Substation equipment bird damage processing method and system based on bird droppings image features
By adopting feature fusion algorithm, deep learning model, improved RCNN model and BERT-FC model in bird damage treatment of substation equipment, the problem of insufficient robustness of bird guano detection in substation equipment in the existing technology is solved, and more accurate and stable bird guano detection and bird recognition are achieved, and scientific bird prevention strategies are supported.
Patent Information
- Application Number
- CN202411842653.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-13
- Publication Date
- 2025-04-29
AI Technical Summary
The prior art lacks robustness in bird dropping detection in substation equipment and is susceptible to image sampling quality, resulting in poor recognition effect and affecting the formulation of bird damage treatment strategies.
The bird damage treatment method of substation equipment based on bird guano image characteristics is adopted. By obtaining the image data of substation equipment, the preset feature fusion algorithm and deep learning model are used to perform equipment positioning and target image feature extraction. Combined with the improved RCNN model and BERT-FC model, bird guano coverage information and bird information are analyzed to generate bird pollution prevention and cleaning strategies.
It improves the accuracy and robustness of bird dropping detection in substation equipment, can operate stably when environmental changes and poor image sampling quality, improves the accuracy of bird recognition, and provides a basis for formulating scientific bird prevention strategies.
Smart Images
Figure CN119904673A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of bird damage treatment for substation equipment, and in particular relates to a method and system for treating bird damage for substation equipment based on bird droppings image features. Background Art
[0002] Substations contain many key equipment, such as transformers, switchgear, busbars, etc. These equipment are usually exposed to the outdoor environment and are often affected by various natural factors. Since substation equipment is usually located in open areas and at a certain height, it is an ideal place for many birds to choose to roost. Birds will excrete feces while roosting on substation equipment, causing the surface of the equipment to be covered with bird droppings. Bird droppings are conductive and contain moisture. Substation equipment covered with bird droppings may cause equipment short circuits, failures, or even power outages, bringing safety hazards and the risk of outages to the power system. Therefore, it is crucial to promptly detect and clean the bird droppings covering substation equipment to ensure the stability and reliability of the power system.
[0003] Intelligent detection of defective images such as bird droppings on substation equipment is divided into two main parts. The first is the identification of substation equipment, and then the detection of defects such as bird droppings on the identified equipment. Traditional machine vision image segmentation algorithms, such as edge detection, template matching, and threshold segmentation, provide reference ideas for the analysis of visible light images of substation equipment. However, these methods have the following main disadvantages: first, they rely on specific detection tasks and working environments, are easily affected by image sampling quality, and lack robustness; second, they rely on manually designed feature engineering; third, the accuracy of complex defect detection is not high. Summary of the invention
[0004] The present invention provides a method and system for handling bird damage in substation equipment based on guano image features, which are used to solve the technical problem that the method and system are dependent on specific detection tasks and working environments, are easily affected by image sampling quality, lack robustness, and result in poor recognition effect, thereby affecting the subsequent generation of bird damage handling strategies.
[0005] In a first aspect, the present invention provides a method for treating bird damage in substation equipment based on bird droppings image features, comprising:
[0006] Acquire image data of substation equipment, locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm, obtain an image of the substation equipment area, and acquire target image features in the image of the substation equipment area according to a preset deep learning model;
[0007] Inputting the target image feature into a preset improved RCNN model, the improved RCNN model outputting guano coverage information corresponding to the target image feature, the guano coverage information including a guano coverage area;
[0008] Performing bird droppings type analysis on the image data to obtain bird information corresponding to the image data;
[0009] An anti-bird pollution strategy and a cleaning strategy are generated according to the substation equipment positioning information, the bird droppings coverage information and the bird information.
[0010] In a second aspect, the present invention provides a bird damage treatment system for substation equipment based on bird droppings image features, comprising:
[0011] An acquisition module is configured to acquire image data of substation equipment, locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm, obtain an image of the substation equipment area, and acquire target image features in the image of the substation equipment area according to a preset deep learning model;
[0012] An output module is configured to input the target image feature into a preset improved RCNN model, and the improved RCNN model outputs guano coverage information corresponding to the target image feature, wherein the guano coverage information includes a guano coverage area;
[0013] An analysis module configured to perform bird droppings type analysis on the image data to obtain bird information corresponding to the image data;
[0014] A generation module is configured to generate an anti-bird pollution strategy and a cleaning strategy according to the substation equipment location information, the bird droppings coverage information and the bird information.
[0015] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the method for handling bird damage in substation equipment based on guano image features according to any embodiment of the present invention.
[0016] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the steps of the method for handling bird damage to substation equipment based on guano image features according to any embodiment of the present invention.
[0017] The bird droppings image feature-based method and system for substation equipment bird damage treatment of the present application ensures more accurate bird droppings detection capability for substation equipment through the fusion of significant features of color, gradient and texture and analysis of spatial position relationship, so that it can still operate stably when the environment changes and the image sampling quality is poor. Through the neural network and BERT-FC model, the system can infer the specific bird species from the bird droppings features and use the cross-entropy loss function to optimize the model, which not only improves the accuracy of bird recognition, but also provides a scientific basis for formulating bird prevention strategies. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0019] Figure 1 A flowchart of a method for handling bird damage to substation equipment based on bird droppings image features provided by an embodiment of the present invention;
[0020] Figure 2 A structural block diagram of a system for handling bird damage in substation equipment based on bird droppings image features provided by an embodiment of the present invention;
[0021] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0022] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0023] See also Figure 1 , which shows a flow chart of a method for handling bird damage to substation equipment based on bird droppings image features of the present application.
[0024] like Figure 1 As shown, the method for handling bird damage to substation equipment based on bird droppings image features specifically includes the following steps:
[0025] Step S101, obtain image data of substation equipment, and locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm to obtain an area image of the substation equipment, and obtain target image features in the area image of the substation equipment according to a preset deep learning model.
[0026] In this step, the pixel block color saliency value, the pixel block gradient saliency value and the pixel block texture saliency value are calculated, and the pixel block color saliency value, the pixel block gradient saliency value and the pixel block texture saliency value are weightedly fused to obtain the target saliency value.
[0027] Specifically, color saliency is a more obvious feature because it has the least dependence on the size, shape and position of the target. In order to describe color features, we first need to select a suitable color space. One of the commonly used color spaces is the HSV space, where H represents hue, S represents saturation, and V represents tone. According to the color characteristics of the insulator, the collected insulator image is converted to the HSV color space for feature extraction.
[0028] In order to represent the color features of the aerial image of the insulator, a color histogram can be used. First, the hue component H is divided into different color intervals, and then each pixel is traversed and its hue value is assigned to the corresponding color interval to obtain a color histogram. Next, the mean shift clustering algorithm can be used to cluster pixels with similar colors in the color histogram, and the pixel block color saliency value is obtained by calculating the saliency value of each pixel block.
[0029] For the gradient saliency of insulators, the gradient value changes at the edge of the image are often large, while in relatively smooth areas, the gradient value changes are small. Therefore, the gradient features of the insulator image can be used for target detection. The directional gradient histogram is used to describe the gradient features of the image. First, the insulator image is normalized in color space and converted into a grayscale image, and then the gradient is calculated to obtain the gradient histogram. By calculating the gradient saliency, the gradient saliency value of the pixel block can be obtained.
[0030] Texture saliency can effectively eliminate the interference of the insulator background in aerial images. Texture is formed by the regular distribution of image grayscale in space. Any two pixels have a specific correspondence in space. The gray level co-occurrence matrix can be used to describe this correspondence, indicating the probability of simultaneous occurrence of pixel pairs. Therefore, the texture features of the image can be extracted through the gray level co-occurrence matrix;
[0031] This part of the calculation first converts the aerial image of the insulator into a grayscale image and forms a grayscale matrix. Then, the grayscale matrix is used to calculate the grayscale co-occurrence matrix, which can express the significant value of the texture feature of the pixel block through contrast statistics.
[0032] Finally, the algorithm will fuse the salient feature values of color, gradient and texture, and consider their spatial relationship to calculate the multi-feature saliency value of each pixel block. Further, this multi-feature saliency value is used to detect the target area of the insulator.
[0033] Furthermore, according to the target significance value, the rotating caliper method is used to divide the area where the substation equipment is located to obtain the substation equipment area image.
[0034] It should be noted that, however, after the insulator area is inspected, there may be some obvious noise interference. Therefore, it is necessary to use the rotating chuck method to further clean up the small holes and noise interference in order to obtain a more prominent and easy-to-identify inspection result.
[0035] In order to accurately locate the boundary of the insulator area and eliminate noise, an algorithm is needed to find the minimum enclosing rectangle: the edge of the minimum enclosing rectangle is located on a convex polygon. Therefore, the convex polygon can be found by algorithms such as the Graham scanning method, and the possible minimum rectangle can be found based on its edge. The four endpoints of the convex polygon (Xmin, Xmax, Ymin, Ymax) are calculated to determine the four tangents (L1, L2, L3, L4), and a caliper range is determined based on each pair of parallel tangents. Then, by rotating the two calipers until one of the lines coincides with one side of the polygon, the algorithm will calculate the area of the new rectangle and continue this process until the minimum enclosing rectangle covering the entire figure is found.
[0036] Furthermore, the target image features in the substation equipment area image are obtained according to the preset Inceptionv4 model. The Inceptionv4 model includes the first 1x1 convolution, the second 1x1 convolution, the third 1x1 convolution, the fourth 1x1 convolution, the first 3x3 convolution, the second 3x3 convolution and the 5x5 convolution. The loss function of the Inceptionv4 model is:
[0037]
[0038] In the formula, p(x) is the true result, q(x) is the predicted result, and H(p, q) is the loss function, which represents the difference between the predicted result and the true result;
[0039] Among them, the substation equipment area image is respectively input into the first 1x1 convolution, the second 1x1 convolution, the third 1x1 convolution and the first 3x3 convolution, and the first result, the second result, the third result and the fourth result are output respectively;
[0040] Input the second result, the third result and the fourth result into the second 3x3 convolution, the 5x5 convolution and the fourth 1x1 convolution respectively, and output the fifth result, the sixth result and the seventh result respectively;
[0041] The first result, the fifth result, the sixth result and the seventh result are concatenated to obtain the final target image feature.
[0042] Step S102: input the target image feature into a preset improved RCNN model, and the improved RCNN model outputs guano coverage information corresponding to the target image feature, wherein the guano coverage information includes a guano coverage area.
[0043] In this step, the improved RCNN model includes a base network, a region proposal network for generating candidate boxes, and a detection network for classifying candidate boxes and regressing bounding boxes;
[0044] The basic network includes a convolutional layer, a ReLU layer, a pooling layer and a fully connected layer;
[0045] The region proposal network is used to predict the target region and target score at each position in the input image. During training, a 3×3 spatial window slides on the convolution feature map output by the last shared convolution layer. At each sliding window position, k anchor boxes are predicted at the same time. The k anchor boxes have the same center, and the length-to-width ratios of the anchor boxes are 1:1, 1:2, and 2:1, respectively. Three types of areas are set for each ratio anchor box.
[0046] The detection network determines the category of each region proposal using a fully connected layer and a Softmax function based on the proposal feature map.
[0047] Step S103: Analyze the bird droppings types on the image data to obtain bird information corresponding to the image data.
[0048] In this step, the image data is analyzed for bird droppings species according to a preset BERT-FC model to obtain bird information corresponding to the image data, wherein the BERT-FC model includes a BERT model and an FC layer, the BERT model includes an input layer, a two-layer Transformer encoding structure and an output layer, and the FC layer includes an input layer for receiving feature vectors, a hidden layer for algorithm optimization and an output layer for mapping;
[0049] The BERT model outputs a feature vector H of length n, and the feature vector H is mapped to a vector Z of length K according to the K-dimensional weight matrix W, where the mapping expression is:
[0050] Z=W T H+b,
[0051] Where b is the bias vector and T is the transposed sign;
[0052] After the mapping is completed, Z is used as input and normalized by the Softmax function to obtain the probability distribution of each category, which is expressed as:
[0053]
[0054] Where P i is the probability of the i-th category, Z i is the i-th vector, Z j is the jth vector;
[0055] Set the prediction result to the category with the highest probability. The expression is:
[0056]
[0057] In the formula, To predict the results, * argmax i The index that selects the maximum value among all class probabilities.
[0058] Step S104: generating a bird pollution prevention strategy and a cleaning strategy according to the substation equipment location information, the bird droppings coverage information and the bird information.
[0059] In this step, the anti-pollution strategy formulation part is as follows: Bird species anti-pollution strategy: The system will formulate corresponding anti-pollution measures based on the bird species identified in real time. For example, for destructive birds, direct intervention methods such as expulsion may be required, and for harmless birds, it may only be necessary to monitor their behavior. Environmental status anti-pollution strategy: Considering the impact of environmental factors on bird behavior, the system also needs to adjust the anti-pollution strategy according to environmental conditions (such as season, climate, geographical location, etc.), such as adjusting the activation frequency of the expulsion equipment or the expulsion method.
[0060] Cleaning frequency adjustment section: Bird species affect cleaning frequency: Different types of birds may cause different levels of pollution. For example, large numbers of birds may produce more pollution and require more frequent cleaning. Anti-pollution strategies affect cleaning frequency: The actual cleaning frequency will also be affected by anti-pollution strategies. Effective anti-pollution strategies may reduce the cleaning frequency, while situations where the strategy is less effective may require an increase in the cleaning frequency. Environmental factors affect cleaning frequency: Environmental conditions (such as seasons, climate change, etc.) may also affect the degree of pollution and cleaning frequency: For example, in the rainy season, the cleaning frequency may need to be increased to prevent pollution accumulation.
[0061] In summary, the method of the present application ensures a more accurate detection capability of bird droppings in substation equipment through the fusion of significant features of color, gradient and texture and analysis of spatial position relationship, so that it can still operate stably when the environment changes and the image sampling quality is poor. Through the neural network and BERT-FC model, the system can infer the specific bird species from the bird droppings characteristics and use the cross-entropy loss function to optimize the model, which not only improves the accuracy of bird recognition, but also provides a scientific basis for the formulation of bird prevention strategies.
[0062] See also Figure 2 , which shows a structural block diagram of a substation equipment bird damage treatment system based on guano image features of the present application.
[0063] like Figure 2 As shown, the substation equipment bird damage treatment system 200 includes an acquisition module 210 , an output module 220 , an analysis module 230 and a generation module 240 .
[0064] Among them, the acquisition module 210 is configured to acquire image data of substation equipment, and locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm to obtain an image of the substation equipment area, and obtain the target image features in the substation equipment area image according to a preset deep learning model; the output module 220 is configured to input the target image features into a preset improved RCNN model, and the improved RCNN model outputs guano coverage information corresponding to the target image features, and the guano coverage information includes the guano coverage area; the analysis module 230 is configured to perform guano species analysis on the image data to obtain bird information corresponding to the image data; the generation module 240 is configured to generate bird pollution prevention strategies and cleaning strategies based on the substation equipment positioning information, the guano coverage information and the bird information.
[0065] It should be understood that Figure 2 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 2 The modules in it will not be described in detail here.
[0066] In some other embodiments, the embodiments of the present invention further provide a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor is caused to execute the method for handling bird damage to substation equipment based on bird droppings image features in any of the above method embodiments;
[0067] As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows:
[0068] Acquire image data of substation equipment, locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm, obtain an image of the substation equipment area, and acquire target image features in the image of the substation equipment area according to a preset deep learning model;
[0069] Inputting the target image feature into a preset improved RCNN model, the improved RCNN model outputting guano coverage information corresponding to the target image feature, the guano coverage information including a guano coverage area;
[0070] Performing bird droppings type analysis on the image data to obtain bird information corresponding to the image data;
[0071] An anti-bird pollution strategy and a cleaning strategy are generated according to the substation equipment positioning information, the bird droppings coverage information and the bird information.
[0072] The computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required by at least one function; the data storage area may store data created according to the use of the substation equipment bird damage treatment system based on guano image features, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the substation equipment bird damage treatment system based on guano image features via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0073] Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3In the example, the bus connection is used. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the method for treating bird damage in substation equipment based on guano image features of the above-mentioned method embodiment is implemented. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the substation equipment bird damage treatment system based on guano image features. The output device 340 may include display devices such as display screens.
[0074] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.
[0075] As an implementation mode, the electronic device is applied to a bird damage treatment system for substation equipment based on bird droppings image features, and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can:
[0076] Acquire image data of substation equipment, locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm, obtain an image of the substation equipment area, and acquire target image features in the image of the substation equipment area according to a preset deep learning model;
[0077] Inputting the target image feature into a preset improved RCNN model, the improved RCNN model outputting guano coverage information corresponding to the target image feature, the guano coverage information including a guano coverage area;
[0078] Performing bird droppings type analysis on the image data to obtain bird information corresponding to the image data;
[0079] An anti-bird pollution strategy and a cleaning strategy are generated according to the substation equipment positioning information, the bird droppings coverage information and the bird information.
[0080] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.
[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for dealing with bird damage in substation equipment based on bird droppings image features, characterized in that: include: Acquire image data of substation equipment, locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm, obtain an image of the substation equipment area, and acquire target image features in the image of the substation equipment area according to a preset deep learning model; Inputting the target image feature into a preset improved RCNN model, the improved RCNN model outputting guano coverage information corresponding to the target image feature, the guano coverage information including a guano coverage area; Performing bird droppings type analysis on the image data to obtain bird information corresponding to the image data; An anti-bird pollution strategy and a cleaning strategy are generated according to the substation equipment positioning information, the bird droppings coverage information and the bird information.
2. A method for dealing with bird damage to substation equipment based on bird droppings image features according to claim 1, characterized in that: The positioning of the substation equipment corresponding to the image data according to the preset feature fusion algorithm to obtain the substation equipment positioning information includes: Calculating a pixel block color significance value, a pixel block gradient significance value, and a pixel block texture significance value, and performing weighted fusion on the pixel block color significance value, the pixel block gradient significance value, and the pixel block texture significance value to obtain a target significance value; According to the target significance value, the area where the substation equipment is located is divided by using the rotating caliper method to obtain a substation equipment area image.
3. The method for treating bird damage in substation equipment based on bird droppings image features according to claim 1 is characterized in that: The step of acquiring target image features in the substation equipment area image according to a preset deep learning model includes: The target image features in the substation equipment area image are obtained according to the preset Inceptionv4 model, wherein the Inceptionv4 model includes a first 1x1 convolution, a second 1x1 convolution, a third 1x1 convolution, a fourth 1x1 convolution, a first 3x3 convolution, a second 3x3 convolution and a 5x5 convolution, and the loss function of the Inceptionv4 model is: In the formula, p(x) is the true result, q(x) is the predicted result, and H(p, q) is the loss function, which represents the difference between the predicted result and the true result; The substation equipment area image is respectively input into the first 1x1 convolution, the second 1x1 convolution, the third 1x1 convolution and the first 3x3 convolution, and the first result, the second result, the third result and the fourth result are output respectively; Input the second result, the third result and the fourth result into a second 3x3 convolution, a 5x5 convolution and a fourth 1x1 convolution respectively, and output a fifth result, a sixth result and a seventh result respectively; The first result, the fifth result, the sixth result and the seventh result are concatenated to obtain a final target image feature.
4. The method for treating bird damage in substation equipment based on bird droppings image features according to claim 1 is characterized in that: The improved RCNN model includes a basic network, a region proposal network for generating candidate boxes, and a detection network for classifying candidate boxes and regressing bounding boxes; The basic network includes a convolutional layer, a ReLU layer, a pooling layer and a fully connected layer; The region proposal network is used to predict the target region and target score at each position in the input image. During training, a 3×3 spatial window slides on the convolution feature map output by the last shared convolution layer. At each sliding window position, k anchor boxes are predicted at the same time. The k anchor boxes have the same center, and the length-to-width ratios of the anchor boxes are 1:1, 1:2, and 2:1, respectively. Three types of areas are set for each ratio anchor box. The detection network determines the category of each region proposal using a fully connected layer and a Softmax function based on the proposal feature map.
5. The method for treating bird damage in substation equipment based on bird droppings image features according to claim 1 is characterized in that: The performing bird droppings type analysis on the image data to obtain bird information corresponding to the image data comprises: Performing bird droppings type analysis on the image data according to a preset BERT-FC model to obtain bird information corresponding to the image data, wherein the BERT-FC model includes a BERT model and an FC layer, the BERT model includes an input layer, a two-layer Transformer encoding structure and an output layer, and the FC layer includes an input layer for receiving feature vectors, a hidden layer for algorithm optimization, and an output layer for mapping; The BERT model outputs a feature vector H of length n, and the feature vector H is mapped to a vector Z of length K according to the K-dimensional weight matrix W, where the mapping expression is: Z=W T .H+b, Where b is the bias vector and T is the transposed sign; After the mapping is completed, Z is used as input and normalized by the Softmax function to obtain the probability distribution of each category, which is expressed as: Where P i is the probability of the i-th category, Z i is the i-th vector, Z j is the jth vector; Set the prediction result to the category with the highest probability. The expression is: In the formula, is the prediction result, *argmax i The index that selects the maximum value among all class probabilities.
6. A bird damage treatment system for substation equipment based on bird droppings image features, characterized in that: include: An acquisition module is configured to acquire image data of substation equipment, locate the substation equipment corresponding to the image data according to a preset feature fusion algorithm, obtain an image of the substation equipment area, and acquire target image features in the image of the substation equipment area according to a preset deep learning model; An output module is configured to input the target image feature into a preset improved RCNN model, wherein the improved RCNN model outputs guano coverage information corresponding to the target image feature, wherein the guano coverage information includes a guano coverage area; An analysis module configured to perform bird droppings type analysis on the image data to obtain bird information corresponding to the image data; A generation module is configured to generate an anti-bird pollution strategy and a cleaning strategy according to the substation equipment location information, the bird droppings coverage information and the bird information.
7. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in any one of claims 1 to 5.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.