A method and system for inspecting power transmission lines

Abstract

This invention proposes a method and system for power transmission line inspection, relating to the field of power transmission line inspection technology. The specific solution includes: real-time acquisition of raw point cloud data collected by UAV lidar; generating clear point clouds from raw point cloud data under different weather conditions using a trained preprocessing model; segmenting and detecting the clear point clouds, obtaining point cloud background types and corresponding background reference images based on the segmented objects; inputting the clear point clouds, point cloud background types, and corresponding background reference images into a trained point cloud generation model to obtain complete point clouds; and performing fault identification based on the complete point clouds. This invention removes the influence of weather factors on point cloud data through adversarial learning, acquires point cloud background types in real time, and introduces background reference images to assist in completing point cloud data, thereby improving the quality and efficiency of completing residual point cloud data for power transmission lines.

Description

Technical Field

[0001] This invention belongs to the field of power transmission line inspection technology, and particularly relates to a method and system for power transmission line inspection. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Regularly inspecting transmission lines to keep abreast of their operational status and changes in the surrounding environment and protection zones is a demanding daily task for power supply companies. Manual inspection is a traditional method, but due to the complex terrain of transmission line corridors and the lack of patrol roads in some harsh conditions, such as crossing rivers or high mountains, this method is labor-intensive, has difficult working conditions, and cannot provide timely information on the operation of the transmission lines.

[0004] Drone inspection is a brand-new inspection technology for inspecting power transmission lines. It has the advantages of being fast and efficient, not affected by geographical location, high inspection quality, and high safety. The application of drones is an effective solution for the intelligent development of line inspection, especially suitable for the power industry. Power inspection has become one of the most promising civilian drone fields in my country.

[0005] The most critical step in drone inspection is identifying power transmission lines. Only by accurately identifying the power transmission lines can further fault detection or intrusion risk identification be carried out. Automatic line identification from lidar point clouds and images during line inspection has long been a focus of academic and industrial attention because the cross-section of power lines is thin, resulting in a small imaging area on the inspection image. In particular, point cloud data acquired by sensors, radar, and other equipment is often incomplete due to occlusion, reflection, noise, and other factors, affecting the accuracy and efficiency of line identification and potentially causing serious consequences for line inspection tasks.

[0006] Existing point cloud completion methods, when applied to the field of power line inspection, do not consider the impact of different natural environments and weather conditions on the completion effect of the inspected lines. For example, point cloud data with different backgrounds such as grassland, sky, and mountains, and adverse weather conditions can cause specific types of noise to the data of sensors such as light detection and ranging, thus affecting the quality of point clouds and the completion effect. Therefore, existing power line inspection methods based on point cloud data are affected by a variety of factors in identifying lines, resulting in low accuracy and robustness. Summary of the Invention

[0007] To overcome the shortcomings of the prior art, the present invention provides a method and system for inspecting power transmission lines. By using adversarial learning to remove the influence of weather factors on point cloud data, the system can obtain the point cloud background type in real time, introduce a background reference image, and assist in the completion of point cloud data, thereby improving the quality and efficiency of point cloud completion for power transmission lines.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0009] The first aspect of this invention provides a method for inspecting power transmission lines.

[0010] A method for inspecting power transmission lines, comprising:

[0011] Real-time acquisition of raw point cloud data collected by the drone's lidar;

[0012] Using a pre-trained preprocessing model, clear point clouds are generated from raw point cloud data under different weather conditions;

[0013] Segmentation and detection are performed on clear point clouds, and the point cloud background type and corresponding background reference image are obtained based on the segmented objects;

[0014] The clear point cloud, the point cloud background type, and the corresponding background reference image are input into the trained point cloud generation model to obtain a complete point cloud.

[0015] Fault identification is performed based on complete point clouds;

[0016] The preprocessing model performs adversarial learning on the relationship between blurry and sharp point clouds to obtain a generator that generates sharp point clouds; the point cloud generation model performs adversarial learning on the differences between different background types to generate point cloud images corresponding to the background types.

[0017] Furthermore, the preprocessing model includes a generator and a discriminator. The generator takes a blurred point cloud as input and outputs a clear point cloud. The discriminator is used to determine whether the input point cloud is a generated point cloud or a real image.

[0018] Furthermore, the generator consists of three parts: an encoder, a converter, and a decoder.

[0019] The encoder consists of convolutional layers, which are used to extract shallow features from the input point cloud to obtain feature vectors in the source domain.

[0020] The converter consists of a residual network and is used to transform feature vectors in the source domain to feature vectors in the target domain.

[0021] The decoder consists of deconvolutional layers used to reconstruct a sharp point cloud from feature vectors in the target domain.

[0022] Furthermore, the discriminator is used to input the sharp point cloud generated by the generator and the real sharp point cloud into the discriminator to perform adversarial training on the preprocessed model until the discriminator can no longer distinguish between the generated sharp point cloud and the real sharp point cloud.

[0023] Furthermore, the segmentation and detection of the clear point cloud is performed by using a pre-trained segmentation and detection model to segment the objects in the point cloud, and then classifying and detecting the segmented objects to obtain the point cloud background type.

[0024] The point cloud background types include grassland, sky, and mountains.

[0025] Furthermore, the background reference image is obtained by searching for the corresponding background reference image from a pre-built background reference image library based on the point cloud background type obtained from classification and detection.

[0026] The background reference image library uses point cloud background type as a tag to search for background reference images.

[0027] Furthermore, the point cloud generation model includes a generator and a discriminator. The generator takes the point cloud to be completed and the background reference image as input to generate a complete point cloud. The discriminator is responsible for determining whether the generated complete point cloud belongs to the current point cloud background type.

[0028] A second aspect of the present invention provides a power transmission line inspection system.

[0029] A power transmission line inspection system includes a data acquisition module, a preprocessing module, a background processing module, a point cloud generation module, and a fault identification module.

[0030] The data acquisition module is configured to acquire raw point cloud data collected by the UAV's lidar in real time.

[0031] The preprocessing module is configured to generate clear point clouds from raw point cloud data under different weather conditions using a trained preprocessing model.

[0032] The background processing module is configured to: segment and detect clear point clouds, and obtain the point cloud background type and the corresponding background reference image based on the segmented objects;

[0033] The point cloud generation module is configured to input a clear point cloud, a point cloud background type, and a corresponding background reference image into a trained point cloud generation model to obtain a complete point cloud.

[0034] The fault identification module is configured to identify faults based on the complete point cloud.

[0035] The preprocessing model performs adversarial learning on the relationship between blurry and sharp point clouds to obtain a generator that generates sharp point clouds; the point cloud generation model performs adversarial learning on the differences between different background types to generate point cloud images corresponding to the background types.

[0036] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a transmission line inspection method as described in the first aspect of the present invention.

[0037] A fourth aspect of the present invention provides an electronic device including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of a transmission line inspection method as described in the first aspect of the present invention.

[0038] The above one or more technical solutions have the following beneficial effects:

[0039] This invention uses an adversarial learning-based preprocessing model to remove the influence of weather factors on point cloud data, resulting in clear point clouds. Then, through detection and segmentation, the background type of the point cloud is obtained in real time, and a background reference image is introduced to assist in the completion of point cloud data, thereby improving the completion quality and efficiency of point cloud for power transmission line defects.

[0040] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0041] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0042] Figure 1 This is a flowchart of the method in the first embodiment.

[0043] Figure 2 This is a structural diagram of the preprocessing model for the first embodiment.

[0044] Figure 3 This is a system structure diagram of the second embodiment. Detailed Implementation

[0045] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0047] UAV measurement technology is now widely used in industries such as surveying, power, water conservancy, mining, and emergency response. UAV lidar is a relatively new measurement technology that combines high-precision laser scanning instruments, GPS, and inertial navigation systems (INS) to achieve incredibly accurate 3D mapping.

[0048] Airborne lidar, as an application system combining lidar and drones, can more directly and effectively measure the three-dimensional real world, and has the advantages of high data accuracy, rich layer details, and all-weather operation.

[0049] In the field of power transmission line inspection, image data of line towers, conductors, line corridors, and surrounding environment can be transformed into 3D point cloud data through spatial 3D calculation. This allows for a more intuitive observation of the spatial location and outline of targets within the line corridor, and determination of the distance between conductors and ground, buildings, vegetation, and other targets. High-precision data information can be obtained from UAV laser point cloud models, and combined with image files, 3D dynamic simulation and analysis can be performed to achieve full coverage of the power transmission line inspection range, including attribute status, location structure, etc., making the inspection results digital, traceable, and analyzable.

[0050] While point cloud data based on UAV lidar can achieve powerful simulation and analysis functions, this is contingent on the integrity and accuracy of the point cloud data. This embodiment provides a method for inspecting power transmission lines by using adversarial learning to remove the influence of weather factors on point cloud data, acquiring the point cloud background type in real time, introducing background reference images, and assisting in the completion of point cloud data, thereby improving the integrity and accuracy of power transmission line point cloud data.

[0051] Example 1

[0052] In one or more embodiments, a method for inspecting power transmission lines is disclosed, such as... Figure 1 As shown, it includes the following steps:

[0053] Step S1: Acquire raw point cloud data collected by the UAV's lidar in real time.

[0054] Step S2: Using the trained preprocessing model, generate clear point clouds from the raw point cloud data under different weather conditions.

[0055] The preprocessing model is used to perform adversarial learning on the relationship between blurry and sharp point clouds, resulting in a generator that produces sharp point clouds. Figure 2 This is a structure diagram of the preprocessing model, such as... Figure 2 As shown, the preprocessing model includes a generator and a discriminator. The generator takes a blurred point cloud as input and outputs a clear point cloud. The discriminator is used to determine whether the input point cloud is a generated point cloud or a real image.

[0056] generator

[0057] To remove the impact of adverse weather factors on point cloud data, including fog, haze, rain, snow, sandstorms, etc., the generator uses a feature extraction, feature transformation, and feature reconstruction approach to generate a clear point cloud that is highly similar to the original input point cloud.

[0058] Specifically, the generator consists of three parts: an encoder, a converter, and a decoder. The encoder is composed of convolutional layers and is used to extract shallow features of the input point cloud to obtain feature vectors in the source domain. The converter is composed of residual networks and is used to transform the feature vectors in the source domain to feature vectors in the target domain. The decoder is composed of deconvolutional layers and is used to reconstruct a sharp point cloud from the feature vectors in the target domain.

[0059] The encoder is responsible for inputting a blurred point cloud affected by weather factors, extracting shallow features, and consists of four downsampling modules connected in sequence. The convolutional kernel sizes of the downsampling modules are 7×7, 4×4, 4×4 and 4×4 respectively. InstanceNormalization is used for normalization and ReLU is used for activation.

[0060] The transformer is responsible for inputting shallow features and transforming them to obtain target features, i.e., feature vectors in the target domain. It consists of four residual connection modules connected in sequence. Each residual connection module consists of two identical convolutional modules. The specific configuration of the convolutional modules is that the kernel size is 3×3, the normalization method is InstanceNormalization, and the activation function is ReLU.

[0061] The decoder is responsible for inputting target features and reconstructing a clear point cloud. The structure of the decoder is the opposite of that of the encoder. It consists of four upsampling modules connected in sequence. The convolution kernel sizes of the upsampling modules are 5×5, 5×5, 5×5 and 5×5, respectively. The normalization method is AdaIN and the activation function is ReLU.

[0062] Discriminator

[0063] The discriminator is used to input both the sharp point cloud generated by the generator and the real sharp point cloud into the discriminator to perform adversarial training on the preprocessed model until the discriminator can no longer distinguish between the generated sharp point cloud and the real sharp point cloud; the discriminator consists of an encoder and a perceptron MLP.

[0064] The generated sharp point cloud is input into an encoder composed of point cloud convolution PointConv to obtain point cloud features. These features are then input into a perceptron layer to output the probability of whether the point cloud is generated or real.

[0065] The goal of adversarial training is to make the preprocessed model output results that are as confusing as possible to the discriminator; therefore, the following loss function is used for optimization:

[0066]

[0067] Where Z represents the original point cloud input to the preprocessing model, τ represents the set of fake samples, ω represents the set of real samples, D represents the discriminator, T represents the preprocessing model, and P represents the true point cloud. It is a penalty term in the WGAN paradigm, used to stabilize the training process.

[0068] After the adversarial training is completed, a generator capable of generating clear point clouds is obtained. This generator is used to preprocess the original point clouds, remove the influence of adverse weather factors, and obtain clear point clouds.

[0069] Step S3: Segment and detect the clear point cloud, and obtain the point cloud background type and the corresponding background reference image based on the segmented objects.

[0070] Different natural environments serve as backgrounds for point clouds, interfering with the point cloud completion effect. Therefore, to reduce the influence of the background, it is necessary to first identify the type of point cloud background and then use the corresponding background reference image to constrain the point cloud completion method. Thus, the background must be processed first.

[0071] Specifically, a pre-trained segmentation and detection model is used to segment objects in the point cloud, and then the segmented objects are classified and detected to obtain the point cloud background type.

[0072] Point cloud background types include grassland, sky, mountains, forests, etc. Different background types have different point cloud shape features. Different shape features are used to classify and detect objects. The segmentation and detection model adopts the YOLOv5 structure, which will not be elaborated on further.

[0073] After obtaining the point cloud background type, the corresponding background reference image is found from the pre-built background reference image library. The background reference image library includes background reference images corresponding to the regular point cloud background types. The background reference image is searched by using the point cloud background type as a label. For example, the background reference images corresponding to the mountain and forest types are searched from the background reference image library.

[0074] Step S4: Input the clear point cloud, the point cloud background type, and the corresponding background reference image into the trained point cloud generation model to obtain the complete point cloud.

[0075] Using a trained point cloud generation model, a high-quality complete point cloud is generated from a clear point cloud and its corresponding background reference image using an end-to-end generation method.

[0076] The point cloud generation model employs a generative adversarial network (GAN), comprising a generator and a discriminator. The generator takes the point cloud to be completed and a background reference image as input to generate a complete point cloud. The discriminator is responsible for determining whether the generated complete point cloud belongs to the current point cloud's background type. Its structure is described in detail below:

[0077] The generator consists of a Point Convolutional (PointConv) layer and a Point Pyramid (Transformer) layer. The Point Convolutional (PointConv) layer is used to extract point cloud features F from the input point cloud, and the Point Pyramid (Transformer) layer is used to predict the point cloud for missing parts.

[0078] To fully learn the point cloud features F, the point cloud features F are input into the point cloud pyramid Transformer, which is composed of alternating Transformer encoders and point cloud convolution PointConv. The Transformer encoder captures long-range relationships in the point cloud, while the point cloud convolution PointConv aggregates local point clouds. During the processing of the point cloud pyramid Transformer, high-level semantics are gradually learned, thereby better predicting the point cloud in the missing parts.

[0079] Specifically, the point cloud pyramid Transformer consists of three Transformer encoders, two point cloud convolutional layers (PointConv), and a perceptron. The extracted point cloud features F are input into the Transformer encoders. A PointConv is connected between every two Transformer encoders to adjust the number of points in the encoded point cloud. The last Transformer encoder is connected to the perceptron to predict and reconstruct missing parts. The reconstructed parts are combined with the missing point cloud data to be filled in to obtain the completed point cloud data.

[0080] The extracted point cloud features F can be regarded as a feature sequence, which conforms to the data input format of Transformer. Since Transformer can learn medium- and long-range relationships, each point cloud feature of the fused features is used as a token and input into the Transformer encoder network. It is worth noting that the Transformer here does not need to add classification tokens and position encoding. The features output by the Transformer encoder are then processed by a point cloud convolution PointConv to aggregate local features and learn higher-level semantics. After processing by three Transformer encoders and two point cloud convolution PointConv, the final output features are obtained. Then, the features output by Transformer are input into a 3-layer perceptron to obtain the final predicted point cloud data of the missing parts, i.e., the complete point cloud.

[0081] The discriminator is responsible for determining whether the generated complete point cloud belongs to the current point cloud background type. The discriminator takes the point cloud as input and outputs the probability that the point cloud is distributed in each point cloud background type. The discriminator consists of a convolutional layer, a downsampling module, two convolutional layers, and a classification layer connected in sequence. The downsampling module consists of an FRN layer and a convolutional layer. FRN (Filter Response Normalization) is a new filter response normalization proposed by Google, used to normalize the trained model for each channel to achieve high accuracy. The classification layer expands the output of the upper layer into K vectors, where K is the number of point cloud background types, and outputs the probability value of belonging to the correct type.

[0082] Adversarial training of a point cloud generation model involves simultaneously training a generator and a discriminator. The generator is responsible for generating a complete point cloud from the input point cloud and a background reference image, while the discriminator is responsible for determining whether the complete point cloud belongs to the current point cloud's background type. The generative adversarial loss L... adv for:

[0083]

[0084] Where is the discriminator, y is the background reference image, and y ′ For the generated complete point cloud, C(*) represents the random cropping operation, Y represents the distribution of the real image, and c represents the point cloud background type.

[0085] Step S5: Based on the complete point cloud, perform fault identification.

[0086] Based on the completed point cloud, line faults and risks are identified, such as identifying external sources of damage, faults, or intrusions.

[0087] Example 2

[0088] In one or more embodiments, a power transmission line inspection system is disclosed, such as Figure 3 As shown, it includes a data acquisition module, a preprocessing module, a background processing module, a point cloud generation module, and a fault identification module:

[0089] The data acquisition module is configured to acquire raw point cloud data collected by the UAV's lidar in real time.

[0090] The preprocessing module is configured to generate clear point clouds from raw point cloud data under different weather conditions using a trained preprocessing model.

[0091] The background processing module is configured to: segment and detect clear point clouds, and obtain the point cloud background type and the corresponding background reference image based on the segmented objects;

[0092] The point cloud generation module is configured to input a clear point cloud, a point cloud background type, and a corresponding background reference image into a trained point cloud generation model to obtain a complete point cloud.

[0093] The fault identification module is configured to identify faults based on the complete point cloud.

[0094] The preprocessing model performs adversarial learning on the relationship between blurry and sharp point clouds to obtain a generator that generates sharp point clouds; the point cloud generation model performs adversarial learning on the differences between different background types to generate point cloud images corresponding to the background types.

[0095] Example 3

[0096] The purpose of this embodiment is to provide a computer-readable storage medium.

[0097] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a transmission line inspection method as described in Embodiment 1 of this disclosure.

[0098] Example 4

[0099] The purpose of this embodiment is to provide an electronic device.

[0100] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in a transmission line inspection method as described in Embodiment 1 of this disclosure.

[0101] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for inspecting power transmission lines, characterized in that, include: Real-time acquisition of raw point cloud data collected by the drone's lidar; Using a pre-trained preprocessing model, clear point clouds are generated from raw point cloud data under different weather conditions; Segmentation and detection are performed on clear point clouds, and the point cloud background type and corresponding background reference image are obtained based on the segmented objects; The clear point cloud, the point cloud background type, and the corresponding background reference image are input into the trained point cloud generation model to obtain a complete point cloud. Fault identification is performed based on complete point clouds; The preprocessing model uses adversarial learning to analyze the relationship between blurry and sharp point clouds, resulting in a generator that produces sharp point clouds. The point cloud generation model uses adversarial learning to analyze the differences between different background types, generating point cloud images corresponding to those background types. The preprocessing model includes a generator and a discriminator. The generator takes blurry point clouds as input and outputs sharp point clouds, while the discriminator determines whether the input point cloud is a generated point cloud or a real image. To remove the influence of adverse weather conditions on the point cloud data, the generator uses a feature extraction-feature transformation-feature reconstruction approach to generate sharp point clouds that are highly similar to the original input point cloud. The segmentation and detection of the clear point cloud involves using a pre-trained segmentation and detection model to segment objects in the point cloud, and then classifying and detecting the segmented objects to obtain the point cloud background type. The point cloud background type includes grassland, sky, and mountains. The background reference image is obtained by searching for a corresponding background reference image from a pre-built background reference image library based on the point cloud background type obtained from the classification and detection. The background reference image library uses the point cloud background type as a label for searching for background reference images.

2. The method for inspecting transmission lines as described in claim 1, characterized in that, The generator consists of three parts: an encoder, a converter, and a decoder. The encoder consists of convolutional layers, which are used to extract shallow features from the input point cloud to obtain feature vectors in the source domain. The converter consists of a residual network and is used to transform feature vectors in the source domain to feature vectors in the target domain. The decoder consists of deconvolutional layers used to reconstruct a sharp point cloud from feature vectors in the target domain.

3. The method for inspecting transmission lines as described in claim 1, characterized in that, The discriminator is used to input the point cloud generated by the generator and the real point cloud into the discriminator to perform adversarial training on the preprocessed model until the discriminator can no longer distinguish between the generated point cloud and the real point cloud.

4. The method for inspecting transmission lines as described in claim 1, characterized in that, The point cloud generation model includes a generator and a discriminator. The generator takes the point cloud to be completed and the background reference image as input to generate a complete point cloud. The discriminator is responsible for determining whether the generated complete point cloud belongs to the current point cloud background type.

5. A power transmission line inspection system, characterized in that, It includes a data acquisition module, a preprocessing module, a background processing module, a point cloud generation module, and a fault identification module. The data acquisition module is configured to acquire raw point cloud data collected by the UAV's lidar in real time. The preprocessing module is configured to generate clear point clouds from raw point cloud data under different weather conditions using a trained preprocessing model. The background processing module is configured to: segment and detect clear point clouds, and obtain the point cloud background type and the corresponding background reference image based on the segmented objects; The point cloud generation module is configured to input a clear point cloud, a point cloud background type, and a corresponding background reference image into a trained point cloud generation model to obtain a complete point cloud. The fault identification module is configured to identify faults based on the complete point cloud. The preprocessing model uses adversarial learning to analyze the relationship between blurry and sharp point clouds, resulting in a generator that produces sharp point clouds. The point cloud generation model uses adversarial learning to analyze the differences between different background types, generating point cloud images corresponding to those background types. The preprocessing model includes a generator and a discriminator. The generator takes blurry point clouds as input and outputs sharp point clouds, while the discriminator determines whether the input point cloud is a generated point cloud or a real image. To remove the influence of adverse weather conditions on the point cloud data, the generator uses a feature extraction-feature transformation-feature reconstruction approach to generate sharp point clouds that are highly similar to the original input point cloud. The segmentation and detection of the clear point cloud involves using a pre-trained segmentation and detection model to segment objects in the point cloud, and then classifying and detecting the segmented objects to obtain the point cloud background type. The point cloud background type includes grassland, sky, and mountains. The background reference image is obtained by searching for a corresponding background reference image from a pre-built background reference image library based on the point cloud background type obtained from the classification and detection. The background reference image library uses the point cloud background type as a label for searching for background reference images.

6. An electronic device, characterized in that it comprises: Memory is used to store computer-readable instructions in a non-transitory manner. as well as Processor, for executing the computer-readable instructions, When the computer-readable instructions are executed by the processor, they perform the method described in any one of claims 1-4.

7. A storage medium, characterized in that, The computer-readable instructions are stored non-transitory, wherein when the non-transitory computer-readable instructions are executed by a computer, the instructions of the method according to any one of claims 1-4 are executed.

Images