GELAN-based pipeline disease detection method, system and device, and medium

Abstract

The invention provides a GELAN-based pipeline disease detection method, system and device, and a medium, and relates to the technical field of pipeline detection. The method comprises the following steps: acquiring a target image containing a to-be-detected pipeline; sending the target image into a deep learning model for pipeline disease detection to obtain a detection result image and disease information; wherein the deep learning model is a network model which performs depth factor and width factor reduction processing on a GELAN network model, takes a deep convolution DWConv as a network backbone and performs up-sampling by a Dysample up-sampling module. According to the method, the accuracy and robustness of pipeline disease detection can be improved, and the number of computing resources and samples is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of pipeline detection, and in particular to a method, system, device and medium for pipeline disease detection based on GELAN. Background Art

[0002] Pipeline defect detection refers to the process of identifying, locating and analyzing possible defects inside or outside the pipeline using advanced technical means. To improve the efficiency of pipeline defect inspection, the current main method is to train deep learning models to automatically identify defects in pipeline images.

[0003] However, existing pipeline defect detection methods based on deep learning still have some problems. For example, the deep learning models used by some methods are too complex and require a lot of computing resources and training data, which makes them difficult to promote in practical applications. In addition, when processing pipeline images, some methods are easily affected by factors such as lighting changes and noise interference, resulting in a decrease in the accuracy of the detection results.

[0004] Therefore, it is urgent to propose a method that can improve the accuracy and robustness of pipeline disease detection and reduce the requirements for computing resources and the number of samples. Summary of the invention

[0005] The purpose of this application is to provide a method, system, device and medium for pipeline defect detection based on GELAN, which can improve the accuracy and robustness of pipeline defect detection and reduce the computing resources and sample quantity.

[0006] This application is implemented as follows:

[0007] In a first aspect, the present application provides a method for pipeline disease detection based on GELAN, comprising the following steps: obtaining a target image containing a pipeline to be detected. Sending the target image to a deep learning model for pipeline disease detection to obtain a detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses a deep convolution DWConv as the network backbone, and uses a Dysample upsampling module for upsampling. In some implementations, the disease information includes at least one of cracks, deformations, sediments, foreign body penetration, misaligned joints, root invasion, or fractures.

[0008] In some implementations, the network framework of the deep learning model includes an input network, a backbone network, a neck network, and a prediction head network. The input network is used to receive a target image; the backbone network is used to extract key features from the received target image through deep convolution DWConv, and introduce linearity through an activation function; the neck network is used to fuse the extracted key features using a feature pyramid network, and use a Dysample upsampling module to combine low-level feature maps with high-level feature maps; the prediction head network is used to receive the output of the neck network to generate an anchor box containing a feature map, and give the category probability corresponding to the anchor box.

[0009] In some implementations, the input network is further used to unify the size of the received target image into a fixed size, and apply random rotation, flipping, scaling, and color transformation for data augmentation processing to enhance sample diversity.

[0010] In some implementations, reducing the depth factor and the width factor of the GELAN network model includes setting the depth factor of the GELAN network model to 0.33 and the width factor to 0.5.

[0011] In some implementations, each convolution kernel of the deep convolution DWConv processes only one input channel, the number of convolution kernels is equal to the number of input channels, and the number of output channels is equal to the number of input channels, so as to independently perform convolution operations on each channel of the input layer.

[0012] In some implementations, the Dysample upsampling module introduces grouped upsampling, and the feature map is divided into multiple groups in the channel dimension, and each group shares the same sampling set; the Dysample upsampling module also introduces a dynamic range factor, including generating a point-by-point dynamic range factor by linearly projecting the input features.

[0013] In a second aspect, the present application provides a system for pipeline disease detection based on GELAN, comprising: an image acquisition module, configured to: acquire a target image containing a pipeline to be detected. A detection and analysis module, configured to: send the target image to a deep learning model for pipeline disease detection to obtain a detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses deep convolution DWConv as the network backbone, and uses the Dysample upsampling module for upsampling.

[0014] In a third aspect, the present application provides an electronic device, comprising a memory for storing one or more programs; a processor; when the one or more programs are executed by the processor, a method as described in any one of the first aspects is implemented.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method as described in any one of the above-mentioned first aspects.

[0016] Compared with the prior art, the present invention has at least the following advantages or beneficial effects:

[0017] This application proposes a pipeline defect detection method based on GELAN, which significantly improves pipeline defect detection technology by optimizing the structure and components of the GELAN network model. This optimization not only reduces the complexity of the model and the demand for computing resources, but also improves the accuracy and robustness of detection. In practical applications, this helps to identify defects in pipelines more quickly and accurately, providing strong technical support for pipeline maintenance and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 This is a flow chart of an embodiment of a method for pipeline disease detection based on GELAN in the present application;

[0020] Figure 2 It is the overall network framework of the deep learning model in one embodiment of the present application;

[0021] Figure 3 Schematic diagram of the structure of an initial dynamic upsampler in an embodiment of the present application;

[0022] Figure 4 This is a schematic diagram of the structure of a dynamic sampler for group upsampling introduced in one embodiment of the present application;

[0023] Figure 5 This is a schematic diagram of the initial sampling position in an embodiment of the present application;

[0024] Figure 6 A schematic diagram of sampling positions for bilinear initialization in an embodiment of the present application;

[0025] Figure 7 A schematic diagram of theoretical boundary conditions satisfying overlapping and non-overlapping situations in one embodiment of the present application;

[0026] Figure 8This is a structural block diagram of an embodiment of a system for pipeline disease detection based on GELAN in this application;

[0027] Fig. 9 A structural block diagram of an electronic device provided in an embodiment of the present application.

[0028] Icon: 101, image acquisition module; 102, detection and analysis module; 201, processor; 202, memory; 203, communication interface. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0030] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0031] Example 1

[0032] See also Figure 1 The pipeline defect detection method based on GELAN includes the following steps: obtaining a target image containing a pipeline to be detected. Sending the target image to a deep learning model for pipeline defect detection to obtain a detection result image and defect information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses deep convolution DWConv as the network backbone, and uses a Dysample upsampling module for upsampling.

[0033] In the above embodiment, first, a target image containing the pipeline to be inspected can be obtained through a camera, a drone or other image acquisition device. These images may contain various diseases of the pipeline, such as cracks, corrosion, sediments, etc. Next, the acquired target image is sent to a specially optimized deep learning model. This deep learning model is improved based on the GELAN network architecture, and the complexity and computational requirements of the model are reduced by reducing the depth factor and the width factor. The network backbone of the deep learning model adopts deep convolution DWConv, which can more effectively extract key features in the image while reducing the number of parameters and the amount of calculation. In the upsampling stage, the Dysample upsampling module is used, which improves the restoration quality of the image resolution by introducing a finer sampling method, thereby helping to more accurately identify the details of the disease. Thus, after processing by the deep learning model, the detection result image and detailed disease information can be output. This information includes the location, type, severity, etc. of the disease, which provides an important reference for subsequent pipeline maintenance and management. Exemplarily, in some implementations of the present application, the disease information includes at least one of cracks, deformations, settled deposits, foreign object penetration, displaced joints, root invasions, or breaks.

[0034] It should be noted that the above embodiment optimizes the traditional GELAN network model in a targeted manner, aiming to solve the shortcomings of the prior art, such as high complexity of deep learning models, large computing resource requirements, and susceptibility to interference from external factors. First, the complexity of the model is effectively reduced by reducing the depth factor and width factor processing of the GELAN network model. This change not only reduces the number of parameters in the model, but also simplifies the structure of the model, thereby reducing the demand for computing resources and storage space. In practical applications, this means that the model can be deployed on a wider range of devices, including those with limited computing power, thereby improving the popularity and practicality of the technology. Exemplarily, in some implementations of the present application, the GELAN network model is processed to reduce the depth factor and width factor, including setting the depth factor of the GELAN network model to 0.33 and the width factor to 0.5. As shown in Table 1, after the depth factor of the GELAN network model is set to 0.33 and the width factor is set to 0.5, the complexity of the model is greatly reduced.

[0035]

[0036] Table 1 Comparison of model parameters and calculation amount

[0037] Secondly, using deep convolution DWConv as the network backbone can improve the deep learning model's ability to learn image features by reducing redundant information between channels. In the pipeline disease detection task, this helps the deep learning model to more accurately capture the disease features in the pipeline image and improve the accuracy and robustness of the detection.

[0038] Furthermore, traditional upsampling methods usually use bilinear interpolation, but it may cause the loss of key detail information. At the same time, the traditional kernel-based upsampling process requires a lot of calculations and parameters, which is not conducive to the lightweight network structure. In the actual drainage pipe disease scenario, relatively small diseases may have problems such as pixel distortion, resulting in the loss of fine-grained details and difficulty in feature learning in the detection task. By introducing the Dysample upsampling module for upsampling, the point-based sampling method and the learning sampling angle are used for upsampling, which avoids time-consuming dynamic convolution operations and additional sub-networks. Improve model efficiency and performance without adding additional computational burden. At the same time, the dynamic factor enables the Dysample upsampling module to adaptively adjust the weight of the convolution kernel according to the feature input, and enhance the detection ability of low-resolution images or smaller pipeline diseases.

[0039] That is, the use of deep convolution DWConv and Dysample upsampling modules effectively enhances the deep learning model's ability to extract image features and capture detail information, so that it can still maintain a high detection performance in complex environments such as lighting changes and noise interference.

[0040] Based on the above scheme, in some implementations of the present application, the network framework of the deep learning model includes an input network, a backbone network, a neck network and a prediction head network. Among them, the input network is used to receive the target image; the backbone network is used to extract key features from the received target image through deep convolution DWConv, and introduce linearity through the activation function; the neck network is used to fuse the extracted key features using the feature pyramid network, and use the Dysample upsampling module to combine the low-level feature map with the high-level feature map; the prediction head network is used to receive the output of the neck network to generate an anchor frame containing a feature map, and give the category probability corresponding to the anchor frame.

[0041] like Figure 2 As shown in the figure, the network framework of the deep learning model is designed to include 4 parts: input network, backbone network, neck network and prediction head network. Among them, the input network is the starting point of the deep learning model, which is responsible for receiving the target images to be processed. These images may be high-resolution images from cameras, image databases or other image sources.

[0042] Next, the backbone network is the core part of the deep learning model, which is responsible for extracting key features from the input target image. These features are usually abstract representations of the shape, texture, color and other information of the target in the image. The backbone network uses deep convolution DWConv, which reduces the amount of calculation by processing only one input channel per convolution kernel while maintaining the effectiveness of feature extraction. In addition, the backbone network also introduces linearity through activation functions (such as ReLU) to enhance the nonlinear expression ability of the model. After being processed by the backbone network, the original image is converted into a series of key feature maps, which contain important information about the disease information in the image.

[0043] The neck network is then responsible for fusing the key features extracted by the backbone network and generating higher-level feature representations. These high-level features are critical for accurate detection of disease information. The neck network uses a feature pyramid network (FPN) for feature fusion. FPN fuses feature maps of different scales through top-down paths and lateral connections to generate feature maps with rich semantic information. In addition, the neck network also uses a Dysample upsampling module, which can efficiently combine low-level feature maps with high-level feature maps to further improve the effect of feature fusion. After processing by the neck network, a feature map with rich semantic information and multi-scale features is generated.

[0044] Finally, the prediction head network is responsible for receiving the output of the neck network and generating anchor boxes containing feature maps. These anchor boxes are areas where disease information may exist, and feature maps are used to describe the characteristics of the disease information in the anchor boxes. The prediction head network further processes the output of the neck network through a series of convolutional layers or fully connected layers to generate the coordinates of the anchor boxes and the corresponding class probabilities. These coordinates and class probabilities are the key outputs of the target detection task. After being processed by the prediction head network, the accurate location and class information of the disease information are obtained, thereby achieving accurate detection and recognition of the disease information in the target image.

[0045] Based on the aforementioned scheme, in some implementations of the present application, the input network is also used to unify the size of the received target image into a fixed size, and apply random rotation, flipping, scaling and color transformation for data enhancement processing to enhance sample diversity.

[0046] It should be noted that the input network first receives the target images from image acquisition devices such as cameras and drones. Since these images may come from different devices and shooting conditions, their sizes and resolutions may vary. In order to ensure that the deep learning model can stably process these images, the input network first unifies their sizes to a fixed size. Then, in order to enhance sample diversity and improve the generalization ability of the deep learning model, the input network also applies data enhancement processing techniques such as random rotation, flipping, scaling, and color transformation. These techniques generate a series of new training samples by randomly transforming the original images. These new samples not only increase the amount of training data, but also improve the diversity and richness of the data. This helps the deep learning model learn more feature information, so as to better adapt to pipeline disease detection tasks in different scenarios. In short, by unifying the target image size to a fixed size and applying data enhancement processing techniques, it can be ensured that the deep learning model can process images of different sizes and qualities, thereby improving detection efficiency and accuracy. This design enables the deep learning model to more accurately identify pipeline diseases and reduce false detections and missed detections caused by inconsistent image sizes or poor quality.

[0047] Based on the above scheme, Figure 2 As shown, in some implementations of the present application, each convolution kernel of the deep convolution DWConv processes only one input channel, the number of convolution kernels is equal to the number of input channels, and the number of output channels is equal to the number of input channels, so as to perform convolution operations on each channel of the input layer independently.

[0048] It should be noted that in traditional convolution operations, a convolution kernel will act on all channels of the input feature map at the same time and generate an output channel. In the above implementation, each convolution kernel of the deep convolution DWConv is tailored for a specific input channel, that is, a convolution kernel only performs convolution operations on one input channel. The number of convolution kernels of the deep convolution DWConv is exactly equal to the number of input channels, which means that each input channel has a corresponding convolution kernel. This design ensures that each channel of the input feature map can be processed independently and efficiently. Since each convolution kernel only processes one input channel, the deep convolution DWConv can significantly reduce the amount of calculation compared to traditional convolution. In addition, by reducing the number of convolution kernels (one convolution kernel for each input channel), the number of parameters of the deep learning model can be reduced. This helps to reduce the storage requirements of the model and may improve the generalization ability of the model to a certain extent.

[0049] Based on the aforementioned scheme, in some implementations of the present application, the Dysample upsampling module introduces grouped upsampling, and the feature map is divided into multiple groups in the channel dimension, and each group shares the same sampling set; the Dysample upsampling module also introduces a dynamic range factor, including generating a point-by-point dynamic range factor by linearly projecting the input features.

[0050] It should be noted that in the Dysample upsampling module, the input features are evenly divided into several groups, each containing the same number of channels. Then, for each group, the Dysample upsampling module uses a shared sampling set (which may include a series of upsampling filters or operations) to upsample the features. This grouping strategy allows the Dysample upsampling module to process a large number of channels in a more efficient manner while maintaining feature consistency within each group. That is, group upsampling can reduce the consumption of computing resources because each group of features can be processed in parallel. And it enables the Dysample upsampling module to learn richer feature representations because different groups can capture different feature patterns.

[0051] Meanwhile, in the Dysample upsampling module, the input features are first processed by linear projection (such as a fully connected layer or a convolutional layer) to generate a dynamic range factor map with the same spatial dimension as the input features. This factor map is then used to adjust the output range of the upsampling operation so that the output features can more accurately reflect the changes in the input features. Since the dynamic range factor allows the upsampling module to adaptively adjust its output according to the content of the input features, the flexibility and accuracy of the model are improved.

[0052] For ease of understanding, the following Figure 3-Figure 7 The Dysample upsampling module of the present application is described. Figure 3 As shown, in the initial version of the dynamic upsampling process, the s of each point in the input image X 2 The sampling positions are fixed at the same initial position, such as Figure 5 This initialization method ignores the positional relationship between adjacent points. When the generated offsets are all zero, the upsampled features are equivalent to the network interpolated features. When the generated offsets are all zero, the upsampled features are equivalent to the network interpolated features. To solve this problem, the Dysample upsampling module of this application changes the initial sampling position to Figure 6 In the "Bilinear Initialization" in, zero offset will produce a bilinear interpolated feature map. In order to reduce the local sampling position overlap that may be caused by the normalization layer, the Dysample upsampling module multiplies the offset by 0.25, which just meets the theoretical boundary conditions for overlapping and non-overlapping cases, such as Figure 7 shown.

[0053] At the same time, the Dysample upsampling module introduces the concept of grouped upsampling, which means that the feature maps are divided into g groups in the channel dimension, and each group shares the same sampling set. Specifically, the feature maps are divided into g groups, and a corresponding offset set is generated for each group. In order to increase the flexibility of the offset, the Dysample upsampling module generates a point-by-point "dynamic range factor" by linearly projecting the input features. This operation is as follows Figure 4 As shown, it uses linear projection to generate s 2 The Dysample upsampling module can process input images or small targets of different resolutions more effectively, enhance detection capabilities, and improve upsampling accuracy and efficiency by introducing group upsampling and dynamic range factors.

[0054] Example 2

[0055] See also Figure 8 The embodiment of the present application provides a pipeline disease detection system based on GELAN, which includes:

[0056] The image acquisition module 101 is configured to: acquire a target image containing a pipeline to be inspected. The detection and analysis module 102 is configured to: send the target image to a deep learning model for pipeline disease detection to obtain a detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses the deep convolution DWConv as the network backbone, and uses the Dysample upsampling module for upsampling.

[0057] For the specific implementation process of the above system, please refer to the pipeline disease detection method based on GELAN provided in Example 1, which will not be described in detail here.

[0058] Example 3

[0059] See also Fig. 9 , the embodiment of the present application provides an electronic device, which includes at least one processor 201 and at least one memory 202; wherein the processor 201 and the memory 202 are directly connected to each other, or communicate with each other through a communication interface 203, or are electrically connected through one or more communication buses or signal lines to achieve data transmission or interaction; the memory 202 stores program instructions that can be executed by the processor 201, and the processor 201 calls the program instructions to execute a method for pipeline disease detection based on GELAN. For example, the following is realized:

[0060] Obtain a target image containing the pipeline to be inspected. Send the target image to a deep learning model for pipeline disease detection to obtain a detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses deep convolution DWConv as the network backbone, and uses the Dysample upsampling module for upsampling.

[0061] Among them, the memory 202 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc.

[0062] The processor 201 may be an integrated circuit chip with signal processing capability. The processor 201 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0063] Understandably, Fig. 9 The structure shown is for illustration only. The electronic device may also include Fig. 9 More or fewer components as shown, or with Fig. 9 Different configurations are shown. Fig. 9 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0064] Example 4

[0065] The present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by the processor 201, a method for pipeline disease detection based on GELAN is implemented. For example, the following is implemented:

[0066] Obtain a target image containing the pipeline to be inspected. Send the target image to a deep learning model for pipeline disease detection to obtain a detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses deep convolution DWConv as the network backbone, and uses the Dysample upsampling module for upsampling.

[0067] If the above functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0068] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the present application is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present application. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A pipeline disease detection method based on GELAN, characterized in that: The following steps are involved: Acquire a target image containing a pipeline to be inspected; The target image is sent to the deep learning model for pipeline disease detection to obtain the detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses deep convolution DWConv as the network backbone, and uses the Dysample upsampling module for upsampling.

2. The method according to claim 1, characterized in that: The disease information includes at least one of cracks, deformations, sediments, foreign body penetration, misaligned joints, root invasion or fractures.

3. The method according to claim 1, characterized in that The network framework of the deep learning model includes an input network, a backbone network, a neck network and a prediction head network; Among them, the input network is used to receive the target image; the backbone network is used to extract key features from the received target image through deep convolution DWConv, and introduce linearity through the activation function; the neck network is used to fuse the extracted key features using the feature pyramid network, and use the Dysample upsampling module to combine the low-level feature map with the high-level feature map; the prediction head network is used to receive the output of the neck network to generate an anchor box containing the feature map, and give the category probability corresponding to the anchor box.

4. The method according to claim 3, characterized in that The input network is also used to unify the size of the received target image into a fixed size, and apply random rotation, flipping, scaling and color transformation to perform data enhancement processing to enhance sample diversity.

5. The method according to claim 1, characterized in that The processing of reducing the depth factor and the width factor of the GELAN network model includes setting the depth factor of the GELAN network model to 0.33 and the width factor to 0.

5.

6. The method according to claim 1, characterized in that Each convolution kernel of the deep convolution DWConv processes only one input channel, the number of convolution kernels is equal to the number of input channels, and the number of output channels is equal to the number of input channels, so as to perform convolution operations on each channel of the input layer independently.

7. The method according to claim 1, characterized in that The Dysample upsampling module introduces grouped upsampling, and the feature map is divided into multiple groups in the channel dimension, and each group shares the same sampling set; the Dysample upsampling module also introduces a dynamic range factor, including generating a point-by-point dynamic range factor by linearly projecting the input features.

8. A pipeline disease detection system based on GELAN, characterized in that: include: The image acquisition module is configured to: acquire a target image including a pipeline to be inspected; The detection and analysis module is configured to: send the target image to the deep learning model for pipeline disease detection to obtain the detection result image and disease information; wherein the deep learning model is a network model that reduces the depth factor and width factor of the GELAN network model, uses the deep convolution DWConv as the network backbone, and uses the Dysample upsampling module for upsampling.

9. An electronic device, characterized in that: include: A memory for storing one or more programs; processor; When the one or more programs are executed by the processor, the method according to any one of claims 1 to 7 is implemented.

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

Images