Method and system for constructing power line defect library

Abstract

This invention relates to the field of power line maintenance technology, and more particularly to a method and system for constructing a power line defect database. The method first acquires multiple image sets, each containing multiple images representing the same power line condition. Then, multiple feature sets and features generated by an extended model are fed into a classification model. The parameters of the extended model are adjusted based on the output of the classification model until the classification result is below a threshold. Next, multiple extended features are generated using the extended model. Finally, these extended features are fed into the classification model, and based on the classification result, they are added to the multiple feature sets. This invention generates a relatively large defect feature database using a relatively small number of defect image samples, requiring fewer defect image samples but generating more defect samples, thus reducing the workload of constructing the defect database.

Description

Technical Field

[0001] This invention relates to the field of power line maintenance technology, and in particular to a method and system for constructing a power line defect database. Background Technology

[0002] Overhead transmission lines typically range in length from several kilometers to hundreds of kilometers. Within such a long and narrow area, the equipment operates exposed to the natural environment, enduring various climatic conditions (such as storms, floods, ice and snow, fog, pollution, and lightning strikes). Furthermore, they are susceptible to damage from other external forces (such as agricultural machinery impacting towers or guy wire foundations, fallen trees and bamboo colliding with conductors, construction work near the line involving soil extraction, blasting, shooting, passing vehicles and cranes breaking conductors, kites getting caught on conductors causing phase-to-phase short circuits, and grounding short circuits caused by birds or animals). All these factors constantly threaten the safe operation of the lines.

[0003] Therefore, the lines are prone to faults, and once a fault occurs, it takes a long time to repair and restore power, resulting in varying degrees of loss. To ensure the safe operation of the lines, the principle of "safety first, prevention foremost" must be implemented during line operation. Line patrols and inspections must be strengthened to promptly identify equipment defects and factors that endanger the safe operation of the lines, so as to eliminate hidden dangers in a timely manner and formulate safety measures.

[0004] Line inspection work has emerged as a result. Line inspection involves inspecting the operation of line equipment in detail along the line, promptly identifying defects and faults in the equipment, and recording them in detail as a basis for line maintenance.

[0005] In existing technologies, line inspection work relies on human experience, and repetitive and high-intensity work is inevitable. Based on this, it is necessary to develop a defect library to enrich existing line inspection experience, digitize and visualize the inspection experience, and thus provide the possibility for developing an automated inspection system.

[0006] Therefore, it is necessary to develop a method and system for constructing a power line defect database. Summary of the Invention

[0007] The present invention provides a method and system for constructing a power line defect database, which solves the problem of insufficient defect database samples in the prior art.

[0008] In a first aspect, embodiments of the present invention provide a method for constructing a power line defect database, comprising:

[0009] Acquire multiple image sets, where each image set contains multiple images representing the state of the same power line;

[0010] Multiple feature sets are extracted and generated from multiple image sets, and these feature sets, along with multiple features generated by the augmented model, are fed into the classification model. The parameters of the augmented model are adjusted based on the output of the classification model until the classification result of the classification model meets predetermined conditions.

[0011] Multiple extended features are generated through the extended model;

[0012] The extended features are fed into the classification model, and the extended features are added to the feature set according to the classification result of the classification model.

[0013] In one possible implementation, the classification model is constructed based on a neural network model, including: a first input layer, a first hidden layer, and a first output layer. The classification model is trained based on the multiple image sets to achieve the expected classification accuracy.

[0014] The training process of the classification model includes:

[0015] Feature extraction is performed on the multiple image sets to obtain multiple feature sets, wherein the multiple feature sets correspond to the multiple image sets, and the multiple image sets include at least one image set representing defects in power lines;

[0016] The features from the multiple feature sets are respectively input into the classification model to obtain multiple output values ​​of the classification model;

[0017] Based on the multiple output values ​​and the power line defects represented by the multiple feature sets, the parameters of the classification model are adjusted until the classification model can determine the power line defects based on the features in the multiple feature sets.

[0018] In one possible implementation, the feature extraction from the plurality of image sets to obtain a plurality of feature sets includes:

[0019] For each of the multiple image sets, perform the following steps:

[0020] Multiple sequences are obtained as multiple features according to a first formula, wherein the multiple sequences are obtained based on multiple images in an image set, and the sequences correspond to the images. The sequences are used to characterize the edges of multiple images in the image set. The first formula is:

[0021]

[0022] In the formula, The number in the sequence k One element, The pixel values ​​of the image. N This represents the total number of pixels in the image. The highest frequency constant is predetermined. i The imaginary unit, e It is a natural constant.

[0023] In one possible implementation, the classification model is used to classify defect-free states and multiple defect states. The step of feeding multiple feature sets and multiple features generated by an extended model into the classification model, and adjusting the parameters of the extended model based on the output of the classification model until the classification result meets predetermined conditions, includes:

[0024] An extended model is constructed, which is based on a neural network model and includes: a second input layer, a second hidden layer, and a second output layer. The second output layer has multiple output nodes, which are used to output multiple elements in the sequence.

[0025] Random number generation steps: Randomly generate multiple input number sequences;

[0026] The multiple input sequences are input into the extended model to obtain multiple output sequences;

[0027] The multiple output sequences and the features in the multiple feature sets are respectively input into the classification model to obtain the number of defect-free states in the classification results of the classification model;

[0028] If the number of defect-free states exceeds the threshold, the parameters of the extended model are adjusted, and the process jumps to the random number generation step.

[0029] In one possible implementation, the generation of multiple augmented features through the augmented model includes:

[0030] Randomly generate multiple input test sequences;

[0031] The multiple input test sequences are input into the expanded model to obtain multiple output test sequences;

[0032] Based on the multiple output test sequences and the second formula, multiple time-domain curves are synthesized. The second formula is:

[0033]

[0034] In the formula, To output the time-domain curve of the corresponding test series, To output the first number in the test sequence m One element, M To test the total number of elements in the output sequence, i The imaginary unit, e It is a natural constant;

[0035] The multiple time-domain curves are sampled and reconstructed to obtain multiple reconstructed images;

[0036] If all of the reconstructed images contain defects in the power lines, then the multiple output test sequences are used as multiple extended features.

[0037] In one possible implementation, prior to generating multiple feature sets from multiple image sets, the process includes:

[0038] For each image in multiple image sets, perform the following steps:

[0039] Desaturate the image to obtain a grayscale image;

[0040] Extract the values ​​of multiple pixels in the grayscale image in a predetermined order to obtain the image sequence;

[0041] The multiple values ​​in the image sequence are normalized according to the third formula, which is:

[0042]

[0043] In the formula, The normalized image sequence is the i-th j One element, The first image sequence before normalization j One element, This is the smallest element in the sequence of images before normalization. It is the largest element in the sequence of images before normalization.

[0044] Secondly, embodiments of the present invention provide a power line defect database construction device, comprising:

[0045] The image set acquisition module is used to acquire multiple image sets, where each image set includes multiple images representing the status of the same power line.

[0046] The extended model parameter adjustment module is used to feed multiple feature sets and multiple features generated by the extended model into the classification model, and adjust the parameters of the extended model according to the output of the classification model until the classification result of the classification model is lower than the threshold. The multiple feature sets are extracted and generated based on multiple image sets.

[0047] An expansion module is used to generate multiple expansion features through the expansion model;

[0048] as well as,

[0049] The classification module is used to input the multiple extended features into the classification model, and add the multiple extended features into the multiple feature sets according to the classification result of the classification model.

[0050] Thirdly, embodiments of the present invention provide a server, including a memory and a processor, wherein the memory stores a computer program executable on the processor, and the processor executes the computer program to implement the steps of the method as described in the first aspect or any possible implementation of the first aspect.

[0051] Fourthly, embodiments of the present invention provide a defect database construction system, the defect database construction system including a terminal and a server as described in the third aspect;

[0052] The terminal is connected to the server via a signal. The terminal is used to capture images of power lines and add these images to corresponding image sets according to the state of the power lines.

[0053] Fifthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method as described in the first aspect or any possible implementation thereof.

[0054] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0055] This invention discloses a method for constructing a power line defect database. First, multiple image sets are acquired, each image set containing multiple images representing the same power line condition. Then, multiple feature sets and features generated by an extended model are fed into a classification model. The parameters of the extended model are adjusted based on the output of the classification model until the classification result is below a threshold. The multiple feature sets are extracted and generated from the multiple image sets. Next, multiple extended features are generated using the extended model. Finally, the multiple extended features are fed into the classification model, and based on the classification result, the multiple extended features are added to the multiple feature sets. In this embodiment of the invention, a feature set is obtained through an image set. Based on the feature set, a classification model is constructed to classify different features. An expansion model is also provided to work in conjunction with the classification model. The expansion model randomly generates a sequence of simulated features. After these sequences are identified by the classification model, the accuracy of the generated sequences simulating defect features is given. The expansion model is adjusted according to this accuracy, ultimately enabling the expansion model to generate defect features. These generated defect features are then classified by the classification model and added to the feature set extracted from the images, thereby expanding the capacity of defect features. This embodiment of the invention generates a relatively large defect feature library with a relatively small number of defect image samples. It requires fewer defect image samples but generates more defect samples, reducing the difficulty of obtaining defect image samples and alleviating the workload of building the defect library. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a flowchart of the method for constructing a power line defect database provided by an embodiment of the present invention;

[0058] Figure 2 This is a functional block diagram of the neural network model provided in the embodiments of the present invention;

[0059] Figure 3 This is a functional block diagram of the power line defect database construction device provided in the embodiments of the present invention;

[0060] Figure 4 This is a functional block diagram of the server provided in an embodiment of the present invention. Detailed Implementation

[0061] In the following description, specific details such as particular system structures and techniques are set forth for illustrative purposes and not for limitation, so as to provide a thorough understanding of embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0062] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0063] The embodiments of the present invention will be described in detail below. This example is implemented based on the technical solution of the present invention, and provides detailed implementation methods and specific operation processes. However, the protection scope of the present invention is not limited to the following embodiments.

[0064] Figure 1 A flowchart illustrating the method for constructing a power line defect database as provided in an embodiment of the present invention.

[0065] like Figure 1 As shown, it illustrates the implementation flowchart of the power line defect database construction method provided by the embodiments of the present invention, which is described in detail below:

[0066] In step 101, multiple image sets are acquired, wherein each image set includes multiple images representing the state of the same power line.

[0067] In step 102, multiple feature sets are extracted and generated from multiple image sets, and the multiple feature sets and multiple features generated by the augmented model are fed into the classification model. The parameters of the augmented model are adjusted according to the output of the classification model until the classification result of the classification model meets the predetermined conditions.

[0068] In some implementations, the classification model is constructed based on a neural network model, including: a first input layer, a first hidden layer, and a first output layer. The classification model is trained based on the multiple image sets to achieve the expected classification accuracy. The training process of the classification model includes:

[0069] Feature extraction is performed on the multiple image sets to obtain multiple feature sets, wherein the multiple feature sets correspond to the multiple image sets, and the multiple image sets include at least one image set representing defects in power lines;

[0070] The features from the multiple feature sets are respectively input into the classification model to obtain multiple output values ​​of the classification model;

[0071] Based on the multiple output values ​​and the power line defects represented by the multiple feature sets, the parameters of the classification model are adjusted until the classification model can determine the power line defects based on the features in the multiple feature sets.

[0072] In some implementations, feature extraction is performed on the plurality of image sets to obtain multiple feature sets, including:

[0073] For each of the multiple image sets, perform the following steps:

[0074] Multiple sequences are obtained as multiple features according to a first formula, wherein the multiple sequences are obtained based on multiple images in an image set, and the sequences correspond to the images. The sequences are used to characterize the edges of multiple images in the image set. The first formula is:

[0075]

[0076] In the formula, The number in the sequence k One element, The pixel values ​​of the image. N This represents the total number of pixels in the image. The highest frequency constant is predetermined. i The imaginary unit, e It is a natural constant.

[0077] For example, multiple image sets are constructed based on the state of power lines. For instance, in one application scenario, there are image sets of power lines in normal condition, image sets of power lines with detached insulation, and image sets of power lines with foreign objects attached.

[0078] For the construction of a power line defect database, if a large number of defect images are taken and used to compare with images collected during inspections, then acquiring these defect images would require a massive amount of work. In addition, there may be some defect images of power lines that were not collected (for example, power lines that are rarely struck by lightning). Therefore, it is necessary to use technical means to obtain the common features of defects and build a defect database based on these common features. That is, to extract defect features from existing defect images and expand the database of these defect features.

[0079] This invention employs a dual-neural network adversarial expansion method. Specifically, one neural network acts as a classification model to identify defects corresponding to features. The other neural network acts as an expansion model to generate features based on certain rules. If these generated features pass the classification result of the classification model, they are added to the existing features according to the classification result, thus expanding the defect features.

[0080] Figure 2 This illustrates a typical artificial neural network structure, which includes an input layer, hidden layers, and an output layer. The data input to the nodes in the input layer is processed by the weight coefficients w. ij The weights are multiplied and then input into the hidden layer neuron u. j The results obtained are then processed by the weighting coefficient w. ’ jk After multiplication, the result is output to output node u. ’ k In neural networks, training involves iteratively adjusting the weight coefficients w based on the deviation between the input node's data and the output node's output. ij and w ’ jk As a result, the relationship between the input nodes and the output nodes conforms to the pattern between the input data and the corresponding results of the input data, which is the process of changing the neural network's fitting samples through iteration.

[0081] In this embodiment of the invention, the classification model takes the features of the image as input and feeds them into the nodes of the input layer. The deviation between the output of the neural network output layer and the defects represented by the image guides the modification of the various weights of the classification model until the classification model can determine the defects based on the features.

[0082] There are various ways to extract image features. For example, image features can be identified through a recognition model. The present invention extracts the boundaries of different depth levels of the image.

[0083] First, the image is converted into a sequence, which means obtaining the values ​​of each pixel in the image in a certain order and arranging them. Then, these values ​​are input into a formula to obtain the coefficients of different level boundaries. These coefficients describe the feature quantities of the boundary at that level. The formula is as follows:

[0084]

[0085] In the formula, The number in the sequence k One element, The pixel values ​​of the image. N This represents the total number of pixels in the image. The highest frequency constant is predetermined. i The imaginary unit, e It is a natural constant.

[0086] The classification model is used to classify defect-free states and various defect states. Step 102 includes:

[0087] An extended model is constructed, which is based on a neural network model and includes: a second input layer, a second hidden layer, and a second output layer. The second output layer has multiple output nodes, which are used to output multiple elements in the sequence.

[0088] Random number generation steps: Randomly generate multiple input number sequences;

[0089] The multiple input sequences are input into the extended model to obtain multiple output sequences;

[0090] The multiple output sequences and the features in the multiple feature sets are respectively input into the classification model to obtain the number of defect-free states in the classification results of the classification model;

[0091] If the number of defect-free states exceeds the threshold, the parameters of the extended model are adjusted, and the process jumps to the random number generation step.

[0092] For example, as mentioned earlier, the augmented model is also built based on a neural network. This augmented model receives a random sequence as input and outputs a sequence. The output sequence, along with the feature sequence extracted from the image, is fed into the classification model. The classification model provides a classification result. If the number of defect-free features in the classification result is much greater than the number of defect-free features in the image-extracted features, it means that the features generated by the augmented model contain a large number of defect-free features. In other words, the augmented model's augmentation effect is poor. Therefore, the parameters of the augmented model need to be adjusted, and then the next sequence generation step is performed. This process is repeated until the sequence generated by the augmented model "deceives" the classification model, making the classification model believe that the sequence output by the augmented model represents defective features. At this point, the augmented model is successfully trained.

[0093] In step 103, multiple extended features are generated using the extended model.

[0094] In some implementations, step 103 includes:

[0095] Randomly generate multiple input test sequences;

[0096] The multiple input test sequences are input into the expanded model to obtain multiple output test sequences;

[0097] Based on the multiple output test sequences and the second formula, multiple time-domain curves are synthesized. The second formula is:

[0098]

[0099] In the formula, To output the time-domain curve of the corresponding test series, To output the first number in the test sequence m One element, M To test the total number of elements in the output sequence, i The imaginary unit, e It is a natural constant;

[0100] The multiple time-domain curves are sampled and reconstructed to obtain multiple reconstructed images;

[0101] If all of the reconstructed images contain defects in the power lines, then the multiple output test sequences are used as multiple extended features.

[0102] For example, as we know, if the generated sequence passes the classification model, it is best to perform further verification, such as reconstructing the sequence generated by the classification model to create an image, and then manually analyzing the image to see if it contains defects.

[0103] Therefore, after generating the sequence through the extended model, since the sequence characterizes the edges of images at different levels, it is necessary to reconstruct the image waveform curve using the feature sequence and apply the second formula:

[0104]

[0105] In the formula, To output the time-domain curve of the corresponding test series, To output the first number in the test sequence m One element, M To test the total number of elements in the output sequence, i The imaginary unit, e It is a natural constant.

[0106] This allows us to obtain a time-domain curve, which represents the arrangement of pixel values ​​in the image. By sampling from this curve and arranging the samples into a matrix according to a certain rule, we obtain the reconstructed image. If this reconstructed image contains defects in electric field lines, then these features obtained through the extended model can be added to the features obtained from the image.

[0107] In step 104, the multiple extended features are fed into the classification model, and the multiple extended features are added to the multiple feature sets according to the classification result of the classification model.

[0108] For example, the expanded features obtained through the above steps can be added to the corresponding feature set according to the classification results of the classification model.

[0109] Furthermore, it should be noted that in some application scenarios, before extracting image features, desaturation and normalization steps are included to make the data easier to process. One normalization formula is:

[0110]

[0111] In the formula, The normalized image sequence is the i-th j One element, The first image sequence before normalization j One element, This is the smallest element in the sequence of images before normalization. It is the largest element in the sequence of images before normalization.

[0112] The present invention discloses an implementation method for constructing a power line defect database. First, multiple image sets are acquired, each image set containing multiple images representing the same power line condition. Then, multiple feature sets and features generated by an extended model are fed into a classification model. The parameters of the extended model are adjusted based on the output of the classification model until the classification result is below a threshold. The multiple feature sets are extracted and generated from the multiple image sets. Next, multiple extended features are generated using the extended model. Finally, the multiple extended features are fed into the classification model, and based on the classification result, the multiple extended features are added to the multiple feature sets. In this embodiment of the invention, a feature set is obtained through an image set. Based on the feature set, a classification model is constructed to classify different features. An expansion model is also provided to work in conjunction with the classification model. The expansion model randomly generates a sequence of simulated features. After these sequences are identified by the classification model, the accuracy of the generated sequences simulating defect features is given. The expansion model is adjusted according to this accuracy, ultimately enabling the expansion model to generate defect features. These generated defect features are then classified by the classification model and added to the feature set extracted from the images, thereby expanding the capacity of defect features. This embodiment of the invention generates a relatively large defect feature library with a relatively small number of defect image samples. It requires fewer defect image samples but generates more defect samples, reducing the difficulty of obtaining defect image samples and alleviating the workload of building the defect library.

[0113] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0114] The following are embodiments of the apparatus of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0115] Figure 3 This is a functional block diagram of the power line defect database construction device provided in the embodiments of the present invention, with reference to... Figure 3 The power line defect database construction device 3 includes: an image set acquisition module 301, an expansion model parameter adjustment module 302, an expansion module 303, and a classification module 304.

[0116] Image set acquisition module 301 is used to acquire multiple image sets, wherein the image set includes multiple images representing the state of the same power line;

[0117] The extended model parameter adjustment module 302 is used to feed multiple feature sets and multiple features generated by the extended model into the classification model, and adjust the parameters of the extended model according to the output of the classification model until the classification result of the classification model is lower than the threshold. The multiple feature sets are extracted and generated based on multiple image sets.

[0118] The expansion module 303 is used to generate multiple expansion features through the expansion model;

[0119] The classification module 304 is used to send the multiple extended features into the classification model, and add the multiple extended features to the multiple feature sets according to the classification result of the classification model.

[0120] Figure 4 This is a functional block diagram of the terminal provided in an embodiment of the present invention. For example... Figure 4 As shown, the server 4 in this embodiment includes a processor 400 and a memory 401, wherein the memory 401 stores a computer program 402 that can run on the processor 400. When the processor 400 executes the computer program 402, it implements the steps of the various power line defect database construction methods and embodiments described above, for example... Figure 1 Steps 101 to 104 are shown.

[0121] Some defect library building systems include terminals and such Figure 4 The terminal is connected to server 4 via a signal. The terminal is used to collect images of power lines and add the images of the power lines to the corresponding image set according to the state of the power lines.

[0122] For example, the computer program 402 may be divided into one or more modules / units, which are stored in the memory 401 and executed by the processor 400 to complete the present invention.

[0123] The server 4 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The server 4 may include, but is not limited to, a processor 400 and a memory 401. Those skilled in the art will understand that... Figure 4 This is merely an example of server 4 and does not constitute a limitation on server 4. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal may also include input / output devices, network access devices, buses, etc.

[0124] The processor 400 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0125] The memory 401 can be an internal storage unit of the server 4, such as a hard drive or memory of the server 4. The memory 401 can also be an external storage device of the server 4, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the server 4. Furthermore, the memory 401 can include both internal storage units and external storage devices of the server 4. The memory 401 is used to store the computer program and other programs and data required by the terminal. The memory 401 can also be used to temporarily store data that has been output or will be output.

[0126] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the aforementioned method embodiments, and will not be repeated here.

[0127] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0128] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0129] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0132] If the integrated module / unit is implemented as 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, all or part of the processes in the above-described embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various methods and apparatus embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0133] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for constructing a power line defect database, characterized in that, include: Acquire multiple image sets, where each image set contains multiple images representing the state of the same power line; Multiple feature sets are extracted and generated from multiple image sets, and these feature sets, along with multiple features generated by the augmented model, are fed into the classification model. The parameters of the augmented model are adjusted based on the output of the classification model until the classification result of the classification model meets predetermined conditions. Multiple extended features are generated through the extended model; The extended features are fed into the classification model, and the extended features are added to the feature set according to the classification result of the classification model. The classification model is constructed based on a neural network model, including a first input layer, a first hidden layer, and a first output layer. The classification model is trained on the multiple image sets to achieve the expected classification accuracy. The training process of the classification model includes: Feature extraction is performed on the multiple image sets to obtain multiple feature sets, wherein the multiple feature sets correspond to the multiple image sets, and the multiple image sets include at least one image set representing defects in power lines; The features from the multiple feature sets are respectively input into the classification model to obtain multiple output values ​​of the classification model; Based on the multiple output values ​​and the power line defects represented by the multiple feature sets, the parameters of the classification model are adjusted until the classification model can determine the power line defects based on the features in the multiple feature sets. The classification model is used to classify defect-free states and multiple defect states. The process of feeding multiple feature sets and features generated by an extended model into the classification model, and adjusting the parameters of the extended model based on the output of the classification model until the classification result meets predetermined conditions, includes: An extended model is constructed, which is based on a neural network model and includes: a second input layer, a second hidden layer, and a second output layer. The second output layer has multiple output nodes, which are used to output multiple elements in the sequence. Random number generation steps: Randomly generate multiple input number sequences; The multiple input sequences are input into the extended model to obtain multiple output sequences; The multiple output sequences and the features in the multiple feature sets are respectively input into the classification model to obtain the number of defect-free states in the classification results of the classification model; If the number of defect-free states exceeds the threshold, the parameters of the extended model are adjusted, and the process jumps to the random number generation step. Prior to generating multiple feature sets from multiple image sets, the process includes: For each image in multiple image sets, perform the following steps: Desaturate the image to obtain a grayscale image; Extract the values ​​of multiple pixels in the grayscale image in a predetermined order to obtain the image sequence; The multiple values ​​in the image sequence are normalized according to the third formula, which is: In the formula, The normalized image sequence is the i-th j One element, The first image sequence before normalization j One element, This is the smallest element in the sequence of images before normalization. It is the largest element in the sequence of images before normalization.

2. A device for constructing a power line defect database, characterized in that, include: The image set acquisition module is used to acquire multiple image sets, where each image set includes multiple images representing the status of the same power line. The extended model parameter adjustment module is used to feed multiple feature sets and multiple features generated by the extended model into the classification model, and adjust the parameters of the extended model according to the output of the classification model until the classification result of the classification model is lower than the threshold. The multiple feature sets are extracted and generated based on multiple image sets. An expansion module is used to generate multiple expansion features through the expansion model; as well as, The classification module is used to feed the multiple extended features into the classification model, and add the multiple extended features into the multiple feature sets according to the classification result of the classification model; The classification model is constructed based on a neural network model, including a first input layer, a first hidden layer, and a first output layer. The classification model is trained on the multiple image sets to achieve the expected classification accuracy. The training process of the classification model includes: Feature extraction is performed on the multiple image sets to obtain multiple feature sets, wherein the multiple feature sets correspond to the multiple image sets, and the multiple image sets include at least one image set representing defects in power lines; The features from the multiple feature sets are respectively input into the classification model to obtain multiple output values ​​of the classification model; Based on the multiple output values ​​and the power line defects represented by the multiple feature sets, the parameters of the classification model are adjusted until the classification model can determine the power line defects based on the features in the multiple feature sets. The classification model is used to classify defect-free states and multiple defect states. The process of feeding multiple feature sets and features generated by an extended model into the classification model, and adjusting the parameters of the extended model based on the output of the classification model until the classification result meets predetermined conditions, includes: An extended model is constructed, which is based on a neural network model and includes: a second input layer, a second hidden layer, and a second output layer. The second output layer has multiple output nodes, which are used to output multiple elements in the sequence. Random number generation steps: Randomly generate multiple input number sequences; The multiple input sequences are input into the extended model to obtain multiple output sequences; The multiple output sequences and the features in the multiple feature sets are respectively input into the classification model to obtain the number of defect-free states in the classification results of the classification model; If the number of defect-free states exceeds the threshold, the parameters of the extended model are adjusted, and the process jumps to the random number generation step. Prior to generating multiple feature sets from multiple image sets, the process includes: For each image in multiple image sets, perform the following steps: Desaturate the image to obtain a grayscale image; Extract the values ​​of multiple pixels in the grayscale image in a predetermined order to obtain the image sequence; The multiple values ​​in the image sequence are normalized according to the third formula, which is: In the formula, The normalized image sequence is the i-th j One element, The first image sequence before normalization j One element, This is the smallest element in the sequence of images before normalization. It is the largest element in the sequence of images before normalization.

3. A server comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in claim 1 above.

4. A defect database construction system, characterized in that, The defect database construction system includes a terminal and a server as described in claim 3; The terminal is connected to the server via a signal. The terminal is used to collect images of power lines and add the images of the power lines to the corresponding image set according to the state of the power lines.

5. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in claim 1 above.

Images