message can be sent through the alarm on the safety helmet to guide the adjustment of the image acquisition direction of the intelligent safety helmet.

[0028] To better illustrate the characteristics of the present invention, the implementation process of the present invention is described in detail below.

[0029] First, introduce the target detection model to be used. To meet the recognition effect of high-value targets in distribution network operations, according to the embodiment of the present invention, the target detection model uses a model of the yolo series for processing. For the convenience of explanation, the yolo v7 model is used as an example for exemplary explanation. The yolo v7 model adds an attention module and a re-parameterization module to the basic network structure of the target detection algorithm. An attention module (CA, Coordinate Attention module) is added after the ELAN module in the backbone network.

[0030] The attention module is the Coordinate Attention (CA) module. By weighting channel features, the CA attention module can encode horizontal and vertical position information into the channel attention, enabling the mobile network to focus on a wide range of position information without incurring excessive computational costs, thereby improving the target recognition effect. The structure of the CA attention module is as follows Figure 4 shown, including a residual layer, an average pooling layer, a concatenation layer (feature connection), a convolutional layer, a batch normalization layer, a non-linear layer, segmentation, an activation function layer, and a Re-weight layer. The data relationships and principles between each layer in the CA attention module are well-known to those skilled in the art and will not be elaborated here.

[0031] In addition, the reparameterization module is a sandwich fusion module with a large receptive field and low model parameters, which can effectively capture the spatial features of the lower layers of the backbone network, enhance the downsampling layers in the network backbone, improve the effectiveness of the downsampling process, and thus enhance the performance of object detection. The structure of the reparameterization module is as Figure 5 shown. (a) is the reparameterization structure during model training, including a convolutional layer, a batch normalization layer, and an activation function layer; (b) is the reparameterization structure during inference (including a convolutional layer and an activation function layer) to enhance the downsampling layers in the network backbone, improving the effectiveness of the downsampling process.

[0032] In addition, the YOLO v7 model can also extract spatial and channel features from the feature maps of three different layers of the network backbone through the Sandwich-fusion (SF) module. The SF structure is as Figure 6 shown, including a depthwise separable convolutional layer, a concatenation layer, and an upsampling layer, which balances the spatial information of low-level features and the semantic information of high-level features, enhancing the network head's recognition and classification of target positions.

[0033] The object detection model is trained in the following way:

[0034] (1) Data collection and annotation: Collect historical video data, including various normal and abnormal operation scenarios, to ensure sample diversity; use annotation tools (such as LabelImg) to annotate high-risk scenarios, including key information such as personnel and equipment, to generate a dataset in VOC format to construct a training set and a validation set, with the ratio of the training set to the validation set being 8:2.

[0035] (2) Model Training and Optimization: Initially train and test the model (the initial number of epochs is 300, batch-size is 32, and the learning rate adopts the recommended value); optimize the model performance by adjusting hyperparameters (learning rate, batch size, etc.) and using data augmentation techniques (such as rotation, flipping, scaling); finally, conduct cross-validation to evaluate the performance of the model on the validation set to ensure high recognition rate and low false alarm rate.

[0036] (3) Actual Environment Testing: Conduct a comprehensive system test in the actual operation environment of the distribution network, observe the working state of the system, and particularly pay attention to the recognition rate, tracking accuracy, and alarm accuracy; collect feedback information during the test, including the usage experience of operation and maintenance personnel and system operation data.

[0037] (4) Algorithm and System Optimization: Analyze the deficiencies of the system based on the test results, adjust the algorithm and system parameters to improve the monitoring effect; regularly update the model and incorporate new data and scenarios to ensure that the system continuously adapts to environmental changes.

[0038] During the above training process, the loss function and convergence conditions adopted are understood by those skilled in the art and will not be elaborated here.

[0039] The following provides a detailed explanation of a power construction operation safety monitoring method provided by the present invention. In the safety monitoring method, according to an embodiment of the present invention, in the S1, the operation line where the intelligent safety helmet is located is determined through the following steps: Based on the operation positioning information of the intelligent safety helmet, obtain the pole tower positions within the preset range where the operation positioning information is located in the power grid map data, and calculate the distance from each pole tower position within the preset range to the operation positioning information of the intelligent safety helmet; The operation line where the pole tower position with the minimum distance to the operation positioning information of the intelligent safety helmet among the obtained pole tower positions within the preset range is located is used as the operation line where the intelligent safety helmet is located.

[0040] It should be noted that in the power system, there are a grid map system, an operation plan system, a work ticket system, a power monitoring system, a distribution automation system, and a dispatching order system. Among them, the ledger information can be obtained from the grid map data in the grid map system, and the ledger information includes the operation line name, pole number, location information of the operation line and the pole, etc.; the plan information can be obtained from the operation plan system, the work ticket system, and the dispatching order system, and the plan information includes the operation line name, the pole numbers to be operated, the number of grounding short-circuit devices required for the operation line, pre-power transmission information, etc. By taking the operation line where the pole position with the smallest operation positioning information from the intelligent safety helmet is located as the operation line where the intelligent safety helmet is located, the determination of the location of the intelligent safety helmet for monitoring the construction site can be improved. For example, if there are three pole positions within 20 meters (preset range) of the operation positioning information of the intelligent safety helmet, calculate the distances from these three poles to the operation positioning information of the intelligent safety helmet, and take the operation line where the pole position corresponding to the smallest of the three obtained positioning information distances is located as the operation line where the intelligent safety helmet is located.

[0041] Further, in S1 of the safety monitoring method, the high-value target image type to be captured by the intelligent safety helmet is determined through the following steps: query the operation plan system according to the identification information of the operation line to obtain the operation plan of the operation line; match the keywords in the operation plan of the operation line in the pre-configured high-value target relationship table to obtain the high-value target image type to be captured by the intelligent safety helmet. Preferably, first detect whether there is an operation plan for the operation line where the intelligent safety helmet is located. When there is an operation plan, then determine the high-value target image type to be captured by the intelligent safety helmet, and subsequently adjust the high-value target images in the captured power construction operation images.

[0042] It should be noted that determining the high-value target image type to be captured by the intelligent safety helmet through keywords can make the high-value targets in the power construction operation images more suitable for the current operation plan of the safety helmet.

[0043] A specific example of the high-value target relationship table is shown in Table 1 below. In addition, classifying the high-value targets facilitates subsequent tracking and identifying the targets according to their characteristics. Among them, the high-value target images include a moving target area and a static target area. The dynamic targets (also called sub-target areas) in the moving target area are the targets that move in the camera frame, such as the operating personnel wearing safety belts, etc.; the static target area is the sub-target area or static target that does not move in the camera frame, such as electric poles, iron towers, iron tower anchor bolts, etc. In addition, the scene image only needs to be acquired once at the beginning of the distribution network operation monitoring task, and no data acquisition will be performed during the entire monitoring process thereafter.

[0044] Table 1 High-Value Target Relationship Table

[0045]

[0046] In the safety monitoring method, the power construction operation images corresponding to the operation lines determined in S1 are collected through an intelligent safety helmet, and the high-value target image types in the power construction operation images are identified by using a pre-trained target detection model. When the high-value target image types in the power construction operation images are inconsistent with the high-value image types to be captured by the intelligent safety helmet, a prompt message is sent to guide the adjustment of the image acquisition direction of the intelligent safety helmet.

[0047] It should be noted that when the high-value target image types in the power construction operation images are inconsistent with the high-value image types to be captured by the intelligent safety helmet, it can be understood that when the number of target labels recognized in the power construction operation images is less than the total number of high-value target image labels to be captured by the intelligent safety helmet. In a specific scenario, it can be understood that the pan-tilt installed on the automatic adjustment camera in the intelligent safety helmet cannot obtain the required high-value target image types within a preset time, and a voice prompt is given through the alarm in the intelligent safety helmet, and the on-site person in charge adjusts the overall shooting orientation of the intelligent safety helmet to make the monitoring more accurate and improve the interactivity between the monitoring method and the on-site personnel.

[0048] Furthermore, there are often multiple high-value targets in the distribution network operation scenario. A single camera cannot effectively handle the tracking and recognition of multiple targets. Moreover, due to the differences between dynamic targets and static targets, in order to reduce problems such as the omission of key information, waste of hardware and software resources caused by the camera switching back and forth between dynamic targets and static targets, and at the same time improve the tracking and recognition efficiency of high-value targets in the operation scenario, according to the embodiments of the present invention, one camera is responsible for tracking dynamic targets, and another camera is responsible for tracking static targets. Through this independent operation mode, the problem of resource waste caused by resource allocation can be effectively reduced. Since the positions of dynamic targets move during the operation process, while static targets hardly move, the execution frequency of the dynamic target tracking mode is higher than that of the static target tracking mode (that is, within the same time period, the number of execution cycles of the dynamic target tracking mode is more than that of the static target tracking mode). According to the embodiments of the present invention, the frequency of dynamic target tracking is 100 times greater than that of static target tracking. Specifically, according to the embodiments of the present invention, the frequency of dynamic target tracking is once every 5 seconds, and the frequency of static target tracking is once every 10 minutes.

[0049] According to an embodiment of the present invention, in S2 of the security monitoring method, both the moving area and the static area include multiple sub-target areas; the adjusted high-value target image is obtained in the following manner: for the position of each sub-target area, the perspective and focal length of the camera are adjusted to obtain the adjusted sub-target area image, and all the adjusted sub-target area images are used as the adjusted high-value target image; wherein, the sub-target area that is at the center of the current screen and the size of the sub-target area reaches a preset ratio relative to the size of the current screen is used as the adjusted sub-target area image. Preferably, for the position of each sub-target area, the perspective and focal length of the camera are adjusted through the following steps: calculating the deviation value based on the center position of the target box of the sub-target area and the center position of the current screen; moving the camera angle based on the deviation value so that the two centers coincide; calculating the ratio of the target box size to the current screen size at this time, and comparing it with the preset ratio range. When both the horizontal ratio and the vertical ratio are within the preset ratio range, stop adjusting the focal length of the camera.

[0050] According to an example of the present invention, the steps of adjusting the perspective and focal length of the camera are as follows.

[0051] 1. Use a pre-trained object detection model to obtain the target of interest in the screen, and calculate the deviation value (x, y) between the center position of the target box (sometimes also called the sub-target area) and the center position of the current screen, as Figure 2 shown, where p1 is the center point of the target box, with the value (x1, y1); p2 is the center point of the image, with the value (x2, y2). Then the deviation value between the center position of the target box and the center position of the image is (x1 - x2, y1 - y2);

[0052] 2. Determine the rotation angle of the camera based on the deviation value (x, y). Regarding the x value, if the x value of the deviation value is positive, the camera moves horizontally to the right by x pixel values; if the x value of the deviation value is negative, the camera moves horizontally to the left by x pixel values; regarding the y value, if the y value of the deviation value is positive, the camera moves vertically down by y pixel values; if the y value of the deviation value is negative, the camera moves vertically up by y pixel values;

[0053] 3. After moving the center of the camera screen to the center position of the target box (as Figure 3 shown), judge the size of the target box in the screen. Taking the image size of 1280*720 as an example, the range of d1 (the distance from the target box in the vertical direction to the image edge) of the adjusted target box is best between [50, 200], and the range of d2 (the distance from the target box in the horizontal direction to the image edge) is best between [50, 300] (for images of other sizes, scale the ranges of d1 and d2 proportionally).

[0054] In addition, since it is difficult to determine the focal length of the camera and the distance to the target, the magnification of the camera is uncertain. Therefore, in this solution, the camera is magnified or reduced by the smallest unit multiple, and the d1 and d2 values of the target box and the image are judged in turn. When d1 and d2 reach the optimal range, it is okay.

[0055] Based on the above content, the following are implementation examples of two target tracking modes (execution steps of one recognition cycle).

[0056] The logical implementation steps for executing the dynamic target tracking mode once are as follows:

[0057] 1. The recognition task starts to execute. The dynamic target recognition camera rotates 360 degrees to inspect the operation scene in the picture. The target recognition algorithm makes a preliminary judgment on the camera picture and locks the area containing the dynamic target in the operation scene picture.

[0058] 2. Determine the number of targets to be tracked and recognized in the dynamic target area, and record the position information of each target in the picture.

[0059] 3. Adjust the viewing angle for each dynamic target so that the target is in the center of the camera picture and the size of the target is in a suitable proportion to the size of the picture. Lock each dynamic target after the viewing angle adjustment for subsequent processing.

[0060] In addition, the implementation steps for executing the static target tracking mode once are as follows:

[0061] 1. The recognition task starts to execute. The static target recognition camera rotates 360 degrees to inspect the operation scene in the picture. The target recognition algorithm makes a preliminary judgment on the camera picture and locks the static target in the operation scene picture.

[0062] 2. Judge the static target, record the position information of the static target, and judge the size and position of the static target in the picture. If the position and size are not appropriate, the viewing angle needs to be adjusted (the processing mode is the same as that of the dynamic target viewing angle adjustment).

[0063] 3. Lock the static target and transmit the picture into the target recognition algorithm for recognition.

[0064] Furthermore, in the safety monitoring method, construction behavior recognition is also performed on the adjusted high-value target images, and the recognition results and the adjusted high-value target images are uploaded for re-construction behavior recognition. In specific applications, the intelligent safety helmet is configured at the ground supervisor. Construction behavior recognition can be performed on the adjusted high-value target images in the intelligent safety helmet. When the result of the construction behavior recognition of the high-value target image is a violation, a voice prompt is issued, such as prompting that the operation of moving target 1 is unsafe.

[0065] Considering that uploading all target scene images (sometimes also referred to as power construction operation images) to a cloud system such as a management server will cause waste of energy consumption and bandwidth, and will increase the workload of staff, the present invention sorts according to the priority of target information (sometimes also referred to as high-value target image types), and uploads high-priority targets in real time to avoid losing important information; while low-priority targets will be selectively uploaded (for example: a total of 5 low-priority target information are recognized, and the target information with the highest recognition rate is selected for upload), to avoid redundancy of too much identical information. When uploading the target image, the system will judge the target image to be uploaded to ensure that the uploaded target image has high clarity and good picture quality, and avoid the situation of uploading blurred pictures.

[0066] When the management server conducts re-construction behavior recognition, it can identify the uploaded high-value target images through a pre-configured target recognition large model. When the uploaded high-value target image belongs to an illegal operation, the illegal operation warning information will be sent to the intelligent safety helmet that uploaded the high-value target image. The target recognition large model can be pre-trained through a public data set and fine-tuned through the above data set. The re-construction behavior recognition can also be secondarily monitored by remote staff, especially for high-value target images belonging to illegal operations, to reduce the work intensity.

[0067] On the above basis, the present invention also provides an intelligent safety helmet, including a controller and a camera group, a positioning module, a communication module, and an alarm connected to the controller. The camera group is used to collect power construction operation images; the controller is used to control the communication module to send the operation positioning information of the intelligent safety helmet determined by the positioning module, and control the communication module to receive the high-value image types to be photographed by the intelligent safety helmet and the operation line where the intelligent safety helmet is located determined by the management server based on the operation positioning information of the intelligent safety helmet; the controller also collects power construction operation images corresponding to the operation line where the safety helmet is located through the camera group; and uses a pre-trained target detection model to identify the high-value target image types in the power construction operation images. When the high-value target image types in the power construction operation images are inconsistent with the high-value image types to be photographed by the intelligent safety helmet, a prompt message is sent to guide the adjustment of the image acquisition direction of the intelligent safety helmet; and the high-value target images in the photographed power construction operation images are adjusted; the construction behavior of the adjusted high-value target images is recognized, and the recognition results and the adjusted high-value target images are uploaded for re-construction behavior recognition.

[0068] On the above basis, the embodiment of the present invention also provides a power construction operation safety monitoring device, and the monitoring device includes: a management server and any one of the above intelligent safety helmets;

[0069] The management server is used to obtain the operation positioning information of the intelligent safety helmet, and determine the operation line where the intelligent safety helmet is located based on the obtained operation positioning information and the pre-configured power grid single map data; and determine the high-value image type to be captured by the intelligent safety helmet according to the operation line where the intelligent safety helmet is located and send the high-value image type to be captured by the intelligent safety helmet to the intelligent safety helmet; the management server is also used to perform re-construction behavior recognition based on the high-value target images uploaded by the intelligent safety helmet.

[0070] The main working principles of the intelligent safety helmet and the safety monitoring device have been described in the safety monitoring method, and will not be elaborated here.

[0071] The present invention triggers the determination of the high-value target image type to be captured by the intelligent safety helmet through the operation positioning information of the intelligent safety helmet, and can improve the accuracy of the monitoring method by increasing the effective data information in the power construction operation images. Furthermore, adjusting the high-value target images in the captured power construction operation images for construction behavior recognition can improve the accuracy of monitoring and the accuracy of secondary monitoring. Among them, by adjusting the high-value target images in the power construction operation images, the utilization of communication resources can be reduced, and the monitoring efficiency can be improved.

[0072] The present invention can be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0073] The computer-readable storage medium can be a tangible device that holds and stores instructions for use by an instruction execution device. The computer-readable storage medium may include, for example, but is not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (non-exhaustive list) of the computer-readable storage medium include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanically encoded devices such as punch cards or raised structures in grooves having instructions stored thereon, and any suitable combination of the foregoing.

[0074] The embodiments of the present invention have been described above. The above description is exemplary and not exhaustive, and is also not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, the practical application, or the improvement of technologies in the market, or to enable other ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A safety monitoring method for electric power construction operations, characterized in that, It includes the following steps: S1. Obtain the operation positioning information of the intelligent safety helmet, and based on the obtained operation positioning information and the pre-configured power grid single map data, determine the operation line where the intelligent safety helmet is located through the following steps: Based on the operation positioning information of the intelligent safety helmet, obtain the pole tower positions within the preset range where the operation positioning information is located in the power grid single map data, and calculate the distances from each pole tower position within the preset range to the operation positioning information of the intelligent safety helmet; take the operation line where the pole tower position with the minimum distance to the operation positioning information of the intelligent safety helmet among the obtained pole tower positions within the preset range as the operation line where the intelligent safety helmet is located; And based on the operation line where the intelligent safety helmet is located, determine the high-value target image type to be captured by the intelligent safety helmet through the following steps: Query the operation plan system according to the identification information of the operation line to obtain the operation plan of the operation line; match the keywords in the operation plan of the operation line in the pre-configured high-value target relationship table to obtain the high-value target image type to be captured by the intelligent safety helmet; S2. Collect the power construction operation images corresponding to the operation line determined in S1, and use the pre-trained target detection model to identify the high-value target image type in the power construction operation images. When the high-value target image type in the power construction operation images is inconsistent with the high-value image type to be captured by the intelligent safety helmet, send a prompt message to guide the adjustment of the image acquisition direction of the intelligent safety helmet; And adjust the high-value target images in the captured power construction operation images; perform construction behavior recognition on the adjusted high-value target images, and upload the recognition results and the adjusted high-value target images for re-construction behavior recognition.

2. The monitoring method according to claim 1, wherein In S2, the high-value target images in the power construction operation images include a moving target area and a static target area, and both the moving target area and the static target area include multiple sub-target areas; Obtain the adjusted high-value target images through the following method: For the position of each sub-target area, obtain the adjusted sub-target area image by adjusting the perspective and focal length of the camera, and use all the adjusted sub-target area images as the adjusted high-value target images; among them, take the sub-target area that is at the center of the current screen and the size of the sub-target area reaches a preset ratio relative to the size of the current screen as the adjusted sub-target area image.

3. The monitoring method according to claim 2, characterized in that, For the position of each sub-target area, adjust the perspective and focal length of the camera through the following steps: Calculate the deviation value based on the center position of the target box of the sub-target area and the center position of the current screen; Move the camera angle based on the deviation value so that the two centers coincide; Calculate the ratio of the target box size to the current screen size at this time, and compare it with the preset ratio range. When both the horizontal ratio and the vertical ratio in the ratio are within the preset ratio range, stop adjusting the focal length of the camera.

4. The monitoring method according to claim 1, wherein In S2, when the result of the construction behavior recognition of the adjusted high-value target image is a violation, a voice prompt is issued.

5. The monitoring method according to claim 1, characterized in that, In S2, the re - construction behavior recognition is performed in the following manner: Using a pre - configured target recognition large - model, recognize the construction behavior of the uploaded high - value target images. When the construction behavior of the uploaded high - value target images belongs to a violation, send the violation operation warning information to the intelligent safety helmet that uploaded the high - value target image.

6. An intelligent safety helmet, comprising a controller, a camera group, a positioning module, a communication module and an alarm connected to the controller, characterized in that: Apply the monitoring method according to any one of claims 1 - 5, wherein, The camera group is used to collect power construction operation images; The controller is used to control the communication module to send the operation positioning information of the intelligent safety helmet determined by the positioning module, and control the communication module to receive the high - value image type to be captured by the intelligent safety helmet and the operation line where the intelligent safety helmet is located determined by the management server based on the operation positioning information of the intelligent safety helmet; The controller also collects the power construction operation images corresponding to the operation line where the safety helmet is located through the camera group; and uses a pre - trained target detection model to identify the high - value target image type in the power construction operation images. When the high - value target image type in the power construction operation images is inconsistent with the high - value image type to be captured by the intelligent safety helmet, send a prompt message to guide the adjustment of the image acquisition direction of the intelligent safety helmet; And adjust the high - value target images in the captured power construction operation images; perform construction behavior recognition on the adjusted high - value target images, and upload the recognition result and the adjusted high - value target images for re - construction behavior recognition.

7. A safety monitoring device for electric power construction operations, characterized in that, The monitoring device includes: A management server and the intelligent safety helmet according to claim 6; The management server is used to obtain the operation positioning information of the intelligent safety helmet, and based on the obtained operation positioning information and the pre - configured power grid single - map data, determine the operation line where the intelligent safety helmet is located; and determine the high - value image type to be captured by the intelligent safety helmet according to the operation line where the intelligent safety helmet is located and send the high - value image type to be captured by the intelligent safety helmet to the intelligent safety helmet; The management server is also used to perform re - construction behavior recognition based on the high - value target images uploaded by the intelligent safety helmet.

Images