Foreign object identification method and device and processor for substation equipment

Abstract

The present invention discloses a foreign body identification method for substation equipment and its device and processor. The method comprises: obtaining an equipment image of a target substation equipment; extracting features from the equipment image to obtain a feature image of the target substation equipment; generating a substation equipment relationship diagram based on the equipment image; performing feature mapping on the embedded feature map corresponding to the feature image and the substation equipment relationship diagram to obtain a feature mapping relationship; identifying foreign bodies on the target substation equipment based on the feature mapping relationship, and obtaining an identification result, wherein the identification result includes the relative position relationship between the target substation equipment and the foreign body, and the category of the foreign body. The present invention solves the technical problem that the foreign body identification means for power equipment in the related art is relatively single, resulting in a low identification accuracy.

Description

Technical Field

[0001] The present invention relates to the field of power grid fault identification, and in particular to a foreign body identification method and a device and a processor of a substation equipment. Background Art

[0002] High-voltage equipment inspection is an effective means to ensure the reliability of power supply in the power grid. As a key technology in automated inspection, the importance of target recognition is self-evident. At present, due to limited training data, the accuracy of identifying the power equipment and foreign objects in the substation is not high. However, there is often a strong correlation between power equipment. If this information can be effectively used, the recognition accuracy will be greatly improved.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiment of the present invention provides a foreign object recognition method and device and a processor for substation equipment, so as to at least solve the technical problem that the foreign object recognition means for power equipment in the related art are relatively single, resulting in low recognition accuracy.

[0005] According to one aspect of an embodiment of the present invention, a method for identifying foreign objects in substation equipment is provided, comprising: obtaining an equipment image of a target substation equipment; performing feature extraction on the equipment image to obtain a feature image of the target substation equipment; generating a substation equipment relationship diagram based on the equipment image; performing feature mapping on an embedded feature map corresponding to the feature image and the substation equipment relationship diagram to obtain a feature mapping relationship; performing foreign object identification on the target substation equipment based on the feature mapping relationship, and obtaining an identification result, wherein the identification result includes a relative position relationship between the target substation equipment and the foreign object, and a category of the foreign object.

[0006] Optionally, feature extraction is performed on the device image to obtain a feature image of the target substation equipment, including: extracting features from the device image through a deep self-attention network model to obtain the feature image, wherein the deep self-attention network model is obtained through machine learning training using multiple sets of training data, and each set of the multiple sets of training data includes: historical device images and historical feature images.

[0007] Optionally, a substation equipment relationship diagram is generated based on the equipment image, including: generating multiple candidate frame nodes corresponding to the equipment image; searching for adjacent nodes of each candidate frame node among the multiple candidate frame nodes; connecting each candidate frame node with its adjacent nodes to obtain the substation equipment relationship diagram.

[0008] Optionally, feature mapping is performed on the embedded feature map corresponding to the feature image and the substation equipment relationship map to obtain a feature mapping relationship, including: decoding the embedded feature map to obtain a decoded embedded feature map; feature mapping is performed on the decoded embedded feature map and the substation equipment relationship map to obtain the feature mapping relationship.

[0009] Optionally, after performing foreign object identification on the target substation equipment based on the feature mapping relationship and obtaining the identification result, the method further includes: performing error analysis on the identification result to obtain an error result; and optimizing using a predetermined algorithm based on the error result to obtain an optimized identification result.

[0010] According to another aspect of an embodiment of the present invention, a foreign object identification device for substation equipment is also provided, including: an acquisition module, used to acquire an equipment image of a target substation equipment; an extraction module, used to perform feature extraction on the equipment image to obtain a feature image of the target substation equipment; a generation module, used to generate a substation equipment relationship diagram based on the equipment image; a mapping module, used to perform feature mapping on the embedded feature map corresponding to the feature image and the substation equipment relationship diagram to obtain a feature mapping relationship; and an identification module, used to perform foreign object identification on the target substation equipment based on the feature mapping relationship and obtain an identification result.

[0011] Optionally, the extraction module includes: an extraction unit, used to perform feature extraction on the device image through a deep self-attention network model to obtain the feature image, wherein the deep self-attention network model is obtained through machine learning training using multiple sets of training data, and each set of the multiple sets of training data includes: historical device images and historical feature images.

[0012] Optionally, the generation module includes: a generation unit, used to generate multiple candidate frame nodes corresponding to the equipment image; a search unit, used to search for adjacent nodes of each of the multiple candidate frame nodes; and a connection unit, used to connect each candidate frame node with its adjacent nodes to obtain the substation equipment relationship diagram.

[0013] Optionally, the mapping module includes: a decoding unit, used to decode the embedded feature map to obtain a decoded embedded feature map; a mapping unit, used to perform feature mapping on the decoded embedded feature map and the substation equipment relationship map to obtain the feature mapping relationship.

[0014] Optionally, the device also includes: an analysis module, which is used to identify foreign objects on the target substation equipment based on the feature mapping relationship and obtain the identification result, and then perform error analysis on the identification result to obtain an error result; an optimization module, which is used to optimize using a predetermined algorithm based on the error result to obtain an optimized identification result.

[0015] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is also provided, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed by a processor, the device where the computer-readable storage medium is located is controlled to execute any of the foreign object identification methods for substation equipment described above.

[0016] According to another aspect of an embodiment of the present invention, a processor is further provided, the processor being used to run a computer program, wherein the computer program executes any one of the above-mentioned foreign object identification methods for substation equipment when running.

[0017] In an embodiment of the present invention, an equipment image of a target substation equipment is obtained; feature extraction is performed on the equipment image to obtain a feature image of the target substation equipment; a substation equipment relationship diagram is generated based on the equipment image; feature mapping is performed on the embedded feature diagram corresponding to the feature image and the substation equipment relationship diagram to obtain a feature mapping relationship; foreign objects are identified on the target substation equipment based on the feature mapping relationship, and an identification result is obtained, wherein the identification result includes the relative position relationship between the target substation equipment and the foreign object, and the category of the foreign object. The foreign object identification method for substation equipment provided by the embodiment of the present invention achieves the purpose of using the feature image equipment relationship diagram and the embedded feature diagram of the target substation equipment to identify foreign objects, thereby achieving the technical effect of improving the accuracy of foreign object identification, and further solving the technical problem that the foreign object identification means for power equipment in the related art are relatively single, resulting in low identification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0019] Figure 1 is a flow chart of a foreign object identification method for a substation device according to an embodiment of the present invention;

[0020] Figure 2 is a schematic diagram of a deep self-attention relationship network structure according to an embodiment of the present invention;

[0021] Figure 3 is a schematic diagram of the relationship between nodes of a device according to an embodiment of the present invention;

[0022] Figure 4 is a timing diagram of a foreign object identification method for a substation device according to an embodiment of the present invention

[0023] Figure 5 is a schematic diagram of a foreign object identification device for a substation device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention 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 invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] Example 1

[0027] According to an embodiment of the present invention, a method embodiment of a foreign object identification method for substation equipment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] Figure 1 is a flow chart of a foreign object identification method for a substation device according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0029] Step S102: Acquire a device image of the target substation device.

[0030] Optionally, in the above steps, the original image of the substation equipment is firstly acquired.

[0031] Step S104: extract features from the equipment image to obtain a feature image of the target substation equipment.

[0032] Optionally, in the above steps, feature extraction is performed on the acquired original image, and the extracted features are combined into a feature image.

[0033] In this embodiment, the ResNet convolutional neural network can be used as the basic network to extract image features from the substation equipment image; the extracted feature map is recorded as f∈R C×H×W , where C is the number of channels, H is the height of the feature image, and W is the width of the feature image; then a convolutional neural network with a size of 1×1 is used to fuse and reduce the dimension of the feature channel. The feature map after fusion and dimensionality reduction is recorded as f o ∈R d×H×W , where d is the number of channels after the change; finally, the feature map after fusion and dimensionality reduction is connected pixel by pixel to compress the spatial dimension into one dimension, that is, f o ′∈R d×HW , as the input of the deep self-attention network encoder program submodule.

[0034] Step S106: generating a substation equipment relationship diagram based on the equipment image.

[0035] Optionally, multiple nodes are generated for the equipment image. The role of the node is to mark each local substation part, such as wires, bushings, insulators, etc. in the substation. Then, the nearest adjacent nodes to each node are found and connected to obtain the equipment relationship diagram.

[0036] In an optional embodiment, the device relationship graph is used to define the relationship between substation devices in the substation device image, which is composed of nodes and edges. For a single substation device image, it is assumed that there are N rectangular candidate box nodes, namely b1, b2, ... b N ; where b i =(x i ,y i ,w i ,h i ), where (x, y), w, and h are the center point, width, and height of the rectangular candidate box respectively; in the substation equipment image, b i The neighborhood of is a rectangular box b nbhd (i)=(x i ,y i ,αw i ,αh i ), α ≥ 1 is an adjustable parameter that controls the size of the neighborhood; if and only if b j and b nbhd (i) When there is an intersection, the rectangular candidate box node bj Yes i When constructing the substation equipment relationship diagram in the substation equipment image, N nodes are first generated, each of which corresponds to a rectangular candidate box, and then the adjacent nodes are searched for each node and connected to the adjacent nodes.

[0037] Among them, the deep self-attention network decoder program submodule fuses the features corresponding to each node in the substation equipment relationship diagram in the substation equipment image and the features of each node's neighboring nodes through the fully connected layer, and uses the fused node features as the input of the deep self-attention network decoder; that is, where z i is the i-th embedding feature map of the input deep self-attention network decoder program submodule, is the weight corresponding to the j-th embedding feature map of the input deep self-attention network decoder program submodule, 1 nbhd (i,j) is an indicator variable, which takes the value 1 when j is a neighbor node of i and 0 otherwise.

[0038] Step S108, performing feature mapping on the embedded feature graph corresponding to the feature image and the substation equipment relationship graph to obtain a feature mapping relationship.

[0039] Optionally, the deep self-attention network encoder consists of a multi-branch self-attention module and a feed-forward network. Fixed position encoding is added to the input layer of the self-attention module to ensure that the multi-branch self-attention module is not affected by the arrangement of input variables. The deep self-attention network encoder program submodule outputs an embedded feature map.

[0040] Optionally, in the above steps, the embedded feature map corresponding to the feature image obtained above is fused and mapped with the device relationship map to obtain a mapping relationship.

[0041] Specifically, the deep self-attention network decoder program submodule is composed of a multi-branch self-attention module and an encoding-decoding attention module. It decodes the embedded feature map generated in step S4. The number of embedded feature maps is N, the size is d, and they are different from each other. The deep self-attention network decoder runs in parallel and uses position encoding as the input of the multi-branch attention module; the deep self-attention network decoder program submodule uses the relationship between substation equipment to identify objects that are related to each other. The importance of the relationship between objects depends on the distance between the objects. The closer the distance, the stronger the relationship; the output is a feature map based on the relationship between substation equipment and the self-attention mechanism.

[0042] Step S110, performing foreign object recognition on the target substation equipment based on the feature mapping relationship, and obtaining a recognition result, wherein the recognition result includes a relative position relationship between the target substation equipment and the foreign object, and a category of the foreign object.

[0043] Specifically, the prediction branch feedforward network program submodule predicts the feature map output by the self-attention network decoder; the feedforward network includes a multi-layer perception hidden layer and a linear layer. The dimension of the multi-layer perception hidden layer is d and the ReLU activation function is used. It is responsible for predicting the position, width and height of the substation equipment body and the foreign object boundary box. The linear layer uses a softmax activation function and is responsible for predicting the substation equipment body and the foreign object type.

[0044] As can be seen from the above, in the embodiment of the present invention, firstly, the equipment image of the target substation equipment can be obtained; then, the feature extraction of the equipment image is performed to obtain the feature image of the target substation equipment; then, the substation equipment relationship diagram is generated based on the equipment image; then, the embedded feature diagram corresponding to the feature image and the substation equipment relationship diagram can be feature mapped to obtain the feature mapping relationship; finally, the target substation equipment can be identified as a foreign object based on the feature mapping relationship, and the identification result can be obtained, wherein the identification result includes the relative position relationship between the target substation equipment and the foreign object, and the foreign object category. The foreign object identification method for substation equipment provided by the embodiment of the present invention achieves the purpose of using the feature image equipment relationship diagram and the embedded feature diagram of the target substation equipment to identify foreign objects, thereby achieving the technical effect of improving the accuracy of foreign object identification, and further solving the technical problem that the foreign object identification means for power equipment in the related technology are relatively single, resulting in low identification accuracy.

[0045] As an optional embodiment, feature extraction is performed on the equipment image to obtain a feature image of the target substation equipment, including: extracting features from the equipment image through a deep self-attention network model to obtain a feature image, wherein the deep self-attention network model is obtained through machine learning training using multiple sets of training data, and each set of training data in the multiple sets of training data includes: historical equipment images and historical feature images.

[0046] Figure 2 is a schematic diagram of a deep self-attention relationship network structure according to an embodiment of the present invention. Figure 2 As shown, the deep self-attention relationship network mainly includes four parts: a basic network, an encoder, a decoder, and a prediction branch. In the embodiment of the present invention, the basic network extracts features from the device image, then uses the encoder to generate nodes and rectangular candidate boxes of the image to obtain a relationship feature map, then uses the decoder to decode the embedded feature map, and finally uses the prediction branch to process the obtained feature mapping relationship to obtain the final substation foreign object recognition result.

[0047] As an optional embodiment, a substation equipment relationship diagram is generated based on a device image, including: generating multiple candidate frame nodes corresponding to the device image; searching for adjacent nodes of each candidate frame node in the multiple candidate frame nodes; connecting each candidate frame node with its adjacent nodes to obtain a substation equipment relationship diagram.

[0048] Figure 3 is a schematic diagram of the relationship between nodes of a device according to an embodiment of the present invention, such as Figure 3 As shown, first, nodes are generated, each of which corresponds to a rectangular candidate box; then for each node, its adjacent nodes are found and connected to the adjacent nodes.

[0049] As an optional embodiment, feature mapping is performed on the embedded feature map corresponding to the feature image and the substation equipment relationship map to obtain a feature mapping relationship, including: decoding the embedded feature map to obtain a decoded embedded feature map; feature mapping is performed on the decoded embedded feature map and the substation equipment relationship map to obtain a feature mapping relationship.

[0050] In the above optional embodiment, converting the feature image into an embedded feature map can convert the data into a fixed-size feature representation for ease of processing and calculation.

[0051] As an optional embodiment, after performing foreign object identification on the target substation equipment based on the feature mapping relationship and obtaining the identification result, the foreign object identification method of the substation equipment also includes: performing error analysis on the identification result to obtain the error result; optimizing using a predetermined algorithm based on the error result to obtain an optimized identification result.

[0052] In the above optional embodiment, the error analysis can use the Hungarian algorithm to match the prediction results (ie, recognition results) and the labeled data one by one, and calculate the errors of specific categories and bounding boxes (ie, candidate boxes).

[0053] In this embodiment, an error function is constructed to calculate the error between a specific substation equipment and the entity type and bounding box, so as to determine whether the prediction result (i.e., the recognition result) matches the actual substation equipment image annotation data, and then adjust the software architecture of the deep self-attention relationship network program module, or further train and test the deep self-attention relationship network program module.

[0054] Specifically: use y to represent the labeled data set, Represents the set of prediction results; Assume that the length of the labeled data set and the prediction result set is the same. If the lengths of the two are inconsistent, they can be solved by padding with zeros. The zero padding of the labeled data set is the background category, and the zero padding of the prediction result set is unpredicted. The best match between the labeled data set and the prediction result set is represented by a permutation σ of the prediction results, corresponding to the smallest error between the labeled data set and the prediction result set: in represents the arrangement of y and σ(i) and When calculating the error between the labeled data set and the predicted result set, the substation equipment body, the type of foreign matter and the bounding box are taken into account. i =(c i ,b i ) includes substation equipment and foreign objects of category c i and position vector b i , for: in It is category c i The corresponding probability is is the predicted bounding box under the σ(i) arrangement, Represents the background category; using the Hungarian algorithm, Rewritten as: in represents the best arrangement of the prediction results obtained by the Hungarian algorithm; in order to eliminate the influence of the bounding box size, the position loss of the substation equipment body and the foreign object bounding box is calculated using the L1 loss function and the generalized IoU loss function L iou The linear combination of is used as the loss function of the substation equipment body and the foreign object bounding box, that is: Among them, λ1 and λ iou is the corresponding weight.

[0055] Figure 4 is a timing diagram of a foreign object identification method for a substation device according to an embodiment of the present invention, such as Figure 4 As shown, the implementation of this method depends on a deep self-attention relationship network program module, which includes an image feature extraction basic network program submodule, a deep self-attention network encoder program submodule, a relationship graph modeling program submodule, a deep self-attention network decoder program submodule, and a prediction branch feedforward network program submodule; and includes the following steps:

[0056] S1,feature extraction basic network program submodule, performs image feature extraction on substation equipment,images;

[0057] S2. The extracted image features are input into the deep self-attention network encoder program submodule, and the deep self-attention network encoder program submodule performs feature fusion based on the self-attention mechanism and outputs an embedded feature map;

[0058] S3, construction of a relationship graph modeling program submodule, a substation equipment relationship graph in a substation equipment image;

[0059] S4, the deep self-attention network decoder program submodule performs feature fusion on the input embedded feature map and the substation equipment relationship map in the substation equipment image based on the fully connected layer;

[0060] S5, the deep self-attention network decoder program submodule performs feature mapping on the embedded feature map generated in step 4 based on the relationship between substation equipment and the self-attention mechanism;

[0061] S6, the prediction branch feedforward network program submodule predicts the position, width, and height of the substation equipment body, the type of foreign matter, and the bounding box based on the feature map.

[0062] In addition, the foreign body identification method of the substation equipment provided in the embodiment of the present invention may also include S0, using the substation equipment image to train and / or test the deep self-attention relationship network program module; and / or S7, prompting or warning according to the substation equipment body and the location, width, and height of the foreign body and the boundary box. Figure 4 Not shown in FIG.

[0063] From the above, it can be seen that in an embodiment of the present invention, a deep self-attention relationship network is used to utilize a basic network to extract features from the original image, which are input into the encoder and decoder in sequence for feature fusion and mapping. At the same time, object relationship modeling is added to the decoder input part to help the model learn features. Finally, the object type and bounding box are obtained through the prediction branch, which has higher computational efficiency, better scalability, flexibility and feature extraction capabilities.

[0064] Example 2

[0065] According to another aspect of an embodiment of the present invention, a foreign body identification device for substation equipment is provided. Figure 5 is a schematic diagram of a foreign object identification device for a substation device according to an embodiment of the present invention. Figure 5 As shown, it includes: an acquisition module 51, an extraction module 53, a generation module 55, a mapping module 57 and an identification module 59. The foreign body identification device of the substation equipment is described below.

[0066] An acquisition module 51 is used to acquire a device image of a target substation device;

[0067] An extraction module 53 is used to extract features from the equipment image to obtain a feature image of the target substation equipment;

[0068] A generating module 55, for generating a substation equipment relationship diagram based on the equipment image;

[0069] A mapping module 57, used to perform feature mapping on the embedded feature map corresponding to the feature image and the substation equipment relationship map to obtain a feature mapping relationship;

[0070] The identification module 59 is used to identify foreign objects in the target substation equipment based on the feature mapping relationship and obtain the identification result.

[0071] It should be noted that the acquisition module 51, extraction module 53, generation module 55, mapping module 57 and identification module 59 correspond to steps S102 to S110 in Embodiment 1, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in Embodiment 1. It should be noted that the modules as part of the apparatus can be executed in a computer system such as a set of computer executable instructions.

[0072] As can be seen from the above, in the embodiment of the present invention, firstly, the equipment image of the target substation equipment can be acquired with the aid of the acquisition module 51; then, the feature extraction of the equipment image can be performed with the aid of the extraction module 53 to acquire the feature image of the target substation equipment; then, the substation equipment relationship diagram can be generated based on the equipment image with the aid of the generation module 55; then, the embedded feature diagram corresponding to the feature image and the substation equipment relationship diagram can be feature mapped with the aid of the mapping module 57 to obtain the feature mapping relationship; finally, the target substation equipment can be identified as a foreign object based on the feature mapping relationship with the aid of the identification module 59, and the identification result can be obtained. The foreign object identification device for substation equipment provided by the embodiment of the present invention achieves the purpose of identifying foreign objects by using the feature image equipment relationship diagram and the embedded feature diagram of the target substation equipment, thereby achieving the technical effect of improving the accuracy of foreign object identification, and further solving the technical problem that the foreign object identification means for power equipment in the related technology are relatively single, resulting in low identification accuracy.

[0073] Optionally, the extraction module includes: an extraction unit, used to extract features of the device image through a deep self-attention network model to obtain a feature image, wherein the deep self-attention network model is obtained through machine learning training using multiple sets of training data, and each set of training data in the multiple sets of training data includes: historical device images and historical feature images.

[0074] Optionally, the generation module includes: a generation unit for generating multiple candidate frame nodes corresponding to the equipment image; a search unit for searching adjacent nodes of each candidate frame node in the multiple candidate frame nodes; and a connection unit for connecting each candidate frame node with its adjacent nodes to obtain a substation equipment relationship diagram.

[0075] Optionally, the mapping module includes: a decoding unit, used to decode the embedded feature map to obtain a decoded embedded feature map; a mapping unit, used to perform feature mapping on the decoded embedded feature map and the substation equipment relationship map to obtain a feature mapping relationship.

[0076] Optionally, the foreign object identification device of the substation equipment also includes: an analysis module, which is used to identify foreign objects in the target substation equipment based on the feature mapping relationship, and after obtaining the identification result, perform error analysis on the identification result to obtain the error result; an optimization module, which is used to optimize based on the error result using a predetermined algorithm to obtain an optimized identification result.

[0077] Example 3

[0078] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein when the computer program is executed by a processor, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned foreign object identification methods for substation equipment.

[0079] Example 4

[0080] According to another aspect of an embodiment of the present invention, a processor is further provided, and the processor is used to run a computer program, wherein the computer program executes any of the above-mentioned foreign object identification methods for substation equipment when running.

[0081] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0082] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0083] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0084] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0085] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0086] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0087] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A foreign body identification method for substation equipment, characterized in that: include: Acquire equipment images of target substation equipment; Performing feature extraction on the device image to obtain a feature image of the target substation device; generating a substation equipment relationship diagram based on the equipment image; Performing feature mapping on the embedded feature graph corresponding to the feature image and the substation equipment relationship graph to obtain a feature mapping relationship; Based on the feature mapping relationship, the target substation equipment is identified as a foreign object, and an identification result is obtained. The equipment relationship diagram is used to define the relationship between substation equipment in the substation equipment image. Generating a substation equipment relationship diagram based on the equipment image includes: Generate multiple candidate frame nodes corresponding to the device image; Searching for adjacent nodes of each candidate frame node among the plurality of candidate frame nodes; Connect each candidate frame node to its adjacent nodes to obtain the substation equipment relationship diagram, The embedded feature map is an image obtained by performing feature fusion of image features in the feature image based on a self-attention mechanism.

2. The method according to claim 1, characterized in that Extracting features from the device image to obtain a feature image of the target substation device includes: The feature image is obtained by performing feature extraction on the device image through a deep self-attention network model, wherein the deep self-attention network model is obtained through machine learning training using multiple sets of training data, and each set of training data in the multiple sets of training data includes: historical device images and historical feature images.

3. The method according to claim 1, characterized in that: Performing feature mapping on the embedded feature graph corresponding to the feature image and the substation equipment relationship graph to obtain a feature mapping relationship includes: Decoding the embedded feature map to obtain a decoded embedded feature map; Feature mapping is performed on the decoded embedded feature graph and the substation equipment relationship graph to obtain the feature mapping relationship.

4. The method according to any one of claims 1 to 3, characterized in that After performing foreign object recognition on the target substation equipment based on the feature mapping relationship and obtaining the recognition result, the method further includes: Performing error analysis on the recognition result to obtain an error result; Based on the error result, optimization is performed using a predetermined algorithm to obtain an optimized recognition result.

5. A foreign body identification device for substation equipment, characterized in that: include: An acquisition module, used to acquire equipment images of target substation equipment; An extraction module, used for performing feature extraction on the equipment image to obtain a feature image of the target substation equipment; A generating module, used for generating a substation equipment relationship diagram based on the equipment image; A mapping module, used for performing feature mapping on the embedded feature map corresponding to the feature image and the substation equipment relationship map to obtain a feature mapping relationship; an identification module, used to identify foreign objects in the target substation equipment based on the feature mapping relationship and obtain an identification result, The equipment relationship diagram is used to define the relationship between substation equipment in the substation equipment image. The generating module comprises: A generating unit, used to generate a plurality of candidate frame nodes corresponding to the device image; A searching unit, configured to search for adjacent nodes of each candidate frame node among the plurality of candidate frame nodes; A connection unit is used to connect each candidate frame node to its adjacent node to obtain the substation equipment relationship diagram, The embedded feature map is an image obtained by performing feature fusion of image features in the feature image based on a self-attention mechanism.

6. The device according to claim 5, characterized in that The extraction module comprises: An extraction unit is used to extract features of the device image through a deep self-attention network model to obtain the feature image, wherein the deep self-attention network model is obtained through machine learning training using multiple sets of training data, and each set of training data in the multiple sets of training data includes: historical device images and historical feature images.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed by a processor, the device where the computer-readable storage medium is located is controlled to execute the foreign object identification method for substation equipment described in any one of claims 1 to 4.

8. A processor, characterized in that: The processor is used to run a computer program, wherein the computer program executes the foreign object identification method for substation equipment described in any one of claims 1 to 4 when running.

Images