Operation site safety supervision method, system, equipment and medium

The method uses edge recognition and deep learning models with AI law enforcement instruments and smart tools to enhance the accuracy and timeliness of safety supervision at work sites, addressing the limitations of manpower-based supervision.

CN119068417APending Publication Date: 2024-12-03HAINAN POWER GRID CO LTD

Patent Information

Application Number
CN202411227987.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing methods for supervising the safety of work sites rely on manpower, leading to supervision blind spots and difficulty in fully and timely controlling the on-site situation, resulting in low accuracy of safety supervision results.

Method used

A method involving edge recognition, intelligent tool operation data analysis, and deep learning models to identify illegal behaviors, using AI law enforcement instruments, smart tools, and wireless ad hoc networks for real-time supervision.

Benefits of technology

The method effectively avoids supervision blind spots, ensures timely control of the on-site situation, and improves the accuracy of safety supervision results by identifying illegal behaviors in real-time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068417A_ABST
    Figure CN119068417A_ABST
Patent Text Reader

Abstract

The invention discloses an operation site safety supervision method, system and device and a medium, and relates to the technical field of power systems. According to the supervision method, an initial monitoring picture and intelligent tool operation data of an operation site are acquired, edge identification is performed on initial intelligent tools corresponding to the initial monitoring picture based on the intelligent tool operation data, and an operation supervision area is generated; adjusting a monitoring angle corresponding to the initial monitoring picture according to the operation supervision area, and taking the collected monitoring picture as a target monitoring picture; and performing violation behavior identification based on the target monitoring picture, and generating operation site safety supervision data. Through carrying out edge identification on the initial intelligent tool corresponding to the initial monitoring picture, the monitoring angle is adjusted in time, the problem of a supervision blind area is avoided, and the field condition is comprehensively mastered. By performing violation behavior recognition on the target monitoring picture, violation behaviors can be found in time, the personal safety of operators is guaranteed, and the accuracy of a safety supervision result is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, system, equipment and medium for safety supervision at a work site. Background Art

[0002] The operational activities of power marketing on-site operations specifically refer to installing electricity meters, connecting electricity, inspecting electricity usage, business expansion surveys, etc. For power companies, only ensuring the safety of equipment, power grid and personal safety is the primary condition for their development. Therefore, safety control of power marketing on-site operations is particularly important.

[0003] With the rapid development of social economy, the safety of on-site operations in electrical marketing has also attracted people's attention. At present, the on-site environment of electrical marketing operations is complex and diverse, and the carriers of violation judgment are easily confused with the background. The existing on-site safety supervision methods rely on manpower for supervision, which has low supervision efficiency and high labor costs. It is easy to have supervision blind spots, and it is difficult to fully and timely control the on-site situation, resulting in low accuracy of safety supervision results. Summary of the invention

[0004] The present invention provides a method, system, equipment and medium for supervising the safety of work sites, which solves the technical problems that the existing methods for supervising the safety of work sites rely on manpower for supervision, are prone to blind spots in supervision, and are difficult to fully and timely control the on-site situation, resulting in low accuracy of safety supervision results.

[0005] The present invention provides a method for safety supervision of a work site, comprising:

[0006] Acquire an initial monitoring screen and intelligent tool operation data of the work site, perform edge recognition on an initial intelligent tool corresponding to the initial monitoring screen based on the intelligent tool operation data, and generate an operation supervision area;

[0007] Adjusting the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and using the collected monitoring screen as the target monitoring screen;

[0008] Based on the target monitoring screen, illegal behaviors are identified to generate work site safety supervision data.

[0009] Optionally, the step of performing edge recognition on the initial smart tool corresponding to the initial monitoring screen based on the smart tool operation data to generate an operation supervision area includes:

[0010] Selecting an initial smart tool corresponding to the operation data of the smart tool in the initial monitoring screen to generate a target smart tool;

[0011] Extracting features from the initial monitoring screen based on the position data corresponding to the target intelligent tool to generate a plurality of original feature points;

[0012] Selecting feature points in the initial monitoring image corresponding to the original feature points to generate a plurality of matching feature points;

[0013] Segmenting the initial monitoring picture based on the matching feature points and performing grayscale processing to generate a plurality of target grayscale unit blocks;

[0014] The region boundaries of the target grayscale unit blocks are connected respectively to generate a job supervision region.

[0015] Optionally, the step of selecting an initial smart tool corresponding to the smart tool operation data in the initial monitoring screen to generate a target smart tool includes:

[0016] When the operation state corresponding to the operation data of the smart tool is an electrical test operation, the smart electrical tester and insulating gloves in the initial monitoring screen are used as target smart tools;

[0017] When the operation state corresponding to the operation data of the smart tool is grounding operation, the smart grounding wire and insulating gloves in the initial monitoring screen are used as target smart tools;

[0018] When the operation state corresponding to the operation data of the smart tool is that the grounding wire is abnormal, the smart grounding wire and the insulating gloves in the initial monitoring screen are used as target smart tools;

[0019] When the operation state corresponding to the operation data of the intelligent tool is a climbing operation, determining whether the safety belt identification data corresponding to the operation data of the intelligent tool is a preset identification data;

[0020] If the seat belt identification data is the preset identification data, the smart bracelet and the seat belt in the initial monitoring screen are used as the target smart tools;

[0021] If the seat belt identification data is not the preset identification data, the smart bracelet, seat belt and escalator in the initial monitoring screen are used as target smart tools.

[0022] Optionally, the step of extracting features from the initial monitoring screen based on the position data corresponding to the target intelligent tool to generate a plurality of original feature points includes:

[0023] The initial monitoring picture is smoothed multiple times according to the position data and preset resolution data corresponding to the target intelligent tool, so as to construct an image pyramid corresponding to the initial monitoring picture;

[0024] Selecting feature points corresponding to image edges of each image in the image pyramid respectively to generate a plurality of image feature points;

[0025] The tangent directions corresponding to the image feature points on the image edge are respectively used as direction vectors of the image feature points to generate original feature points corresponding to the image feature points.

[0026] Optionally, the step of segmenting the initial monitoring picture based on the matching feature points and performing grayscale processing to generate a plurality of target grayscale unit blocks includes:

[0027] Segmenting the initial monitoring image according to all the matching feature points and the preset segmentation size to generate a plurality of unit blocks;

[0028] Performing grayscale processing on the unit blocks respectively to generate a plurality of initial grayscale unit blocks;

[0029] Comparing the grayscale values ​​of the initial grayscale unit block with those of adjacent unit blocks respectively, and generating a plurality of comparison data corresponding to the initial grayscale unit block;

[0030] When the comparison data has a preset comparison threshold, the initial grayscale unit block is used as a target grayscale unit block.

[0031] Optionally, the step of identifying illegal behaviors based on the target monitoring screen and generating work site safety supervision data includes:

[0032] Using the work site violation video image set corresponding to the work site for classification and annotation, and building a typical violation behavior sample library;

[0033] Using the typical violation behavior sample library to train the initial deep learning model to generate a target deep learning model;

[0034] Violation behaviors are identified on the target monitoring screen through the target deep learning model to generate work site safety supervision data.

[0035] Optionally, the step of identifying illegal behaviors on the target monitoring screen through the target deep learning model to generate work site safety supervision data includes:

[0036] Performing personal protection identification of the operating personnel in the target monitoring screen by using the target deep learning model and the personal protection standard corresponding to the target monitoring screen to generate personal protection identification data;

[0037] When the personal protection identification data is abnormal, constructing first work site safety supervision data using the personal protection identification data;

[0038] When the personal protection identification data is normal, the target deep learning model is used to identify the working state of the intelligent tool corresponding to the target monitoring screen, and the working state data of the tool is generated;

[0039] When the tool working state data is abnormal, constructing second work site safety supervision data using the tool working state data;

[0040] When the working state data of the device is normal, the operating behavior in the target monitoring screen is compared with the corresponding typical violation behavior through the target deep learning model to generate behavior comparison data;

[0041] When the behavior comparison data are the same, the behavior comparison data are used to construct the third work site safety supervision data;

[0042] When the behavior comparison data are different, the fourth work site safety supervision data is set to no illegal behavior at the work site.

[0043] The present invention also provides a work site safety supervision system, comprising:

[0044] An operation supervision area generation module is used to obtain an initial monitoring screen and intelligent tool operation data of the operation site, perform edge recognition on the initial intelligent tool corresponding to the initial monitoring screen based on the intelligent tool operation data, and generate an operation supervision area;

[0045] A target monitoring screen determination module, used to adjust the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and use the collected monitoring screen as the target monitoring screen;

[0046] The operation site safety supervision data generation module is used to identify illegal behaviors based on the target monitoring screen and generate operation site safety supervision data.

[0047] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of implementing any of the above-mentioned methods for safety supervision of a work site.

[0048] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, the computer program implements any of the above-mentioned methods for safety supervision of a work site.

[0049] It can be seen from the above technical solutions that the present invention has the following advantages:

[0050] The present invention performs edge recognition on the initial intelligent tool corresponding to the initial monitoring screen, adjusts the monitoring angle in time, avoids the problem of supervision blind spots, and fully controls the on-site situation. By identifying illegal behaviors on the target monitoring screen acquired after adjusting the monitoring angle, illegal behaviors can be discovered in time, the personal safety of the operators can be guaranteed, and the accuracy of the safety supervision results can be improved. This solves the technical problem that the existing on-site operation safety supervision method relies on manpower for supervision, which is prone to supervision blind spots, making it difficult to fully and timely control the on-site situation, resulting in low accuracy of safety supervision results. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0052] Figure 1 A flowchart of a method for safety supervision of a work site provided in the first embodiment of the present invention;

[0053] Figure 2 A structural block diagram of a visual monitoring system provided in Embodiment 1 of the present invention;

[0054] Figure 3 An information table corresponding to the information stored in the smart electrical tester, smart grounding wire, and smart bracelet provided in the first embodiment of the present invention;

[0055] Figure 4 A flowchart of a method for safety supervision of a work site provided in Embodiment 2 of the present invention;

[0056] Figure 5 A schematic diagram of the structure of an image pyramid provided in the second embodiment of the present invention;

[0057] Figure 6 A structural block diagram of a work site safety supervision system provided by Embodiment 6 of the present invention;

[0058] Figure 7 This is a structural block diagram of an electronic device provided in Embodiment 7 of the present invention. DETAILED DESCRIPTION

[0059] The embodiments of the present invention provide a method, system, device and medium for supervising the safety of work sites, which are used to solve the technical problems that the existing methods for supervising the safety of work sites rely on manpower for supervision, are prone to supervision blind spots, and are difficult to fully and timely control the site situation, resulting in low accuracy of safety supervision results.

[0060] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] Embodiment 1

[0062] See also Figure 1 , Figure 1 A flowchart of the steps of a work site safety supervision method provided in Example 1 of the present invention.

[0063] like Figure 2 As shown, the present invention can be applied to a visual supervision system constructed by an AI law enforcement instrument, intelligent tools and video terminals. The intelligent tools interact with the AI ​​law enforcement instrument through a wireless ad hoc network. After the intelligent tools are started, instructions are transmitted to the AI ​​law enforcement instrument. After receiving the instructions, the AI ​​law enforcement instrument identifies the on-site violations. The video terminal communicates with the AI ​​law enforcement instrument through the Internet of Things, and then visually supervises the work site. The video terminal can not only display the monitoring screen collected by the AI ​​law enforcement instrument, but also display the on-site safety supervision data obtained after the monitoring screen identifies the violations, so that the supervisor can timely control the on-site situation.

[0064] The AI ​​body camera is a combination of a general law enforcement recorder and a computer. It can replace the human eye to identify, track and measure targets, and further process images to make the processed images more suitable for human observation. The AI ​​body camera has image recognition capabilities and can process, analyze and understand images to identify targets and objects in various different modes. On this basis, the AI ​​body camera is selected to identify violations on site.

[0065] The AI ​​law enforcement device is used to record the work site and obtain the initial monitoring screen. It can also be used to perform edge recognition on the initial intelligent tools corresponding to the initial recording screen and the initial monitoring screen, adjust the recording angle so that the intelligent tools are in the preset area of ​​the recording screen, and obtain the target recording screen, that is, the target monitoring screen. It can also be used to perform image recognition on the target monitoring screen through target deep learning methods to identify violations on site.

[0066] Smart tools are equipped with smart sensor technology to obtain the motion status of smart tools and operators at the work site. When the smart tools are started, the motion status of the smart tools and operators, that is, the operation data of the smart tools, are sent to the AI ​​law enforcement instrument. Smart sensors are sensors with information processing functions. Smart sensors are equipped with microprocessors and have the ability to collect, process, and exchange information. They are the product of the combination of sensor integration and microprocessors. Compared with general sensors, smart sensors have the following three advantages: high-precision information collection can be achieved through software technology at a low cost; they have certain programming automation capabilities; and they have diverse functions.

[0067] Intelligent tools can realize automatic data collection, analysis and processing. They can sense various physical quantities in the surrounding environment and transmit the sensed information to computers or other controllers in a digital way.

[0068] Optionally, the smart tool is equipped with a Bluetooth module to communicate with the AI ​​law enforcement instrument, and is also equipped with an RFID tag to synchronize the information of the smart tool to the AI ​​law enforcement instrument. The smart tool communicates with the AI ​​law enforcement instrument through wireless Bluetooth technology, and is also equipped with an RFID (Radio Frequency Identification) tag to synchronize the information of the smart tool to the AI ​​law enforcement instrument. The smart tool can be a smart tester, a smart grounding wire, a smart bracelet, personal protective equipment worn by operators, etc. The information stored in the smart tester, smart grounding wire, and smart bracelet is as follows: Figure 3 Specifically, for the two on-site operations of electrical testing and grounding and climbing, the personal protective equipment worn by the workers includes: insulating gloves, escalators, safety belts, etc. Among them, the escalators are non-slip escalators and the safety belts are five-point safety belts. These personal protective equipment can be equipped with RFID tags to assist the AI ​​law enforcement device in intelligent identification.

[0069] Since smart tools are not a single device, each smart tool needs to be linked to the AI ​​body camera through wireless ad hoc networking. Wireless ad hoc networking is a special self-organizing, peer-to-peer, multi-hop, wireless mobile network that does not rely on preset infrastructure and has the characteristics of temporary networking, rapid deployment, no control center, and strong anti-destruction.

[0070] As a remote monitoring platform, the video terminal communicates with the AI ​​law enforcement device through the Internet of Things to conduct visual real-time supervision of the work site. As a front-end acquisition device, the AI ​​law enforcement device 10 uploads the collected video to the transmission network. The video terminal reads the video according to the video transmission protocol. The video terminal can be a smart phone, computer monitor, tablet and other networked display device. The AI ​​law enforcement device can also choose wireless video transmission technology, use radio waves as the transmission medium, and build a transmission link in the air to transmit video, sound, data and other signals.

[0071] The video terminal monitors the target monitoring screen and the safety supervision data of the work site through the Internet of Things. As the front-end acquisition device, the AI ​​law enforcement instrument uploads the collected video and the safety supervision data of the work site obtained after the violation behavior is identified to the transmission network. The video terminal reads the video and the safety supervision data of the work site according to the video transmission protocol. The video terminal can be a smart phone, computer monitor, tablet and other networked display devices. The AI ​​law enforcement instrument can also choose wireless video transmission technology, use radio waves as the transmission medium, and build a transmission link in the air to transmit video, sound, data and other signals. Through the video terminal, the supervisor can remotely manage it, understand the safety and efficiency of the on-site operation in real time, and provide solutions for the on-site staff when encountering technical obstacles on the site. Supervisors can evaluate the on-site workers in response to the violations on the site, thereby strengthening the safety management and training of power marketing workers, improving their understanding and awareness of violations, and reducing the occurrence of violations.

[0072] A method for safety supervision of a work site provided in Example 1 of the present invention comprises:

[0073] Step 101: Acquire the initial monitoring screen and intelligent tool operation data of the work site, perform edge recognition on the initial intelligent tool corresponding to the initial monitoring screen based on the intelligent tool operation data, and generate an operation supervision area.

[0074] In an embodiment of the present invention, the work site is recorded by an AI law enforcement instrument to obtain an initial monitoring screen. The intelligent tools are used to obtain the motion status of the intelligent tools and workers on the site. When the intelligent tools are started, the motion status of the intelligent tools and workers is sent to the AI ​​law enforcement instrument. That is, each intelligent tool on the work site synchronizes its corresponding operation data and the collected worker motion data to the AI ​​law enforcement instrument through a Bluetooth module or RFID tag, thereby constructing the operation data of the intelligent tools.

[0075] Edge computing is to provide edge intelligent services nearby on the edge of the network near the source of objects or data through a distributed open platform that integrates the core capabilities of network, computing, storage, and application. Simply put, edge computing is to analyze the data collected from the terminal directly in the local device or network near the data generation, without transmitting the data to the cloud data processing center. The specific edge recognition process is: select the initial intelligent tools corresponding to the operation data of the intelligent tools in the initial monitoring screen to generate the target intelligent tools. Extract features from the initial monitoring screen based on the position data corresponding to the target intelligent tools to generate multiple original feature points. Select feature points corresponding to the original feature points in the initial monitoring screen to generate multiple matching feature points. Segment the initial monitoring screen based on the matching feature points and perform grayscale processing to generate multiple target grayscale unit blocks. Connect the regional boundaries of the target grayscale unit blocks respectively to generate the operation supervision area.

[0076] Step 102: adjust the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and use the collected monitoring screen as the target monitoring screen.

[0077] In an embodiment of the present invention, the recording angle of the AI ​​body camera is adjusted according to the operation supervision area, thereby changing the monitoring angle corresponding to the initial monitoring screen, and the monitoring screen collected by the monitoring angle is used as the target monitoring screen.

[0078] Step 103: Identify illegal behaviors based on the target monitoring screen and generate work site safety supervision data.

[0079] In the embodiment of the present invention, the set of video images of violations at the work site corresponding to the work site is used for classification and annotation to build a typical violation behavior sample library. The initial deep learning model is trained using the typical violation behavior sample library to generate a target deep learning model. Violations are identified based on the target monitoring screen to generate work site safety supervision data.

[0080] In an embodiment of the present invention, an operation supervision area is generated by obtaining the initial monitoring screen and the operation data of the intelligent tool at the operation site, and edge recognition is performed on the initial intelligent tool corresponding to the initial monitoring screen based on the operation data of the intelligent tool. The monitoring angle corresponding to the initial monitoring screen is adjusted according to the operation supervision area, and the collected monitoring screen is used as the target monitoring screen. Based on the target monitoring screen, illegal behaviors are identified to generate safety supervision data for the operation site. By edge recognition of the initial intelligent tool corresponding to the initial monitoring screen and timely adjustment of the monitoring angle, the problem of supervision blind spots is avoided, and the on-site situation is fully and timely controlled. By identifying illegal behaviors on the target monitoring screen, illegal behaviors can be discovered in time, the personal safety of the operators is guaranteed, and the accuracy of the safety supervision results is improved. The existing on-site operation safety supervision method relies on manpower for supervision, which is prone to supervision blind spots, making it difficult to fully and timely control the on-site situation, resulting in low accuracy of safety supervision results.

[0081] Embodiment 2

[0082] See also Figure 4 , Figure 4 A flowchart of the steps of a work site safety supervision method provided in Embodiment 2 of the present invention.

[0083] Another work site safety supervision method provided by Example 2 of the present invention includes:

[0084] Step 401: Acquire an initial monitoring screen and intelligent tool operation data of the work site, select an initial intelligent tool corresponding to the intelligent tool operation data in the initial monitoring screen, and generate a target intelligent tool.

[0085] Further, step 401 may include the following sub-steps S11-S16:

[0086] S11. When the operation status corresponding to the operation data of the intelligent tool is an electrical test operation, the intelligent electrical tester and insulating gloves in the initial monitoring screen are used as target intelligent tools.

[0087] S12. When the operation state corresponding to the operation data of the intelligent tool is grounding operation, the intelligent grounding wire and insulating gloves in the initial monitoring screen are used as target intelligent tools.

[0088] S13. When the operation status corresponding to the operation data of the smart tool is that the grounding wire is abnormal, the smart grounding wire and insulating gloves in the initial monitoring screen are used as target smart tools.

[0089] S14. When the operation state corresponding to the operation data of the intelligent tool is climbing operation, determine whether the safety belt identification data corresponding to the operation data of the intelligent tool is the preset identification data. If so, execute step S15; if not, execute step S16.

[0090] S15. The smart bracelet and the safety belt in the initial monitoring image are used as target smart tools.

[0091] S16. The smart bracelet, seat belt and escalator in the initial monitoring image are taken as target smart tools.

[0092] The preset identification data means that the operation status corresponding to the operation data of the intelligent tool is high-altitude operation, and the safety belt identification has not yet been performed.

[0093] In an embodiment of the present invention, for different operation scenarios, such as: stations, distribution lines, box transformers, racks, distribution rooms, meter boxes, etc., business analysis is performed on each scenario to determine whether it is necessary to test electricity-grounding and whether it is necessary to climb. According to the operation status corresponding to the operation data of the intelligent tool, the target intelligent tool that should be safely identified in the initial monitoring screen is determined. Specifically: when the operation status corresponding to the operation data of the intelligent tool is an electricity test operation, the intelligent electrician and insulating gloves in the initial monitoring screen need to be used as the target intelligent tool. When the operation status corresponding to the operation data of the intelligent tool is a grounding operation, the intelligent grounding wire and insulating gloves in the initial monitoring screen need to be used as the target intelligent tool. When the operation status corresponding to the operation data of the intelligent tool is an abnormal grounding wire, the intelligent grounding wire and insulating gloves in the initial monitoring screen need to be used as the target intelligent tool. When the operation status corresponding to the operation data of the intelligent tool is an abnormal grounding wire, the intelligent grounding wire and insulating gloves in the initial monitoring screen need to be used as the target intelligent tool. When the operation status corresponding to the operation data of the intelligent tool is a climbing operation, it is necessary to determine whether the safety belt identification data corresponding to the operation data of the intelligent tool is the preset identification data. If the seat belt identification data is the preset identification data, the smart bracelet and seat belt in the initial monitoring screen are used as the target smart tools. If the seat belt identification data is not the preset identification data, the smart bracelet, seat belt and escalator in the initial monitoring screen are used as the target smart tools.

[0094] Step 402: extract features from the initial monitoring image based on the position data corresponding to the target intelligent tool to generate a plurality of original feature points.

[0095] Further, step 402 may include the following sub-steps S21-S23:

[0096] S21. Perform multiple smoothing processes on the initial monitoring picture according to the position data and preset resolution data corresponding to the target intelligent tool, and construct an image pyramid corresponding to the initial monitoring picture.

[0097] S22, respectively selecting feature points corresponding to the image edge of each image in the image pyramid to generate a plurality of image feature points.

[0098] S23, respectively using the tangent directions corresponding to the image feature points on the image edge as the direction vectors of the image feature points to generate original feature points corresponding to the image feature points.

[0099] The preset resolution data includes the resolution corresponding to each layer of the image pyramid.

[0100] In an embodiment of the present invention, the target intelligent tool here can be multiple or single, which is determined by the corresponding working state of the working site. The position data corresponding to the target intelligent tool is used to determine the position of the target intelligent tool in the initial monitoring screen, and then the target intelligent tool in the initial monitoring screen is subjected to scale-invariant feature transformation processing according to the preset resolution data to construct an image pyramid composed of images of different scales. Feature points corresponding to images of different scales in the image pyramid are selected, and feature points on the edge of the image are selected as image feature points. Then, the tangent direction of the image feature point on the edge of the image is used as the direction vector of the image feature point to obtain the original feature point corresponding to the image feature point.

[0101] An image pyramid is a structure that interprets images at different scales and resolutions. It smoothes the original image to varying degrees, places the original image with the highest resolution at the bottom of the pyramid, and then gradually reduces its size.

[0102] like Figure 5 As shown in the figure, the bottom layer is the image of the original target intelligent tool, which has the largest scale and the highest resolution. The images of other layers are obtained through scale-invariant feature transformation. The top layer image has the smallest scale and low resolution. The Gaussian pyramid is composed of multiple groups of image pyramids of different scales. Each group of image pyramids has many layers, and the smoothness of each layer is different.

[0103] Step 403: Select feature points in the initial monitoring image that correspond to the original feature points to generate multiple matching feature points.

[0104] In an embodiment of the present invention, the image of the original target intelligent tool is processed by scale-invariant feature transformation, and each layer has feature points corresponding to the feature points of the original image. By matching the feature points of each layer with the video screen, including the pixel values ​​and direction vectors of the matching feature points, the position of the matching feature points in the video screen can be found, and multiple matching feature points can be obtained. Through these matching feature points, the size and position of the target intelligent tool in the video screen can be roughly locked.

[0105] Step 404: segment the initial monitoring image based on the matching feature points and perform grayscale processing to generate a plurality of target grayscale unit blocks.

[0106] Further, step 404 may include the following sub-steps S31-S34:

[0107] S31 . Segment the initial monitoring image according to all matching feature points and a preset segmentation size to generate a plurality of unit blocks.

[0108] S32, performing grayscale processing on the unit blocks respectively to generate a plurality of initial grayscale unit blocks.

[0109] S33 , comparing the grayscale values ​​of the initial grayscale unit block with those of the adjacent unit blocks respectively, and generating a plurality of comparison data corresponding to the initial grayscale unit block.

[0110] S34: When the comparison data has a preset comparison threshold, the initial grayscale unit block is used as the target grayscale unit block.

[0111] The preset segmentation size refers to the size of the unit block obtained by segmentation that is set in advance based on actual needs.

[0112] The preset comparison threshold refers to the difference between the grayscale value of the initial grayscale unit block area and any grayscale value of the four adjacent upper, lower, left and right unit blocks.

[0113] In the embodiment of the present invention, grayscale processing is combined with scale-invariant feature transformation processing to obtain the edge of the target intelligent tool more accurately. Scale-invariant feature transformation can only identify the rough outline due to limited feature points, while grayscale recognition can make up for this shortcoming and connect the feature points according to the grayscale change to obtain the image edge.

[0114] According to all matching feature points and preset segmentation size, the initial monitoring image is divided into several unit blocks and grayscale processing is performed to obtain initial grayscale unit blocks of different grayscales. To determine whether the grayscale of the initial grayscale unit block area where the matching feature point is located has changed, it is necessary to determine the grayscale value of the unit block where the matching feature point is located, that is, the initial grayscale unit block area, and the grayscale values ​​of the four adjacent upper, lower, left, and right unit blocks. If the grayscale values ​​have different values, it is determined that the unit block area where the matching feature point is located is at the edge of the image, and the initial grayscale unit block is used as the target grayscale unit block.

[0115] Step 405: Connect the region boundaries of the target grayscale unit blocks respectively to generate a job supervision region.

[0116] In the embodiment of the present invention, the changed unit block area boundaries are connected as the edge of the intelligent tool in the video image, that is, the area boundaries of each target grayscale unit block are connected to construct the operation supervision area.

[0117] Step 406: adjust the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and use the collected monitoring screen as the target monitoring screen.

[0118] In the embodiment of the present invention, the specific implementation process of step 407 is similar to step 102 and will not be repeated here.

[0119] Step 407: Identify illegal behaviors based on the target monitoring screen and generate work site safety supervision data.

[0120] Further, step 407 may include the following sub-steps S41-S43:

[0121] S41. Classify and annotate the video image set of violations at the work site corresponding to the work site to build a sample library of typical violations.

[0122] S42. Use a typical violation behavior sample library to train the initial deep learning model and generate a target deep learning model.

[0123] S43. Identify illegal behaviors in the target monitoring images through the target deep learning model and generate work site safety supervision data.

[0124] Further, step S43 may include the following sub-steps S431-S437:

[0125] S431. Use the target deep learning model to identify the personal protection of the operating personnel in the target monitoring screen using the personal protection standards corresponding to the target monitoring screen to generate personal protection identification data.

[0126] S432. When the personal protection identification data is abnormal, the personal protection identification data is used to construct the first work site safety supervision data.

[0127] S433. When the personal protection identification data is normal, the target deep learning model is used to identify the working status of the intelligent tool working data corresponding to the target monitoring screen to generate the tool working status data.

[0128] S434. When the equipment working status data is abnormal, the equipment working status data is used to construct the second work site safety supervision data.

[0129] S435. When the working status data of the equipment is normal, the operating behavior in the target monitoring screen is compared with the corresponding typical violation behavior through the target deep learning model to generate behavior comparison data.

[0130] S436. When the behavior comparison data are the same, the behavior comparison data are used to construct the third work site safety supervision data.

[0131] S437. When the behavior comparison data are different, the fourth work site safety supervision data is set to no illegal behavior at the work site.

[0132] The initial deep learning model refers to a pre-set deep learning model that can perform image AI recognition. The initial deep learning model can also use the survey method to collect business manuals, combine historical accident information and other materials, analyze the influencing factors of on-site operation violations, and use hierarchical analysis theory to analyze the risk causes of on-site operation violations from multiple perspectives of human factors and material factors.

[0133] The target deep learning model refers to the deep learning model obtained by training the initial deep learning model with the violation behavior sample library. The target deep learning model includes detection algorithms such as the height detection algorithm, the seat belt detection algorithm and the escalator algorithm. Among them, the height detection algorithm refers to an algorithm that can find the height-climbing personnel from the target monitoring screen and compare them with the corresponding samples in the sample library. The seat belt detection algorithm refers to an algorithm that can identify the wearing status of the seat belt from the target monitoring screen. The escalator algorithm refers to an algorithm that can identify whether there is an unmanned escalator phenomenon from the target monitoring screen.

[0134] In an embodiment of the present invention, the AI ​​law enforcement instrument identifies on-site violations of the target monitoring screen. First, the set of video images of violations of the work site corresponding to the work site is collected, and the video images of violations of the work site are processed and classified according to the different work states corresponding to the work site to form a typical violation sample library, where typical violations include failure to wear as required, failure to use safety tools in a standardized manner, and non-standard operations. Specifically, typical violations in power marketing operations, such as not wearing a safety helmet, climbing without a safety belt, not wearing work clothes, working without a ticket, etc.; collect and organize image data of various violations in power marketing operations, and establish a typical violation sample library for training and verifying the accuracy of the initial learning model.

[0135] The initial deep learning model is trained using a typical violation sample library, and a typical violation image recognition algorithm is developed, so that the target deep learning model has high accuracy in identifying violations. The AI ​​law enforcement instrument serves as a computing center, which obtains video data in real time and processes the video data to determine whether there are typical violations. The judgment process is: determine whether the operator wears personal protective equipment correctly; determine whether the intelligent tools are working properly; and determine whether there are typical violations.

[0136] Specifically, the AI ​​body camera uses the personal protection standards corresponding to the target monitoring screen through the target deep learning model to identify the personal protection of the workers in the target monitoring screen, and generates personal protection identification data. Among them, the personal protection standard refers to the personal protection requirements that workers in this working state should meet. When the personal protection identification data is abnormal, the personal protection identification data is used to construct the first work site safety supervision data. When the personal protection identification data is normal, the target deep learning model is used to identify the working status of the smart tools corresponding to the target monitoring screen, and generate tool working status data. Among them, whether the smart tools corresponding to the target monitoring screen are working normally is combined Figure 3 The data in Table 1 are used for judgment.

[0137] When the equipment working status data is abnormal, the equipment working status data is used to construct the second work site safety supervision data. When the equipment working status data is normal, the target deep learning model is used to compare the operating behavior in the target monitoring screen with the corresponding typical violation behavior to generate behavior comparison data. When the behavior comparison data are the same, the behavior comparison data is used to construct the third work site safety supervision data. When the behavior comparison data are different, the fourth work site safety supervision data is set to no violation of the work site.

[0138] In an embodiment of the present invention, an initial monitoring screen and intelligent tool operation data of a work site are obtained, and an initial intelligent tool corresponding to the intelligent tool operation data in the initial monitoring screen is selected to generate a target intelligent tool. Based on the position data corresponding to the target intelligent tool, feature extraction is performed on the initial monitoring screen to generate a plurality of original feature points. Feature points corresponding to the original feature points in the initial monitoring screen are selected to generate a plurality of matching feature points. The initial monitoring screen is segmented and grayscale processed based on the matching feature points to generate a plurality of target grayscale unit blocks. The regional boundaries of the target grayscale unit blocks are connected respectively to generate a work supervision area. The monitoring angle corresponding to the initial monitoring screen is adjusted according to the work supervision area, and the collected monitoring screen is used as the target monitoring screen. Based on the target monitoring screen, illegal behavior is identified to generate work site safety supervision data. The present invention is based on image AI recognition, intelligent sensing, wireless ad hoc network and other technologies to construct a work site safety supervision method, combined with an AI law enforcement instrument, real-time monitoring of the power marketing work site, linkage of intelligent tools, timely discovery and alarm of illegal behavior and upload of video terminals, multi-dimensional monitoring of the power marketing work site, and improve the accuracy of illegal behavior recognition.

[0139] Embodiment 3

[0140] When the operation status corresponding to the operation data of the intelligent tool is an electrical test operation, it means that the operation site is an electrical test-grounding operation site. The supervision process of the electrical test-grounding operation site is as follows: (1) Turn on the electrical tester and perform the insulating glove test; (2) After the electrical test is completed, the grounding operation can be performed; (3) Stop the insulating glove test after grounding.

[0141] In terms of anti-electric shock applications, the electric test-grounding measures can be managed and controlled. When the smart electric tester starts self-testing, the electric test command is transmitted to the AI ​​law enforcement instrument via Bluetooth. After receiving the electric test command, the AI ​​law enforcement instrument uses the smart electric tester and insulating gloves in the initial monitoring screen as the target smart tools, and obtains the position of the insulating gloves and the smart electric tester through the RFID tags embedded in the insulating gloves and the smart electric tester, thereby obtaining the position data corresponding to the target smart tools. By performing edge detection on the insulating gloves and smart electric tester in the initial monitoring screen, the operation supervision area is determined, and the video angle is adjusted to collect the target supervision screen corresponding to the operation supervision area, and then the target supervision screen is identified for violation behavior. The operator is judged to have violated the rules in the following order: first judge personal protection, and when it is judged that personal protection is normal, then judge the working status of the target smart tool, and when it is judged that the working status is also normal, then judge whether the operation is standardized, so as to obtain the safety supervision data of the operation site.

[0142] When the operation state corresponding to the operation data of the smart tool is grounding operation, the smart grounding wire and insulating gloves in the initial monitoring screen are used as the target smart tools, and the specific identification process is the same as the identification process of the above-mentioned smart tester self-test. Among them, the smart grounding wire uses smart sensor technology to automatically judge the state of hanging, falling off, and removal, and sends the state information to the AI ​​law enforcement instrument via Bluetooth. When the three-phase grounding is completed, the AI ​​law enforcement instrument stops the identification and detection of insulating gloves.

[0143] When the grounding wire falls off or becomes loose, that is, when the operating status corresponding to the operating data of the smart tool is that the grounding wire is abnormal, the smart grounding wire and insulating gloves in the initial monitoring screen are used as target smart tools, and the specific identification process is the same as the identification process performed by the above-mentioned smart tester to start the self-test. Among them, the smart grounding wire makes its own status judgment and triggers the AI ​​law enforcement device again to identify the smart grounding wire and insulating gloves. The smart grounding wire is suitable for 380V-10kV grounding applications. It has a built-in trigger wake-up mechanism, which is triggered when the hanging status changes. It can monitor the hanging status of the grounding wire and has Bluetooth communication function. The status information of the grounding wire hanging, falling off, and removal is synchronized in the AI ​​law enforcement device to realize the linkage interaction of products, such as whether insulating gloves are worn during operation.

[0144] Embodiment 4

[0145] When the operation status corresponding to the operation data of the smart tool is high-altitude operation, the operation site is a high-altitude operation site. The supervision process of the high-altitude operation site is as follows: (1) The smart bracelet automatically detects the movement status of the human body to determine whether to climb; (2) When it is determined that the human body is climbing, illegal behaviors such as not wearing a safety belt / no one using the ladder are identified.

[0146] In terms of anti-fall applications, it can be used to control high-altitude operations such as tower climbing and ladder climbing. The operators wear smart bracelets, which automatically detect the operator's horizontal height and determine whether there is any high-altitude operation. When the operation status corresponding to the smart tool operation data is high-altitude operation, the smart tool operation data includes the operator's motion status and the position data of the safety belt and escalator. The specific acquisition process is to send the operator's motion status to the AI ​​law enforcement device, and the AI ​​law enforcement device obtains the position of the safety belt and escalator through the RFID tags embedded in the safety belt and escalator. When the safety belt identification data is the preset identification data, that is, the safety belt identification has not been performed, the smart bracelet and safety belt in the initial monitoring screen are used as the target smart tools, and the recording angle is adjusted to obtain the corresponding target monitoring screen. The safety belt is identified through the target deep learning model to determine whether the high-altitude operator is wearing a safety belt. When the seat belt recognition data is the preset recognition data, that is, the seat belt recognition has been performed, the smart bracelet, seat belt and escalator in the initial monitoring screen are used as target smart tools, the recording angle is adjusted to obtain the corresponding target monitoring screen, and the escalator is recognized through the target deep learning model to determine whether someone is taking the escalator.

[0147] Embodiment 5

[0148] The AI ​​body camera is also equipped with voice interaction technology. The specific implementation process is as follows: when a voice command is received, the voice command is recognized and converted into text to obtain the corresponding command text; the command text is structurally split and converted into a language that the machine can understand to generate command intent data; expected feedback is given based on the command intent data to generate a reply text; and the reply text is converted into voice output.

[0149] Specifically, the AI ​​law enforcement device can receive voice commands from on-site personnel. For example, on-site staff can issue a command to "record the scene", at which point the AI ​​law enforcement device replies "recording has started" and turns on the recording function; or, the on-site staff issues a command to "identify the location of the smart electric tester", and the AI ​​law enforcement device sends radio electromagnetic waves to the RFID tag of the smart electric tester. After receiving the smart electric tester information fed back by the RFID tag, it replies "the smart electric tester has been identified"; or, the on-site staff issues a command to "determine whether there are any illegal operations at the scene", and the AI ​​law enforcement device replies "intelligent identification has been turned on", and during the recording process, it identifies on-site violations of regulations by on-site workers, etc.

[0150] Furthermore, voice warnings for preventing electric shock and falling from height can be stored in advance for the work scenarios, and intelligent voice warnings can be provided for on-site work electrical testing, grounding or climbing nodes.

[0151] Furthermore, the AI ​​law enforcement device is also equipped with an alarm device: in response to the risks in power marketing operations, the buzzer in the alarm device sounds an alarm, that is, the corresponding alarm sounds according to the generated work site safety supervision data. For example: at the work site, for electrical testing and grounding, an alarm is issued for the risk of electric shock, including but not limited to: whether insulating gloves are worn, whether the smart grounding wire is securely connected, whether the smart electrical tester is measuring electricity, etc. For climbing, an alarm is issued for the risk of falling from a height, including but not limited to: whether a safety belt is worn, whether a smart bracelet is worn, whether someone is holding the ladder, etc.

[0152] Embodiment 6

[0153] See also Figure 6 , Figure 6 This is a structural block diagram of a work site safety supervision system provided in Example 6 of the present invention.

[0154] A work site safety supervision system provided in Example 6 of the present invention includes:

[0155] The operation supervision area generation module 601 is used to obtain the initial monitoring screen and intelligent tool operation data of the operation site, perform edge recognition on the initial intelligent tool corresponding to the initial monitoring screen based on the intelligent tool operation data, and generate an operation supervision area.

[0156] The target monitoring screen determination module 602 is used to adjust the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and use the collected monitoring screen as the target monitoring screen.

[0157] The work site safety supervision data generation module 603 is used to identify illegal behaviors based on the target monitoring screen and generate work site safety supervision data.

[0158] Optionally, the operation supervision area generation module 601 includes:

[0159] The target intelligent tool generation module is used to select the initial intelligent tool corresponding to the intelligent tool operation data in the initial monitoring screen to generate the target intelligent tool.

[0160] The original feature point generation module is used to extract features from the initial monitoring screen based on the position data corresponding to the target intelligent tool and generate multiple original feature points.

[0161] The matching feature point generation module is used to select feature points corresponding to the original feature points in the initial monitoring image and generate multiple matching feature points.

[0162] The target grayscale unit block generation module is used to segment the initial monitoring picture based on the matching feature points and perform grayscale processing to generate multiple target grayscale unit blocks.

[0163] The operation supervision area generation submodule is used to connect the area boundaries of the target grayscale unit blocks respectively to generate the operation supervision area.

[0164] Optionally, the target intelligent tool generation module may perform the following steps:

[0165] When the operation state corresponding to the operation data of the intelligent tool is an electrical test operation, the intelligent electrical tester and insulating gloves in the initial monitoring screen are used as target intelligent tools;

[0166] When the operation state corresponding to the operation data of the intelligent tool is grounding operation, the intelligent grounding wire and insulating gloves in the initial monitoring screen are used as target intelligent tools;

[0167] When the operation status corresponding to the operation data of the intelligent tool is that the grounding wire is abnormal, the intelligent grounding wire and insulating gloves in the initial monitoring screen are used as target intelligent tools;

[0168] When the operation state corresponding to the operation data of the intelligent tool is climbing operation, determining whether the safety belt identification data corresponding to the operation data of the intelligent tool is the preset identification data;

[0169] If the seat belt identification data is the preset identification data, the smart bracelet and seat belt in the initial monitoring screen are used as the target smart tools;

[0170] If the seat belt identification data is not the preset identification data, the smart bracelet, seat belt and escalator in the initial monitoring screen will be used as the target smart tools.

[0171] Optionally, the original feature point generation module may perform the following steps:

[0172] The initial monitoring picture is smoothed multiple times according to the position data and preset resolution data corresponding to the target intelligent tool, and an image pyramid corresponding to the initial monitoring picture is constructed;

[0173] Select feature points corresponding to the image edge of each image in the image pyramid respectively to generate multiple image feature points;

[0174] The tangent directions corresponding to the image feature points on the image edge are respectively used as the direction vectors of the image feature points to generate the original feature points corresponding to the image feature points.

[0175] Optionally, the target grayscale unit block generation module may perform the following steps:

[0176] The initial monitoring image is segmented according to all matching feature points and preset segmentation sizes to generate multiple unit blocks;

[0177] Performing grayscale processing on the unit blocks respectively to generate a plurality of initial grayscale unit blocks;

[0178] Comparing the grayscale values ​​of the initial grayscale unit block with those of the adjacent unit blocks respectively, and generating a plurality of comparison data corresponding to the initial grayscale unit block;

[0179] When the comparison data has a preset comparison threshold, the initial grayscale unit block is used as the target grayscale unit block.

[0180] Optionally, the work site safety supervision data generation module 603 includes:

[0181] The typical violation behavior sample library construction module is used to classify and annotate the work site violation video image set corresponding to the work site to construct a typical violation behavior sample library.

[0182] The target deep learning model generation model is used to train the initial deep learning model using a typical violation behavior sample library to generate a target deep learning model.

[0183] The work site safety supervision data generation submodule is used to identify illegal behaviors in the target monitoring screen through the target deep learning model and generate work site safety supervision data.

[0184] Optionally, the job site safety supervision data generation submodule may perform the following steps:

[0185] The target deep learning model uses the personal protection standard corresponding to the target monitoring screen to identify the personal protection of the operating personnel in the target monitoring screen and generate personal protection identification data;

[0186] When the personal protection identification data is abnormal, the personal protection identification data is used to construct the first work site safety supervision data;

[0187] When the personal protection identification data is normal, the target deep learning model is used to identify the working status of the intelligent tool corresponding to the target monitoring screen, and the tool working status data is generated;

[0188] When the equipment working status data is abnormal, the equipment working status data is used to construct the second work site safety supervision data;

[0189] When the working status data of the equipment is normal, the target deep learning model is used to compare the operating behavior in the target monitoring screen with the corresponding typical violation behavior to generate behavior comparison data;

[0190] When the behavior comparison data are the same, the behavior comparison data are used to construct the third work site safety supervision data;

[0191] When the behavior comparison data are different, the fourth work site safety supervision data is set to no illegal behavior at the work site.

[0192] Embodiment 7

[0193] See also Figure 7 , Figure 7 This is a structural block diagram of an electronic device provided in Embodiment 7 of the present invention.

[0194] An electronic device according to an embodiment of the present invention includes: a memory 701 and a processor 702, wherein the memory 701 stores a computer program; when the computer program is executed by the processor 702, the processor 702 executes a work site safety supervision method as described in any of the above embodiments.

[0195] The memory 701 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk or a ROM. The memory 701 has a storage space 703 for a program code 713 for executing any method step in the above method. For example, the storage space 703 for the program code may include individual program codes 713 for implementing the various steps in the above method, respectively. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as a hard disk, a compact disk (CD), a memory card or a floppy disk. The program code may be compressed, for example, in an appropriate form. When these codes are run by a computing and processing device, the computing and processing device performs the various steps in the above-described method for safety supervision of the work site.

[0196] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for supervising safety at a work site as described in any of the above embodiments is implemented.

[0197] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0198] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0199] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0200] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0201] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0202] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for safety supervision at a work site, characterized in that: include: Acquire an initial monitoring screen and intelligent tool operation data of the work site, perform edge recognition on an initial intelligent tool corresponding to the initial monitoring screen based on the intelligent tool operation data, and generate an operation supervision area; Adjusting the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and using the collected monitoring screen as the target monitoring screen; Based on the target monitoring screen, illegal behaviors are identified to generate work site safety supervision data.

2. The method for safety supervision of a work site according to claim 1, characterized in that: The step of performing edge recognition on the initial smart tool corresponding to the initial monitoring screen based on the smart tool operation data to generate an operation supervision area includes: Selecting an initial smart tool corresponding to the operation data of the smart tool in the initial monitoring screen to generate a target smart tool; Extracting features from the initial monitoring screen based on the position data corresponding to the target intelligent tool to generate a plurality of original feature points; Selecting feature points in the initial monitoring image corresponding to the original feature points to generate a plurality of matching feature points; Segmenting the initial monitoring picture based on the matching feature points and performing grayscale processing to generate a plurality of target grayscale unit blocks; The region boundaries of the target grayscale unit blocks are connected respectively to generate a job supervision region.

3. The method for safety supervision of a work site according to claim 2, characterized in that: The step of selecting an initial smart tool corresponding to the operation data of the smart tool in the initial monitoring screen to generate a target smart tool includes: When the operation state corresponding to the operation data of the smart tool is an electrical test operation, the smart electrical tester and insulating gloves in the initial monitoring screen are used as target smart tools; When the operation state corresponding to the operation data of the smart tool is grounding operation, the smart grounding wire and insulating gloves in the initial monitoring screen are used as target smart tools; When the operation state corresponding to the operation data of the smart tool is that the grounding wire is abnormal, the smart grounding wire and the insulating gloves in the initial monitoring screen are used as target smart tools; When the operation state corresponding to the operation data of the intelligent tool is a climbing operation, determining whether the safety belt identification data corresponding to the operation data of the intelligent tool is a preset identification data; If the seat belt identification data is the preset identification data, the smart bracelet and the seat belt in the initial monitoring screen are used as the target smart tools; If the seat belt identification data is not the preset identification data, the smart bracelet, seat belt and escalator in the initial monitoring screen are used as target smart tools.

4. The method for safety supervision of a work site according to claim 2, characterized in that: The step of extracting features from the initial monitoring screen based on the position data corresponding to the target intelligent tool to generate a plurality of original feature points includes: The initial monitoring picture is smoothed multiple times according to the position data and preset resolution data corresponding to the target intelligent tool, so as to construct an image pyramid corresponding to the initial monitoring picture; Selecting feature points corresponding to image edges of each image in the image pyramid respectively to generate a plurality of image feature points; The tangent directions corresponding to the image feature points on the image edge are respectively used as direction vectors of the image feature points to generate original feature points corresponding to the image feature points.

5. The method for safety supervision of a work site according to claim 2, characterized in that: The step of segmenting the initial monitoring picture based on the matching feature points and performing grayscale processing to generate a plurality of target grayscale unit blocks includes: Segmenting the initial monitoring image according to all the matching feature points and the preset segmentation size to generate a plurality of unit blocks; Performing grayscale processing on the unit blocks respectively to generate a plurality of initial grayscale unit blocks; Comparing the grayscale values ​​of the initial grayscale unit block with those of adjacent unit blocks respectively, and generating a plurality of comparison data corresponding to the initial grayscale unit block; When the comparison data has a preset comparison threshold, the initial grayscale unit block is used as a target grayscale unit block.

6. The method for safety supervision of a work site according to claim 1, characterized in that: The step of identifying illegal behaviors based on the target monitoring screen and generating work site safety supervision data includes: Using the work site violation video image set corresponding to the work site for classification and annotation, and building a typical violation behavior sample library; Using the typical violation behavior sample library to train the initial deep learning model to generate a target deep learning model; Violation behaviors are identified on the target monitoring screen through the target deep learning model to generate work site safety supervision data.

7. The method for safety supervision of a work site according to claim 6, characterized in that: The step of identifying illegal behaviors on the target monitoring screen through the target deep learning model and generating work site safety supervision data includes: Performing personal protection identification of the operating personnel in the target monitoring screen by using the target deep learning model and the personal protection standard corresponding to the target monitoring screen to generate personal protection identification data; When the personal protection identification data is abnormal, constructing first work site safety supervision data using the personal protection identification data; When the personal protection identification data is normal, the target deep learning model is used to identify the working state of the intelligent tool corresponding to the target monitoring screen, and the working state data of the tool is generated; When the tool working state data is abnormal, constructing second work site safety supervision data using the tool working state data; When the working state data of the device is normal, the operating behavior in the target monitoring screen is compared with the corresponding typical violation behavior through the target deep learning model to generate behavior comparison data; When the behavior comparison data are the same, the behavior comparison data are used to construct the third work site safety supervision data; When the behavior comparison data are different, the fourth work site safety supervision data is set to no illegal behavior at the work site.

8. A work site safety supervision system, characterized in that: include: An operation supervision area generation module is used to obtain an initial monitoring screen and intelligent tool operation data of the operation site, perform edge recognition on the initial intelligent tool corresponding to the initial monitoring screen based on the intelligent tool operation data, and generate an operation supervision area; A target monitoring screen determination module, used to adjust the monitoring angle corresponding to the initial monitoring screen according to the operation supervision area, and use the collected monitoring screen as the target monitoring screen; The operation site safety supervision data generation module is used to identify illegal behaviors based on the target monitoring screen and generate operation site safety supervision data.

9. An electronic device, characterized in that: It comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the work site safety supervision method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the work site safety supervision method according to any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Intelligent personal identification monitoring tracking transmission method and system

    CN108769595A

  • Electric power field operation safety monitoring method and device based on video image analysis

    CN112347916A

  • Defect analysis and diagnosis system, method and device based on artificial intelligence

    CN112862770A

  • Power grid operation and maintenance operation violation behavior monitoring and early warning management system

    CN115375287A

  • Operation site digital safety management and control method, device and equipment and storage medium

    CN115471865A

Cited By

  • Real-time monitoring method and system for business expansion process of power supply enterprise

    CN121258099A