CN112200030A - Power system field operation action risk identification method based on graph convolution - Google Patents
Power system field operation action risk identification method based on graph convolution Download PDFInfo
- Publication number
- CN112200030A CN112200030A CN202011033289.9A CN202011033289A CN112200030A CN 112200030 A CN112200030 A CN 112200030A CN 202011033289 A CN202011033289 A CN 202011033289A CN 112200030 A CN112200030 A CN 112200030A
- Authority
- CN
- China
- Prior art keywords
- graph
- power system
- field
- behavior
- action
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
- G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a dynamic action risk identification method for field workers of an electric power system. The invention can realize the risk identification of the dynamic behavior of the field operating personnel and can identify the dynamic violation behavior and risk of the field operating personnel in real time, thereby providing a technical means for the safety risk early warning and management and control of the power production field, reducing the accident occurrence probability of a power system and improving the intrinsic safety level of power production.
Description
Technical Field
The invention belongs to the field of electric power system operation safety control, and particularly relates to a dynamic action risk identification method for electric power system field operation personnel based on graph convolution.
Background
The safe production is the basic guarantee of the stable operation of the power system. Once a safety accident occurs in the power system, huge economic loss and adverse social influence can be caused. The real-time identification and management and control of the site operation risk of the electric power system are of great significance to ensure the personal safety of operators and the safe and stable operation of the power grid.
Currently, the safety risk management method for field operation mainly includes an artificial safety supervision method and a video monitoring method. The manual safety supervision method is mainly used for supervising the behavior operation of the operators and the like by specially arranging guardians. But the guardian cannot guarantee the all-round supervision of the operator. And the supervisors and the operators are easily influenced by external factors, and the attention may not be concentrated, so that safety accidents may be caused. Particularly, the operation of the power system has the characteristics of multiple sites, multiple devices, complex operation and the like, so that the manual supervision method cannot realize real-time supervision and risk early warning of all operation processes. The video monitoring system provides effective assistance for safety supervision, but the actual monitoring task still needs more manual work to be completed, and the monitoring system usually only records video images for later evidence collection. In addition, some students propose monitoring video-based intelligent analysis methods, including personnel information checking of field operating personnel, fire detection of safety helmets and safety belt detection substations, billboard detection and the like, and thus, the existing methods mainly focus on static safety risk identification. However, the actual field operation in a power system is a continuous and dynamic process where there are many dynamic safety risks and violations. The existing security risk identification methods for security supervision cannot identify dynamic risks and violation behaviors of operators.
Therefore, the invention introduces a risk identification method for dynamic behaviors of field operators of an electric power system, so that dynamic action risk identification and real-time safety control of the field operators are realized.
Disclosure of Invention
The invention aims to provide a risk identification method for dynamic behaviors of field operators of an electric power system, which can realize dynamic risk identification of the field operators of the electric power system. The method mainly comprises the following steps:
step 1: and monitoring the behavior of field operators of the power system and transmitting images. The monitoring cameras arranged on the electric power system operation site monitor the behaviors of the operators and upload the monitoring videos to the central cloud platform.
Step 2: and estimating the behavior posture of the operator. And estimating the behavior posture of the field operator in the monitoring video by using the human body posture estimation model openposition, and acquiring skeleton information of the behavior action of the field operator.
And step 3: connecting the obtained human skeleton information, and constructing an undirected graph G ═ v, A and X containing the action information of the field operator, whereinIs a set of N vertices, a is a weighted adjacency matrix, and X is a signal matrix on the vertices. The set of vertices v corresponds to all N joints of the skeleton, and the adjacency matrix A represents the link relationship between two vertices if the vertex v corresponds to all N joints of the skeletoniAnd vertex vjAnd with a direct connection, then Aij1, otherwise Aij=0。
And 4, step 4: and transforming the constructed undirected graph into a spectral domain, and filtering the undirected graph containing the human skeleton information in the spectral domain. For undirected graph G ═ (v, a, X), the laplacian matrix L of the undirected graph is defined as follows:
L=D-A
where D is the degree matrix of the vertices and the diagonal elements are, in turn, the degrees of each vertex.
The laplacian matrix L of the undirected graph is a symmetric positive definite matrix. This laplacian matrix L can be decomposed as follows:
L=UΛUT
wherein Λ ═ diag ([ λ ])1,λ2,…,λN]) Is a diagonal matrix of eigenvalues, U ═ U1,u2,…,uN]Is an orthogonal matrix corresponding to the eigenvalue vector.
Given the filter function G (-) of the undirected graph G, the frequency domain filter response of the input signal X can be defined asThe inverse fourier transform of the graph is defined as follows:
thus, the matrix description of the undirected graph filtering can be defined as follows:
and 5: an undirected Graph including skeleton information of a human body is identified by using a Graph Convolution Network (GCN).
Step 6: a human body key skeleton node attention module is added between graph volume networks, key skeleton nodes relevant to actions are highlighted, and therefore feature expression and action identification accuracy of the networks are improved. For the violation behaviors of operators in the field of the power system, the influence of each skeleton joint on the action recognition is different. Since an action of the operator is usually strongly related to only a few skeletal joints, but weakly related to other joints. Therefore, the patent proposes that the information of the action key bone joint is enhanced by using the human body key bone node attention module, so that the accuracy of identifying the illegal action of an operator is improved. Assuming that the feature map obtained by graph convolution is H, the output feature map H' obtained by the key bone node attention module is as follows.
H′=W(H)H
Wherein WijNot less than 0 andWijthe larger the number of the nodes, the more important the ith joint node is for violation identification.
And 7: and (5) identifying and classifying actions. And for the characteristic diagrams which are extracted by the graph convolution network and contain human body posture information, converting all the characteristic diagrams into one-dimensional matrixes by using a full connection layer, and then identifying and classifying the violation behaviors and the risk actions of different operators by using a softmax function.
The invention is characterized in that:
the method and the device convert the video information into the undirected graph containing the skeleton information by extracting the skeleton information of the field operating personnel of the power system, and then realize the action recognition of the field operating personnel by utilizing the graph convolution network. The invention can realize the risk identification of the dynamic behavior of the field operating personnel and can identify the violation behavior and the risk of the field operating personnel in real time, thereby providing a technical means for the early warning and the management and control of the safety risk of the power production field, reducing the accident occurrence probability of a power system and improving the intrinsic safety level of power production.
Drawings
Fig. 1 is a flowchart of a power system field work action risk identification method based on graph convolution according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and the following detailed description, but the present invention is not limited to these embodiments.
The specific implementation mode is as follows:
a flowchart of a power system field operation risk identification method based on graph convolution is shown in fig. 1, and includes the following specific steps:
step 1: and monitoring the behavior of field operators of the power system and transmitting images. The monitoring cameras arranged on the electric power system operation site monitor the behaviors of the operators and upload the monitoring videos to the central cloud platform.
Step 2: and estimating the behavior posture of the operator. And estimating the behavior posture of the field operator in the monitoring video by using the human body posture estimation model openposition, and acquiring skeleton information of the behavior action of the field operator.
And step 3: connecting the obtained human skeleton information to construct an undirected graph G ═ v, A and X containing the action information of the field operator, whereinIs a set of N vertices, a is a weighted adjacency matrix, and X is a signal matrix on the vertices. The set of vertices v corresponds to all N joints of the skeleton, and the adjacency matrix A represents the link relationship between two vertices if the vertex v isiAnd vertex vjAnd with a direct connection, then Aij1, otherwise Aij=0。
And 4, step 4: and then transforming the constructed undirected graph into a spectral domain, and filtering the undirected graph containing the human skeleton information in the spectral domain. For undirected graph G ═ (v, a, X), the laplacian matrix L of the undirected graph is defined as follows:
L=D-A
where D is the degree matrix of the vertices and the diagonal elements are, in turn, the degrees of each vertex.
The laplacian matrix L of the undirected graph is a symmetric positive definite matrix. This laplacian matrix L can be decomposed as follows:
L=UΛUT
wherein Λ ═ diag ([ λ ])1,λ2,…,λN]) Is a diagonal matrix of eigenvalues, U ═ U1,u2,…,uN]Is an orthogonal matrix corresponding to the eigenvalue vector.
Given the filter function G (-) of the undirected graph G, the frequency domain filter response of the input signal X can be defined asThe inverse fourier transform of the graph is defined as follows:
thus, the matrix description of the undirected graph filtering can be defined as follows:
and 5: an undirected Graph including skeleton information of a human body is identified by using a Graph Convolution Network (GCN).
And 5.1, calculating a degree matrix D, an adjacent matrix A and a Laplace matrix L for the obtained human body architecture undirected graph G (V, E), and obtaining an output result for the input undirected graph G through graph convolution operation as follows:
wherein l represents the first layer input, and W is the weight matrix, which is the quantity to be solved;
and 5.2, inputting the undirected graph G into 4 graph convolution neural network modules, namely Block1_3x3, Block2_64x3, Block3_128x3 and Block4_256x3 for feature extraction.
Step 6: a human body key skeleton node attention module is added between graph volume networks, key skeleton nodes relevant to actions are highlighted, and therefore feature expression and action identification accuracy of the networks are improved.
For the violation behaviors of operators in the field of the power system, the influence of each skeleton joint on the action recognition is different. Since an action of the operator is usually strongly related to only a few skeletal joints, but weakly related to other joints. Therefore, the invention utilizes the human body key bone node attention module to enhance the information of the action key bone joint, thereby improving the accuracy of illegal action identification of operators. Assuming that the feature map obtained by graph convolution is H, the output feature map H' obtained by the key bone node attention module is as follows.
H′=W(H)H
Wherein WijNot less than 0 andWijthe larger the number of the nodes, the more important the ith joint node is for violation identification.
And 7: and (3) action identification and classification: for the feature maps containing human body posture information extracted by the graph convolution network, all the feature maps are converted into a one-dimensional matrix by using a full connection layer with 256 neuron numbers, and then identification and classification of violation behaviors and risk actions of different operators are realized by using a softmax function.
Claims (1)
1. A power system field operation action risk identification method based on graph convolution comprises the following steps:
step 1, behavior monitoring and image transmission of field operating personnel of the power system: monitoring the behavior of an operator through a monitoring camera arranged on the operation site of the power system, and uploading a monitoring video to a central cloud platform;
step 2, estimating the behavior attitude of the operator: estimating the behavior attitude of the field operator in the monitoring video by using a human body attitude estimation model openposition, and acquiring skeleton information of the behavior action of the field operator;
step 3, constructing an undirected graph containing human skeleton information: connecting the obtained human skeleton information, and constructing an undirected graph G ═ v, A and X containing the action information of the field operator, whereinIs a set of N vertices, A is a weighted adjacency matrix, X is a signal matrix on the vertices, the set of vertices v corresponds to all N joints of the skeleton, the adjacency matrix A represents the linkage between two vertices, where if vertex v isiAnd vertex vjAnd with a direct connection, then Aij1, otherwise Aij=0;
Step 4, transforming the constructed undirected graph into a spectral domain, and filtering the undirected graph containing human skeleton information in the spectral domain;
step 5, performing action recognition on an undirected graph containing human skeleton information by utilizing a space-time graph convolutional network;
step 6, adding a human body key skeleton node attention module between graph volume networks: the feature graph obtained by graph convolution is H, and the output feature graph H' obtained by the key bone node attention module is as follows:
H′=W(H)H
wherein, WijNot less than 0 andWijthe larger the number is, the more important the ith joint node is for violation identification;
step 7, action identification and classification: and for the characteristic diagrams which are extracted by the graph convolution network and contain human body posture information, converting all the characteristic diagrams into one-dimensional matrixes by using a full connection layer, and then identifying and classifying the violation behaviors and the risk actions of different operators by using a softmax function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011033289.9A CN112200030A (en) | 2020-09-27 | 2020-09-27 | Power system field operation action risk identification method based on graph convolution |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011033289.9A CN112200030A (en) | 2020-09-27 | 2020-09-27 | Power system field operation action risk identification method based on graph convolution |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112200030A true CN112200030A (en) | 2021-01-08 |
Family
ID=74008480
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011033289.9A Pending CN112200030A (en) | 2020-09-27 | 2020-09-27 | Power system field operation action risk identification method based on graph convolution |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112200030A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113537051A (en) * | 2021-07-14 | 2021-10-22 | 安徽炬视科技有限公司 | Marine personnel safety operation monitoring algorithm under complex large scene |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109858406A (en) * | 2019-01-17 | 2019-06-07 | 西北大学 | A kind of extraction method of key frame based on artis information |
-
2020
- 2020-09-27 CN CN202011033289.9A patent/CN112200030A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109858406A (en) * | 2019-01-17 | 2019-06-07 | 西北大学 | A kind of extraction method of key frame based on artis information |
Non-Patent Citations (5)
Title |
---|
关晨至: "《深度学习与人体姿态动作识别》", 30 November 2019, 武汉大学出版社 * |
吴晨晖: "基于图卷积神经网络的人脸表情识别研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
李超龙: "基于图卷积神经网络的人体动作识别研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
李龙: "融合注意力机制的人体骨骼点动作识别方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
杨杰: "《人工智能基础》", 30 April 2020, 机械工业出版社 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113537051A (en) * | 2021-07-14 | 2021-10-22 | 安徽炬视科技有限公司 | Marine personnel safety operation monitoring algorithm under complex large scene |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110674772B (en) | Intelligent safety control auxiliary system and method for electric power operation site | |
CN107730017A (en) | Intelligent metallurgical equipment operational diagnostics system | |
CN110379036A (en) | Intelligent substation patrol recognition methods, system, device and storage medium | |
CN113947555A (en) | Infrared and visible light fused visual system and method based on deep neural network | |
CN112257500A (en) | Intelligent image recognition system and method for power equipment based on cloud edge cooperation technology | |
CN113642403B (en) | Crowd abnormal intelligent safety detection system based on edge calculation | |
CN115585731A (en) | Space-air-ground integrated hydropower station space state intelligent monitoring management system and method thereof | |
CN112200030A (en) | Power system field operation action risk identification method based on graph convolution | |
Tang et al. | Research on intelligent substation monitoring by image recognition method | |
CN116485135A (en) | Wisdom commercial building big data management platform | |
CN116933626A (en) | Data monitoring method and device based on digital twinning | |
LU501401B1 (en) | Action risk identification method for power system field operation based on graph convolution | |
Chang et al. | Safety risk assessment of electric power operation site based on variable precision rough set | |
CN113284103B (en) | Substation equipment defect online detection method based on space transformation fast R-CNN model | |
CN115859344A (en) | Secret sharing-based safe sharing method for data of federal unmanned aerial vehicle group | |
CN111985625A (en) | Infrared-visible light fused deep neural network and modeling method thereof | |
Zheng et al. | Research on optimization of agricultural machinery fault monitoring system based on artificial neural network algorithm. | |
CN113780224B (en) | Unmanned inspection method and system for transformer substation | |
Menghao et al. | Research on the Application of Artificial Intelligence Technology in Safety Control of Electric Power Operation Site | |
Yu et al. | Disaster Intelligent Perception and Emergency Command of Power Grid | |
Xu et al. | A Method of Multi-Information Perception Collection for Power Equipment Disaster Damage | |
Zhang et al. | A Power Operation Site Safety Assistant System Based on Convolutional Neural Network | |
Li | Design of Power Intelligent Auxiliary Control and Monitoring System Based on IoT 3D Image Processing Technology | |
Yuan et al. | Research on anomaly detection technology of power equipment based on AI vision analysis technology | |
Feng | Design of the safety Monitoring system for Civil engineering Construction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |