Electric power part defect image generation method, system, equipment and medium

Abstract

The invention discloses a power component defect image generation method, system and device and a medium, and the method comprises the following steps: obtaining real-time long-shot image data of a monitoring area, and carrying out the preprocessing of the data, so as to obtain the preprocessed real-time long-shot image data, performing target detection on the preprocessed real-time long-shot image data by adopting a constructed power equipment target detection model to obtain real-time power equipment target image data; performing electric power part defect identification on the real-time electric power equipment target image data by adopting a constructed electric power part defect identification model to obtain a real-time electric power part defect identification result; and generating a real-time electric power part defect image according to the real-time electric power part defect identification result. According to the method, the power equipment in the long-range image can be detected, the power equipment can be accurately positioned, defect identification is carried out on a single power component, and a targeted color power component defect image is generated and visually displayed.

Description

Technical Field

[0001] The present invention relates to the technical field of electric component defect recognition, and in particular to an electric component defect image generation method, system, equipment and medium. Background Art

[0002] Power equipment is an important part of the power system, and its health is directly related to the safe and stable operation of the power system. Power components are a component of power equipment. Power equipment is usually composed of multiple components, each of which has a specific function. Identifying defects in power components is an important measure to ensure the normal operation of the power system.

[0003] With the development and progress of monitoring technology and image recognition algorithms, more and more defect recognition algorithms have been proposed. However, since power equipment often includes multiple power components, the existing defect recognition algorithms only target the overall power equipment and cannot accurately and efficiently identify defects in a single power component, resulting in low practicality. In addition, although the existing defect recognition algorithms can detect and identify defects, they cannot intuitively display defective targets, especially power components, resulting in the need to consume a lot of manpower and material resources to check them one by one. Summary of the invention

[0004] The purpose of the present invention is to provide a method, system, device and medium for generating defect images of power components, which can detect power equipment in a distant image, accurately locate the power equipment, identify defects of individual power components, generate targeted color power component defect images, and display them intuitively.

[0005] The present invention is achieved through the following technical solutions:

[0006] In a first aspect, an embodiment of the present invention provides a method for generating a defect image of an electric component, comprising the following steps:

[0007] Acquire real-time long-range shooting image data of the monitoring area, perform preprocessing to obtain preprocessed real-time long-range shooting image data, use the constructed electric power equipment target detection model to perform target detection on the preprocessed real-time long-range shooting image data, and obtain real-time electric power equipment target image data;

[0008] Using the constructed power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data to obtain a real-time power component defect recognition result;

[0009] A real-time power component defect image is generated based on the real-time power component defect recognition result.

[0010] Furthermore, before the step of acquiring the real-time long-range shooting image data of the monitoring area, it also includes: constructing a power equipment target detection model and a power component defect recognition model, and the specific method includes:

[0011] Acquire a number of historical perspective shooting image data and perform preprocessing to obtain a number of preprocessed historical perspective shooting image data;

[0012] According to the pre-processed historical perspective shooting image data, a weakly supervised image recognition algorithm is used to build a power equipment target detection model, and the power equipment target detection model is used to perform target detection on the pre-processed historical perspective shooting image data to obtain a number of historical power equipment target image data;

[0013] Based on the historical target image data of power equipment, a deep learning algorithm is used to construct a power component defect recognition model.

[0014] Furthermore, the specific method of constructing a power equipment target detection model using a weakly supervised image recognition algorithm based on the pre-processed historical perspective image data includes:

[0015] The multi-color channel edge detection algorithm is used to extract the historical edge features of the pre-processed historical perspective shooting image data;

[0016] A feature point detection algorithm is used to extract several historical feature points from historical edge features;

[0017] Taking the extreme value points in the historical feature points as corresponding relaxed edge points to obtain a number of relaxed edge points;

[0018] A foreground seed point generation model is constructed using a deep learning algorithm, and several pre-processed historical distant view image data with relaxed edge points are predicted to obtain several historical foreground seed points.

[0019] Expanding the historical foreground seed points to obtain a corresponding historical seed cluster set, and calculating the geodesic distance from each pixel point in the preprocessed historical distant view image data to the historical seed cluster set;

[0020] A pseudo-label for each preprocessed historical perspective image data is generated according to the geodesic distance, and a power equipment target detection model is constructed using a deep learning algorithm. The power equipment target detection model is used to perform target detection on the preprocessed historical perspective image data with pseudo-labels to obtain several historical power equipment target image data.

[0021] Furthermore, the specific method of using the constructed power equipment target detection model to perform target detection on real-time long-range shooting image data includes:

[0022] Collecting real-time long-range shooting image data of the monitoring area and preprocessing it to obtain preprocessed real-time long-range shooting image data;

[0023] The pre-processed real-time distant view shooting image data is processed by using a foreground seed point generation model to obtain a number of real-time foreground seed points, and corresponding real-time foreground seed point regions are obtained according to the number of real-time foreground seed points;

[0024] Using the electric power equipment target detection model, target detection is performed on the real-time foreground seed point area to obtain a real-time detection frame of the real-time electric power equipment target;

[0025] According to the real-time detection frame of the real-time power equipment target, the pre-processed real-time distant view shooting image data is cropped to obtain the real-time power equipment target image data.

[0026] Furthermore, the specific method of using the constructed power component defect recognition model to perform power component defect recognition on real-time power equipment target image data includes:

[0027] The power component defect recognition model is used to extract real-time high-level image features and real-time low-level image features of real-time power equipment target image data;

[0028] According to the preset attention weight value, the real-time high-level features and the real-time low-level features are fused to obtain the real-time fused image features;

[0029] The real-time fusion image features are used to identify defects in power components, and real-time power component defect identification results are obtained.

[0030] Furthermore, the specific method for generating a real-time power component defect image according to the real-time power component defect recognition result includes:

[0031] Determine whether there is a power component defect in the real-time power component defect identification result;

[0032] If so, the real-time RGB values ​​of all pixels in the real-time foreground seed point area of ​​the real-time power equipment target image data are retained to obtain a color real-time power component defect target;

[0033] Graying all pixel points in the real-time background area except the color real-time power component defect target in the real-time power equipment target image data to obtain a real-time power component defect image;

[0034] If not, continue to collect and monitor the real-time distant view shooting image data at the next moment.

[0035] Furthermore, the specific method of expanding the historical foreground seed points to obtain the corresponding historical seed cluster set includes:

[0036] Performing gradient calculation on the grayscale values ​​of adjacent pixel points around each foreground seed point of the preprocessed historical distant shot image data, and generating an extended seed point list for each foreground seed point based on the gradient calculation result;

[0037] De-duplicate the extended seed point list of each foreground seed point to obtain the historical seed cluster of each foreground seed point;

[0038] All foreground seed points are traversed to obtain a set of historical seed clusters of preprocessed historical distant view image data.

[0039] In a second aspect, an embodiment of the present invention provides a power component defect image generation system, comprising: a data acquisition module, a target detection module, a defect recognition module and a defect image generation module, wherein the data acquisition module is used to acquire real-time long-range shooting image data of a monitoring area and perform preprocessing to obtain preprocessed real-time long-range shooting image data;

[0040] The target detection module is used to perform target detection on the pre-processed real-time distant view shooting image data using the electric power equipment target detection model to obtain real-time electric power equipment target image data;

[0041] The defect recognition module is used to use the power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data to obtain a real-time power component defect recognition result;

[0042] The defect image generation module is used to generate a real-time power component defect image according to the real-time power component defect recognition result.

[0043] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor, an input device, an output device and a memory, wherein the processor is connected to the input device, the output device and the memory, respectively, the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute the method described in the above embodiment.

[0044] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the method described in the above embodiment.

[0045] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0046] A method for generating a defect image of a power component provided by an embodiment of the present invention detects the power equipment contained in a distant image through a power equipment target detection model, accurately locates the power equipment, and then realizes pixel-level power component defect recognition of a single power component through a power component defect recognition model. While ensuring the recognition efficiency, the accuracy of defect recognition of a single power component is improved, thereby improving the overall practicality. Targeted color power component defect images are generated in real time, which intuitively display the location and condition of the power component defect target, facilitate timely repair and replacement of damaged components, avoid troubleshooting, reduce the investment of manpower and material resources, and ensure the stable operation of power equipment.

[0047] An electric component defect image generation system, electronic device, and computer-readable storage medium provided by the embodiments of the present invention have the same inventive concept and the same beneficial effects as an electric component defect image generation method, which will not be described in detail herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative work. In the drawings:

[0049] Figure 1 A flowchart of a method for generating a defect image of an electric component provided by the first embodiment of the present invention;

[0050] Figure 2 A structural block diagram of a power component defect image generation system provided by another embodiment of the present invention;

[0051] Figure 3 A structural block diagram of an electronic device provided in another embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0053] Example 1

[0054] like Figure 1 FIG. 1 is a flowchart of a method for generating a defect image of an electric component according to a first embodiment of the present invention. The method comprises the following steps:

[0055] S1: Construct power equipment target detection model and power component defect recognition model;

[0056] S2: acquiring real-time long-range shooting image data of the monitoring area, and performing preprocessing to obtain preprocessed real-time long-range shooting image data, and using the constructed electric power equipment target detection model to perform target detection on the preprocessed real-time long-range shooting image data to obtain real-time electric power equipment target image data;

[0057] S3: using the constructed power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data to obtain a real-time power component defect recognition result;

[0058] S4: Generate a real-time power component defect image according to the real-time power component defect recognition result.

[0059] Among them, the construction of the power equipment target detection model and the power component defect recognition model specifically includes the following steps:

[0060] S11: Collect a number of historical long-range image data and perform preprocessing to obtain a number of preprocessed historical long-range image data. The preprocessing includes data cleaning, Gaussian denoising, contrast enhancement, size adjustment, etc., to improve data quality and provide data support for subsequent model training.

[0061] S12: Based on a number of pre-processed historical perspective image data, a weakly supervised image recognition algorithm is used to construct a power equipment target detection model and generate a number of historical power equipment target image data, including the following steps:

[0062] S121: Use a multi-color channel Canny edge detection algorithm to extract historical edge features of the pre-processed historical perspective shooting image data. The multi-color channel Canny edge detection algorithm performs Canny edge detection on each color channel (RGB) independently to capture edge information on different color channels, and merges edge information from different channels to obtain more comprehensive edge features.

[0063] S122: extracting a number of historical feature points of the preprocessed historical perspective image data using the Shi-Tomasi corner detection algorithm according to the historical edge features. The Shi-Tomasi corner detection algorithm is used to identify key feature points in an image, which will represent significant structure or shape information in the image.

[0064] S123: taking the extreme points among the feature points as corresponding relaxed edge points, and obtaining a number of relaxed edge points of the preprocessed historical distant shot image data; identifying the extreme points from the detected feature points, which may be located at prominent positions on the edge, and marking these extreme points as relaxed edge points as a reference for subsequent steps.

[0065] S124: Based on a number of pre-processed historical distant view image data with relaxed edge points, a foreground seed point generation model is constructed using an FCN algorithm, and a number of historical foreground seed points are generated for each pre-processed historical distant view image data. The foreground seed point generation model is constructed based on a fully convolutional network (FCN), and a convolutional layer is used to replace a traditional fully connected layer, so that the network can accept input images of any size. In practical applications, the fusion strategy and model structure are adjusted according to specific data and application scenarios to construct a foreground seed point generation model, which can predict the position of the foreground seed point according to the image features of the image data, thereby avoiding the problems of low efficiency and poor accuracy of manually annotating foreground seed points in traditional technologies.

[0066] S125: Expanding a number of historical foreground seed points of the preprocessed historical distant view image data to obtain a corresponding historical seed cluster set, and obtaining a geodesic distance from each pixel point in the preprocessed historical distant view image data to the historical seed cluster set, specifically comprising the following steps:

[0067] S1251: performing gradient calculation on the grayscale values ​​of the adjacent pixel points around each foreground seed point of the preprocessed historical distant shot image data, and generating an extended seed point list for each foreground seed point according to the gradient calculation result.

[0068] S1252: De-duplicate the extended seed point list of each foreground seed point to obtain a historical seed cluster of each foreground seed point.

[0069] S1253: Traverse all foreground seed points to obtain a set of historical seed clusters of pre-processed historical distant view image data.

[0070] S1254: Calculate the geodesic distance from each pixel in the preprocessed historical distant shot image data to the historical seed cluster set. The calculation formula of the geodesic distance is:

[0071]

[0072] In formula (1), G(i",S,I) is the distance from the ith pixel to the bounding box point set S(S∈{S f ,S b}) of the geodesic distance set; S f is the foreground seed point set; S is the bounding box point set; Sb is the outer frame pixel set; D geo(i", j", I) is the geodesic distance from the i-th pixel to the j-th pixel; I is the pixel set of the preprocessed historical perspective image data; i" and j" are both pixel indicators, and the calculation formula for the geodesic distance from the i-th pixel to the j-th pixel is:

[0073]

[0074] In formula (2), P i",j" is the set of all paths between the i-th pixel and the j-th pixel; p(s) is a feasible path; s is a path parameter, s∈[0,1]; u(s)=p′(s) / ||p′(s)|| is a unit vector tangent to the path direction; p′(s) is a path tangent to the feasible path p(s); is the geodesic distance weight.

[0075] S126: Generate a pseudo label for each pre-processed historical perspective image data according to the geodesic distance, and use the YOLO-RF algorithm to construct a power equipment target detection model based on a plurality of pre-processed historical perspective image data with pseudo labels, and generate a plurality of historical power equipment target image data, including the following steps:

[0076] S1261: If the geodesic distance of any pixel point of the historical distant view image data after preprocessing is greater than the geodesic distance threshold, the pixel point is discarded and replaced with the next pixel point, and the relationship between the geodesic distance of the next pixel point and the geodesic distance threshold is continued to be determined. If it is less than, the next pixel point is directly replaced, and this step is repeated until the comparison of all pixels is completed and then the next step is entered.

[0077] S1262: Traverse all pixel points in the pre-processed historical distant shot image data to obtain a corresponding geodesic distance map, and perform pseudo-label mapping according to the geodesic distance map to obtain a corresponding pseudo-label.

[0078] S1263: Based on several pre-processed historical perspective image data with pseudo labels, a YOLO-RF algorithm is used to construct a power equipment target detection model, and several historical power equipment target image data are generated. The power equipment target detection model includes a feature extraction and prediction bounding box output module constructed based on the YOLOv4 algorithm and a target classification module constructed based on the Random Forest (RF) algorithm; the feature extraction and prediction bounding box output module quickly extracts features in the image, which can represent different objects and scenes in the image, and directly predicts the position of the bounding box on the image, which can not only identify the object, but also reflect the specific position in the image; the target classification module can capture the complex relationship between features, which helps to improve performance in multi-category and multi-scale detection, identify and classify the features in the predicted bounding box, and output the detection box of the power equipment target.

[0079] S13: Based on several historical power equipment target image data, a power component defect recognition model is constructed using the LFTCNN-Attention-Elman algorithm. The power component defect recognition model includes a dual-channel feature extraction module constructed based on the low level feature fusion two-channel convolutional neural network (LFTCNN) algorithm, an attention weight module constructed based on the Attention mechanism, and a defect recognition module constructed based on the Elman algorithm. The dual-channel feature extraction module includes a first channel and a second channel in parallel constructed using a convolutional neural network (CNN). The first channel is used to process the original color image and capture high-level features such as shape, texture, etc. The second channel focuses on the extraction of low-level features, such as edges, corners, etc. This can be achieved by using a convolution kernel similar to the Sobel operator or the Laplacian operator. The attention weight module fuses the features of the two channels through a preset attention weight value. The defect recognition module realizes the recognition prediction of power component defects through a trained network structure.

[0080] Among them, S2: collecting real-time long-range shooting image data of the monitoring area, and using the power equipment target detection model to perform target detection on the real-time long-range shooting image data to obtain real-time power equipment target image data, including the following steps:

[0081] S21: collecting real-time long-range shooting image data of the monitoring area, performing preprocessing, and obtaining preprocessed real-time long-range shooting image data;

[0082] S22: using a foreground seed point generation model to generate a number of real-time foreground seed points of the pre-processed real-time distant view shooting image data, and obtaining corresponding real-time foreground seed point areas according to the number of real-time foreground seed points;

[0083] S23: using the electric power equipment target detection model, performing target detection on the pre-processed real-time distant view shooting image data with the real-time foreground seed point area set, and obtaining a real-time detection frame of the real-time electric power equipment target;

[0084] S24: According to the real-time detection frame of the real-time electric power equipment target, the pre-processed real-time distant view shooting image data is cropped to obtain the real-time electric power equipment target image data with the real-time foreground seed point area set.

[0085] Among them, the specific method of using the power equipment target detection model to perform target detection on the pre-processed real-time distant view shooting image data with a real-time foreground seed point area includes the following steps:

[0086] S231: using a feature extraction and prediction bounding box output module of the electric power equipment target detection model to extract real-time fast features of the pre-processed real-time distant view shooting image data having a real-time foreground seed point area set;

[0087] S232: Detect objects based on the real-time fast features, and output predicted bounding boxes of all real-time detected objects;

[0088] S233: using the target classification module, extracting the real-time detection target image features in the predicted boundary box of the real-time detection target, performing image recognition, and outputting the real-time image recognition result as a real-time detection box of the real-time power equipment target.

[0089] S3: Using the power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data, and obtaining the real-time power component defect recognition result includes the following steps:

[0090] S31: using a dual-channel feature extraction module of a power component defect recognition model to extract real-time high-level image features and real-time low-level image features of real-time power equipment target image data;

[0091] S32: performing feature fusion on the real-time high-level features and the real-time low-level features according to the attention weight value preset by the attention weight module to obtain a real-time fused image feature;

[0092] S33: Based on the real-time fusion image features, a defect recognition module is used to perform power component defect recognition to obtain a real-time power component defect recognition result.

[0093] S4: generating a power component defect image for the real-time power equipment target image data according to the real-time power component defect recognition result, and obtaining a real-time power component defect image, including the following steps:

[0094] S41: If the real-time power component defect recognition result is that there is a power component defect, proceed to the next step, otherwise, continue to collect real-time long-range shooting image data of the monitoring area at the next moment;

[0095] S42: retaining the real-time RGB values ​​of all pixels in the real-time foreground seed point area of ​​the real-time power equipment target image data to obtain a color real-time power component defect target;

[0096] S43: grayscale all pixels in the real-time background area except the color real-time power component defect target in the real-time power equipment target image data to obtain a real-time power component defect image. The real-time power component defect image displays the real-time power component defect target with power component defects in color, and displays other power components of the background power equipment in grayscale.

[0097] The present invention discloses a method for generating defective images of electric components. The method detects the electric equipment contained in a distant image through an electric equipment target detection model, accurately locates the electric equipment, and then realizes pixel-level electric component defect recognition of a single electric component through an electric component defect recognition model. While ensuring the recognition efficiency, the accuracy of defect recognition of a single electric component is improved, and the overall practicality is improved. The method generates targeted color electric component defect images in real time, intuitively displays the location and condition of the electric component defect target, facilitates timely repair and replacement of damaged components, avoids investigation, reduces the investment of manpower and material resources, and ensures the stable operation of the electric equipment.

[0098] Embodiment 2:

[0099] like Figure 2As shown, a structural block diagram of a power component defect image generation system provided by another embodiment of the present invention is shown, the system is used to implement a power component defect image generation method, the system includes a model construction module, a data acquisition module, a target detection module, a defect recognition module and a defect image generation module, the model construction module is used to construct a power equipment target detection model and a power component defect recognition model based on a number of historical perspective shooting image data using an image recognition algorithm, the data acquisition module is used to obtain real-time perspective shooting image data of the monitoring area and perform preprocessing to obtain preprocessed real-time perspective shooting image data, the target detection module is used to use the power equipment target detection model to perform target detection on the preprocessed real-time perspective shooting image data to obtain real-time power equipment target image data, the defect recognition module is used to use the power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data to obtain real-time power component defect recognition results, and the defect image generation module is used to generate a real-time power component defect image based on the real-time power component defect recognition results.

[0100] The present invention discloses a power component defect image generation system. Through a power equipment target detection model, the power equipment contained in a distant image is detected to accurately locate the power equipment. Then, through a power component defect recognition model, pixel-level power component defect recognition of a single power component is realized. While ensuring the recognition efficiency, the accuracy of defect recognition of a single power component is improved, and the overall practicality is improved. Targeted color power component defect images are generated in real time, and the location and condition of the power component defect target are intuitively displayed, which is convenient for timely repair and replacement of damaged components, avoids investigation, reduces the investment of manpower and material resources, and ensures the stable operation of power equipment.

[0101] like Figure 3 As shown, a structural block diagram of an electronic device according to another embodiment of the present invention is shown, the device includes a processor, an input device, an output device and a memory, the processor is connected to the input device, the output device and the memory respectively, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute a method for generating a defect image of an electric component described in the first embodiment above.

[0102] It should be understood that in the embodiments of the present invention, the processor referred to may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0103] Input devices may include a touchpad, a fingerprint sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and output devices may include a display (LCD, etc.), a speaker, etc.

[0104] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0105] In a specific implementation, the processor, input device, and output device described in the embodiments of the present invention may execute the implementation described in the method embodiment provided in the embodiments of the present invention, or may execute the implementation described in the system embodiment described in the embodiments of the present invention, which will not be described in detail here.

[0106] The present invention also provides an embodiment of a computer-readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, the processor executes a method for generating a defect image of an electric component described in the first embodiment.

[0107] The computer-readable storage medium may be an internal storage unit of the terminal described in the foregoing embodiments, such as a hard disk or memory of the terminal. The computer-readable storage medium may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the computer-readable storage medium may also include both an internal storage unit of the terminal and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the terminal. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0108] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0109] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the terminals and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here.

[0110] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0111] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for generating a defect image of an electric component, characterized in that: The following steps are involved: Acquire real-time long-range shooting image data of the monitoring area, perform preprocessing to obtain preprocessed real-time long-range shooting image data, use the constructed electric power equipment target detection model to perform target detection on the preprocessed real-time long-range shooting image data, and obtain real-time electric power equipment target image data; Using the constructed power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data to obtain a real-time power component defect recognition result; A real-time power component defect image is generated based on the real-time power component defect recognition result.

2. The method for generating a defect image of an electric component according to claim 1, characterized in that: Before the step of acquiring the real-time long-range shooting image data of the monitoring area, the method further includes: constructing a power equipment target detection model and a power component defect recognition model, and the specific method includes: Acquire a number of historical perspective shooting image data and perform preprocessing to obtain a number of preprocessed historical perspective shooting image data; According to the pre-processed historical perspective shooting image data, a weakly supervised image recognition algorithm is used to build a power equipment target detection model, and the power equipment target detection model is used to perform target detection on the pre-processed historical perspective shooting image data to obtain a number of historical power equipment target image data; Based on the historical target image data of power equipment, a deep learning algorithm is used to construct a power component defect recognition model.

3. The method for generating a defect image of an electric component according to claim 2, characterized in that: The specific method of constructing a power equipment target detection model using a weakly supervised image recognition algorithm based on pre-processed historical perspective image data includes: The multi-color channel edge detection algorithm is used to extract the historical edge features of the pre-processed historical perspective shooting image data; A feature point detection algorithm is used to extract several historical feature points from historical edge features; Taking the extreme value points in the historical feature points as corresponding relaxed edge points to obtain a number of relaxed edge points; A foreground seed point generation model is constructed using a deep learning algorithm, and several pre-processed historical distant view image data with relaxed edge points are predicted to obtain several historical foreground seed points. Expanding the historical foreground seed points to obtain a corresponding historical seed cluster set, and calculating the geodesic distance from each pixel point in the preprocessed historical distant view image data to the historical seed cluster set; A pseudo-label for each preprocessed historical perspective image data is generated according to the geodesic distance, and a power equipment target detection model is constructed using a deep learning algorithm. The power equipment target detection model is used to perform target detection on the preprocessed historical perspective image data with pseudo-labels to obtain several historical power equipment target image data.

4. The method for generating a defect image of an electric component according to claim 3, characterized in that: The specific method of using the constructed power equipment target detection model to perform target detection on real-time long-range shooting image data includes: Collecting real-time long-range shooting image data of the monitoring area and preprocessing it to obtain preprocessed real-time long-range shooting image data; The pre-processed real-time distant view shooting image data is processed by using a foreground seed point generation model to obtain a number of real-time foreground seed points, and corresponding real-time foreground seed point regions are obtained according to the number of real-time foreground seed points; Using the electric power equipment target detection model, target detection is performed on the real-time foreground seed point area to obtain a real-time detection frame of the real-time electric power equipment target; According to the real-time detection frame of the real-time power equipment target, the pre-processed real-time distant view shooting image data is cropped to obtain the real-time power equipment target image data.

5. The method for generating a defect image of an electric component according to claim 4, characterized in that: The specific method of using the constructed power component defect recognition model to perform power component defect recognition on real-time power equipment target image data includes: The power component defect recognition model is used to extract real-time high-level image features and real-time low-level image features of real-time power equipment target image data; According to the preset attention weight value, the real-time high-level features and the real-time low-level features are fused to obtain the real-time fused image features; The real-time fusion image features are used to identify defects in power components, and real-time power component defect identification results are obtained.

6. The method for generating a defect image of an electric component according to claim 5, characterized in that: The specific method for generating a real-time power component defect image according to the real-time power component defect recognition result comprises: Determine whether there is a power component defect in the real-time power component defect identification result; If so, the real-time RGB values ​​of all pixels in the real-time foreground seed point area of ​​the real-time power equipment target image data are retained to obtain a color real-time power component defect target; Graying all pixel points in the real-time background area except the color real-time power component defect target in the real-time power equipment target image data to obtain a real-time power component defect image; If not, continue to collect and monitor the real-time distant view shooting image data at the next moment.

7. The method for generating a defect image of an electric component according to claim 6, characterized in that: The specific method of expanding the historical foreground seed points to obtain the corresponding historical seed cluster set includes: Performing gradient calculation on the grayscale values ​​of adjacent pixel points around each foreground seed point of the preprocessed historical distant shot image data, and generating an extended seed point list for each foreground seed point based on the gradient calculation result; De-duplicate the extended seed point list of each foreground seed point to obtain the historical seed cluster of each foreground seed point; All foreground seed points are traversed to obtain a set of historical seed clusters of preprocessed historical distant view image data.

8. A power component defect image generation system, characterized in that: include: A data acquisition module, a target detection module, a defect recognition module and a defect image generation module, wherein the data acquisition module is used to acquire the real-time distant view shooting image data of the monitoring area and perform preprocessing to obtain the real-time distant view shooting image data after preprocessing; The target detection module is used to perform target detection on the pre-processed real-time distant view shooting image data using the electric power equipment target detection model to obtain real-time electric power equipment target image data; The defect recognition module is used to use the power component defect recognition model to perform power component defect recognition on the real-time power equipment target image data to obtain a real-time power component defect recognition result; The defect image generation module is used to generate a real-time power component defect image according to the real-time power component defect recognition result.

9. An electronic device, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor is connected to the input device, the output device and the memory respectively, the memory is used to store a computer program, and the computer program comprises program instructions. The method is characterized in that the processor is configured to call the program instructions to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 7.

Images