Image bubble noise removal method and system based on dual-channel characteristics

Abstract

The invention discloses an image bubble noise removal method and system based on dual-channel features, and relates to the technical field of image processing, an improved neural network model is constructed, and dual channels comprise a first channel used for extracting color features related to image recognition and a second channel used for extracting bubble interference features; the feature processing unit is used for superposing the bubble interference features and color features related to image recognition and then extracting; obtaining a picture without interference and with a concentration label and a picture with part of bubbles as an interference-free image and a bubble-containing interference image, training and testing the improved neural network model, and obtaining a trained improved neural network model; and acquiring a to-be-processed image, inputting the to-be-processed image into the trained improved neural network model, performing processing such as convolution mask and the like, removing bubble noise, expanding the image into a vector form after convolution, and outputting a target recognition result of the input image by using a final output node 1.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for removing image bubble noise based on dual-channel features. Background Art

[0002] Existing image recognition technology often suffers from reduced recognition accuracy when processing images subject to specific interference. For example, in industrial inspection, medical image processing, and daily visual recognition tasks, interference on the image such as bubbles, reflections, water stains, etc. will affect the classification model's correct recognition of image targets.

[0003] Convolutional neural networks (CNNs) commonly used in image processing are mostly trained without noise interference, and fail to fully consider the impact of interfering objects on model feature extraction, resulting in unstable performance of the model in an interfering environment. Even if interfering images are added for training, the model still cannot fully extract effective information of interference and non-interference. At the same time, no method for correctly processing interference information is added later. Therefore, this hybrid training method increases the accuracy of the model to a certain extent, but the accuracy is still very poor on unfamiliar data sets.

[0004] For example, in the process of microalgae cultivation, bubbles have a significant interference effect on concentration monitoring. The generation of bubbles is closely related to processes such as aeration and stirring. These bubbles are randomly distributed in the culture medium and are constantly generated and burst. The reflection and refraction effects of the bubbles will produce uneven brightness areas, destroy the color consistency in the image, increase the noise of the image, and make feature extraction based on color and brightness more difficult. Secondly, bubbles will cause local dynamic changes in the image, thereby affecting the prediction accuracy of the model.

[0005] In recent years, some researchers have proposed some methods to alleviate bubble interference. They use image filtering and denoising techniques to reduce the interference of bubbles on image features. However, such algorithms are usually used for images with obvious themes and rich feature information color channels. If they are applied to target recognition in images with relatively uniform colors and unclear bubble colors, the accuracy and robustness of the model will be poor. Summary of the invention

[0006] In order to overcome the shortcomings of the above-mentioned prior art in the recognition of image targets with relatively uniform colors and unclear bubble colors, the main purpose of the present invention is to provide an image bubble noise removal method and system based on dual-channel features.

[0007] To achieve the above object, the present invention adopts the following technical scheme: a method for removing bubble noise from an image based on dual-channel features, specifically comprising: Construct an improved neural network model based on CNN, including a first channel, a second channel, multiple sequentially connected feature processing units, a mask module, and a fully connected layer; the first channel is used to extract color features related to image recognition, and the second channel is used to extract bubble interference features; the feature processing unit includes four convolutional layers and a feature superposition extraction module, and the feature superposition extraction module is used to superpose the bubble interference features and the color features related to image recognition and then extract them; the mask module is used to perform mask processing on the output of the feature processing unit, and the fully connected layer is used to process the masked features after unfolding and output the target recognition result; Obtain pictures without interference and with annotation information and some pictures with bubbles as interference-free images and images with bubble interference, construct a training set, and input them into the first channel and the second channel respectively to train the improved neural network model to obtain the trained improved neural network model; Obtain the image to be processed and input it into the trained improved neural network model to output the corresponding target recognition result.

[0008] Preferably, the convolutional kernels of each convolutional layer in each feature processing unit are used for extracting bubble interference features and color features related to image recognition.

[0009] Preferably, the step of obtaining the image to be processed and inputting it into the trained improved neural network model to output the corresponding target recognition result includes the following steps: Input the image to be processed into the improved neural network model to obtain color features related to image recognition and bubble interference features; Input the color features related to image recognition and the extracted bubble interference features into multiple sequentially connected feature processing units for feature extraction, where: In each feature processing unit, use four convolutional layers to perform convolutional processing on the input in sequence, use a filter to extract features from the convolved input to obtain the re-extracted convolutional input features, add the re-extracted convolutional input features element by element, obtain the superimposed features and then extract features to obtain the intermediate enhanced bubble interference features; Generate a low-value matrix with the same size as the bubble interference features according to the intermediate enhanced bubble interference features; Cover the generated low-value matrix to the bubble interference features for mask processing to obtain the masked features, perform superimposed feature processing on the masked features and the color features related to image recognition to remove bubble noise and obtain the result of removing bubble noise; Use the fully connected layer to unfold the result of removing bubble noise into a vector form, set the output node of the fully connected layer to 1, and output the corresponding target recognition result.

[0010] A system for an image bubble noise removal method based on dual-channel features, comprising: A model improvement module, configured to build an improved neural network model based on CNN, including a first channel, a second channel, a plurality of sequentially connected feature processing units, a mask module, and a fully connected layer; the first channel is used to extract color features related to image recognition, and the second channel is used to extract bubble interference features; the feature processing unit includes four convolutional layers and a feature superposition extraction module, and the feature superposition extraction module is used to superpose the bubble interference features and the color features related to image recognition and then extract them; the mask module is used to perform mask processing on the output of the feature processing unit, and the fully connected layer is used to process the unfolded masked features and output a target recognition result; A model training module, configured to obtain pictures without interference and with annotation information and some pictures with bubbles as interference-free images and bubble-interference-containing images, construct a training set, and input them into the first channel and the second channel respectively to train the improved neural network model to obtain a trained improved neural network model; A bubble noise removal module, which obtains the image to be processed and inputs it into the trained improved neural network model, and outputs the corresponding target recognition result.

[0011] Compared with the prior art, the beneficial effects of the present invention are: 1. The dual-channel input of the present invention allows the simultaneous processing of interference-free images and bubble-interference-containing images, and respectively extracts pure color features and bubble interference features. By overlapping the feature maps of the dual-channel input data, and superposing the feature maps of the two channels, the neural network can better identify and distinguish bubble noise and the target to be recognized itself. Using a low-value matrix to perform mask processing on the bubble interference features reduces the weight of the interference area, thereby reducing the influence of bubble noise in subsequent processing. The method for improving the online monitoring accuracy in an interference environment can effectively improve the fitting accuracy in an unfamiliar and complex environment.

[0012] 2. The present invention adopts feature superposition and extraction after convolution, thus solving to a certain extent the problem of highly consistent requirements for corresponding pixel points of the original image and the image with interference. It only requires that the two pictures belong to the same concentration, reducing the complexity of data collection and annotation. By dividing the convolution kernel into two parts, bubble interference features and color features can be extracted more precisely. The final output layer performs a regression operation, which can finely adjust the prediction result to obtain a higher-quality image with bubble noise removed.

[0013] 3. The present invention aims at a concentration fitting model mainly based on machine vision, and improves the fitting accuracy by processing images, reducing the difficulty and cost of data collection.

[0014] 4. The present invention can effectively improve the sensitivity of partial convolution to areas that do not belong to the main features, ensuring the effective extraction of the main features and the accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute improper limitations on the present application.

[0016] Figure 1 is a heat map of the original image and the degree of attention paid by the visualization model to the original image in the embodiment of the present invention; wherein, Figure 1 a is the heat map of the original image, Figure 1 b is a heat map of the degree of attention paid by the basic model to the original image. Figure 1 c is a heat map of the degree of attention paid by the model to the original image after the introduction of the attention mechanism. Figure 1 d is a heat map of the degree of attention paid by the model of the present invention to the original image; Figure 2 is a heat map of the degree of attention of the visualization model after the original image and the initial layer and four pooling layers in the model architecture in the embodiment of the present invention are processed; wherein, Figure 2 a is the heat map of the original image, Figure 2 b is the heat map after the initial layer processing, Figure 2 c is the heat map after the first pooling layer processing, Figure 2 d is the heat map after the second pooling layer processing, Figure 2 e is the heat map after the third pooling layer processing. Figure 2 f is the heat map after processing by the fourth pooling layer; Figure 3 It is a schematic diagram of the process structure of the present invention. DETAILED DESCRIPTION

[0017] During the cultivation of microalgae, bubbles have a significant interference effect on concentration monitoring. The generation of bubbles is closely related to processes such as aeration and stirring. These bubbles are randomly distributed in the culture solution and are constantly generated and burst. The reflection and refraction effects of the bubbles will produce uneven brightness areas, destroy the color consistency in the image, increase the noise of the image, and make feature extraction based on color and brightness more difficult. Secondly, bubbles will cause local dynamic changes in the image, thereby affecting the prediction accuracy of the model.

[0018] In recent years, researchers have proposed some methods to alleviate bubble interference. They use image filtering and denoising techniques to reduce the interference of bubbles on image features. However, such algorithms are usually used for images with more obvious themes and richer color channels of feature information. When applied to the concentration identification of microalgae solutions with more uniform colors and less obvious bubble colors, the robustness of the model still needs further optimization and evaluation.

[0019] This mixed training method has increased the accuracy of the model to a certain extent, but the accuracy is still very poor on unfamiliar data sets. In particular, in the process of microalgae cultivation, bubbles have a significant interference effect on concentration monitoring. The generation of bubbles is mainly closely related to processes such as aeration and stirring. These bubbles are randomly distributed in the culture solution and are constantly generated and burst. The reflection and refraction effects of the bubbles will produce uneven brightness areas, destroy the color consistency in the image, increase the noise of the image, and make feature extraction based on color and brightness more difficult. Secondly, bubbles will cause local dynamic changes in the image, thereby affecting the prediction accuracy of the model. In recent years, some researchers have proposed some methods to alleviate bubble interference. Image filtering and denoising techniques are used to reduce the interference of bubbles on image features. However, such algorithms are usually used for images with more obvious themes and richer feature information color channels. Applying them to the concentration recognition of microalgae solutions with more uniform colors and less obvious bubble colors, the robustness of the model still needs further optimization and evaluation.

[0020] The existing technology is specifically shown as follows: (1) Microalgae identification and biomass prediction based on convolutional neural network algorithm: The article explains the feasibility of convolutional neural network in fitting the concentration of microalgae solution, but does not consider the bubble interference. The introduction of bubbles will have a huge interference on the feature extraction of the model, resulting in a decrease in recognition accuracy.

[0021] (2) Microalgae concentration prediction method and system based on big data: This patent proposes a method to establish a prediction model based on historical sampling and environmental data. However, this method of using data to build a prediction model requires a large amount of data support, and for certain specific culture environments, parameter measurement requires the support of a large number of sensors. Prediction using machine vision does not require a large amount of data and is easy to obtain.

[0022] (3) A bubble recognition image processing method for bubble flow in a gas-liquid reactor: This patent proposes a method for shadow monitoring and bubble compensation using background subtraction. However, this method requires the background and the image with bubbles to be differentiated, which is easily affected by the inconsistent matching of the original image pixels, resulting in a decrease in accuracy. In addition, it has certain requirements for the contrast between the background and the bubbles.

[0023] (4)FFDNet: Toward a Fast and Flexible Solution for CNN based Image Denoising: Most of the existing denoising methods mainly focus on noise removal for object detection and recognition applications. The target objects in the data sources detected by these models generally have rich colors. However, in the field of microalgae concentration prediction, the target objects to be detected have relatively uniform and single colors, and the introduced bubbles will have different color states with the change of algae solution concentration. Therefore, the effect of introducing an additional denoising module to remove bubbles is not obvious. To overcome the above problems, the present invention proposes an image bubble noise removal method based on dual-channel features. Refer to Figure 3 , which specifically includes: Construct an improved neural network model based on CNN, including a first channel, a second channel, multiple sequentially connected feature processing units, a mask module, and a fully connected layer; the first channel is used to extract color features related to image recognition, and the second channel is used to extract bubble interference features; the feature processing unit includes four convolutional layers and a feature superposition extraction module, and the feature superposition extraction module is used to superpose the bubble interference features and the color features related to image recognition and then extract them; the mask module is used to perform mask processing on the output of the feature processing unit, and the fully connected layer is used to process the unfolded masked features and output the target recognition result; Obtain pictures without interference and with annotation information and some pictures with bubbles as interference-free images and images with bubble interference, construct a training set, and input them into the first channel and the second channel respectively to train the improved neural network model to obtain a trained improved neural network model; Obtain the image to be processed and input it into the trained improved neural network model to output the corresponding target recognition result.

[0024] The present invention will be further described below in conjunction with the drawings and embodiments.

[0025] Example 1: This example takes microalgae pictures as an example. To overcome the shortcomings of the prior art, a method for removing bubble noise in uniform color images by performing mask processing based on the overlap of dual-channel feature maps is proposed. The specific content is as follows: 1. Dataset construction: Place the collected microalgae concentrations without interference and with concentration annotations and some microalgae pictures with bubbles in two different folders respectively. And make remarks on the folders.

[0026] 2. Model input channel design: Modify the input layer structure of the adopted neural network model to expand the single-channel input to a dual-channel input. The above two types of data, interference-free images and images with bubble interference, are input into the model through different input channels respectively to effectively distinguish the data sources.

[0027] 3. Feature extraction module optimization: After the initial convolution layers of the interference-free images and images with bubble interference in the model, add a feature map superposition and extraction module. It is used to extract bubble interference features and color features related to microalgae concentration, and realize the extraction of bubble and concentration features from the feature maps respectively.

[0028] As the model comes into contact with real interference and background, it will be trained, so that the convolution kernels in the model architecture have a certain recognition ability for color features and bubble interference features.

[0029] 4. Introduction of masking operation: After the feature extraction module, add a masking operation to mask the extracted bubble interference features and reduce their weight influence in the model prediction process. Effectively reduce the negative impact of bubble interference on the concentration prediction accuracy.

[0030] 5. Model structure adjustment and optimization: In the model structure, insert a feature extraction and masking module as a feature processing unit every four convolution layers to continuously suppress bubble interference. At the same time, set the number of nodes in the final output layer of the model to 1, and convert the model from a classification model to a regression model to achieve accurate fitting and continuous numerical prediction of microalgae concentration.

[0031] 6. Model training and generalization ability improvement: Train the improved model to significantly improve its generalization ability and prediction accuracy for the working conditions with bubble interference without reducing the concentration prediction accuracy under normal working conditions.

[0032] Table 1 Performance of the basic model and two improvement methods on two unfamiliar datasets

[0033] Beneficial effects achieved: (1) Verify the basic model, the attention mechanism mentioned in the existing literature, and the two models formed by adding the method proposed in the present invention to the basic model on two unfamiliar datasets. The verification results are shown in Table 1. The two unfamiliar datasets include: Dataset a is the pictures with bubbles in the concentration that the model has come into contact with during training, and Dataset b is the photos with bubbles in the concentration that the model has not come into contact with during training. From the various evaluation parameters of the model, regardless of whether the model has over-learned the concentration gradient in the unfamiliar dataset, the superposition of feature maps and masking processing shows higher prediction accuracy.

[0034] (2)The heat map after visualizing the attention of the model to the original image area is as Figure 1 shown, where Figure 1 a is the heat map of the original image, Figure 1 b is the heat map of the attention degree of the basic model to the original image, Figure 1 c is the heat map of the attention degree of the model after introducing the attention mechanism to the original image, Figure 1 d is the heat map of the attention degree of the model of the present invention to the original image. By comparing the heat maps of the original image and each model, it can be seen that the masked model has a higher attention degree to the non-bubble area, that is, the area containing more concentration information, which is also the reason for the higher accuracy of the model.

[0035] (3)Draw the heat map of the visualization model attention degree after the initial layer and the four pooling layers in the model architecture are processed, as Figure 2 shown, where Figure 2 a is the heat map of the original image, Figure 2 b is the heat map after the initial layer is processed, Figure 2 c is the heat map after the first pooling layer is processed, Figure 2 d is the heat map after the second pooling layer is processed, Figure 2 e is the heat map after the third pooling layer is processed, Figure 2 f is the heat map after the fourth pooling layer is processed. It can be seen that as the number of layers deepens, the attention degree to the bubble area gradually decreases, and the masking module plays a role.

[0036] Example 2: The microalgae concentration prediction method based on masking processing by overlapping dual-channel feature maps to remove bubble noise in uniform color images specifically includes the following steps First is the dataset construction, including the interference-free image set and the image set with bubble interference: Collect 1000 microalgae concentration images without bubble interference, and each image has an accurate concentration annotation, ranging from 0.1 g / L to 1.0 g / L, with an interval of 0.1 g / L.

[0037] Collect 500 microalgae concentration images with different numbers and sizes of bubbles, also with concentration annotations, covering the same concentration range.

[0038] Modify the input layer of the neural network model to expand the single-channel input to a dual-channel input; the interference-free image is input through channel A, and the image with bubble interference is input through channel B.

[0039] Add a feature map superposition and extraction module after the initial convolution layer to extract bubble interference features and microalgae concentration-related color features from the feature map respectively.

[0040] Add a masking operation after the feature extraction module. Mask the extracted bubble interference features to reduce their weight to 10% of the original.

[0041] Insert a feature extraction and masking module as a feature processing unit every four convolutional layers in the model. Set the final output layer node to 1 to convert it into a regression model. Use the interference-free image set and the image set with bubble interference to jointly train the model. During the training process, the ratio of interference-free images to images with bubble interference is 2:1. Adopt cross-validation and early stopping strategies to prevent overfitting. The training cycle is 100 epochs, the initial learning rate is 0.001, and it decays by half every 20 epochs.

[0042] Concentration prediction accuracy under normal working conditions: On the interference-free image set, the mean absolute error of the model's predicted concentration is 0.02 g / L, and the accuracy rate reaches 95%.

[0043] Generalization ability under the working condition with bubble interference: On the image set with bubble interference, the mean absolute error of the model's predicted concentration is 0.04 g / L, and the accuracy rate is increased to 90%, which is 15 percentage points higher than the model without using the overlapping masking process of dual-channel feature maps.

[0044] Overall performance of the model: Combining the two data sets, the overall prediction accuracy rate of the model reaches 92%, meeting the requirements of real-time monitoring of microalgae concentration.

[0045] From the above embodiments, it can be seen that the proposed method effectively reduces the influence of bubble interference on the prediction of microalgae concentration, and improves the generalization ability and prediction accuracy of the model under complex working conditions.

[0046] Embodiment 3: A method for removing bubble noise in uniform color images based on overlapping masking of dual-channel feature maps specifically includes the following: Collect and annotate microalgae concentration data, a total of 1000 pictures.

[0047] Put 520 interference-free microalgae concentration pictures with the concentration annotation range of 100 - 1000 mg / L in the folder "bubble-free", and put 480 microalgae concentration pictures with bubbles and the same concentration annotation range in the folder "with bubbles".

[0048] The folders are respectively marked as "bubble-free" and "with bubbles", and the naming format of each picture is "concentration_XX mg / L_sequence number.jpg" for subsequent tracking and analysis.

[0049] Based on the ResNet18 architecture, modify the input layer structure to expand from single-channel input to dual-channel input. The first channel inputs "bubble-free" pictures, and the second channel inputs "with bubbles" pictures.

[0050] Add a feature map superposition and extraction module after the initial convolutional layer of the model.

[0051] Add a masking operation after the feature extraction module. The bubble interference weight calculated by the masking matrix will affect the prediction and be updated in each iteration. Insert a feature extraction and masking module as a feature processing unit after every four convolutional layers. Set the number of nodes in the final output layer to 1, convert the model from a classification model to a regression model, and output the continuous value of the microalgae concentration.

[0052] Use the Adam optimizer, set the initial learning rate to 0.001, and adopt a training strategy with a batch size of 32. Train for 50 epochs. By means of cross-validation, divide the data into 80% for the training set and 20% for the test set to verify the performance of the model. On the test set containing bubbles, evaluate the prediction accuracy of the model, and use the root mean square error RMSE as the index. The goal is to control the RMSE within 10 mg / L.

[0053] After the training is completed, evaluate the model and find that: On the "bubble-free" image set, the prediction error of the model is 5 mg / L, RMSE.

[0054] On the "bubble-containing" image set, the prediction error of the model is 8 mg / L, RMSE, showing strong robustness.

[0055] Further observe the prediction effect of the model in the case of more bubbles. The prediction concentration error of 80% is controlled within 10 mg / L, and the concentration of 10% of the samples exceeds this range.

[0056] Through the above implementation, a model that can effectively handle bubble interference and predict the microalgae concentration is constructed, providing precise technical support for microalgae cultivation, monitoring and related research.

[0057] It should be noted that in the present invention, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0058] The above embodiments are only illustrative examples of the present invention and do not constitute a limitation on the protection scope of the present invention. Any design identical or similar to the present invention falls within the protection scope of the present invention.

Claims

1. An image bubble noise removal method based on dual-channel features, characterized in that, Including the following steps: Construct an improved neural network model based on CNN, including a first channel, a second channel, a plurality of sequentially connected feature processing units, a mask module, and a fully connected layer; the first channel is used to extract color features related to image recognition, and the second channel is used to extract bubble interference features of the image; the feature processing unit includes four convolutional layers and a feature superposition extraction module, and the feature superposition extraction module is used to superpose the bubble interference features and the color features related to image recognition and then extract them; the mask module is used to perform mask processing on the output of the feature processing unit, and the fully connected layer is used to process the unfolded masked features and output the target recognition result; Obtain pictures without interference and with annotation information and some pictures with bubbles as interference-free images and images with bubble interference, construct a training set, and input them into the first channel and the second channel respectively to train the improved neural network model to obtain the trained improved neural network model; Obtain the image to be processed and input it into the trained improved neural network model to output the corresponding target recognition result.

2. The method for removing image bubble noise based on dual-channel features according to claim 1, wherein In each of the feature processing units, the convolutional kernels of each layer of convolutional layer are used for extracting bubble interference features and color features related to image recognition.

3. The method for removing image bubble noise based on dual-channel features according to claim 1, characterized in that The step of obtaining the image to be processed and inputting it into the trained improved neural network model to output the corresponding target recognition result includes the following steps: Input the image to be processed into the improved neural network model to obtain color features related to image recognition and bubble interference features; Input the color features related to image recognition and the extracted bubble interference features into a plurality of sequentially connected feature processing units for feature extraction, where: In each of the feature processing units, use four convolutional layers to perform convolutional processing on the input in sequence, use a filter to extract features from the convolved input to obtain the re-extracted convolutional input features, add the re-extracted convolutional input features element by element, obtain the superimposed features and then extract features to obtain the intermediate enhanced bubble interference features; Generate a low-value matrix with the same size as the bubble interference features according to the intermediate enhanced bubble interference features; Cover the generated low-value matrix to the bubble interference features for mask processing to obtain the masked features, and perform superimposed feature processing on the masked features and the color features related to image recognition to remove bubble noise and obtain the result of removing bubble noise; Use the fully connected layer to expand the result of removing bubble noise into a vector form, set the output node of the fully connected layer to 1, and output the corresponding target recognition result.

4. A system for an image bubble noise removal method based on dual-channel features, characterized in that, Including: The model improvement module is used to construct an improved neural network model based on CNN, including a first channel, a second channel, multiple sequentially connected feature processing units, a mask module, and a fully connected layer; the first channel is used to extract color features related to image recognition, and the second channel is used to extract bubble interference features; the feature processing unit includes four convolutional layers and a feature superposition extraction module, and the feature superposition extraction module is used to superpose the bubble interference features and the color features related to image recognition and then extract them; the mask module is used to perform mask processing on the output of the feature processing unit, and the fully connected layer is used to process the unfolded masked features and output the target recognition result; The model training module is used to obtain pictures without interference and with annotation information and some pictures with bubbles as interference-free images and images with bubble interference, construct a training set, and input them into the first channel and the second channel respectively to train the improved neural network model and obtain the trained improved neural network model; The bubble noise removal module obtains the image to be processed, inputs it into the trained improved neural network model, and outputs the corresponding target recognition result.

5. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and when the computer program is executed by a processor, it implements the method according to any one of claims 1 to 3 above.

6. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the method according to any one of claims 1 to 3 above.

Images