Power equipment image processing method, device, equipment and medium

Abstract

The application relates to a power equipment image processing method, device, equipment and medium. The method comprises the following steps: acquiring criterion data corresponding to each candidate scheme, wherein the criterion data is used for representing the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing, and the candidate scheme is a preset algorithm or a preset model used for power equipment image processing; acquiring a judgment matrix corresponding to a target scene, wherein the non-diagonal line elements in the judgment matrix are determined based on the first number and the importance of each dimension under the target scene; and determining a target scheme from each candidate scheme according to the judgment matrix and the criterion data of each dimension, wherein the target scheme is used for performing power equipment image processing based on the target scheme. The method can obtain a target scheme for performing power equipment image processing in a suitable target scene.

Description

Technical Field

[0001] This application relates to the field of power equipment technology, and in particular to a method, apparatus, device and medium for image processing of power equipment. Background Technology

[0002] With the rapid development of deep learning and machine learning technologies, a large number of algorithms or models for power equipment image processing have emerged. However, each algorithm or model exhibits its own advantages and disadvantages in terms of efficiency, accuracy, and resource consumption on different tasks and datasets, such as power equipment image processing, power equipment target detection, and power equipment semantic segmentation.

[0003] Current methods for image processing of power equipment mainly focus on optimizing and improving existing algorithms to better perform image processing. However, selecting the most suitable model to adapt to specific tasks and application scenarios for power equipment image processing remains a pressing issue. Summary of the Invention

[0004] Therefore, it is necessary to provide a power equipment image processing method, apparatus, device, and medium that can select a suitable model to adapt to specific tasks and application scenarios, addressing the aforementioned technical problems.

[0005] In a first aspect, this application provides a method for image processing of power equipment, including:

[0006] Obtain the target criterion data corresponding to each candidate scheme. The criterion data is used to characterize the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing.

[0007] Obtain the judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension in the target scene.

[0008] Based on the judgment matrix and the target criteria data, the target scheme is determined from the candidate schemes. The target scheme is used for power equipment image processing.

[0009] In one embodiment, the criteria data corresponding to each candidate solution are obtained, including:

[0010] Obtain the initial criterion data for each candidate scheme;

[0011] The initial criterion data are normalized to obtain intermediate criterion data;

[0012] The intermediate criterion data are processed by integral mapping to obtain the target criterion data.

[0013] In one embodiment, the initial criterion data are normalized to obtain intermediate criterion data, including:

[0014] The initial criterion data are standardized by zero mean to obtain standardized data.

[0015] The standardized data are forward standardized to obtain the intermediate criterion data.

[0016] In one embodiment, obtaining the judgment matrix corresponding to the target scene includes:

[0017] Generate importance scale values ​​based on the first quantity;

[0018] Obtain the importance ranking data for each dimension in the target scenario;

[0019] The judgment matrix is ​​obtained by comparing each dimension pairwise based on the importance ranking data and the importance scale values.

[0020] In one embodiment, generating importance scale values ​​based on the dimensions of the target criterion data includes:

[0021] The first quantity is processed by the natural logarithm to obtain the scale denominator value;

[0022] Based on the importance of each dimension, a scale numerator value is generated. The magnitude of each scale numerator value is positively correlated with the magnitude of each dimension. The maximum value among the scale numerator values ​​is the first quantity, and the minimum value among the scale numerator values ​​is 1.

[0023] The importance scale values ​​are obtained based on the numerator and denominator values ​​of each scale.

[0024] In one embodiment, the process of performing pairwise comparisons of each dimension based on the importance ranking data and each importance scale value to obtain the judgment matrix further includes:

[0025] Perform consistency verification on the judgment matrix to obtain the verification result;

[0026] If the verification result is that the verification fails, the process of obtaining and re-obtaining the importance data of each dimension in the target scenario will be repeated until the verification result is obtained, until the verification result is that the verification passes.

[0027] In one embodiment, the target solution is determined from each candidate solution based on the judgment matrix and the target criterion data, including:

[0028] The judgment matrix is ​​normalized to obtain the judgment weight vector;

[0029] The data for each objective criterion are multiplied by the judgment weight vector to obtain the evaluation score corresponding to each candidate scheme.

[0030] The candidate solution with the highest evaluation score among all evaluation scores is determined as the target solution.

[0031] Secondly, this application also provides an image processing apparatus for power equipment, comprising:

[0032] The criterion data acquisition module is used to acquire the target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing.

[0033] The weighted data acquisition module is used to obtain the judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension under the target scene.

[0034] The target scheme determination module is used to determine the target scheme from each candidate scheme based on the judgment matrix and each target criterion data. The target scheme is used for power equipment image processing based on the target scheme.

[0035] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0036] Obtain the target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing.

[0037] Obtain the judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension in the target scene.

[0038] Based on the judgment matrix and the target criteria data, the target scheme is determined from the candidate schemes. The target scheme is used for power equipment image processing.

[0039] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0040] Obtain the target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing.

[0041] Obtain the judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension in the target scene.

[0042] Based on the judgment matrix and the target criteria data, the target scheme is determined from the candidate schemes. The target scheme is used for power equipment image processing.

[0043] The aforementioned power equipment image processing method, apparatus, device, and medium acquire target criterion data corresponding to each candidate scheme. This criterion data characterizes the performance of the candidate schemes in a first number of dimensions during power equipment image processing. The candidate schemes are preset algorithms or models used for power equipment image processing. A judgment matrix corresponding to the target scene is acquired. The off-diagonal elements in the judgment matrix are determined based on the first number and the importance of each dimension under the target scene. Based on the judgment matrix and the target criterion data, a target scheme is determined from the candidate schemes. This target scheme is used for power equipment image processing. Thus, the performance of each candidate scheme in power equipment image processing is first quantified using the target criterion data corresponding to each candidate scheme. Then, the judgment matrix is ​​acquired to quantify the relative importance of the performance in each dimension under the target scene. Based on the target criterion data and the judgment matrix corresponding to the target scene, a suitable target scheme for power equipment image processing can be obtained. Furthermore, in this embodiment, the off-diagonal elements in the judgment matrix are determined based on the first number and the importance of each dimension under the target scene, which can omit the determination of the relative importance between dimensions during the acquisition of the judgment matrix, improving processing efficiency. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is an application environment diagram of the power equipment image processing method in one embodiment;

[0046] Figure 2 This is a flowchart illustrating an image processing method for power equipment in one embodiment;

[0047] Figure 3 This is a flowchart illustrating the steps for obtaining data for each target criterion in one embodiment;

[0048] Figure 4 This is a flowchart illustrating the steps for obtaining intermediate criterion data in one embodiment;

[0049] Figure 5 This is a flowchart illustrating the steps for obtaining the judgment matrix in one embodiment;

[0050] Figure 6 This is a flowchart illustrating the steps for obtaining each importance scale value in one embodiment;

[0051] Figure 7 This is a flowchart illustrating the steps for determining the target solution in one embodiment;

[0052] Figure 8 This is a flowchart illustrating the image processing method for power equipment in another embodiment;

[0053] Figure 9 This is a structural block diagram of an image processing device for power equipment in one embodiment;

[0054] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] The power equipment image processing method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0057] Terminal 102 acquires target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in a first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing. It acquires a judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first number and the importance of each dimension in the target scene. Based on the judgment matrix and each target criterion data, the target scheme is determined from each candidate scheme. The target scheme is used to perform power equipment image processing based on the target scheme.

[0058] In one exemplary embodiment, such as Figure 2 As shown, an image processing method for power equipment is provided, which can be applied to... Figure 1 Taking terminal 102 as an example, the explanation includes the following steps 202 to 206. Wherein:

[0059] Step 202: Obtain the target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing.

[0060] Among them, the candidate schemes can be preset algorithms for power equipment image processing, such as K-Nearest Neighbor (KNN), Naive Bayes model (NBM), Support Vector Machine (SVM), etc., or preset models for power equipment image processing, such as Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Generative Adversarial Networks (GAN), Residual Network (ResNet), etc.

[0061] Among them, power equipment may include transformers, power switchgear, and GIS (Gas Insulated Switchgear) equipment.

[0062] The target criterion data of the candidate scheme is used to characterize the performance of the candidate scheme in a first number of dimensions when performing power equipment image processing; for example, the target criterion data includes the accuracy, resource consumption, and computing speed of the candidate scheme in power equipment image processing, and the corresponding dimension of the target criterion data is 3. For ease of description, N is used to represent the first number in subsequent embodiments.

[0063] In the process of determining the target solution from multiple candidate solutions, the dimensions and quantity of the target criterion data corresponding to each candidate solution are consistent.

[0064] In one possible implementation, obtaining the target criterion data corresponding to each candidate solution may include: using a sample database corresponding to the target scenario to test the performance of each candidate solution in N dimensions, thereby obtaining the target criterion data corresponding to each candidate solution.

[0065] For example, the specific performance dimensions of the candidate solutions in the target criterion data can be determined based on the target scenario; and for another example, the specific performance dimensions of the candidate solutions in the target criterion data can be pre-defined fixed dimensions.

[0066] Step 204: Obtain the judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension of the target scene.

[0067] The judgment matrix is ​​used to characterize the relative importance of each dimension in the target scenario. The order of the judgment matrix is ​​consistent with the dimension of the target criterion data; in this embodiment, the order of the judgment criterion is N. Each element in the judgment matrix represents the relative importance between two dimensions. Generally, the importance of each dimension in the target scenario is determined based on expert judgment or experience. Then, a certain scale is used to compare the importance of each dimension pairwise, and the comparison results are used as the corresponding elements in the judgment matrix.

[0068] The judgment matrix can be represented as shown in Formula 1:

[0069]

[0070] Among them, a ij This indicates the relative importance of the factor in the i-th row compared to the factor in the j-th column.

[0071] Traditional methods for determining the elements of a judgment matrix employ a nine-level scaling method, which uses numerical values ​​from 1 to 9 and their reciprocals to assign values ​​to each element in the judgment matrix. For example, the meaning of each numerical value in the nine-level scaling method is shown in Table 1.

[0072] In this embodiment, the elements on the off-diagonal of the judgment matrix are determined based on the number of dimensions of the target criterion data (i.e., the first quantity N) and the importance of each dimension of the target scene. This avoids the traditional method of directly assigning fixed scale values ​​to the off-diagonal elements in the judgment matrix, omitting the process of confirming the importance between dimensions, and enabling faster determination of the judgment matrix. The judgment matrix can be easily and quickly constructed simply by obtaining the differences in the importance of each dimension of the target scene as determined by experts or experience. In this embodiment, the value of each element on the diagonal of the judgment matrix is ​​1.

[0073] Table 1

[0074]

[0075] Step 206: Based on the judgment matrix and the target criteria data, determine the target scheme from each candidate scheme. The target scheme is used for power equipment image processing based on the target scheme.

[0076] In this process, the judgment matrix is ​​used as the weight matrix for each target criterion data to obtain the evaluation score corresponding to each candidate model. This evaluation score can quantify the processing capability of the candidate model for the corresponding power equipment image in the target scenario. For example, the candidate scheme with the largest evaluation score is selected as the target scheme, and then power equipment image processing is performed based on the target scheme.

[0077] In the above-described power equipment image processing method, target criterion data corresponding to each candidate scheme is obtained. The criterion data is used to characterize the performance of the candidate scheme in a first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing. A judgment matrix corresponding to the target scene is obtained. The off-diagonal elements in the judgment matrix are determined based on the first number and the importance of each dimension under the target scene. According to the judgment matrix and each target criterion data, a target scheme is determined from each candidate scheme. The target scheme is used to perform power equipment image processing. In this way, the performance of each candidate scheme in performing power equipment image processing is first quantified by the target criterion data corresponding to each candidate scheme. Then, the judgment matrix is ​​obtained to quantify the relative importance of the performance in each dimension under the target scene. According to each target criterion data and the judgment matrix corresponding to the target scene, the target scheme most suitable for power equipment image processing under the target scene can be obtained. At the same time, in this embodiment, the off-diagonal elements in the judgment matrix are determined based on the first number and the importance of each dimension under the target scene. The determination of the relative importance between each dimension can be omitted in the process of obtaining the judgment matrix, thereby improving processing efficiency and processing reliability.

[0078] In an exemplary embodiment, step 202, obtaining the target criterion data corresponding to each candidate scheme, may include: preprocessing the initial criterion data of each candidate scheme to obtain the target criterion data. This unifies the distribution range of the criterion sub-data corresponding to different dimensions, making the distribution of the target criterion data more intuitive and effective. In this embodiment, the performance data corresponding to each dimension can be referred to as criterion sub-data.

[0079] In one exemplary embodiment, such as Figure 3 As shown, step 202 includes steps 302 to 306.

[0080] in:

[0081] Step 302: Obtain the initial criterion data corresponding to each candidate scheme.

[0082] In this embodiment, the initial criterion data refers to data that directly describes the performance of each candidate scheme. This embodiment uses five candidate schemes as examples to illustrate the principle and beneficial effects of the method provided in this embodiment: the original ResNet34 (i.e., the original model in Table 2), the ResNet34 model that uses transfer learning to obtain pre-trained weights and performs fine-tuning (i.e., the pre-linked model in Table 2), the ResNet34 model that uses depthwise separable convolution for lightweighting (i.e., the lightweight model in Table 2), the ResNet34 model that uses the multi-head output structure in network depth adaptive adjustment technology (i.e., the multi-head output model in Table 2), and the ResNet34 model that uses a multi-branch structure (i.e., the multi-branch model in Table 2).

[0083] For example, the initial criterion data corresponding to each candidate scheme are shown in Table 2.

[0084] Table 2

[0085]

[0086] Step 304: Normalize the initial criterion data to obtain intermediate criterion data.

[0087] As shown in Table 2, the distribution range of the data in each dimension of the target criterion data is not uniform, making it difficult to compare. Therefore, in this embodiment, the initial accurate data are normalized to unify the distribution range of the data in each dimension, so as to accurately quantify the performance of each candidate model.

[0088] In one possible implementation, please refer to Figure 4 Step 304 includes steps 402 and 404, wherein:

[0089] Step 402: Perform zero-mean standardization on each initial criterion data to obtain standardized data.

[0090] Zero-mean standardization, also known as z-score standardization, involves subtracting the mean from the original data and then dividing by the standard deviation, resulting in data with a mean of 0 and a standard deviation of 1. For example, zero-mean standardization can be represented by the following formula 2:

[0091]

[0092] Where x′ is the standardized data, x is the original data, μ is the mean of the original data, and δ is the standard deviation of the original data.

[0093] For example, please refer to Table 3, which shows the standardized data after zero-mean standardization of the initial criterion data corresponding to each candidate scheme shown in Table 2.

[0094] Table 3

[0095]

[0096] Step 404: Perform forward standardization on each standardized data to obtain each intermediate criterion data.

[0097] In this context, higher performance data for a particular dimension does not always indicate better performance. For example, as shown in Table 2, resource consumption and computation speed are indicators where lower is better, while accuracy is an indicator where higher is better. To ensure that the sub-data for each dimension in the target criterion data are represented according to the positive standard that higher values ​​indicate better candidate models, the standardized data are subjected to positive standardization to obtain intermediate criterion data.

[0098] For example, the data for resource consumption and computing speed in criterion 3 are inverted, while the data for accuracy remains unchanged, to obtain the intermediate criterion data.

[0099] Step 306: Perform integral mapping processing on each intermediate criterion data to obtain each target criterion data.

[0100] The intermediate criterion data are obtained through zero-mean standardization and conform to a normal distribution. However, normally distributed data cannot intuitively represent the differences between individual data points. Therefore, in this embodiment, the intermediate criterion data are mapped using normal integrals based on Formula 3:

[0101]

[0102] In Formula 3, x is the standardized x′ in Formula 1. Formula 2 calculates the integral of the normal distribution of x, where P is the probability that the probability value is less than x, denoted as the function Φ(x), and μ and δ are the mean and standard deviation of the data, respectively. The calculated probability value P is between 0 and 1. Magnifying P by 100 times allows it to be mapped to a percentage system.

[0103] For example, Table 4 provides the target criterion data obtained after integral mapping processing of the standardized data corresponding to each candidate scheme in Table 3.

[0104] Table 4

[0105]

[0106] In the power equipment image processing method provided in the above embodiments, the initial criterion data corresponding to each candidate scheme is first obtained, and then the initial criterion data is normalized to obtain intermediate criterion data with unified step-by-step value range; then the intermediate criterion data is integrally mapped to obtain target criterion data that can characterize the differences in data of each dimension, thereby improving the reliability of the target scheme obtained subsequently.

[0107] In one exemplary embodiment, such as Figure 5 As shown, step 204 includes steps 502 to 506, wherein:

[0108] Step 502: Generate each importance scale value based on the first quantity.

[0109] In this embodiment, each importance scale value is determined based on the number of dimensions corresponding to the target criterion data. For example, if the target criterion data corresponds to 3 dimensions, then each importance scale value is determined based on the value 3.

[0110] In another possible implementation, please refer to Figure 6 Step 502 includes steps 602 to 606, wherein:

[0111] Step 602: Perform natural logarithmic processing on the first quantity to obtain the scaling denominator value.

[0112] Step 604: Generate each scale molecule value according to the importance of each dimension. The size of each scale molecule value is positively correlated with the size of each dimension. The maximum value among the scale molecules is the first quantity, and the minimum value among the scale molecules is 1.

[0113] Step 606: Obtain the importance scale values ​​based on the numerator and denominator values ​​of each scale.

[0114] Through steps 502 to 506 above, the importance scale values ​​of the judgment matrix can be directly determined as follows: This implementation uses the natural logarithm ln(N) as the scaling factor, which can effectively reduce the variance between elements in the judgment matrix and increase the stability and rationality of the judgment matrix.

[0115] Please refer to Table 5 for an example illustrating the meaning of the logarithmic scaling method provided in this embodiment. The reciprocal is used when the importance relationship between elements is reversed.

[0116] Table 5

[0117]

[0118] Step 504: Obtain the importance ranking data of each dimension in the target scenario.

[0119] For example, when processing images of power equipment in a target scenario, the target criterion data includes accuracy, resource consumption, and computing speed, with a first dimension having 3 elements and a criterion layer having 3 indicators. Expert evaluation determined that accuracy is the most important among the target criterion data, followed by resource consumption, and computing speed is the least important. Therefore, the importance ranking of each dimension can be obtained using importance ranking data.

[0120] Step 506: Based on the importance ranking data and the importance scale values, perform pairwise comparisons on each dimension to obtain the judgment matrix.

[0121] For example, when the first number of dimensions is 3, the importance scale values ​​obtained by the logarithmic scaling method are 2.73, 1.82, and 0.91 from high to low.

[0122] Table 6 provides an example of a judgment matrix obtained using the logarithmic scaling method. Accuracy is considered the most important factor, followed by resource consumption, and then computation speed. Since the importance of accuracy and resource consumption differs by only one level, 1.82 (the second largest importance scale value) is used as the element value for accuracy relative to resource consumption in the judgment matrix. If the importance of accuracy and resource consumption differs by two levels, 2.73 (the first largest importance scale value) is used as the element value for accuracy relative to computation speed in the judgment matrix.

[0123] Table 6

[0124]

[0125] In this embodiment, importance scale values ​​are generated based on a first quantity, and then combined with importance ranking data. This omits the confirmation process for each dimension in the target criterion data, enabling a simple and quick determination of the judgment matrix. For example, if the traditional nine-level scaling method is used to obtain the judgment matrix, it is necessary to determine whether the relative importance between two dimensions corresponds to a proportion where one is significantly more important than the other, or is strongly more important, slightly more important, or the median of adjacent judgments. This makes the judgment matrix acquisition process lengthy and inefficient. However, using the method provided in this embodiment, the judgment matrix can be determined simply based on the order difference between the two dimensions in the importance ranking data, resulting in high efficiency. Furthermore, the method provided in this embodiment improves the reliability of the obtained judgment matrix, avoiding the low reliability problem caused by using fixed values ​​in the traditional nine-level scaling method.

[0126] In an exemplary embodiment, during the execution of step 406, the process further includes: performing a consistency check on the judgment matrix to obtain a check result; if the check result is a check failure, then re-execute the process of obtaining and re-obtaining the importance ranking data of each dimension in the target scenario until a check result is obtained, until the check result is a check success.

[0127] Consistency refers to the stability and rationality of the comparison matrix; the more consistent the matrix, the more reliable the decision result. Decision-makers need to verify the consistency of the matrix, and if the consistency requirement is not met, a new comparison is required.

[0128] The consistency index CI is calculated as shown in Formula 4.

[0129]

[0130] Where λ max is the largest eigenvalue of the judgment matrix, and n is the order of the judgment matrix. The smaller the value of the consistency index CI calculated by Formula 3, the more stable the judgment matrix is, and the more consistent the pairwise comparisons of decision-makers are.

[0131] In one possible implementation, the random consistency index RI can be obtained directly from Table 7.

[0132] Table 7

[0133]

[0134] The random consistency index RI can be directly obtained from the order n of the matrix. The consistency ratio CR can be calculated from CI and RI, as shown in Formula 5.

[0135]

[0136] If the CR is less than or equal to the threshold of 0.1, it indicates that the consistency of the judgment matrix has reached a good level and is acceptable. If the CR is greater than 0.1, it indicates that the consistency of the judgment matrix is ​​insufficient, and there may be some degree of subjective bias or contradiction, requiring re-comparison. If the judgment matrix fails the consistency test, it can be adjusted by re-acquiring the importance ranking data of each dimension in the target scenario until the consistency requirements are met.

[0137] It can be seen that, compared with the traditional nine-level scaling method for determining the judgment matrix, the aforementioned embodiment uses the method of generating each importance scale value based on the first quantity to determine the judgment matrix, which can pass the matrix consistency check more quickly and improve the efficiency of the power equipment image processing method.

[0138] In one exemplary embodiment, such as Figure 7 As shown, step 206 includes steps 702 to 706, wherein:

[0139] Step 702: Normalize the judgment matrix to obtain the judgment weight vector.

[0140] For example, the judgment matrix shown in Table 6 is first normalized by column to obtain the intermediate matrix data shown in Table 8. Then, the matrix data shown in Table 8 is normalized by column to obtain the judgment weight vector a as shown in Formula 6.

[0141] Table 8

[0142]

[0143]

[0144] Step 704: Multiply each target criterion data with the judgment weight vector to obtain the evaluation score corresponding to each candidate scheme.

[0145] For example, the judgment weight vector shown in Formula 5 and the target criterion data shown in Table 6 are multiplied together to obtain the evaluation scores corresponding to each candidate scheme, as shown in the "Score (Normal Distribution)" column of Table 9.

[0146] Table 9

[0147]

[0148] As shown in Table 9, the scores of the normal distribution can also be processed by integral mapping to obtain the evaluation scores shown in "Scores (out of 100)" in Table 9.

[0149] Step 706: The candidate solution corresponding to the target evaluation data with the largest value among all target evaluation data is determined as the target solution.

[0150] For example, as shown in Table 9, the multi-head output model has the highest score, meaning that this model has the highest accuracy and fastest computing speed in the target scenario, and can be used preferentially.

[0151] In one exemplary embodiment, such as Figure 8 As shown, the provided image processing method for power equipment includes:

[0152] Step 801: Obtain the initial criterion data corresponding to each candidate scheme. The candidate scheme is a preset algorithm or preset model for power equipment image processing. The initial criterion data refers to the data that directly describes the performance of each candidate scheme.

[0153] Step 802: Normalize the initial criterion data to obtain intermediate criterion data.

[0154] Optionally, step 802 includes: performing zero-mean standardization on each initial criterion data to obtain standardized data; and performing forward standardization on each standardized data to obtain intermediate criterion data.

[0155] Step 803: Perform integral mapping processing on each intermediate criterion data to obtain each target criterion data.

[0156] Step 804: Generate each importance scale value based on the first quantity.

[0157] Step 804 includes: performing natural logarithmic processing on the first quantity to obtain the scale denominator value; generating each scale numerator value according to the importance of each dimension, wherein the size of each scale numerator value is positively correlated with the size of each dimension, and the maximum value among the scale numerator values ​​is the first quantity, and the minimum value among the scale numerator values ​​is 1; obtaining each importance scale value according to each scale numerator value and scale denominator value.

[0158] Step 805: Obtain the importance ranking data of each dimension in the target scenario.

[0159] Step 806: Based on the importance ranking data and the importance scale values, perform pairwise comparisons on each dimension to obtain the judgment matrix.

[0160] Optionally, the process of executing step 806 may also include: performing consistency verification on the judgment matrix to obtain the verification result; if the verification result is that the verification fails, then re-execute the process of obtaining and re-obtaining the importance data of each dimension in the target scenario until the verification result is obtained, until the verification result is that the verification passes.

[0161] Step 807: Normalize the judgment matrix to obtain the judgment weight vector.

[0162] Step 808: Multiply each target criterion data with the judgment weight vector to obtain the evaluation score corresponding to each candidate scheme.

[0163] Step 809: The candidate solution corresponding to the highest evaluation score among all evaluation scores is determined as the target solution.

[0164] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0165] Based on the same inventive concept, this application also provides a power equipment image processing apparatus for implementing the power equipment image processing method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more power equipment image processing apparatus embodiments provided below can be found in the limitations of the power equipment image processing method described above, and will not be repeated here.

[0166] In one exemplary embodiment, such as Figure 9 As shown, an image processing device for power equipment is provided, comprising: a criterion data acquisition module 902, a weight data acquisition module 904, and a target scheme determination module 906, wherein:

[0167] The criterion data acquisition module 902 is used to acquire the target criterion data corresponding to each candidate scheme. The criterion data is used to characterize the performance of the candidate scheme in the first number of dimensions when performing power equipment image processing. The candidate scheme is a preset algorithm or preset model for power equipment image processing.

[0168] The weighted data acquisition module 904 is used to obtain the judgment matrix corresponding to the target scene. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension under the target scene.

[0169] The target scheme determination module 906 is used to determine the target scheme from each candidate scheme according to the judgment matrix and each target criterion data. The target scheme is used for power equipment image processing based on the target scheme.

[0170] In one exemplary embodiment, the criterion data acquisition module 902 includes:

[0171] The initial criterion acquisition unit is used to acquire the initial criterion data corresponding to each candidate scheme.

[0172] The criterion normalization unit is used to normalize the initial criterion data to obtain intermediate criterion data.

[0173] The integral mapping unit is used to perform integral mapping processing on the intermediate criterion data to obtain the target criterion data.

[0174] In an exemplary embodiment, the criterion normalization unit is used to perform zero-mean standardization on each initial criterion data to obtain standardized data, and to perform forward standardization on each standardized data to obtain intermediate criterion data.

[0175] In one exemplary embodiment, the weight data acquisition module 904 includes:

[0176] A scale value generation unit is used to generate scale values ​​for each degree of importance based on a first quantity.

[0177] The sorting data acquisition unit is used to acquire sorting data of importance of each dimension in the target scenario;

[0178] The matrix determination unit is used to perform pairwise comparisons of each dimension based on the data sorted by importance and the importance scale values ​​to obtain the judgment matrix.

[0179] In an exemplary embodiment, the scaling value generation unit is used to process the first quantity using the natural logarithm to obtain a scaling denominator value; generate each scaling numerator value according to the importance corresponding to each dimension, wherein the size of each scaling numerator value is positively correlated with the size of each dimension, and the maximum value among the scaling numerator values ​​is the first quantity, and the minimum value among the scaling numerator values ​​is 1; and obtain each importance scaling value according to each scaling numerator value and the scaling denominator value.

[0180] In an exemplary embodiment, the matrix determination unit is used to perform consistency verification processing on the judgment matrix to obtain a verification result; if the verification result is that the verification fails, the process of obtaining and re-obtaining the importance data of each dimension in the target scenario is repeated until the verification result is obtained, until the verification result is that the verification passes.

[0181] In one exemplary embodiment, the target solution determination module 906 includes:

[0182] The weight vector acquisition unit is used to normalize the judgment matrix to obtain the judgment weight vector;

[0183] The evaluation score acquisition unit is used to multiply the data of each target criterion with the judgment weight vector to obtain the evaluation score corresponding to each candidate scheme.

[0184] The target solution determination unit determines the candidate solution corresponding to the highest evaluation score among all evaluation scores as the target solution.

[0185] Each module in the aforementioned power equipment image processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0186] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an image processing method for power equipment. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0187] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0188] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0189] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0190] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0192] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0193] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for image processing of power equipment, characterized in that, The method includes: Obtain target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in a first number of dimensions when performing power equipment image processing. The candidate scheme is a preset model for power equipment image processing. The preset model is any one of convolutional neural network, recurrent neural network and residual network. The first quantity is processed by natural logarithm to obtain the scaling denominator value; according to the importance of each dimension, each scaling numerator value is generated, wherein the size of each scaling numerator value is positively correlated with the size of each dimension, and the maximum value among the scaling numerator values ​​is the first quantity, and the minimum value among the scaling numerator values ​​is 1; according to the scaling numerator values ​​and the scaling denominator value, each importance scaling value is obtained; importance ranking data of each dimension in the target scene is obtained; each dimension is compared pairwise according to the importance ranking data and each importance scaling value to obtain a judgment matrix; consistency verification is performed on the judgment matrix to obtain a verification result; if the verification result is a verification failure, the process of obtaining and re-obtaining the importance data of each dimension in the target scene is repeated until a verification result is obtained, until the verification result is a verification success. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension in the target scene; Based on the judgment matrix and the target criterion data, a target scheme is determined from the candidate schemes, and the target scheme is used for power equipment image processing.

2. The method according to claim 1, characterized in that, The acquisition of target criterion data corresponding to each candidate scheme includes: Obtain the initial criterion data corresponding to each candidate scheme; The initial criterion data are normalized to obtain intermediate criterion data; The intermediate criterion data are subjected to integral mapping to obtain the target criterion data.

3. The method according to claim 2, characterized in that, The normalization process for each of the initial criterion data to obtain intermediate criterion data includes: The initial criterion data are standardized by zero mean to obtain standardized data. The standardized data are forward standardized to obtain the intermediate criterion data.

4. The method according to claim 1, characterized in that, The step of determining the target solution from the candidate solutions based on the judgment matrix and each of the target criterion data includes: The judgment matrix is ​​normalized to obtain the judgment weight vector; The target criteria data are multiplied by the judgment weight vector to obtain the evaluation score corresponding to each candidate scheme. The candidate solution corresponding to the highest evaluation score among all the evaluation scores is determined as the target solution.

5. An image processing device for power equipment, characterized in that, The device includes: The criterion data acquisition module is used to acquire target criterion data corresponding to each candidate scheme. The target criterion data is used to characterize the performance of the candidate scheme in a first number of dimensions when performing power equipment image processing. The candidate scheme is a preset model for power equipment image processing. The preset model is any one of convolutional neural network, recurrent neural network and residual network. The weighted data acquisition module is used to process the first quantity using the natural logarithm to obtain a scaling denominator value; generate scaling numerator values ​​according to the importance of each dimension, wherein the size of each scaling numerator value is positively correlated with the size of each dimension, and the maximum value among the scaling numerator values ​​is the first quantity, and the minimum value among the scaling numerator values ​​is 1; obtain each importance scaling value based on the scaling numerator values ​​and the scaling denominator value; acquire importance ranking data of each dimension in the target scenario; perform pairwise comparison processing on each dimension based on the importance ranking data and each importance scaling value to obtain a judgment matrix; perform consistency verification processing on the judgment matrix to obtain a verification result; if the verification result is a verification failure, the process of acquiring and re-acquiring the importance data of each dimension in the target scenario is repeated until a verification result is obtained, until the verification result is a verification success. The off-diagonal elements in the judgment matrix are determined based on the first quantity and the importance of each dimension in the target scenario. The target scheme determination module is used to determine a target scheme from the candidate schemes based on the judgment matrix and the target criterion data, and the target scheme is used for power equipment image processing based on the target scheme.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images