A method for identifying product surface defects based on multi-domain feature fusion networks, equipment and media

Abstract

This invention relates to a method, device, and medium for identifying product surface defects based on a multi-domain feature fusion network. The method involves inputting a defect image obtained from a product surface defect identification dataset into the first layer of an A-ACN backbone network to obtain a first feature map; inputting the defect image into a peak region auxiliary feature learning module to obtain a peak feature map, and adding the peak feature map to the first feature map to obtain a second feature map; inputting the second feature map into the second layer of the backbone network to obtain a third feature map; inputting the third feature map and the defect image into a multi-domain hybrid attention module to obtain a fourth feature map; inputting the fourth feature map into the third layer of the backbone network and passing it through a gating feature selection module to obtain a fifth feature map; and inputting the fifth feature map into the fourth and fifth layers of the backbone network to obtain the final output defect category. Compared with existing technologies, this invention has advantages such as high accuracy, strong generalization ability, and strong robustness.

Description

Technical Field

[0001] This invention relates to the field of product surface defect identification, and in particular to a method, device and medium for product surface defect identification based on a multi-domain feature fusion network. Background Technology

[0002] As the manufacturing industry shifts from quantity expansion to quality improvement, enhancing product quality has become crucial for boosting competitiveness. Since most products are mass-produced, the manufacturing process involves numerous complex steps, and any abnormality during manufacturing can lead to surface defects. However, surface damage to certain critical components can not only severely impair product performance but also pose safety risks to users. Surface defect identification methods can promptly detect problems, eliminate defective products, improve corresponding production processes, and proactively maintain production machinery. Through automated and intelligent surface defect detection methods, the causes of defects can be analyzed, making production line maintenance more targeted, effectively reducing product defect rates, improving product quality, and ensuring user safety.

[0003] Traditional methods for identifying product surface defects mostly rely on manual feature extraction and classifier design, typically using interpretable mathematical theories and mature expert experience to analyze representative features of product surface defects. These algorithms, depending on rich expert knowledge and theoretical models, often struggle to adapt to complex and ever-changing scenarios and the demands of intelligent identification. In recent years, research on product surface defect identification based on deep learning methods has gradually become mainstream in this field, with convolutional neural network models being applied. While this data-driven approach can automatically learn hidden patterns and characteristics from large-scale sample data, offering higher accuracy, faster speed, and simpler design compared to traditional methods, practical applications of product surface defect identification often face challenges such as difficulty in interpreting learned features, large datasets with small sample sizes, and limitations in visual cognition. A deep integration of theory-driven and data-driven methods is particularly important in product surface defect identification. Methods that improve the accuracy and interpretability of product surface defect identification by fusing traditional and deep features have been proposed. However, some problems still need to be addressed in real-world product surface defect scenarios. First, product surface defect images typically possess complex and varied shapes and textures, necessitating consideration of how to better capture features of key areas during limited sample learning. Second, many studies integrate physical models into deep neural networks for feature fusion; however, because model performance is heavily influenced by fixed input physical features, its generalization ability may be limited when faced with new samples whose distribution differs significantly from the training data, resulting in poor adaptability. Furthermore, these algorithms are usually complex to construct and require substantial computational resources. Additionally, while fine-grained modeling methods can directly improve the recognition performance of deep learning models, achieving a good balance between model accuracy and speed is challenging in practical applications. Therefore, improving defect recognition capabilities, enhancing adaptability and generalization, and increasing classification accuracy in product surface defect identification are key technical challenges that need to be addressed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology by providing a product surface defect recognition method, device and medium based on a multi-domain feature fusion network. By integrating the peak region auxiliary feature learning module, the multi-domain hybrid attention module and the gated feature selection module with the A-CAN backbone network, the defect recognition capability is improved, the features of the image target are comprehensively described, and the output of the multi-domain feature fusion is used for product surface defect recognition.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] According to one aspect of the present invention, a method for identifying product surface defects based on a multi-domain feature fusion network is provided, the specific steps of which include:

[0007] S1. Collect the product surface defect identification dataset and preprocess it to obtain defect images with uniform pixel size;

[0008] S2. Input the defect image into the first layer of the A-ACN backbone network to obtain the first feature map;

[0009] S3. Input the defect image into the peak region auxiliary feature learning module to obtain the peak feature map, and add the peak feature map to the first feature map to obtain the second feature map;

[0010] S4. Input the second feature map into the second layer of the A-ACN backbone network to obtain the third feature map. Input the third feature map and the defect image into the multi-domain hybrid attention module respectively, and obtain the fourth feature map through the wavelet driving branch and the attention branch.

[0011] S5. Input the fourth feature map into the third layer of the A-ACN backbone network, and obtain the fifth feature map through the gated feature selection module;

[0012] S6. Input the fifth feature map into the fourth and fifth layers of the A-ACN backbone network to obtain the final output defect category.

[0013] Furthermore, in step S3, the step of obtaining the peak feature map using the peak region auxiliary feature learning module includes: performing Gaussian filtering on the defect image to retain the main features; adaptively selecting the optimal threshold using the Otsu method; obtaining the largest connected region, i.e. the target local peak region, using the binary threshold method; and generating a binary map of the target local peak region as the peak feature map.

[0014] Furthermore, the expression for the Gaussian filter is:

[0015] ,

[0016] in,( x , y () represents the coordinates of the point. σ Standard deviation k Here is the window template size parameter; the window template size is (2 k +1)×(2 k +1).

[0017] Furthermore, the expression for the binary threshold method is:

[0018] ,

[0019] Among them, dst( x , y ) is at coordinate position ( x , yThe pixel values ​​of the target image, src( x , y ) represents the pixel of the source image at coordinate position (x, y), thresh is the threshold automatically determined by Otsu's method, and maxval is the maximum pixel value.

[0020] Furthermore, in S4, the feature mapping values ​​are reweighted based on global visual information by the frequency domain attention module; the low-frequency information A, horizontal high-frequency information H, vertical high-frequency information V, and diagonal high-frequency information D of the defect image are obtained through the two-dimensional discrete wavelet transform of the wavelet-driven branch. i , j The eigenvalue expressions for the 10 wavelet feature maps are:

[0021] ，

[0022] ，

[0023] ，

[0024] ，

[0025] in, For the ( i , j Low-frequency information of wavelet feature maps, For the ( i , j ) horizontal high-frequency information of wavelet feature maps For the ( i , j Vertical high-frequency information of wavelet feature maps, For the ( i , j ( ) Diagonal high-frequency information of wavelet feature maps.

[0026] Furthermore, the features of the obtained wavelet feature maps are processed through grouped convolution to preserve the local semantics of each wavelet feature, generating wavelet activation feature maps. The third feature map is convolved to make its dimension consistent with that of the wavelet activation feature map, and the element-wise product of the third feature map and the wavelet activation feature map is obtained as the wavelet activation group feature emphasizing each wavelet component. The compressed features of each subgroup are obtained through a selective average pooling layer, and the output is obtained by processing by an MLP layer. After the subgroup features are concatenated, the channel weights are recalibrated by a Softmax layer. The subgroup features are multiplied channel-wise with the wavelet activation group feature to obtain the local semantic enhancement features of the wavelet-driven branch based on wavelet information.

[0027] Further, in step S5, the fourth feature map is input into the gated feature selection module, and the channel is divided into a gated path and a convolutional path. The gated path obtains a weight ranging from 0 to 1 through the sigmoid function. The higher the weight, the more important the feature is at that position. The convolutional path non-linearly activates the feature map through a LeakyReLU layer. The outputs of the gated path and the convolutional path are multiplied to obtain the fifth feature map.

[0028] Furthermore, during the execution of the method, the training parameters of the model are updated through backpropagation using the cross-entropy loss function.

[0029] According to a second aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.

[0030] According to a third aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) Improve the accuracy of product surface defect identification: By introducing the peak region auxiliary feature learning module, the network's learning of key target areas, especially the target shape and structure features, is effectively strengthened, and the model's ability to capture product surface defect images with complex and varied shapes and textures is improved, so that key features can be accurately identified even in limited sample learning. In addition, the wavelet-driven multi-domain hybrid attention module adaptively fuses semantic prior knowledge, wavelet domain and frequency domain attention features, and then uses a gated feature selection mechanism to filter and enhance the features, activate and reweight the features, and strengthen the learning of local semantics, so as to understand the defect information in the image more deeply, enhance the model's generalized feature extraction ability, and improve the accuracy of defect identification.

[0033] (2) Improve the robustness of recognition: Through the gating feature selection module, the features are effectively filtered and enhanced. The gating feature selection mechanism can focus on relevant features and suppress irrelevant features, thereby enhancing the meaning of the learned features. This helps the model to focus more on key defect features during the recognition process, thus improving the accuracy and robustness of recognition.

[0034] (3) Enhance generalization ability: The peak region auxiliary feature learning module, multi-domain hybrid attention module and gated feature selection module are fully integrated with the A-ACN backbone network to construct a multi-domain feature fusion network. The spatial domain, wavelet domain and frequency domain features of the image are fully extracted, and the high-order deep features and low-order local artificial features are optimized and fused to comprehensively describe the features of the image target. This makes the model exhibit higher performance, stronger generalization ability and credibility in product surface defect recognition. Attached Figure Description

[0035] Figure 1 This is a flowchart of the steps in a product surface defect identification method based on a multi-domain feature fusion network.

[0036] Figure 2 This is a schematic diagram of the network structure for product surface defect identification based on a multi-domain feature fusion network. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0038] like Figure 1 The image shows a method for identifying product surface defects based on a multi-domain feature fusion network. The specific steps include:

[0039] S1. Collect the product surface defect identification dataset and preprocess it to obtain defect images with uniform pixel size;

[0040] S2. Input the defect image into the first layer of the A-ACN backbone network to obtain the first feature map;

[0041] S3. Input the defect image into the peak region auxiliary feature learning module to obtain the peak feature map, and add the peak feature map to the first feature map to obtain the second feature map;

[0042] S4. Input the second feature map into the second layer of the A-ACN backbone network to obtain the third feature map. Input the third feature map and the defect image into the multi-domain hybrid attention module respectively, and obtain the fourth feature map through the wavelet driving branch and the attention branch.

[0043] S5. Input the fourth feature map into the third layer of the A-ACN backbone network, and obtain the fifth feature map through the gated feature selection module;

[0044] S6. Input the fifth feature map into the fourth and fifth layers of the A-ACN backbone network to obtain the final output defect category.

[0045] This embodiment uses the WM-811K wafer surface image dataset, collected during actual semiconductor manufacturing processes, for experimental verification. The WM-811K dataset contains 811,457 wafer surface images labeled with defect patterns, encompassing nine wafer patterns: center defects, rings, edge local defects, edge ring defects, local defects, near-full defects, random defects, scratches, and normal wafer surface patterns. 32,077 wafer images were randomly selected from the WM-811K dataset to form the data used in this paper. The images were preprocessed to 128×128 pixels, and each class of data was divided into training and test sets in a 7:3 ratio as network input. The training set was used to train the MDFFN and adjust the network parameters. Finally, the test set was used to evaluate the classification performance of the MDFFN. The backbone network was used to extract features from the input product surface defect images. A-ConvNets (ACN) is a classic deep learning algorithm in the field of object recognition. Its backbone network is a fully convolutional structure. This embodiment utilizes a backbone network with adjusted parameters and structure based on the original ACN algorithm, denoted as A-ACN. The input is a product surface defect image I∈ℝ128×128×1. The backbone network extracts features from the product surface defect image through five stages {F1, F2, F3, F4, F5}. Each stage… i Each stage includes a convolutional layer for feature extraction and a max-pooling layer for feature map dimensionality reduction, with the output feature maps represented as {E1, E2, E3, E4, E5}. Furthermore, each convolutional layer output undergoes batch normalization and activation function operations with rectified linear units, increasing network non-linearity. The feature map space size in the later stage is reduced to half that of the previous stage. As the network depth increases, the convolutional kernel channel width expands by a factor of 2, thus enabling the extraction of features with stronger abstract semantics.

[0046] In S3, a peak region-assisted feature learning mechanism is constructed to extract peak features from the image and fuse them with deep features. After embedding the target peak map into the first layer of the network structure, the network encourages neurons to respond positively to peak regions during the learning and optimization process, which can enhance the learning of key areas of the target, especially the shape and structural features of the target. The steps of obtaining the peak feature map using the peak region-assisted feature learning module include: performing Gaussian filtering on the defect image to retain the main features; adaptively selecting the optimal threshold using the Otsu method; obtaining the largest connected region, i.e., the local peak region of the target, using the binary thresholding method; and generating a binary map of the local peak region of the target as the peak feature map.

[0047] The expression for Gaussian filtering is:

[0048] ,

[0049] in,( x , y () represents the coordinates of the point. σ Standard deviation k Here is the window template size parameter; the window template size is (2 k +1)×(2 k +1).

[0050] The expression for the binary threshold method is:

[0051] ,

[0052] Among them, dst( x , y ) is at coordinate position ( x , y The pixel values ​​of the target image, src( x , y ) represents the pixel of the source image at coordinate position (x, y), thresh is the threshold automatically determined by Otsu's method, and maxval is the maximum pixel value.

[0053] After the first layer of the network outputs the first feature map and the peak feature map, each is processed by a 3×3 convolution layer to transform them into the same dimension. The local peak feature map is subject to the same stable and slowly decreasing adjustment factor α. t After multiplication, the result is added to the network feature map and then passed to the next layer. The expression is:

[0054] ,

[0055] Where Q is the peak image after convolution. The first feature map after convolution is obtained by adding the two feature maps together to get the output feature map. α(t) is based on the number of training rounds t The changing adjustment factor gradually decreases during training, and its expression is:

[0056] ,

[0057] in, t This is the current training round number. t max This represents the maximum number of training rounds.

[0058] In S4, the feature mapping values ​​are reweighted based on global visual information through a frequency domain attention module; the low-frequency information A, horizontal high-frequency information H, vertical high-frequency information V, and diagonal high-frequency information D of the defect image are obtained through a two-dimensional discrete wavelet transform with wavelet-driven branches. i , j The eigenvalue expressions for the 10 wavelet feature maps are:

[0059] ，

[0060] ，

[0061] ，

[0062] ，

[0063] in, For the ( i , j Low-frequency information of wavelet feature maps, For the ( i , j ) horizontal high-frequency information of wavelet feature maps For the ( i , j Vertical high-frequency information of wavelet feature maps, For the ( i , j ( ) Diagonal high-frequency information of wavelet feature maps.

[0064] The obtained wavelet feature maps are processed by grouped convolution to preserve the local semantics of each wavelet feature, generating wavelet activation feature maps. The third feature map is convolved to make its dimension consistent with that of the wavelet activation feature map, and the element-wise product of the third feature map and the wavelet activation feature map is obtained as the wavelet activation group feature emphasizing each wavelet component. The compressed features of each subgroup are obtained through selective average pooling layers, and the output is obtained by MLP layer. After the subgroup features are concatenated, the channel weights are recalibrated by Softmax layer. Multiplying them channel-wise with the wavelet activation group feature yields the local semantic enhancement features of the wavelet-driven branch based on wavelet information.

[0065] Wavelet activation feature maps are generated based on the feature paths of the wavelet feature maps. The expression is:

[0066] ,

[0067] in, Conv i For the firsti Each group of convolutional layers, This is the i-th wavelet feature;

[0068] The element-wise product expression of the third feature map and the wavelet activation feature map is:

[0069] ,

[0070] in, The third feature map, after convolution processing, is unified with the wavelet activation feature map into a feature map of the same dimension.

[0071] In S5, the fourth feature map is input into the gated feature selection module, which divides the channel into a gated path and a convolutional path. The gated path uses the sigmoid function to obtain weights ranging from 0 to 1; the higher the weight, the more important the feature at that position. The convolutional path non-linearly activates the feature map through a LeakyReLU layer. The outputs of the gated path and the convolutional path are multiplied to obtain the fifth feature map. The expression is:

[0072] ,

[0073] ,

[0074] ,

[0075] in, E 4 This is the fourth feature map. C g and C f For convolution operations, The sigmoid function transforms the gated output value to a range between 0 and 1. This is the LeakyReLU activation function.

[0076] During the execution of the method, the training parameters of the model are updated through backpropagation using the cross-entropy loss function.

[0077] like Figure 2 The diagram shows a network structure for product surface defect recognition based on a multi-domain feature fusion network. The peak region auxiliary feature learning module, multi-domain hybrid attention module, and gated feature selection module are integrated with A-ACN to fully extract spatial, wavelet, and frequency domain features of the image, thereby optimizing the fusion of high-order deep features and low-order local artificial features to comprehensively describe the features of the image target.

[0078] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0079] The electronic device of this invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0080] Multiple components in the device are connected to an I / O interface, including: input units such as a keyboard, mouse, etc.; output units such as various types of displays, speakers, etc.; storage units such as disks, optical disks, etc.; and communication units such as network interface cards, modems, wireless transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks. The processing unit performs the various methods and processes described above, such as the method of the present invention. For example, in some embodiments, the method of the present invention may be implemented as a computer software program tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or the communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of the method of the present invention described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute the method of the present invention by any other suitable means (e.g., by means of firmware).

[0081] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0082] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0083] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0084] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying product surface defects based on a multi-domain feature fusion network, characterized in that, The specific steps include: S1. Collect the product surface defect identification dataset and preprocess it to obtain defect images with uniform pixel size; S2. Input the defect image into the first layer of the A-ACN backbone network to obtain the first feature map; S3. Input the defect image into the peak region auxiliary feature learning module to obtain the peak feature map, and add the peak feature map to the first feature map to obtain the second feature map; S4. Input the second feature map into the second layer of the A-ACN backbone network to obtain the third feature map. Input the third feature map and the defect image into the multi-domain hybrid attention module respectively, and obtain the fourth feature map through the wavelet driving branch and the attention branch. S5. Input the fourth feature map into the third layer of the A-ACN backbone network, and obtain the fifth feature map through the gated feature selection module; S6. Input the fifth feature map into the fourth and fifth layers of the A-ACN backbone network to obtain the final output defect category.

2. The product surface defect identification method based on a multi-domain feature fusion network according to claim 1, characterized in that, In step S3, the step of obtaining the peak feature map using the peak region auxiliary feature learning module includes: performing Gaussian filtering on the defect image to retain the main features; adaptively selecting the optimal threshold using the Otsu method; obtaining the largest connected region, i.e. the target local peak region, using the binary threshold method; and generating a binary map of the target local peak region as the peak feature map.

3. The product surface defect identification method based on a multi-domain feature fusion network according to claim 2, characterized in that, The expression for the Gaussian filter is: ， in,( x , y () represents the coordinates of a point. σ Standard deviation k Here is the window template size parameter; the window template size is (2 k +1)×(2 k +1).

4. The product surface defect identification method based on a multi-domain feature fusion network according to claim 2, characterized in that, The expression for the binary threshold method is: ， Among them, dst( x , y ) is at coordinate position ( x , y The pixel values ​​of the target image, src( x , y ) represents the pixel of the source image at coordinate position (x, y), thresh is the threshold automatically determined by Otsu's method, and maxval is the maximum pixel value.

5. The product surface defect identification method based on a multi-domain feature fusion network according to claim 1, characterized in that, In step S4, the feature mapping values ​​are reweighted based on global visual information using a frequency domain attention module; the low-frequency information A, horizontal high-frequency information H, vertical high-frequency information V, and diagonal high-frequency information D of the defect image are obtained through a two-dimensional discrete wavelet transform driven by wavelet branches. i , j The eigenvalue expressions for the 10 wavelet feature maps are: ， ， ， ， in, For the ( i , j Low-frequency information of wavelet feature maps, For the ( i , j ) horizontal high-frequency information of wavelet feature maps For the ( i , j Vertical high-frequency information of wavelet feature maps, For the ( i , j ( ) Diagonal high-frequency information of wavelet feature maps.

6. The product surface defect identification method based on a multi-domain feature fusion network according to claim 5, characterized in that, The features of the obtained wavelet feature map are processed by grouped convolution to preserve the local semantics of each wavelet feature and generate a wavelet activation feature map. The third feature map is processed by convolution to make it consistent with the dimension of the wavelet activation feature map, and the element-wise product of the third feature map and the wavelet activation feature map is obtained as the wavelet activation group feature that emphasizes each wavelet component. The compressed features of each subgroup are obtained through a selective average pooling layer, and the output is obtained by processing by an MLP layer. After the subgroup features are concatenated, the channel weights are recalibrated by a Softmax layer. The features are then multiplied channel by channel with the wavelet activation group features to obtain the wavelet-driven branch local semantic enhancement features based on wavelet information.

7. The product surface defect identification method based on a multi-domain feature fusion network according to claim 1, characterized in that, In step S5, the fourth feature map is input into the gated feature selection module, and the channel is divided into a gated path and a convolutional path. The gated path obtains a weight ranging from 0 to 1 through the sigmoid function. The higher the weight, the more important the feature. The convolutional path non-linearly activates the feature map through a LeakyReLU layer. The outputs of the gated path and the convolutional path are multiplied to obtain the fifth feature map.

8. The product surface defect identification method based on a multi-domain feature fusion network according to claim 1, characterized in that, During the execution of the method, the training parameters of the model are updated through backpropagation using the cross-entropy loss function.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 8.

Images