Hidden danger identification method and device based on knowledge graph, computer equipment and readable storage medium

Abstract

The invention relates to a knowledge graph-based hidden danger identification method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining an electric power knowledge graph according to equipment attribute feature information corresponding to different equipment types in an electric power scene; the electric power knowledge graph takes each equipment type and each equipment attribute feature as nodes, and obtains a first sample feature vector representation, a second sample feature vector representation and a third sample feature vector representation corresponding to the electromechanical equipment sample image; fusing the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation to obtain a fused sample feature vector representation; and obtaining a pre-trained equipment type identification model according to the fused sample feature vector representation and the electromechanical equipment sample image, so as to judge whether a hidden danger point exists in a power scene associated with the to-be-detected image. By adopting the method, the hidden danger identification reliability of the power system can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a knowledge graph-based hidden danger identification method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] Power system monitoring and operation and maintenance refers to the real-time monitoring, analysis and management of the power system operating status to ensure the safe, stable and efficient operation of the power grid.

[0003] In order to ensure the stable operation of the power system, it is necessary to monitor the power system for abnormalities to determine whether there are hidden dangers in the power system. In related technologies, the main monitoring method is to rely on image / video for picture recognition to determine whether there are hidden dangers. However, the power scene is complex and changeable. The obstruction of obstacles and the blurring of the picture caused by the environment will affect the accuracy of image recognition, resulting in untimely identification of hidden dangers or frequent false alarms, affecting the reliability of hidden danger identification in the power system.

[0004] Therefore, there is a problem in the related art that the reliability of hidden danger identification of the power system is low. Summary of the invention

[0005] Based on this, it is necessary to provide a knowledge graph-based hidden danger identification method, device, computer equipment, computer-readable storage medium and computer program product that can improve the reliability of hidden danger identification in the power system in response to the above-mentioned technical problems.

[0006] In the first aspect, the present application provides a hidden danger identification method based on knowledge graph, including:

[0007] According to the device attribute feature information corresponding to different device types in the power scenario, an electric power knowledge graph is obtained; the device attribute feature information includes device attribute features; the electric power knowledge graph uses each device type and each device attribute feature as a node, and uses the association between the device type and the device attribute feature as an edge between the nodes;

[0008] Obtaining a first sample feature vector representation, a second sample feature vector representation, and a third sample feature vector representation corresponding to a sample image of an electromechanical device;

[0009] The first sample feature vector is represented by extracting attribute features from the electromechanical equipment sample image according to the attribute factors associated with the equipment attribute features; the second sample feature vector is represented by the attribute feature vector associated with the corresponding node in the electric power knowledge graph to which the electromechanical equipment in the electromechanical equipment sample image belongs; the third sample feature vector is represented by the image semantic feature vector corresponding to the electromechanical equipment sample image;

[0010] Fusing the first sample feature vector representation, the second sample feature vector representation, and the third sample feature vector representation to obtain a fused sample feature vector representation;

[0011] The device type recognition model to be trained is trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained device type recognition model; the pre-trained device type recognition model is used to output the device type recognition result of the image to be detected; the device type recognition result is used to determine whether there is a hidden danger point in the power scene associated with the image to be detected.

[0012] In one embodiment, obtaining the power knowledge graph according to the device attribute feature information corresponding to different device types in the power scenario includes:

[0013] Acquire device attribute characteristic information corresponding to the different device types; the device attribute characteristic information includes device attribute characteristics corresponding to each preset attribute factor;

[0014] The power knowledge graph is constructed by taking the device attribute characteristics corresponding to each of the device types and each of the attribute factors as nodes, and taking the association relationship between the device types and the device attribute characteristics as edges between the nodes.

[0015] In one embodiment, obtaining a first sample feature vector representation corresponding to a sample image of an electromechanical device includes:

[0016] According to the feature extraction rules corresponding to each of the attribute factors, attribute features are extracted from the electromechanical equipment sample images to obtain image attribute sample feature vector representations corresponding to each of the attribute factors;

[0017] The image attribute sample feature vector representations corresponding to the attribute factors are fused to obtain the first sample feature vector representation.

[0018] In one embodiment, the process of obtaining the second sample feature vector representation includes:

[0019] Obtaining a node feature vector representation corresponding to each of the nodes in the electric power knowledge graph; the node feature vector representation is obtained by performing graph representation learning on the electric power knowledge graph;

[0020] In the node feature vector representation corresponding to each of the nodes, the attribute feature vector representation associated with the node representing the sample device type is determined as the second sample feature vector representation; the attribute feature vector representation includes the node feature vector representation corresponding to the attribute node associated with the node representing the sample device type; the attribute node is a node representing the attribute characteristics of the device.

[0021] In the second aspect, the present application provides a hidden danger identification method based on knowledge graph, including:

[0022] Obtaining a first eigenvector representation, a second eigenvector representation, and a third eigenvector representation corresponding to an image to be detected; the image to be detected is an image obtained by photographing an electric power scene to be detected;

[0023] The first feature vector is represented by extracting attribute features from the image to be detected according to preset attribute factors; the preset attribute factors are attribute factors associated with device attribute features of electromechanical devices of different device types in the power scene;

[0024] The second feature vector is represented as an attribute feature vector representation in the electric power knowledge graph whose similarity with the first feature vector representation satisfies a preset condition; the electric power knowledge graph uses each of the equipment types and each of the equipment attribute features as nodes, and uses the association between the equipment type and the equipment attribute features as edges between the nodes; the attribute feature vector is represented as a node feature vector representation corresponding to the node representing the equipment attribute features;

[0025] The third feature vector representation is an image semantic feature vector representation corresponding to the image to be detected;

[0026] fusing the first feature vector representation, the second feature vector representation, and the third feature vector representation to obtain a fused feature vector representation;

[0027] The fused feature vector representation and the image to be detected are input into a pre-trained device type recognition model, and a device type recognition result of the image to be detected is output; the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected; the pre-trained device type recognition model is obtained according to the knowledge graph-based hidden danger recognition method as described in the first aspect or any possible implementation method of the first aspect.

[0028] In a third aspect, the present application also provides a hidden danger identification device based on a knowledge graph, comprising:

[0029] A graph acquisition module, for acquiring a power knowledge graph according to device attribute feature information corresponding to different device types in a power scenario; the device attribute feature information includes device attribute features; the power knowledge graph has each device type and each device attribute feature as a node, and has an association relationship between the device type and the device attribute feature as an edge between the nodes;

[0030] A sample vector acquisition module is used to acquire a first sample feature vector representation, a second sample feature vector representation and a third sample feature vector representation corresponding to a sample image of an electromechanical device; the first sample feature vector representation is obtained by extracting attribute features of the sample image of the electromechanical device according to attribute factors associated with the device attribute features; the second sample feature vector representation is an attribute feature vector representation of the sample device type to which the electromechanical device in the sample image of the electromechanical device belongs, and the corresponding node in the electric power knowledge graph; the third sample feature vector representation is an image semantic feature vector representation corresponding to the sample image of the electromechanical device;

[0031] A sample vector fusion module, used for fusing the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation to obtain a fused sample feature vector representation;

[0032] A training module is used to train the equipment type recognition model to be trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained equipment type recognition model; the pre-trained equipment type recognition model is used to output the equipment type recognition result of the image to be detected; the equipment type recognition result is used to determine whether there are hidden dangers in the power scene associated with the image to be detected.

[0033] In a fourth aspect, the present application also provides a hidden danger identification device based on a knowledge graph, including:

[0034] A vector acquisition module, used to acquire a first feature vector representation, a second feature vector representation and a third feature vector representation corresponding to an image to be detected; the image to be detected is an image obtained by photographing the power scene to be detected;

[0035] The first feature vector is represented by extracting attribute features from the image to be detected according to preset attribute factors; the preset attribute factors are attribute factors associated with device attribute features of electromechanical devices of different device types in the power scene;

[0036] The second feature vector is represented as an attribute feature vector representation in the electric power knowledge graph whose similarity with the first feature vector representation satisfies a preset condition; the electric power knowledge graph uses each of the equipment types and each of the equipment attribute features as nodes, and uses the association between the equipment type and the equipment attribute features as edges between the nodes; the attribute feature vector is represented as a node feature vector representation corresponding to the node representing the equipment attribute features;

[0037] The third feature vector representation is an image semantic feature vector representation corresponding to the image to be detected;

[0038] A vector fusion module, used for fusing the first feature vector representation, the second feature vector representation and the third feature vector representation to obtain a fused feature vector representation;

[0039] A detection module, used to input the fused feature vector representation and the image to be detected into a pre-trained device type recognition model, and output a device type recognition result of the image to be detected; the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected; the pre-trained device type recognition model is obtained according to the hidden danger recognition method based on the knowledge graph as described in the first aspect or any possible implementation method of the first aspect.

[0040] In a fifth aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the hidden danger identification method based on the knowledge graph as described in the first aspect or any possible implementation method of the first aspect, and / or the steps of the hidden danger identification method based on the knowledge graph as described in the second aspect.

[0041] In the sixth aspect, the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the hidden danger identification method based on the knowledge graph as described in the first aspect or any possible implementation of the first aspect, and / or the steps of the hidden danger identification method based on the knowledge graph as described in the second aspect.

[0042] In the seventh aspect, the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the hidden danger identification method based on the knowledge graph as described in the first aspect or any possible implementation of the first aspect, and / or the steps of the hidden danger identification method based on the knowledge graph as described in the second aspect.

[0043] The above-mentioned hidden danger identification method, device, computer equipment, computer-readable storage medium and computer program product based on knowledge graph obtain the power knowledge graph according to the equipment attribute feature information corresponding to different equipment types in the power scene; the equipment attribute feature information includes equipment attribute features; the power knowledge graph uses each equipment type and each equipment attribute feature as a node, and the association relationship between the equipment type and the equipment attribute feature as the edge between the nodes; the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation corresponding to the electromechanical equipment sample image are obtained; the first sample feature vector representation is obtained by extracting the attribute features of the electromechanical equipment sample image according to the attribute factors associated with the equipment attribute features; the second sample feature vector representation The sample equipment type to which the electromechanical equipment in the electromechanical equipment sample image belongs is represented by the attribute feature vector associated with the corresponding node in the electric power knowledge graph; the third sample feature vector representation is the image semantic feature vector representation corresponding to the electromechanical equipment sample image; the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation are fused to obtain a fused sample feature vector representation; the equipment type recognition model to be trained is trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained equipment type recognition model; the pre-trained equipment type recognition model is used to output the equipment type recognition result of the image to be detected; the equipment type recognition result is used to determine whether there are hidden dangers in the electric power scene associated with the image to be detected.

[0044] In this way, after constructing the power knowledge graph according to the device attribute feature information corresponding to different equipment types in the power scene, the attribute feature of the electromechanical equipment sample image is extracted according to the attribute factors associated with the device attribute features to obtain the first sample feature vector representation, the sample device type to which the electromechanical equipment in the electromechanical equipment sample image belongs, the attribute feature vector representation associated with the corresponding node in the power knowledge graph, and the image semantic feature vector representation corresponding to the electromechanical equipment sample image are obtained, and these feature vector representations are fused to achieve the fusion of the image attribute features extracted by the knowledge graph and the high-level semantic features extracted by the neural network, so that the fused vector contains the device category and semantic information, and a more reliable device type recognition model is trained based on the fused vector and the electromechanical equipment sample image, so that the device type recognition model can accurately detect the device type recognition results of the image to be detected in complex scenes, reduce false positives and false negatives in the device recognition process, thereby improving the accuracy of the device type recognition results, and then based on the device type recognition results output by the model, it can be more accurately judged whether there are hidden danger points in the power scene, which can improve the accuracy of hidden danger recognition and achieve the technical effect of improving the reliability of hidden danger recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A schematic diagram of a process of a hidden danger identification method based on a knowledge graph in an embodiment;

[0047] Figure 2 A flowchart of another hidden danger identification method based on a knowledge graph in an embodiment;

[0048] Figure 3 It is a flowchart of a hidden danger identification method based on knowledge graph in another embodiment;

[0049] Figure 4 A process framework diagram of yet another hidden danger identification method based on knowledge graph in another embodiment;

[0050] Figure 5 It is a structural block diagram of a hidden danger identification device based on knowledge graph in one embodiment;

[0051] Figure 6 It is a structural block diagram of another hidden danger identification device based on knowledge graph in one embodiment;

[0052] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0054] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0055] In one embodiment, Figure 1 As shown, a method for identifying hidden dangers based on a knowledge graph is provided. This embodiment takes the method applied to a computer device as an example. It can be understood that the computer device can be a terminal, a server, or a system including a terminal and a server. In this embodiment, the method includes the following steps:

[0056] Step S110, obtaining a power knowledge graph based on device attribute feature information corresponding to different device types in the power scenario.

[0057] Among them, the power scenario can be the specific circumstances or situations of various power applications in the power system, such as household power scenarios, industrial power scenarios, renewable energy scenarios or power dispatch scenarios.

[0058] Among them, the device attribute characteristic information includes device attribute characteristics.

[0059] Among them, the power knowledge graph takes each equipment type and each equipment attribute feature as a node, and the association relationship between equipment type and equipment attribute features as the edge between nodes.

[0060] Among them, the equipment type can be the equipment type corresponding to different electromechanical equipment in the power scene, such as electromechanical equipment including wires, cranes, hooks, etc.

[0061] Among them, the device attribute characteristics may include device attribute characteristics corresponding to the preset attribute factors. For example, if the preset attribute factors include at least one of color attribute, structure attribute, detail attribute, etc., then the corresponding device attribute characteristics may include at least one of device color feature, device structure feature and device detail feature.

[0062] In practical applications, the equipment attribute feature information of different electromechanical equipment can be collected. For example, common wires in power scenarios have slender appearance characteristics; cranes have obvious tower or arm structures; hook machines have hooks, etc. Specifically, the equipment attribute feature information corresponding to different equipment types includes equipment attribute features of different equipment types corresponding to various preset attribute factors. For example, the equipment attribute feature information corresponding to different equipment types may include equipment color features, equipment structure features, and equipment detail features of different equipment types. These features help to accurately identify equipment even in the case of partial occlusion, so as to improve the stability and reliability of hidden danger identification in complex scenarios.

[0063] In this way, computer equipment can obtain the power knowledge graph based on the device attribute feature information corresponding to different device types in the power scenario; the power knowledge graph uses each device type and each device attribute feature as a node, and the association relationship between the device type and the device attribute feature as the edge between the nodes.

[0064] In some embodiments, the construction of an electric power knowledge graph may include the following steps: obtaining device attribute feature information corresponding to different device types; the device attribute feature information includes device attribute features corresponding to preset attribute factors; in this way, the computer device may construct an electric power knowledge graph with the device attribute features corresponding to each device type and each attribute factor as nodes, and the association relationship between the device type and the device attribute features as edges between the nodes.

[0065] Among them, the nodes representing the equipment types in the power knowledge graph can be named equipment type nodes, and the nodes representing the equipment attribute characteristics can be named attribute nodes.

[0066] Exemplarily, the points in the power knowledge graph include equipment types Equipment attribute characteristics corresponding to each attribute factor .in Represents a collection of device type nodes, Indicates the number of attribute nodes.

[0067] Step S120 , obtaining a first sample feature vector representation, a second sample feature vector representation, and a third sample feature vector representation corresponding to the electromechanical device sample image.

[0068] The electromechanical equipment sample images may be sample images containing electromechanical equipment in an electric power scene.

[0069] The first sample feature vector is represented by the attribute factor associated with the device attribute feature, and is obtained by extracting the attribute features of the electromechanical device sample image.

[0070] In some embodiments, the process of obtaining the first sample feature vector representation includes the following steps: the computer device can extract attribute features of the electromechanical equipment sample image according to the feature extraction rules corresponding to each attribute factor, and obtain the image attribute sample feature vector representation corresponding to each attribute factor; then, the image attribute sample feature vector representation corresponding to each attribute factor is fused to obtain the first sample feature vector representation corresponding to the electromechanical equipment sample image.

[0071] Exemplarily, the attribute factors may include color attributes, structural attributes, and detail attributes. The computer device may convert the color information of the electromechanical device sample image into a feature vector representation, and obtain the image color feature vector representation corresponding to the electromechanical device sample image as the image attribute sample feature vector representation corresponding to the color attribute, recorded as vector ; Use edge detection algorithm (such as Sobel edge detection algorithm) to extract the structural features of the electromechanical equipment sample image, and obtain the image structural feature vector representation corresponding to the electromechanical equipment sample image, as the image attribute sample feature vector representation corresponding to the structural attribute, recorded as vector ; Use detail feature extraction algorithms (such as Gray-level Co-occurrence Matrix (GLCM) or Local Binary Pattern (LBP)) to extract image detail features of the electromechanical equipment sample image, and obtain the image detail feature vector representation corresponding to the electromechanical equipment sample image, which is used as the image attribute sample feature vector representation corresponding to the detail attribute, recorded as vector Finally, the feature fusion of the image attribute sample feature vector representation corresponding to each attribute factor is performed to obtain the first sample feature vector representation as follows: .

[0072] In practical applications, the first sample feature vector representation can be named as a manual attribute feature vector representation.

[0073] Among them, the second sample feature vector is represented by the sample equipment type to which the electromechanical equipment in the electromechanical equipment sample image belongs, and is represented by the attribute feature vector associated with the corresponding node in the power knowledge graph.

[0074] In some embodiments, the computer device may use a graph convolutional network (e.g., GCT (Graph Convolutional Tracking)) to perform graph representation learning on the electric power knowledge graph, map the electric power knowledge graph into a concept feature vector representation, and obtain a node feature vector representation corresponding to each node in the electric power knowledge graph. Among them, the nodes include device type nodes and attribute nodes, so that the node feature vector representation corresponding to each node may include a node feature vector representation corresponding to each device type node and a node feature vector representation corresponding to each attribute node. Thus, the computer device may index in the electric power knowledge graph based on the device type label (Label) corresponding to the sample image of the electromechanical device, and obtain an attribute feature vector representation associated with the device type node representing the sample device type in the electric power knowledge graph, and the associated attribute feature vector representation may include a node feature vector representation corresponding to the attribute node associated with the device type node representing the sample device type in the electric power knowledge graph, wherein the attribute node associated with the device type node representing the sample device type may be an attribute node connected to the device type node representing the sample device type. Specifically, when there are multiple attribute nodes connected to the device type node, the node feature vector representations corresponding to the attribute nodes connected to the device type node are fused, and the fused node feature vector representation is used as the attribute feature vector representation associated with the device type node.

[0075] In this way, the attribute feature vector representation of the node association representing the sample device type can be used as the second sample feature vector representation, which is recorded as .

[0076] Among them, the similarity loss between the manual attribute feature vector representation and the concept feature vector representation mapped to the power knowledge graph is:

[0077] .

[0078] The goal is to minimize the similarity loss , so that the manual attribute feature vector representation is consistent with the concept feature vector representation obtained by mapping the power knowledge graph.

[0079] The third sample feature vector representation is an image semantic feature vector representation corresponding to the electromechanical equipment sample image.

[0080] In a specific implementation, the computer device may input the electromechanical device sample image into a pre-trained semantic feature extraction model to obtain an image semantic feature vector representation corresponding to the electromechanical device sample image as the third sample feature vector representation.

[0081] In practical applications, Resnet (Residual Neural Network) can be used to extract semantic features. (Size is ),in is the height, is the width, is the number of channels. After the convolution layer, multiple residual blocks and pooling layers, the feature map is converted into a fixed-length feature vector to obtain the image semantic feature vector representation: .

[0082] Step S130: fusing the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation to obtain a fused sample feature vector representation.

[0083] In a specific implementation, the computer device may represent the first sample feature vector as , the second sample feature vector represents And the third sample feature vector represents Perform feature fusion to obtain the fused sample feature vector representation: .

[0084] Step S140 , training the device type recognition model to be trained according to the fused sample feature vector representation and the electromechanical device sample image to obtain a pre-trained device type recognition model.

[0085] The pre-trained device type recognition model is used to output the device type recognition result of the image to be detected.

[0086] The image to be detected is an image that needs to be detected by device type.

[0087] Among them, the device type recognition result is used to determine whether there are hidden danger points in the power scene associated with the image to be detected.

[0088] The device type recognition model to be trained may be a deep learning model built on the basis of a YOLO (You Only Look Once) framework.

[0089] In the specific implementation, the computer device can train the equipment type recognition model to be trained based on the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained equipment type recognition model. Specifically, the fused sample feature vector representation is combined with the YOLO framework, a fusion layer is inserted before the detection head of YOLO, and the detection results of YOLO and the classification results of the fusion model are combined to obtain the final more refined equipment type recognition results. The equipment type recognition results are used to determine whether there are hidden dangers in the power scene associated with the image to be detected.

[0090] In the above-mentioned hidden danger identification method based on knowledge graph, the power knowledge graph is obtained according to the equipment attribute feature information corresponding to different equipment types in the power scene; the equipment attribute feature information includes equipment attribute features; the power knowledge graph uses each equipment type and each equipment attribute feature as a node, and the association relationship between the equipment type and the equipment attribute feature as the edge between the nodes; the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation corresponding to the electromechanical equipment sample image are obtained; the first sample feature vector representation is obtained by extracting the attribute features of the electromechanical equipment sample image according to the attribute factors associated with the equipment attribute features; the second sample feature vector representation is the electromechanical equipment in the electromechanical equipment sample image The sample equipment type to which the equipment belongs is represented by the attribute feature vector associated with the corresponding node in the electric power knowledge graph; the third sample feature vector is represented as the image semantic feature vector representation corresponding to the electromechanical equipment sample image; the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation are fused to obtain the fused sample feature vector representation; the equipment type recognition model to be trained is trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained equipment type recognition model; the pre-trained equipment type recognition model is used to output the equipment type recognition result of the image to be detected; the equipment type recognition result is used to determine whether there are hidden dangers in the electric power scene associated with the image to be detected.

[0091] In this way, after constructing the power knowledge graph according to the device attribute feature information corresponding to different equipment types in the power scene, the attribute feature of the electromechanical equipment sample image is extracted according to the attribute factors associated with the device attribute features to obtain the first sample feature vector representation, the sample device type to which the electromechanical equipment in the electromechanical equipment sample image belongs, the attribute feature vector representation associated with the corresponding node in the power knowledge graph, and the image semantic feature vector representation corresponding to the electromechanical equipment sample image are obtained, and these feature vector representations are fused to achieve the fusion of the image attribute features extracted by the knowledge graph and the high-level semantic features extracted by the neural network, so that the fused vector contains the device category and semantic information, and a more reliable device type recognition model is trained based on the fused vector and the electromechanical equipment sample image, so that the device type recognition model can accurately detect the device type recognition results of the image to be detected in complex scenes, reduce false positives and false negatives in the device recognition process, thereby improving the accuracy of the device type recognition results, and then based on the device type recognition results output by the model, it can be more accurately judged whether there are hidden danger points in the power scene, which can improve the accuracy of hidden danger recognition and achieve the technical effect of improving the reliability of hidden danger recognition.

[0092] In one embodiment, Figure 2 As shown, a flow chart of another hidden danger identification method based on knowledge graph is provided. This embodiment takes the method applied to a computer device as an example. The method includes the following steps:

[0093] Step S210, obtaining a first feature vector representation, a second feature vector representation and a third feature vector representation corresponding to the image to be detected.

[0094] The image to be detected is an image obtained by photographing the power scene to be detected.

[0095] Among them, the power scene to be detected may be a power scene that requires hidden danger identification.

[0096] The first feature vector is represented by extracting attribute features from the image to be detected according to preset attribute factors.

[0097] Among them, the preset attribute factors are attribute factors associated with device attribute characteristics of electromechanical equipment of different device types in the power scenario.

[0098] The attribute factors may include at least one attribute factor such as color attribute, structure attribute, detail attribute, etc.

[0099] Specifically, the computer device can extract attribute features of the image to be detected according to the feature extraction rules corresponding to each attribute factor, and obtain the image attribute feature vector representation corresponding to each attribute factor; then, the image attribute feature vector representation corresponding to each attribute factor is merged to obtain the first feature vector representation corresponding to the image to be detected.

[0100] Among them, the second feature vector representation is an attribute feature vector representation in the power knowledge graph whose similarity with the first feature vector representation meets preset conditions.

[0101] Among them, the power knowledge graph takes each equipment type and each equipment attribute feature as a node, and the association relationship between equipment type and equipment attribute features as the edge between nodes.

[0102] The attribute feature vector is represented as a node feature vector corresponding to a node representing the attribute feature of the device.

[0103] In a specific implementation, each device type node in the electric power knowledge graph is connected to a corresponding attribute node. According to the node feature vector representation corresponding to the attribute node connected to each device type node, the attribute feature vector representation associated with the device type node can be obtained. When there are multiple attribute nodes connected to the device type node, the node feature vector representation corresponding to each attribute node connected to the device type node can be fused, and the node feature vector representation obtained by fusion is used as the attribute feature vector representation associated with the device type node. In this way, there are multiple device type nodes in the electric power knowledge graph, and each device type node is associated with a corresponding attribute feature vector representation.

[0104] Thus, the computer device can calculate the feature similarity between the first feature vector representation and the attribute feature vector representation associated with each device type node in the power knowledge graph, and use the attribute feature vector representation whose feature similarity meets the preset conditions as the second feature vector representation corresponding to the image to be detected. In some embodiments, the attribute feature vector representation with the greatest feature similarity can be used as the second feature vector representation corresponding to the image to be detected.

[0105] The third feature vector is represented as an image semantic feature vector corresponding to the image to be detected.

[0106] In a specific implementation, the computer device can input the image to be detected into a pre-trained semantic feature extraction model to obtain an image semantic feature vector representation corresponding to the image to be detected as the third feature vector representation.

[0107] Step S220, fusing the first feature vector representation, the second feature vector representation and the third feature vector representation to obtain a fused feature vector representation.

[0108] In a specific implementation, the computer device may fuse the first feature vector representation, the second feature vector representation, and the third feature vector representation to obtain a fused feature vector representation.

[0109] Step S230, inputting the fused feature vector representation and the image to be detected into a pre-trained device type recognition model, and outputting a device type recognition result of the image to be detected.

[0110] Among them, the equipment type identification result is used to determine whether there are hidden danger points in the power scene to be tested.

[0111] Among them, the pre-trained equipment type recognition model is obtained according to the hidden danger recognition method based on the knowledge graph as described above.

[0112] In a specific implementation, the computer device can input the fused feature vector representation and the image to be detected into a pre-trained device type recognition model, perform target detection on the image to be detected through the device type recognition model combined with the fused feature vector representation, locate the electromechanical equipment in the image to be detected and predict the device type corresponding to the detected electromechanical equipment, thereby outputting the device type recognition result of the image to be detected, and the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected. In this way, after the accuracy of electromechanical equipment recognition is improved by combining the pre-trained device type recognition model of the YOLO framework, it is possible to determine whether there are hidden danger points by judging the recognized device type. For example, if there are both wires and cranes, a safety hazard prompt will be triggered.

[0113] In the above-mentioned hidden danger identification method based on knowledge graph, the first feature vector representation, the second feature vector representation and the third feature vector representation corresponding to the image to be detected are obtained; the image to be detected is an image obtained by photographing the power scene to be detected; the first feature vector representation is obtained by extracting attribute features of the image to be detected according to preset attribute factors; the preset attribute factors are attribute factors associated with the device attribute features of electromechanical equipment of different equipment types in the power scene; the second feature vector representation is an attribute feature vector representation in the power knowledge graph whose similarity with the first feature vector representation meets the preset conditions; the power knowledge graph uses each equipment type and each equipment attribute feature as a node, and the relationship between the equipment type and the equipment attribute feature is expressed as a node. The association relationship between them is the edge between the nodes; the attribute feature vector is represented as the node feature vector representation corresponding to the node representing the attribute characteristics of the device; the third feature vector is represented as the image semantic feature vector representation corresponding to the image to be detected; the first feature vector representation, the second feature vector representation and the third feature vector representation are fused to obtain a fused feature vector representation; the fused feature vector representation and the image to be detected are input into a pre-trained device type recognition model, and the device type recognition result of the image to be detected is output; the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected; the pre-trained device type recognition model is obtained by the hidden danger recognition method based on the knowledge graph such as the first aspect or any possible implementation method of the first aspect.

[0114] In this way, by fusing the image attribute features corresponding to each attribute factor of the image to be detected, the attribute features extracted based on the power knowledge graph, and the high-level semantic features extracted by the neural network, the fused vector contains the equipment category and semantic information. Combined with the detection capability of the pre-trained equipment type recognition model, it reduces false alarms and missed alarms in the equipment identification process, improves the reliability and practicality of the model, and thus improves the accuracy of hidden danger identification.

[0115] Figure 3 A flowchart of another hidden danger identification method based on knowledge graph is provided, such as Figure 3 As shown, the following steps are included:

[0116] Step S302: Acquire device attribute feature information corresponding to different device types.

[0117] Step S304, constructing an electric power knowledge graph with the device attribute features corresponding to each device type and each attribute factor as nodes, and the association relationship between the device type and the device attribute features as edges between the nodes.

[0118] Step S306 , extracting attribute features from the electromechanical equipment sample images according to the feature extraction rules corresponding to each attribute factor, and obtaining image attribute sample feature vector representations corresponding to each attribute factor.

[0119] Step S308: fusing the image attribute sample feature vector representations corresponding to the attribute factors to obtain a first sample feature vector representation.

[0120] Step S310, obtaining the node feature vector representation corresponding to each node in the power knowledge graph.

[0121] Step S312: Determine, from the node feature vector representations corresponding to each node, a node-associated attribute feature vector representation that characterizes the sample device type as a second sample feature vector representation.

[0122] Step S314: fuse the first sample feature vector representation, the second sample feature vector representation, and the third sample feature vector representation to obtain a fused sample feature vector representation.

[0123] Step S316, training the device type recognition model to be trained according to the fused sample feature vector representation and the electromechanical device sample image to obtain a pre-trained device type recognition model.

[0124] Step S318, obtaining a first feature vector representation, a second feature vector representation and a third feature vector representation corresponding to the image to be detected.

[0125] Step S320: fuse the first feature vector representation, the second feature vector representation, and the third feature vector representation to obtain a fused feature vector representation.

[0126] Step S322, input the fused feature vector representation and the image to be detected into a pre-trained device type recognition model, and output the device type recognition result of the image to be detected.

[0127] It should be noted that the specific limitations of the above steps can be found in the specific limitations of a hidden danger identification method based on a knowledge graph above, which will not be repeated here.

[0128] To facilitate understanding by those skilled in the art, in another embodiment, Figure 4 Another process framework diagram of a hidden danger identification method based on knowledge graph is provided. Figure 4As shown, for the input electromechanical equipment sample image, the manual feature vector representation (i.e., manual attribute feature vector representation) of the electromechanical equipment sample image can be extracted by traditional image processing methods. That is, the manual feature vector representation is a first sample feature vector representation obtained by extracting attribute features of the electromechanical equipment sample image according to the attribute factors associated with the equipment attribute features. And, based on the electric power knowledge graph, the conceptual feature vector representation of the electromechanical equipment sample image is obtained, that is, the attribute feature vector representation associated with the corresponding node in the electric power knowledge graph to which the electromechanical equipment in the electromechanical equipment sample image belongs is obtained as the conceptual feature vector representation of the electromechanical equipment sample image (equivalent to the second sample feature vector representation). In addition, it is also necessary to extract the image feature vector representation of the electromechanical equipment sample image through a neural network, that is, to extract the image semantic feature vector representation corresponding to the electromechanical equipment sample image through a pre-trained semantic feature extraction model as the image feature vector representation (equivalent to the third sample feature vector representation).

[0129] In this way, the computer device can optimize the parameters by calculating the similarity loss between the manual feature vector representation and the concept feature vector representation, so that the manual feature vector representation is consistent with the concept feature vector representation obtained by mapping the power knowledge graph.

[0130] In addition, the computer device can fuse the manual feature vector representation, the concept feature vector representation and the image feature vector representation, combine the fused feature vector representation with the YOLO framework, and insert a fusion layer before the detection head of YOLO to realize hidden danger identification.

[0131] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0132] Based on the same inventive concept, the embodiment of the present application also provides a knowledge graph-based hidden danger identification device for implementing the knowledge graph-based hidden danger identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the hidden danger identification device based on the knowledge graph provided below can refer to the limitations of the hidden danger identification method based on the knowledge graph above, and will not be repeated here.

[0133] In an exemplary embodiment, Figure 5 As shown, a hidden danger identification device based on knowledge graph is provided, including: a graph acquisition module 510, a sample vector acquisition module 520, a sample vector fusion module 530 and a training module 540, wherein:

[0134] The graph acquisition module 510 is used to acquire the power knowledge graph according to the device attribute feature information corresponding to different device types in the power scenario; the device attribute feature information includes device attribute features; the power knowledge graph uses each device type and each device attribute feature as a node, and the association relationship between the device type and the device attribute feature as an edge between the nodes.

[0135] The sample vector acquisition module 520 is used to obtain a first sample feature vector representation, a second sample feature vector representation and a third sample feature vector representation corresponding to a sample image of electromechanical equipment; the first sample feature vector representation is obtained by extracting attribute features of the sample image of electromechanical equipment according to attribute factors associated with the device attribute features; the second sample feature vector representation is an attribute feature vector representation associated with a corresponding node in the electric power knowledge graph to which the electromechanical equipment in the sample image of electromechanical equipment belongs; the third sample feature vector representation is an image semantic feature vector representation corresponding to the sample image of electromechanical equipment.

[0136] The sample vector fusion module 530 is used to fuse the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation to obtain a fused sample feature vector representation.

[0137] The training module 540 is used to train the equipment type recognition model to be trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained equipment type recognition model; the pre-trained equipment type recognition model is used to output the equipment type recognition result of the image to be detected; the equipment type recognition result is used to determine whether there are hidden dangers in the power scene associated with the image to be detected.

[0138] In one of the embodiments, the graph acquisition module 510 is specifically used to obtain device attribute feature information corresponding to the different device types; the device attribute feature information includes device attribute features corresponding to each preset attribute factor; the device attribute features corresponding to each device type and each attribute factor are used as nodes, and the association relationship between the device type and the device attribute features is used as the edge between the nodes to construct the power knowledge graph.

[0139] In one of the embodiments, the sample vector acquisition module 520 is specifically used to extract attribute features of the electromechanical equipment sample image according to the feature extraction rules corresponding to each of the attribute factors, and obtain the image attribute sample feature vector representation corresponding to each of the attribute factors; and fuse the image attribute sample feature vector representation corresponding to each of the attribute factors to obtain the first sample feature vector representation.

[0140] In one of the embodiments, the sample vector acquisition module 520 is specifically used to obtain the node feature vector representation corresponding to each of the nodes in the power knowledge graph; the node feature vector representation is obtained by performing graph representation learning on the power knowledge graph; in the node feature vector representation corresponding to each of the nodes, the attribute feature vector representation associated with the node representing the sample device type is determined as the second sample feature vector representation; the attribute feature vector representation includes the node feature vector representation corresponding to the attribute node associated with the node representing the sample device type; the attribute node is a node representing the attribute characteristics of the device.

[0141] Based on the same inventive concept, the embodiment of the present application further provides a knowledge graph-based hidden danger identification device for implementing the knowledge graph-based hidden danger identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the hidden danger identification device based on the knowledge graph provided below can refer to the limitations of the hidden danger identification method based on the knowledge graph above, and will not be repeated here.

[0142] In an exemplary embodiment, Figure 6 As shown, a hidden danger identification device based on knowledge graph is provided, including: a vector acquisition module 610, a vector fusion module 620 and a detection module 630, wherein:

[0143] The vector acquisition module 610 is used to obtain the first feature vector representation, the second feature vector representation and the third feature vector representation corresponding to the image to be detected; the image to be detected is an image obtained by photographing the power scene to be detected; the first feature vector representation is obtained by extracting attribute features of the image to be detected according to preset attribute factors; the preset attribute factors are attribute factors associated with the device attribute features of electromechanical equipment of different equipment types in the power scene; the second feature vector representation is an attribute feature vector representation in the power knowledge graph whose similarity with the first feature vector representation meets preset conditions; the power knowledge graph uses each of the equipment types and each of the equipment attribute features as nodes, and the association relationship between the equipment types and the equipment attribute features as edges between the nodes; the attribute feature vector representation is a node feature vector representation corresponding to the node representing the device attribute features; the third feature vector representation is an image semantic feature vector representation corresponding to the image to be detected.

[0144] The vector fusion module 620 is used to fuse the first feature vector representation, the second feature vector representation and the third feature vector representation to obtain a fused feature vector representation.

[0145] The detection module 630 is used to input the fused feature vector representation and the image to be detected into a pre-trained device type recognition model, and output the device type recognition result of the image to be detected; the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected; the pre-trained device type recognition model is obtained according to the knowledge graph-based hidden danger recognition method as described in the first aspect or any possible implementation method of the first aspect.

[0146] Each module in the above-mentioned knowledge graph-based hidden danger identification device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0147] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 7As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be realized through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a hidden danger identification method based on a knowledge graph is realized. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0148] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0149] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0150] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0151] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0152] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0153] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0154] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0155] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A hidden danger identification method based on knowledge graph, characterized in that: The method comprises: According to the device attribute feature information corresponding to different device types in the power scenario, an electric power knowledge graph is obtained; the device attribute feature information includes device attribute features; the electric power knowledge graph uses each device type and each device attribute feature as a node, and uses the association between the device type and the device attribute feature as an edge between the nodes; Obtaining a first sample feature vector representation, a second sample feature vector representation, and a third sample feature vector representation corresponding to a sample image of an electromechanical device; The first sample feature vector is represented by extracting attribute features from the electromechanical equipment sample image according to the attribute factors associated with the equipment attribute features; the second sample feature vector is represented by the attribute feature vector associated with the corresponding node in the electric power knowledge graph to which the electromechanical equipment in the electromechanical equipment sample image belongs; the third sample feature vector is represented by the image semantic feature vector corresponding to the electromechanical equipment sample image; Fusing the first sample feature vector representation, the second sample feature vector representation, and the third sample feature vector representation to obtain a fused sample feature vector representation; The device type recognition model to be trained is trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained device type recognition model; the pre-trained device type recognition model is used to output the device type recognition result of the image to be detected; the device type recognition result is used to determine whether there is a hidden danger point in the power scene associated with the image to be detected.

2. The method according to claim 1, characterized in that The step of obtaining a power knowledge graph according to device attribute feature information corresponding to different device types in the power scenario includes: Acquire device attribute characteristic information corresponding to the different device types; the device attribute characteristic information includes device attribute characteristics corresponding to each preset attribute factor; The power knowledge graph is constructed by taking the device attribute characteristics corresponding to each of the device types and each of the attribute factors as nodes, and taking the association relationship between the device types and the device attribute characteristics as edges between the nodes.

3. The method according to claim 2, characterized in that The obtaining of a first sample feature vector representation corresponding to the electromechanical device sample image includes: According to the feature extraction rules corresponding to each of the attribute factors, attribute features are extracted from the electromechanical equipment sample images to obtain image attribute sample feature vector representations corresponding to each of the attribute factors; The image attribute sample feature vector representations corresponding to the attribute factors are fused to obtain the first sample feature vector representation.

4. The method according to claim 1, characterized in that: The process of obtaining the second sample feature vector representation includes: Obtaining a node feature vector representation corresponding to each of the nodes in the electric power knowledge graph; the node feature vector representation is obtained by performing graph representation learning on the electric power knowledge graph; In the node feature vector representation corresponding to each of the nodes, the attribute feature vector representation associated with the node representing the sample device type is determined as the second sample feature vector representation; the attribute feature vector representation includes the node feature vector representation corresponding to the attribute node associated with the node representing the sample device type; the attribute node is a node representing the attribute characteristics of the device.

5. A hidden danger identification method based on knowledge graph, characterized in that: The method comprises: Obtaining a first eigenvector representation, a second eigenvector representation, and a third eigenvector representation corresponding to an image to be detected; the image to be detected is an image obtained by photographing an electric power scene to be detected; The first feature vector is represented by extracting attribute features from the image to be detected according to preset attribute factors; the preset attribute factors are attribute factors associated with device attribute features of electromechanical devices of different device types in the power scene; The second feature vector is represented as an attribute feature vector representation in the electric power knowledge graph whose similarity with the first feature vector representation satisfies a preset condition; the electric power knowledge graph uses each of the equipment types and each of the equipment attribute features as nodes, and uses the association between the equipment type and the equipment attribute features as edges between the nodes; the attribute feature vector is represented as a node feature vector representation corresponding to the node representing the equipment attribute features; The third feature vector representation is an image semantic feature vector representation corresponding to the image to be detected; fusing the first feature vector representation, the second feature vector representation, and the third feature vector representation to obtain a fused feature vector representation; The fused feature vector representation and the image to be detected are input into a pre-trained device type recognition model, and the device type recognition result of the image to be detected is output; the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected; the pre-trained device type recognition model is obtained according to the hidden danger recognition method based on the knowledge graph as described in any one of claims 1 to 4.

6. A hidden danger identification device based on knowledge graph, characterized in that: The device comprises: A graph acquisition module, for acquiring a power knowledge graph according to device attribute feature information corresponding to different device types in a power scenario; the device attribute feature information includes device attribute features; the power knowledge graph has each device type and each device attribute feature as a node, and has an association relationship between the device type and the device attribute feature as an edge between the nodes; A sample vector acquisition module is used to acquire a first sample feature vector representation, a second sample feature vector representation and a third sample feature vector representation corresponding to a sample image of an electromechanical device; the first sample feature vector representation is obtained by extracting attribute features of the sample image of the electromechanical device according to attribute factors associated with the device attribute features; the second sample feature vector representation is an attribute feature vector representation of the sample device type to which the electromechanical device in the sample image of the electromechanical device belongs, and the corresponding node in the electric power knowledge graph; the third sample feature vector representation is an image semantic feature vector representation corresponding to the sample image of the electromechanical device; A sample vector fusion module, used for fusing the first sample feature vector representation, the second sample feature vector representation and the third sample feature vector representation to obtain a fused sample feature vector representation; A training module is used to train the equipment type recognition model to be trained according to the fused sample feature vector representation and the electromechanical equipment sample image to obtain a pre-trained equipment type recognition model; the pre-trained equipment type recognition model is used to output the equipment type recognition result of the image to be detected; the equipment type recognition result is used to determine whether there are hidden dangers in the power scene associated with the image to be detected.

7. A hidden danger identification device based on knowledge graph, characterized in that: The device comprises: A vector acquisition module, used to acquire a first feature vector representation, a second feature vector representation and a third feature vector representation corresponding to an image to be detected; the image to be detected is an image obtained by photographing the power scene to be detected; The first feature vector is represented by extracting attribute features from the image to be detected according to preset attribute factors; the preset attribute factors are attribute factors associated with device attribute features of electromechanical devices of different device types in the power scene; The second feature vector is represented as an attribute feature vector representation in the electric power knowledge graph whose similarity with the first feature vector representation satisfies a preset condition; the electric power knowledge graph uses each of the equipment types and each of the equipment attribute features as nodes, and uses the association between the equipment type and the equipment attribute features as edges between the nodes; the attribute feature vector is represented as a node feature vector representation corresponding to the node representing the equipment attribute features; The third feature vector representation is an image semantic feature vector representation corresponding to the image to be detected; A vector fusion module, used for fusing the first feature vector representation, the second feature vector representation and the third feature vector representation to obtain a fused feature vector representation; A detection module, used to input the fused feature vector representation and the image to be detected into a pre-trained device type recognition model, and output a device type recognition result of the image to be detected; the device type recognition result is used to determine whether there are hidden danger points in the power scene to be detected; the pre-trained device type recognition model is obtained according to the hidden danger recognition method based on the knowledge graph as described in any one of claims 1 to 4.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images