A method and device for intelligent driving data mining, and a computer program product

Abstract

The present application relates to an intelligent driving data mining method and its device, and a computer program product, including: obtaining multiple frames of sample images, and extracting image feature vectors and text feature vectors of the multiple frames of sample images; using the image feature vectors and the text feature vectors as training samples to train a cross-modal model; during the training process, the cross-modal model learns the correlation between the image feature vectors and the text feature vectors; obtaining a driving scene image, extracting the image feature vectors of the driving scene image and inputting them into the trained cross-modal model for processing, and obtaining scene description information based on the text feature vectors output by the cross-modal model; and constructing an intelligent driving scene knowledge base based on the scene description information. Through this application, the comprehensiveness and accuracy of the feature representation of the driving scene image during the mining process can be improved, and more efficient and accurate intelligent driving data mining can be achieved based on this.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent driving data mining, and specifically to an intelligent driving data mining method and its device, and a computer program product. Background Art

[0002] Intelligent driving data mining refers to the process of using artificial intelligence and machine learning technologies to process and analyze large amounts of data collected by intelligent driving vehicles to extract valuable information and knowledge. The large amount of data collected by vehicles usually includes vehicle sensor data (such as cameras, radars, lidars, etc.), vehicle status information, environmental perception data, driving behavior records, etc. Among them, image-based intelligent driving data mining is particularly important.

[0003] Image-based intelligent driving data mining refers to the use of image processing and machine learning techniques to analyze and understand image data captured by intelligent driving vehicles in order to extract valuable information and knowledge from them. This information and knowledge can be used to improve the performance and safety of intelligent driving systems, such as assisting intelligent driving vehicles in object detection, scene understanding, behavior prediction, and decision making.

[0004] In the process of image-based intelligent driving data mining, comprehensiveness and accuracy are one of the challenges facing current technology when processing driving scene images. Driving scene images contain a huge amount of information and are highly complex. How to effectively represent and extract image features is crucial for subsequent object detection, scene understanding and other tasks. Summary of the invention

[0005] The purpose of this application is to propose an intelligent driving data mining method and its device, and a computer program product to improve the comprehensiveness and accuracy of the feature representation of driving scene images during the mining process, and based on this, to achieve more efficient and accurate intelligent driving data mining.

[0006] To achieve the above-mentioned purpose, according to the first aspect of the present application, a method for intelligent driving data mining is provided, the method comprising:

[0007] Acquire multiple frames of sample images, and extract image feature vectors and text feature vectors from the multiple frames of sample images;

[0008] The image feature vector and the text feature vector are used as training samples to train a cross-modal model; during the training process, the cross-modal model learns the correlation relationship between the image feature vector and the text feature vector;

[0009] Acquire a driving scene image, extract an image feature vector of the driving scene image and input the image feature vector into a trained cross-modal model for processing, and obtain scene description information according to a text feature vector output by the cross-modal model;

[0010] Construct an intelligent driving scenario knowledge base based on the scenario description information.

[0011] According to the method described in the first aspect of the present application, it is proposed to obtain scene description information based on a cross-modal model. The cross-modal model learns the mapping relationship between image features and text features through training, and can process and associate two different types of data, image and text, and can convert images into meaningful text descriptions in subsequent processing. Based on the cross-modal model, the driving scene images to be mined can be continuously input into the cross-modal model for processing, and the text feature vectors output by the cross-modal model are used as scene description information. In this way, not only can objects and scenes in the image be identified, but also these objects and scenes can be described in natural language, which improves the comprehensiveness and accuracy of the feature representation of the driving scene image. Finally, based on the scene description information, an intelligent driving scene knowledge base containing multiple driving scenes is constructed or updated. This intelligent driving scene knowledge base can help the intelligent driving system better understand the surrounding environment and make decisions adapted to different scenes.

[0012] According to a second aspect of the present application, an intelligent driving data mining device is provided, comprising a module for executing the method according to the first aspect.

[0013] According to a third aspect of the present application, there is provided an intelligent driving data mining device, comprising:

[0014] A communication interface, used to communicate with other electronic devices;

[0015] a memory for storing computer program instructions;

[0016] A processor is used to execute the computer program instructions to support the intelligent driving data mining device according to the first aspect to implement the method according to the first aspect.

[0017] According to a fourth aspect of the present application, a computer program product is provided, comprising computer program instructions, wherein the computer program instructions instruct a computer device to perform operations corresponding to the method according to the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 This is a flow chart of an intelligent driving data mining method in an embodiment of the present application.

[0020] Figure 2 This is a schematic diagram of a smart driving data mining method in an embodiment of the present application. DETAILED DESCRIPTION

[0021] The detailed description of the drawings is intended as an illustration of the current embodiment of the present application, and is not intended to represent the only form in which the present application can be implemented. It should be understood that the same or equivalent functions can be accomplished by different embodiments intended to be included in the spirit and scope of the present application.

[0022] See also Figure 1 An embodiment of the present application provides a method for intelligent driving data mining, the method comprising:

[0023] Step S10, acquiring multiple frames of sample images, and extracting image feature vectors and text feature vectors from the multiple frames of sample images.

[0024] Specifically, step S10 describes the process of preparing training samples from image data collected from intelligent driving vehicles, which includes two main parts: obtaining multiple frames of sample images and extracting image feature vectors and text feature vectors.

[0025] Smart driving vehicles are equipped with cameras. During driving, the cameras continuously capture images of the vehicle's surroundings. These images exist in the form of continuous frames, and each frame is a snapshot of the scene seen by the vehicle at a specific moment. In order to train the model, a large number of these image frames need to be collected as samples.

[0026] An image feature vector is a mathematical representation of the image content. It contains key information about the image, such as color, shape, texture, etc. In intelligent driving, image feature vectors can be used to identify and classify objects in the image, such as pedestrians, vehicles, traffic signs, etc.

[0027] A text feature vector is a mathematical representation of text data, which contains the semantic information of the text. In this embodiment, the text feature vector is a description of the image content. For example, the description of the image scene by humans is converted into a text feature vector through natural language processing (NLP) technology. These descriptions include, for example, the names of objects in the scene, the behavior of the objects, the scene, and other information.

[0028] Step S20, using the image feature vector and the text feature vector as training samples to train a cross-modal model; during the training process, the cross-modal model learns the correlation relationship between the image feature vector and the text feature vector.

[0029] Specifically, the cross-modal model is a model that can process and associate different types of data (such as images and text). In this embodiment, the goal of the cross-modal model is to understand the content of the image and to be able to describe the content in natural language. In step S20, the image feature vector and text feature vector prepared in step S10 are used to train the cross-modal model. In this process, the model learns the intrinsic connection between image and text data so that new image data can be converted into corresponding text descriptions in the future.

[0030] Specifically, during the training process, the model will try to find the mapping relationship between the image feature vector and the text feature vector, that is, the model will learn how to "predict" the corresponding text description from the image features, or vice versa, how to "imagine" the image content from the text description. Through this training process, the cross-modal model can gradually establish the association between images and texts, so that it can generate accurate text descriptions when faced with new, unseen images.

[0031] Step S30, acquiring a driving scene image, extracting an image feature vector of the driving scene image and inputting the image feature vector into a trained cross-modal model for processing, and obtaining scene description information according to a text feature vector output by the cross-modal model.

[0032] Specifically, after the cross-modal model is trained, the trained cross-modal model can be used to process actual driving scene images and generate scene description information. Specifically, when the intelligent driving vehicle is driving, the on-board camera will capture images of the vehicle's surroundings in real time. These images represent the current driving scene and are input data that the intelligent driving system needs to understand and analyze. Similar to the training phase, these real-time captured driving scene images need to use image processing technology to extract image feature vectors. These feature vectors are mathematical representations of the image content and contain important information in the image. The extracted image feature vectors are then input into the trained cross-modal model. The cross-modal model processes these feature vectors using the image-to-text mapping relationship learned during the training process, and outputs the corresponding text feature vectors. These text feature vectors represent the cross-modal model's understanding and description of the input image content. The text feature vectors output by the cross-modal model are converted into scene description information, which is a natural language description, such as "a car is driving on the road" or "a pedestrian is crossing the road". These descriptions provide the intelligent driving system with detailed information about the current driving scene, helping the system make more accurate and safe driving decisions.

[0033] Step S40: construct an intelligent driving scenario knowledge base based on the scenario description information.

[0034] Specifically, the scene description information output by the cross-modal model is organized to build an intelligent driving scene knowledge base to make it easy to query and use. This knowledge base is a database that stores and manages the knowledge and information required by the intelligent driving system. It can help the system better identify and understand different driving scenarios, thereby improving the accuracy and efficiency of decision-making.

[0035] The construction process of the intelligent driving scene knowledge base may include but is not limited to: organizing the scene description information according to different categories, such as classification by object type, behavior, etc.; for rapid retrieval, it is necessary to index the information in the knowledge base so that relevant scene descriptions can be quickly found when needed; integrating new scene description information with existing information in the knowledge base to enrich and improve the content of the knowledge base; and, as intelligent driving vehicles continue to collect new data, the knowledge base needs to be updated regularly to include the latest scene information.

[0036] In summary, the method of this embodiment proposes to obtain scene description information based on a cross-modal model. The cross-modal model learns the mapping relationship between image features and text features through training, and can process and associate two different types of data, image and text, and can convert images into meaningful text descriptions in subsequent processing. Based on the cross-modal model, the driving scene images to be mined can be continuously input into the cross-modal model for processing, and the text feature vectors output by the cross-modal model are used as scene description information, which can not only identify objects and scenes in the image, but also describe these objects and scenes in natural language, thereby improving the comprehensiveness and accuracy of the feature representation of the driving scene image. Finally, based on the scene description information, an intelligent driving scene knowledge base containing multiple driving scenes is constructed or updated. This intelligent driving scene knowledge base can help the intelligent driving system better understand the surrounding environment and make decisions adapted to different scenes.

[0037] In some embodiments, the step S10 of acquiring multiple frames of sample images further comprises:

[0038] A video sequence related to a driving scene is obtained, and key frames are extracted from the video sequence as sample images according to similarities between video frames in the video sequence.

[0039] Specifically, a video sequence is composed of continuous video frames, each of which is a snapshot of the driving scene. In order to effectively select representative sample images for training, key frames can be extracted by calculating the similarity between video frames. The similarity here can be calculated by the structural similarity index (SSIM) algorithm. The structural similarity index (SSIM) is a measure of the similarity between two images, which takes into account changes in brightness, contrast, and structure. In video processing, the SSIM algorithm can be used to compare the similarity between consecutive frames in a video sequence. By calculating the SSIM value between frames, it can be determined which frames are most sensitive to changes in the scene. These frames usually contain the most important information.

[0040] Specifically, the lower the SSIM value, the greater the difference between the two frames, which may contain important scene changes. According to the calculation results of the SSIM algorithm, those video frames with lower similarity are selected as key frames. These key frames represent significant scene changes in the video sequence, such as vehicle turns, pedestrians crossing the road, traffic light changes, etc., so they are ideal samples for training models.

[0041] In this way, the most representative and informative sample images can be efficiently selected from a large amount of video data for training the cross-modal model, which helps improve the performance of the model because it can learn more diverse scene changes and thus better understand and describe driving scenarios in practical applications.

[0042] In some embodiments, the extraction of image feature vectors of multiple frames of sample images in step S20 further includes:

[0043] The multiple frames of sample images are input into a pre-trained Dinov2 model for processing to obtain image feature vectors of the multiple frames of sample images.

[0044] Specifically, the Dinov2 (DiNoVe II) model is a deep learning model used for image recognition and classification tasks that can extract high-level features from images. A pre-trained model means that the model has been trained on a large image dataset and has learned a general image feature representation.

[0045] The multiple frames of sample images obtained in step S10 are input into the pre-trained Dinov2 model. The Dinov2 model processes the input sample images and extracts high-level features of the images through its internal convolutional neural network (CNN) structure. These feature vectors are abstract representations of the image content and contain key information of the image, such as shape, texture, color, etc. After being processed by the Dinov2 model, each sample image will output an image feature vector.

[0046] It should be noted that the method of this embodiment uses the Dinov2 model to extract the image feature vector of the sample image. Compared with the method of extracting image feature vectors based on hash codes or Reset series models commonly used in the industry, the Dinov2 model is a self-supervised advanced visual model based on contrastive learning. The higher-dimensional embedding vector features extracted by Dinov2 have richer image visual coding information. Compared with the hash code and Reset series models that extract image embedding features, this model contains richer image information.

[0047] In some embodiments, the extraction of text feature vectors of multiple frames of sample images in step S20 further includes:

[0048] Step S201, detecting multiple targets in the multiple frames of sample images based on a target detection model;

[0049] Specifically, the method of this embodiment uses a target detection model to identify targets in an image. These targets include static targets and dynamic targets, such as vehicles, pedestrians, lane lines, crosswalks, etc.

[0050] Step S202, identifying the relationship between the targets based on the machine learning model;

[0051] Specifically, after detecting the objects, the relationships between these objects are analyzed. For example, a vehicle may be following another vehicle, or a pedestrian may be crossing the road. Machine learning models based on rules and clustering algorithms can help identify and understand these relationships.

[0052] Step S203, generating a semantic graph according to the multiple targets and the relationships between the targets; wherein the multiple targets in the semantic graph are nodes, and the relationships between the targets are used as connecting edges between the nodes;

[0053] Specifically, a semantic graph is a graphical representation in which each object is regarded as a node and the relationships between objects are connected as edges between nodes. Such a graphical representation helps to understand and represent the complex interactions of objects in the scene.

[0054] Step S204: converting the semantic graph into a text feature vector based on a natural language model.

[0055] Specifically, semantic graphs can be converted into text form through natural language models, which are artificial intelligence models that can understand and generate human language. These models are usually based on machine learning algorithms, especially deep learning, and are trained to process text data, understand patterns and structures in language, and generate coherent and meaningful text output. After training, natural language models can convert semantic graphs into natural language descriptions, such as "a car is driving on the road, followed by another car", which are used as text feature vectors for training cross-modal models.

[0056] In some embodiments, the step S20 uses the image feature vector and the text feature vector as training samples to train a cross-modal model, further comprising:

[0057] During the training process, the cross-modal model maps the image feature vectors and text feature vectors of the multiple frames of sample images to the same feature space, pairs the image feature vectors and text feature vectors in the same feature space, and learns the correlation between the paired image feature vectors and text feature vectors.

[0058] Specifically, the goal of the cross-modal model is to transform data from different modalities (images and text) into the same feature space, that is, the cross-modal model needs to find a common representation so that image features and text features can be compared and associated in the same feature space. During the training process, the cross-modal model will pair the image feature vectors of multiple sample frames with the corresponding text feature vectors. These paired feature vectors are the basis for cross-modal model learning because they represent the direct association between image content and text description. By learning the intrinsic connection between paired feature vectors, the cross-modal model can learn the semantic association between cross-modalities, understand how the visual content in the image corresponds to the semantic information in the text description, and improve the comprehensiveness and accuracy of image data representation.

[0059] In this way, the cross-modal model can not only understand the content of the image, but also describe it in natural language, which is very important for intelligent driving systems. For example, an intelligent driving system can not only know that there is an obstacle ahead, but also know that the obstacle is a pedestrian, and predict the possible behavior of the pedestrian, so as to make safer and more reasonable driving decisions.

[0060] In some embodiments, the step S50 further comprises:

[0061] Acquire driving common sense information, and build an intelligent driving scenario knowledge base of the RAG system based on the scenario description information and the driving common sense information.

[0062] Specifically, driving common sense information refers to general knowledge and rules related to driving, such as traffic regulations, the meaning of road signs, driving habits, safety rules, etc. This information is crucial for intelligent driving systems because it helps the system better understand and comply with traffic rules, thereby making safer driving decisions.

[0063] The RAG system, or Retrieval-Augmented Generation system, is an AI technology that combines retrieval models and generation models. It is mainly used to improve the quality and relevance of text generation, especially when dealing with specific areas or problems that require the latest information. Scene description information is extracted from driving scene images through a cross-modal model. They represent the real-time driving environment. Continuously combining this real-time information with the latest driving common sense information can build a more comprehensive and rich knowledge base of the RAG system. This knowledge base not only includes the current driving scene, but also incorporates the basic rules and common sense of driving. Intelligent driving systems can apply the RAG system to achieve intelligent driving scene retrieval and intelligent decision-making when dealing with complex situations.

[0064] In some embodiments, the RAG system includes a retrieval component, a generation component, and the intelligent driving scenario knowledge base;

[0065] The retrieval component is used to receive retrieval information, and retrieve the intelligent driving scenario knowledge base according to the retrieval information to obtain a retrieval result;

[0066] The generating component is used to generate and output a driving decision according to the retrieval result.

[0067] Specifically, the retrieval component receives the information to be retrieved, such as the current vehicle environment, road conditions, traffic rules, etc. Then, it searches the intelligent driving scenario knowledge base based on this information to find relevant information or data. These retrieval results may be the best driving decisions, safe driving guidelines or other useful information in similar scenarios in the past.

[0068] After receiving the results provided by the retrieval component, the generation component will use this information to generate driving decisions for the current intelligent driving scenario. These decisions may include how to accelerate, decelerate, change lanes, stop, etc. The purpose of the generation component is to ensure that the vehicle can make safe and reasonable driving decisions based on the retrieved knowledge and information.

[0069] The intelligent driving scenario knowledge base is constructed based on step S50 and stores a large amount of knowledge and information related to intelligent driving. The role of the knowledge base is to provide rich information resources for the retrieval component to help the generation component make better driving decisions.

[0070] like Figure 2 The following is a schematic diagram of the technical principle of the method of this embodiment. Figure 2 The training of the cross-modal model, the construction of the intelligent driving scenario knowledge base, and the use of the intelligent driving scenario knowledge base are shown. Preprocessing the video sequence can obtain a series of key frame images, and high-dimensional image features of these key frame images are extracted to obtain image feature vectors. Image semantic understanding is performed to obtain text feature vectors, thereby obtaining the training sample <image, text> embedding of the cross-modal model; the actual driving scene image is used as a query input to the trained cross-modal model to obtain scene description information, and the intelligent driving scenario knowledge base of the RAG system can be constructed based on the scene description information. The RAG system can respond to the user's retrieval request and output the corresponding driving decision or other driving-related information.

[0071] Another embodiment of the present application also provides an intelligent driving data mining device, including a module for executing the intelligent driving data mining method described in the above embodiment. This device can be a hardware device, a software device, or a combination of software and hardware.

[0072] Specifically, this device includes a series of modules that can execute each step in the intelligent driving data mining method described above, for example, obtaining multiple frames of sample images and extracting image feature vectors and text feature vectors of the multiple frames of sample images; using the image feature vectors and the text feature vectors as training samples to train a cross-modal model; during the training process, the cross-modal model learns the association between the image feature vectors and the text feature vectors; obtaining a driving scene image, extracting the image feature vectors of the driving scene image and inputting them into the trained cross-modal model for processing, and obtaining scene description information based on the text feature vectors output by the cross-modal model; and constructing an intelligent driving scene knowledge base based on the scene description information.

[0073] Another embodiment of the present application also provides another intelligent driving data mining device, including:

[0074] A communication interface, used to communicate with other electronic devices;

[0075] a memory for storing computer program instructions;

[0076] The processor is used to execute the computer program instructions to support the intelligent driving data mining device to implement the method described in the above embodiment.

[0077] In this embodiment, the memory mainly includes a program storage area and a data storage area, wherein the program storage area can store an operating device, an application required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, and a flash card (FlashCard), etc., or the memory can also be other volatile solid-state storage devices.

[0078] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor is the control center of the intelligent driving data mining device, and uses various interfaces and lines to connect the various parts of the intelligent driving data mining device.

[0079] An embodiment of the present application also provides a computer program product, including computer program instructions, wherein the computer program instructions instruct a computer device to perform operations corresponding to the method described in the above embodiment.

[0080] Specifically, the computer program product includes a series of computer program instructions that can instruct a computer device to execute the intelligent driving data mining method described in this application. These instructions are codes written in a computer program that define how to perform specific operations. In this embodiment, these instructions are used to execute the intelligent driving data mining method of the above embodiment.

[0081] These program instructions are designed to be loaded onto a computer device and instruct the device to perform specific operations, which refer to the various steps in the intelligent driving data mining method described in the above embodiments, for example, obtaining multiple frames of sample images and extracting image feature vectors and text feature vectors of the multiple frames of sample images; using the image feature vectors and the text feature vectors as training samples to train a cross-modal model; during the training process, the cross-modal model learns the association between the image feature vectors and the text feature vectors; obtaining a driving scene image, extracting the image feature vectors of the driving scene image and inputting them into the trained cross-modal model for processing, and obtaining scene description information based on the text feature vectors output by the cross-modal model; and constructing an intelligent driving scene knowledge base based on the scene description information.

[0082] In this way, the computer program product provides a complete software solution that can run on various computer devices to implement the intelligent driving data mining method of the above embodiment.

[0083] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for intelligent driving data mining, characterized in that: The method comprises: Acquire multiple frames of sample images, and extract image feature vectors and text feature vectors from the multiple frames of sample images; The image feature vector and the text feature vector are used as training samples to train a cross-modal model; during the training process, the cross-modal model learns the correlation relationship between the image feature vector and the text feature vector; Acquire a driving scene image, extract an image feature vector of the driving scene image and input the image feature vector into a trained cross-modal model for processing, and obtain scene description information according to a text feature vector output by the cross-modal model; Building an intelligent driving scenario knowledge base according to the scenario description information; The extraction of text feature vectors of the multiple frames of sample images further includes: Detecting multiple targets in the multiple frames of sample images based on a target detection model; Identify relationships between objects based on machine learning models; Generate a semantic graph according to the multiple targets and the relationships between the targets; wherein the multiple targets in the semantic graph are nodes, and the relationships between the targets are used as connecting edges between the nodes; The semantic graph is converted into a text feature vector based on a natural language model.

2. The method according to claim 1, characterized in that The acquiring of multiple frames of sample images further comprises: A video sequence related to a driving scene is obtained, and key frames are extracted from the video sequence as sample images according to similarities between video frames in the video sequence.

3. The method according to claim 2, characterized in that The extraction of image feature vectors of the multiple frames of sample images further includes: The multiple frames of sample images are input into a pre-trained Dinov2 model for processing to obtain image feature vectors of the multiple frames of sample images.

4. The method according to claim 1, characterized in that: The using the image feature vector and the text feature vector as training samples to train the cross-modal model further includes: During the training process, the cross-modal model maps the image feature vectors and text feature vectors of the multiple frames of sample images to the same feature space, pairs the image feature vectors and text feature vectors in the same feature space, and learns the correlation between the paired image feature vectors and text feature vectors.

5. The method according to claim 1, characterized in that The method further comprises: Acquire driving common sense information, and build an intelligent driving scenario knowledge base of the RAG system based on the scenario description information and the driving common sense information.

6. The method according to claim 5, characterized in that The RAG system includes a retrieval component, a generation component and the intelligent driving scenario knowledge base; The retrieval component is used to receive retrieval information, and retrieve the intelligent driving scenario knowledge base according to the retrieval information to obtain a retrieval result; The generating component is used to generate and output a driving decision according to the retrieval result.

7. An intelligent driving data mining device, characterized in that: Comprising modules for executing the method according to any one of claims 1 to 6.

8. An intelligent driving data mining device, characterized in that: include: A communication interface, used to communicate with other electronic devices; a memory for storing computer program instructions; A processor, configured to execute the computer program instructions to support the apparatus to implement the method according to any one of claims 1 to 6.

9. A computer program product, characterized in that The method comprises computer program instructions, wherein the computer program instructions instruct a computer device to execute operations corresponding to the method according to any one of claims 1 to 6.

Images