Zero sample transformer fault identification method based on multiple modes

Abstract

The invention relates to a zero sample transformer fault identification method based on multiple modes, and belongs to the field of transformer fault identification. According to the method, on the basis of a cross-modal attention technology and a multi-modal feature fusion technology, the concept of multi-modal data fusion processing and zero sample detection is fused into transformer fault diagnosis, a series of information such as pictures, sound, vibration, oil chromatography and humidity are comprehensively utilized, the transformer is comprehensively supervised from multiple dimensions, and the diagnosis result is more accurate. According to the method, the problem that a fault single-modal sample is insufficient in diagnosis model accuracy can be relieved through multi-modal data fusion, and the universality and diagnosis precision of the model are improved.

Description

Technical Field

[0001] The present invention belongs to the field of transformer fault identification, and in particular relates to a zero-sample transformer fault identification method based on multi-modality. Background Art

[0002] In recent years, with the optimization of power systems and the explosive growth of a large number of new energy systems, the installed capacity of transformers has also doubled. At the same time, the occurrence of transformer failures has also increased accordingly, which may affect equipment safety and the economic benefits of enterprises, or cause large-scale power outages, seriously endangering the economic development of society and the nation. As large-scale industrial equipment, power transformers are expensive and their health status gradually deteriorates over time. In addition, the difficulty and high cost of collecting fault samples have brought challenges to the research in the field of transformer fault diagnosis. At the same time, transformer faults are highly complex and correlated, and even local faults can cause global abnormalities. Therefore, how to accurately diagnose and locate faults using the multimodal data generated by the transformer when it is online has become a difficult problem to be solved in the current field of transformer fault diagnosis. Summary of the invention

[0003] The purpose of the present invention is to solve the problem that single modal data cannot achieve accurate power equipment fault diagnosis, and to provide a zero-sample transformer fault identification method based on multi-modality.

[0004] To achieve the above objectives, the technical solution of the present invention is: a zero-sample transformer fault identification method based on multimodal, which integrates the concepts of multimodal data fusion processing and zero-sample detection into transformer fault diagnosis, and performs all-round and accurate diagnosis of the transformer from multiple dimensions.

[0005] In one embodiment of the present invention, the method comprises the following steps:

[0006] S1. Feature extraction: Obtain information data set of transformer and perform feature extraction;

[0007] S2, semantic embedding: semantically embed the data features after feature extraction and map them to the attribute feature space to achieve fusion between various modalities;

[0008] S3, attribute learning: classify the data features based on semantics and determine whether each fault type has the attribute characteristics of the corresponding sample data;

[0009] S4. Fault classification: Construct a transformer fault attribute matrix to realize transformer fault identification.

[0010] In one embodiment of the present invention, in step S1, the transformer information data set includes a one-dimensional data set obtained by a one-dimensional sensor, and a video data set and an infrared thermal imaging data set obtained by a multi-dimensional sensor.

[0011] In one embodiment of the present invention, the infrared thermal imaging data set obtained by the multi-dimensional sensor needs to be normalized and then feature extracted using the HSV color segmentation method. The specific formula is as follows:

[0012] R′=R / 255

[0013] G′=G / 255

[0014] B′=B / 255

[0015] V=max(R′,G′,B′)

[0016]

[0017] Among them, R represents the brightness of the red channel of the image, R′ represents the brightness of the red channel of the image after normalization, G represents the brightness of the green channel of the image, G′ represents the brightness of the green channel of the image after normalization, B represents the brightness of the blue channel of the image, B′ represents the brightness of the blue channel of the image after normalization, H represents the color of the image, S represents the saturation of the image, and V represents the brightness of the image.

[0018] In one embodiment of the present invention, the video data set obtained by the multi-dimensional sensor uses the gray level co-occurrence matrix GLCM to extract features, as follows:

[0019]

[0020] In the formula, ASM represents the energy of the image, reflecting the uniformity of its grayscale distribution and the degree of texture coarseness; P(i,j) represents the probability of grayscale j appearing with grayscale starting from point i;

[0021]

[0022] Where Corr is homogeneity, which means the grayscale difference between pixel pairs at a predetermined distance and angle; μ x is the mean gray value in the x direction, μ y is the mean gray value in the y direction, σ x is the mean square error of grayscale values ​​in the x direction, σ y is the mean square error of the grayscale values ​​in the y direction; when the homogeneity value is higher, it means that the grayscale values ​​of the pixel pairs in the corresponding direction are closer and the texture of the image is more uniform.

[0023] In one embodiment of the present invention, in step S2, in the process of semantic embedding of the data features after feature extraction, the embedded semantics include transformer fault features, transformer fault description, transformer fault location, and transformer fault status. Semantic embedding is to convert data or text into representations in a high-dimensional vector space, and these vectors can capture the semantic information of the data or text. In sensor data fusion, semantic embedding can be used to convert data from different sources and in different formats into a unified representation, thereby facilitating subsequent processing and analysis. Multidimensional sensor data fusion and semantic embedding include data synchronization and calibration, feature fusion, and semantic embedding. In terms of data synchronization and calibration, since multidimensional sensors may have different sampling frequencies and accuracies, it is necessary to synchronize and calibrate the data. This is achieved by methods such as timestamp matching and interpolation. In terms of feature fusion, features from different sensors are fused to form a more comprehensive data representation. Fusion methods include random methods such as weighted average, Kalman filtering, multi-Bayesian estimation, and artificial intelligence methods such as fuzzy logic reasoning and artificial neural networks. In terms of semantic embedding, on the basis of feature fusion, an embedding algorithm is used to convert the fused features into representations in a high-dimensional vector space. These vectors should be able to capture the overall semantic information of multi-dimensional sensor data, including the correlation and complementarity between different sensor data.

[0024] In one embodiment of the present invention, in step S3, before the data features are classified for attribute learning based on semantics, an attribute learner is first constructed. The training method of the attribute learner adopts the classic ID3 algorithm of the decision tree, that is, a top-down decision tree is constructed, and the maximum information entropy and conditional entropy contained in the data are calculated in each layer of nodes. The information gain in the data is calculated according to the difference between the two entropies. When the information gain drops to 0, it means that the training of the high entropy data is completed, and the corresponding decision tree is successfully constructed, that is, the attribute learning is completed. The specific formula is as follows:

[0025]

[0026] Where I is the information entropy of the transformer node, P j is the proportion of samples in the jth category; i is the starting number of sample counting;

[0027]

[0028] In the formula, U is the number of samples, n is the number of branches of the decision tree, and V m is the number of the mth sample branch, I m is the information entropy of the mth sample branch;

[0029] G=IN

[0030] The information gain G of the corresponding decision tree is calculated by the above formula. When G drops to 0, the attribute learner training is completed.

[0031] In one embodiment of the present invention, in step S4, the transformer fault attribute matrix is ​​constructed in a fine-grained manner by combining the transformer fault description table and the transformer common fault type table. The row vector in the transformer fault attribute matrix represents the fault type, and the column vector in the transformer fault attribute matrix represents the fault attribute feature, which is obtained by semantic embedding of the data features after feature extraction. The attribute value of the transformer fault attribute matrix consists of only 0 and 1, which represents the correlation between rows and columns.

[0032] The present invention also provides a multi-modal zero-sample transformer fault identification system, comprising a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0033] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0034] Compared with the prior art, the present invention has the following beneficial effects: Based on cross-modal attention technology and multimodal feature fusion technology, the present invention integrates the concept of multimodal data fusion processing and zero-sample detection into transformer fault diagnosis, and comprehensively utilizes a series of information such as pictures, sounds, vibrations, oil chromatography, humidity, etc. to conduct all-round supervision of the transformer from multiple dimensions, making the diagnosis result more accurate. The present invention can alleviate the problem of insufficient accuracy of the single modality sample of the fault on the diagnostic model through multimodal data fusion, and improve the versatility and diagnostic accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Schematic diagram of transformer semantic embedding.

[0036] Figure 2 is the transformer fault attribute matrix. DETAILED DESCRIPTION

[0037] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0038] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0039] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0040] The present invention provides a zero-sample transformer fault identification method based on multimodality, which integrates the concepts of multimodal data fusion processing and zero-sample detection into transformer fault diagnosis, and performs all-round and accurate diagnosis of the transformer from multiple dimensions.

[0041] The following is the specific implementation process of the present invention.

[0042] The present invention provides a multimodal zero-sample transformer fault identification method, which integrates the concept of multimodal data fusion processing and zero-sample detection into transformer fault diagnosis, and comprehensively utilizes a series of information such as pictures, sounds, vibrations, oil chromatography, humidity, etc. to conduct all-round supervision of the transformer from multiple dimensions, so that the diagnosis result is more accurate. The fault attribute matrix is ​​constructed by combining the supervised attributes with the transformer knowledge base, and the diagnosis of a certain specific difficult-to-collect fault (invisible fault) is realized by means of zero-sample transfer learning, which can effectively improve the flexibility of the system; the specific steps are as follows:

[0043] Step 1: Feature extraction

[0044] The information data set of the transformer is obtained from various sensors, and the data and image features are extracted using machine learning methods to obtain the damage and deterioration of the transformer shape, bushing, and oil quality.

[0045] In the feature extraction module, we first preprocess the data. The one-dimensional sensors used in this model include temperature, vibration, current, partial discharge, audio, and leakage sensors; multidimensional sensors include video, infrared thermal imaging, and photoacoustic spectroscopy sensors. For the infrared thermal imaging data set in the multidimensional sensor, this chapter uses the HSV color segmentation method for feature extraction. This method can extract color information from infrared thermal imaging and oil color video images to judge the health status of the transformer by the transformer box temperature, local temperature, oil color, etc. In the HSV model, color is composed of hue (H), saturation (S), and brightness (V). It is necessary to convert the RGB model of the image into the HSV model. The specific conversion formula is shown below.

[0046] R′=R / 255

[0047] G′=G / 255

[0048] B′=B / 255

[0049] V=max(R′,G′,B′)

[0050]

[0051] R represents the brightness of the red channel of the image, R′ represents the brightness of the red channel of the image after normalization, G represents the brightness of the green channel of the image, G′ represents the brightness of the green channel of the image after normalization, B represents the brightness of the blue channel of the image, and B′ represents the brightness of the blue channel of the image after normalization. H represents the color of the image, S represents the saturation of the image, and V represents the brightness of the image.

[0052] The H, S, and V characteristics extracted according to the above formula serve as training samples for the attribute learner. The classification results of the attribute learner can be divided into three categories: normal temperature, low temperature overheating, and medium and high temperature overheating.

[0053] For the video appearance dataset, this chapter uses the gray level co-occurrence matrix (GLCM) to extract the texture and shape features of the image and detect whether the transformer bushing is broken. The gray level co-occurrence matrix algorithm is a commonly used algorithm in the field of image processing and analysis. The algorithm describes the texture characteristics of the image based on the adjacency relationship between gray levels. Specifically, the algorithm divides the gray value of the image into several levels, and then calculates the number of occurrences between each gray level to obtain a gray level co-occurrence matrix. Each element in the matrix represents the number of occurrences between two gray levels. This algorithm is easy to implement for transformer fault diagnosis and has strong adaptability and robustness. The specific principle formula is as follows.

[0054]

[0055] In the formula, ASM represents the energy of the image, reflecting the uniformity of its grayscale distribution and the degree of texture coarseness. P(i,j) represents the probability of grayscale j (column) appearing with grayscale starting from i (row).

[0056]

[0057] In the formula, Corr is homogeneity, which indicates the grayscale difference between pixel pairs at a specific distance and angle. x is the mean gray value in the x direction, μ y is the mean gray value in the y direction, σ x is the mean square error of grayscale values ​​in the x direction, σ y is the mean square error of the grayscale values ​​in the y direction; when the homogeneity value is higher, it means that the grayscale values ​​of the pixel pairs in the corresponding direction are closer and the texture of the image is more uniform.

[0058] This chapter selects the most representative historical image data as a data set and performs GLCM calculations on it to obtain two transformer video feature values, Asm and Corr, which serve as training samples for the attribute learner. In image training, i is 1394 and j is 1955.

[0059] Step 2: Semantic Embedding

[0060] The multimodal fusion method proposed in the present invention is to embed the extracted data features semantically and map them to the attribute feature space to achieve fusion between the modalities. The embedded semantics is not limited to the transformer fault features but can also be fault description, fault location or fault status. Figure 1 As shown in Figure 1. Semantic embedding is the process of converting data or text into representations in a high-dimensional vector space. These vectors can capture the semantic information of the data or text. In sensor data fusion, semantic embedding can be used to convert data from different sources and formats into a unified representation, which is convenient for subsequent processing and analysis. Multidimensional sensor data fusion and semantic embedding include data synchronization and calibration, feature fusion, and semantic embedding. In terms of data synchronization and calibration, since multidimensional sensors may have different sampling frequencies and accuracies, data synchronization and calibration are required. This is achieved through methods such as timestamp matching and interpolation. In terms of feature fusion, features from different sensors are fused to form a more comprehensive data representation. Fusion methods include random methods such as weighted averaging, Kalman filtering, and multi-Bayesian estimation, as well as artificial intelligence methods such as fuzzy logic reasoning and artificial neural networks. In terms of semantic embedding, based on feature fusion, an embedding algorithm is used to convert the fused features into representations in a high-dimensional vector space. These vectors should be able to capture the overall semantic information of multidimensional sensor data, including the correlation and complementarity between different sensor data.

[0061] Referring to the daily transformer operation status and its fault characteristics, this application generates 16 types of fault descriptions after semantic embedding of the data collected by the sensor, as shown in Table 1.

[0062] Table 1 Transformer fault description

[0063]

[0064] Step 3: Attribute Learning

[0065] In the attribute learning module, the present application classifies the data for attribute learning based on semantics, and determines whether each fault type has the attribute characteristics of the sample data (the relevant attribute is 1, and the irrelevant attribute is 0). The training method of the attribute learner adopts the classic ID3 algorithm of the decision tree. The algorithm constructs a top-down decision tree, calculates the maximum information entropy and conditional entropy contained in the data in each layer of nodes, and calculates the information gain in the data based on the difference between the two entropies. When the information gain drops to 0, it means that the training of the high entropy data is completed, and the tree is successfully constructed (that is, the attribute learning is completed). This method is more efficient when dealing with samples such as transformers with a small number of samples, small noise impact, or a small amount of segmentation between attributes and attribute values. The specific formula is as follows.

[0066]

[0067] Where I is the information entropy of the transformer node, P j is the proportion of samples in the jth category.

[0068]

[0069] In the formula, U is the number of samples, n is the number of branches of the decision tree, and V m is the number of the mth sample branch, I m is the information entropy of the mth sample branch.

[0070] G=IN

[0071] The information gain G of the tree can be calculated by the above formula. When G drops to 0, the algorithm can complete the training classification.

[0072] Step 4: Fault classification

[0073] This application classifies 12 common fault types, as shown in Table 2.

[0074] Table 2 Common fault types of transformers

[0075]

[0076] Combined with the transformer fault description table and the transformer common fault type table, the transformer fault attribute matrix is ​​constructed in a fine-grained manner. The fault attribute matrix is ​​as follows: Figure 2 As shown in the figure, the row vector of the matrix represents the fault type, which is composed of common fault types. The column vector represents the fault attribute characteristics, which are obtained by semantic embedding of sensor data. The attribute values ​​of the matrix are composed of only 0 and 1, which represent the correlation between rows and columns.

[0077] In the attribute transfer phase, the constructed fault attribute matrix is ​​used as an auxiliary knowledge source to achieve generalization from visible class faults to invisible class faults by transfer learning. The method only needs to embed the matrix into the training set when training samples, and the system can make accurate diagnosis when encountering zero-sample faults during testing. The fault attribute matrix is ​​a special matrix that contains attribute information of various fault types. These attributes can be the characteristics, manifestations, and impact range of the fault. The purpose of constructing such a matrix is ​​to provide a comprehensive and structured fault knowledge base as an auxiliary knowledge source for subsequent transfer learning. Transfer learning is a machine learning method that can transfer knowledge learned on one task to another related but different task. In this scenario, the purpose of transfer learning is to use the existing fault attribute knowledge (i.e., the information in the fault attribute matrix) to help the system identify or diagnose new and unseen fault types. "Visible class faults" refer to the fault types that have been encountered and learned during the training process, while "invisible class faults" refer to those fault types that have not been encountered during the training process but may occur in actual applications. Through transfer learning, the system can apply the knowledge learned about visible class faults to the diagnosis of invisible class faults, thereby achieving generalization capabilities. When training a machine learning model, the constructed fault attribute matrix is ​​embedded into the training set. This means that during the training process, the model not only learns how to extract features from the input data to identify faults, but also learns how to use the knowledge in the fault attribute matrix to enhance its understanding of the fault type.

[0078] Diagnosis of zero-sample faults: When the system encounters a fault type not included in the training set (i.e., zero-sample fault) in actual applications, the system can still use this knowledge to make an accurate diagnosis because the knowledge of the fault attribute matrix has been incorporated into the model through transfer learning.

[0079] The present invention also provides a multi-modal zero-sample transformer fault identification system, comprising a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0080] The present invention also provides a computer-readable storage medium, on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0081] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0082] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0085] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A zero-sample transformer fault identification method based on multi-modality, characterized in that: The concept of multimodal data fusion processing and zero-sample detection is integrated into transformer fault diagnosis to conduct all-round and accurate diagnosis of the transformer from multiple dimensions.

2. A multi-modal zero-sample transformer fault identification method according to claim 1, characterized in that: The steps include: S1. Feature extraction: Obtain information data set of transformer and perform feature extraction; S2, semantic embedding: semantically embed the data features after feature extraction and map them to the attribute feature space to achieve fusion among various modalities; S3, attribute learning: classify the data features based on semantics and determine whether each fault type has the attribute characteristics of the corresponding sample data; S4. Fault classification: Construct a transformer fault attribute matrix to realize transformer fault identification.

3. The multi-modal zero-sample transformer fault identification method according to claim 2 is characterized in that: In step S1, the transformer information data set includes a one-dimensional data set obtained by a one-dimensional sensor, and a video data set and an infrared thermal imaging data set obtained by a multi-dimensional sensor.

4. The multi-modal zero-sample transformer fault identification method according to claim 3 is characterized in that: The infrared thermal imaging data set obtained by the multi-dimensional sensor needs to be normalized and the HSV color segmentation method is used for feature extraction. The specific formula is as follows: R′=R / 255 G′=G / 255 B′=B / 255 V=max(R′,G′,B′) Among them, R represents the brightness of the red channel of the image, R' represents the brightness of the red channel of the image after normalization, G represents the brightness of the green channel of the image, G' represents the brightness of the green channel of the image after normalization, B represents the brightness of the blue channel of the image, B' represents the brightness of the blue channel of the image after normalization, H represents the color of the image, S represents the saturation of the image, and V represents the brightness of the image.

5. The multi-modal zero-sample transformer fault identification method according to claim 3 is characterized in that: The video dataset obtained by the multi-dimensional sensor uses the gray level co-occurrence matrix GLCM for feature extraction, as follows: In the formula, ASM represents the energy of the image, reflecting the uniformity of its grayscale distribution and the degree of texture coarseness; P(i,j) represents the probability of grayscale j appearing with grayscale starting from point i; In the formula, Corr is homogeneity, which means the grayscale difference between pixel pairs at a predetermined distance and angle, μ x is the mean gray value in the x direction, μ y is the mean gray value in the y direction, σ x is the mean square error of grayscale values ​​in the x direction, σ y is the mean square error of the grayscale values ​​in the y direction; when the homogeneity value is higher, it means that the grayscale values ​​of the pixel pairs in the corresponding direction are closer and the texture of the image is more uniform.

6. The multi-modal zero-sample transformer fault identification method according to claim 2, characterized in that: In step S2, in the process of semantic embedding of the data features after feature extraction, the embedded semantics include transformer fault features, transformer fault description, transformer fault location, and transformer fault status.

7. The multi-modal zero-sample transformer fault identification method according to claim 2, characterized in that: In step S3, before the attribute learning of data features is classified based on semantics, an attribute learner is first constructed. The training method of the attribute learner adopts the classic ID3 algorithm of decision tree, that is, a top-down decision tree is constructed, and the maximum information entropy and conditional entropy contained in the data are calculated in each layer of nodes. The information gain in the data is calculated according to the difference between the two entropies. When the information gain drops to 0, it means that the training of high entropy data is completed, and the corresponding decision tree is successfully constructed, that is, the attribute learning is completed. The specific formula is as follows: Where I is the information entropy of the transformer node, P j is the proportion of samples of the jth category, and i is the starting number of sample counting; In the formula, U is the number of samples, n is the number of branches of the decision tree, and V m is the number of the mth sample branch, I m is the information entropy of the mth sample branch; G=IN The information gain G of the corresponding decision tree is calculated by the above formula. When G drops to 0, the attribute learner training is completed.

8. The multi-modal zero-sample transformer fault identification method according to claim 2, characterized in that: In step S4, the transformer fault attribute matrix is ​​constructed in a fine-grained manner by combining the transformer fault description table and the transformer common fault type table. The row vector in the transformer fault attribute matrix represents the fault type, and the column vector in the transformer fault attribute matrix represents the fault attribute feature, which is obtained by semantic embedding of the data features after feature extraction. The attribute value of the transformer fault attribute matrix consists of only 0 and 1, which represents the correlation between rows and columns.

9. A zero-sample transformer fault identification system based on multi-modality, characterized in that: The method comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps as claimed in any one of claims 1 to 8 can be implemented.

10. A computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, and when the processor executes the computer program instructions, the method steps according to any one of claims 1 to 8 can be implemented.

Images