Wireless private network coverage map generation method, device, equipment, and storage medium

Abstract

This application discloses a method, device, apparatus, and storage medium for generating a wireless private network signal coverage map, belonging to the field of computer technology. The embodiment of this application uses an artificial intelligence model to flexibly, accurately, and efficiently solve the problem that wireless network planning usually depends on the design schemes of professional teams or the output of professional simulation tools. It helps to promote the popularization and development of private network construction. The embodiment of this application can process input data to obtain a three-dimensional structure diagram of the interior of a building, and determine the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram through the target feature vector obtained from the input data, to generate a wireless private network signal coverage map inside the building. The entire process is automated, and users only need to input building data or demand data to automatically implement wireless private network planning, which reduces the threshold for network planning and design schemes, improves the efficiency and versatility of wireless private network planning, and reduces its cost.

Description

[0001] Wireless Private Network Coverage Map Generation Method, Device, Equipment, and Storage Medium

[0002] Technical Field

[0003] This application relates to the field of computer technology, particularly to a method, device, and storage medium for generating a wireless private network signal coverage map.

[0004] Background Technology

[0005] With the widespread deployment of 4G and 5G cellular wireless technologies, their applications are progressively extending into the core areas of various industries. An increasing number of industries arc adopting wireless private networks to support their business operations, thereby imposing higher demands on network design and planning.

[0006] However, current wireless network planning ty pically relies on design schemes from professional teams or outputs from specialized simulation tools. For small and medium-sized enterprises, this approach is not only costly but also has a relatively high technical barrier.

[0007] Technical Disclosure

[0008] This application provides a method, device, and storage medium for generating a wireless private network signal coverage map, achieving automated execution, reducing technical barriers and costs, and enhancing efficiency and versatility. The technical solution is as follows:

[0009] On the one hand, a method for generating a wireless private network signal coverage map is provided, which includes:

[0010] • Preprocessing input data to obtain standardized data in the target fonnat, said input data including target building data and / or user requirement data for wireless private network planning, said input data including one or more modalities;

[0011] • Extracting features from the standardized data in the target fonnat and processing to obtain a three-dimensional structure diagram and target feature vectors of the interior of the target building:

[0012] • Matching the target feature vectors with the three-dimensional structure diagram, and estimating the signal strength corresponding to the base station for each grid point in the three- dimensional structure diagram based on the matching results and the feature values of each dimension in the target feature vectors;

[0013] • Mapping the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram to generate a wireless private network signal coverage map inside the target building.

[0014] In some embodiments, extracting features from the standardized data in the target format to obtain a three-dimensional structure diagram of the interior of tire target building includes any of the following: • Extracting a three-dimensional structure diagram of the interior of the target building from the input data;

[0015] • Extracting feature information of the building space from the standardized data in the target format, inputting the building space information into an image generation model, and generating a three-dimensional structure diagram of the interior of the target building by the image generation model based on the building space information.

[0016] In some embodiments, inputting the building space information into the image generation model, and generating a three-dimensional structure diagram of the interior of the building by the image generation model based on the building space information, includes:

[0017] • Inputting the building space information into the image generation model to generate a predicted two-dimensional design plan of the interior of the target building, and generating a three-dimensional structure diagram of the interior of the target building by a generative neural network based on the two-dimensional design plan;

[0018] • Dividing the three-dimensional structure diagram into different grid point areas according to the target distance, each grid point area corresponding to a spatial area with physical significance in the target building space.

[0019] In some embodiments, the input data includes at least one of architectural images, architectural videos, two-dimensional structure diagrams, or three-dimensional structure diagrams of the interior of the building;

[0020] • Preprocessing the input data to obtain standardized data in the target format; extracting features from the standardized data in the target format to obtain building space information, including:

[0021] • Preprocessing the input data to obtain images in the target format:

[0022] • Converting the images in the target format into a unified dot matrix format and resolution, performing edge detection on the converted images to obtain edge feature information;

[0023] • Performing image segmentation and linear projection mapping on the converted images to obtain spatial vectors corresponding to the image block sequence of each image;

[0024] • Using a convolutional neural network to recognize and encode features of the converted images to obtain embedded vectors containing building feature information for each image;

[0025] • Recording positional information and / or relative positional information of the interior space of the target building to obtain positional encoding of the image block sequence;

[0026] • Adding the embedded vectors and positional encodings of each image to the spatial vectors corresponding to the image block sequence to obtain a sequence containing building space information.

[0027] In some embodiments, matching the target feature vectors with the three-dimensional structure diagram, and estimating the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram based on the matching results and the feature values of each dimension in the target feature vectors, includes:

[0028] • Matching the target feature vectors with the three-dimensional structure diagram, combining different feature components matched for each grid point in the matching results;

[0029] • Processing the values of different feature components matched for each grid point according to the corresponding relationship between the type of feature component and the signal strength estimation method;

[0030] • Synthesizing the processing results of different feature components matched for each grid point to estimate the signal strength for each grid point, obtaining the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram.

[0031] In some embodiments, the different feature components include at least two of material characteristics, signal spatial characteristics, base station characteristics, and propagation path characteristics. Processing the values of different feature components matched for each grid point according to the corresponding relationship between the type of feature component and the signal strength estimation method includes at least two of the following:

[0032] • For material characteristic components, determining the electromagnetic parameters and reflection / absorption coefficients of the material indicated by the material characteristic component; query ing the signal strength coefficient of the material under different conditions based on the electromagnetic parameters and rcflcction / absorption coefficients;

[0033] • For signal spatial characteristic components, extracting the signal spatial characteristic component from the target feature vector, using interpolation technology to fit the signal spatial characteristic component into a continuous numcncal vector, and adding the numerical vector to the material characteristic component, base station characteristic component, or propagation path characteristic component for signal strength estimation;

[0034] • For base station characteristic components, for grid points where the base station is installed, normalizing the numerical features in the base station characteristic component, converting the categorical features in the base station characteristic component into integer encoding, normalizing the integer encoding, and merging the normalized numerical features and categorical features into a first feature vector;

[0035] • For propagation path characteristic components, converting the phase angles in the propagation path characteristic components into sine and / or cosine values, standardizing the sine and / or cosine values, path length, and loss coefficients, and merging the standardized sine and / or cosine values, path length, and loss coefficients into a second feature vector.

[0036] Tn some embodiments, synthesizing the processing results of different feature components matched for each grid point to estimate the signal strength for each grid point, obtaining the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram, includes:

[0037] • For the first grid point where the base station is installed, estimating the signal strength of the first grid point based on the first feature vector of the first grid point and the signal strength coefficient of the material under different conditions;

[0038] • For the second grid point where the base station is not installed, iterating through each grid point that serves as a receiving point, determining the signal strength components of the second grid point through different propagation paths based on the signal strength coefficient of the material under different conditions and the propagation path characteristic components of the second grid point, and determining the signal strength corresponding to tire base station at the second grid point based on the signal strength components.

[0039] In some embodiments, estimating the signal strength for each grid point in the three- dimensional structure diagram is implemented based on an artificial intelligence model; the method also includes:

[0040] • Using the signal strength measured inside the target building after the deployment of the base station as the label for the target feature vector and the three-dimensional structure diagram, training the artificial intelligence model based on the target feature vector, the three- dimensional structure diagram, and the label, and updating the weights of the artificial intelligence model.

[0041] On the one hand, a wireless private network signal coverage map generation device is provided, which includes:

[0042] • A preprocessing module for preprocessing input data to obtain standardized data in the target format. The input data includes target building data and / or user requirement data for wireless private network planning, and includes one or more modalities;

[0043] • A processing module for feature extraction and processing of the standardized data in the target format to obtain a three-dimensional structure diagram and target feature vectors of the interior of the target building;

[0044] • An estimation module for matching the target feature vectors with the three- dimensional structure diagram, and estimating the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram based on the matching results and the feature values of each dimension in the target feature vectors;

[0045] • A generation module for mapping the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram to generate a wireless private network signal coverage map inside the target building.

[0046] In some embodiments, the processing module is used to perform any of the following:

[0047] • Extracting a three-dimensional structure diagram of the interior of the target building from the input data;

[0048] • Extracting feature information of the building space from the standardized data in the target format, inputting the building space information into an image generation model, and generating a three-dimensional structure diagram of the interior of the target building by the image generation model based on the building space information.

[0049] In some embodiments, the processing module is used to: • Input the building space information into the image generation model to generate a predicted two-dimensional design plan of the interior of the target building, and generate a three-dimensional structure diagram of the interior of the target building by a generative neural network based on the two-dimensional design plan;

[0050] • Divide the three-dimensional structure diagram into different grid point areas according to tire taiget distance, each grid point area corresponding to a spatial area with physical significance in the target building space.

[0051] In some embodiments, the input data includes at least one of architectural images, architectural videos, two-dimensional structure diagrams, or three-dimensional structure diagrams of the interior of the building;

[0052] • The preprocessing module and the processing module arc used to: o Preprocess the input data to obtain images in the taiget fomiat; o Convert the images in the target format into a unified dot matrix format and resolution, perform edge detection on the converted images to obtain edge feature information; o Perfonn image segmentation and linear projection mapping on the converted images to obtain spatial vectors corresponding to the image block sequence of each image; o Use a convolutional neural network to recognize and encode features of the converted images to obtain embedded vectors containing building feature information for each image; o Record positional information and / or relative positional information of the interior space of the taiget building to obtain positional encoding of tire image block sequence; o Add the embedded vectors and positional encodings of each image to the spatial vectors corresponding to the image block sequence to obtain a sequence containing building space information.

[0053] In some embodiments, the estimation module is used to:

[0054] • Match the target feature vectors with the three-dimensional structure diagram, combine different feature components matched for each grid point in the matching results;

[0055] • Process the values of different feature components matched for each grid point according to the corresponding relationship between the type of feature component and the signal strength estimation method;

[0056] • Synthesize the processing results of different feature components matched for each grid point to estimate the signal strength for each grid point, obtaining the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram.

[0057] In some embodiments, the different feature components include at least two of material characteristics, signal spatial characteristics, base station characteristics, and propagation path characteristics:

[0058] • The estimation module is used to perform at least two of the following: o For material characteristic components, determine the electromagnetic parameters and reflection / absorption coefficients of the material indicated by the material characteristic component; query the signal strength coefficient of the material under different conditions based on the electromagnetic parameters and reflection / absorption coefficients; o For signal spatial characteristic components, extract the signal spatial characteristic component from the target feature vector, use interpolation technology to fit the signal spatial characteristic component into a continuous numerical vector, and add the numerical vector to the material characteristic component, base station characteristic component, or propagation path characteristic component for signal strength estimation; o For base station characteristic components, for grid points where the base station is installed, normalize the numerical features in the base station characteristic component, convert the categorical features in the base station characteristic component into integer encoding, normalize the integer encoding, and merge the normalized numerical features and categorical features into a first feature vector; o For propagation path characteristic components, convert the phase angles in the propagation path characteristic components into sine and / or cosine values, standardize the sine and / or cosine values, path length, and loss coefficients, and merge the standardized sine and / or cosine values, path length, and loss coefficients into a second feature vector

[0059] In some embodiments, the estimation module is used to:

[0060] • For the first grid point where the base station is installed, estimate the signal strength of the first grid point based on the first feature vector of the first grid point and the signal strength coefficient of the material under different conditions;

[0061] • For the second grid point where the base station is not installed, iterate through each grid point that serves as a receiving point, determine the signal strength components of the second grid point through different propagation paths based on the signal strength coefficient of the material under different conditions and the propagation path characteristic components of the second grid point, and determine the signal strength corresponding to the base station at the second grid point based on the signal strength components.

[0062] In some embodiments, estimating the signal strength for each grid point in the three- dimensional structure diagram is implemented based on an artificial intelligence model; the device also includes a training module, which is used to:

[0063] • Use the signal strength measured inside the target building after the deployment of the base station as the label for the target feature vector and the three-dimensional structure diagram, train the artificial intelligence model based on the target feature vector, the three-dimensional structure diagram, and tire label, and update the weights of the artificial intelligence model.

[0064] On the one hand, an electronic device is provided, which includes one or more processors and one or more memories, where the one or more memories store at least one computer program, and the at least one computer program is loaded and executed by the one or more processors to implement various optional implementations of the above wireless private network signal coverage map generation method.

[0065] On the one hand, a computer-readable storage medium is provided, which stores at least one computer program, and the at least one computer program is loaded and executed by a processor to implement various optional implementations of the above wireless private network signal coverage map generation method.

[0066] On the one hand, a computer program product or computer program is provided, which includes one or more program codes, and the one or more program codes are stored in a computer-readable storage medium. One or more processors of an electronic device read the one or more program codes from the computer-readable storage medium, and the one or more processors execute the one or more program codes, enabling the electronic device to implement any possible implementation method of the wireless private network signal coverage map generation method as described above.

[0067] The embodiments of this application use artificial intelligence models to flexibly, accurately, and efficiently solve the problems that wireless network planning usually relies on the design schemes of professional teams or the output of professional simulation tools, helping to promote the popularization and development of private network construction; the embodiments of this application can process input data to obtain a three-dimensional structure diagram of the interior of a building, and determine the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram through the target feature vectors obtained from the input data, generating a wireless private network signal coverage map inside the building. The entire process is automated, and users only need to input building data or requirement data to automatically implement wireless private network planning, reducing the threshold for network planning and design, improving the efficiency and versatility' of wireless private network planning, and reducing its costs.

[0068] Brief description of the drawings:

[0069] Figure 1 is a schematic diagram of the implementation environment for generating a wireless private network signal coverage map according to an embodiment of the application.

[0070] Figure 2 is a flowchart of the method for generating a wireless private network signal coverage map according to an embodiment of the application.

[0071] Figure 3 is a structural schematic diagram of the device for generating a wireless private network signal coverage map according to an embodiment of the application. Figure 4 is a block diagram of an electronic device according to an embodiment of the application.

[0072] Figure 5 is a block diagram of a terminal according to an embodiment of the application.

[0073] Figure 6 is a schematic diagram of a server according to an embodiment of the application.

[0074] Detailed method of implementation

[0075] To more clearly convey the objectives, technical solutions, and advantages of this application, further detailed descriptions of the implementation of this application will be provided in conjunction with the accompanying drawings.

[0076] In this application, terms such as "first," "second," etc., are used to distinguish between items or similar items that have essentially the same functions and roles. It should be understood that there is no logical or temporal dependency between "first," "second," "nth," nor do they limit the quantity or execution order. It should also be understood that even though the following description uses terms like "first," "second," etc., to describe various elements, these elements should not be restricted by these terms. These terms are merely used to differentiate one element from another. For example, without departing from the scope of the various examples given, the "first image" could be referred to as the "second image," and similarly, the "second image" could be referred to as the "first image." Both the first and second images are images and, in certain contexts, are separate and distinct images.

[0077] The term "at least one" in this application means one or more, and the term "multiple" means two or more, for example, "multiple data packets" refers to two or more data packets.

[0078] It should be understood that the terminology used in the description of various examples in this document is intended merely to describe specific examples and is not intended to be limiting. As used in the description of various examples and the appended claims, the singular forms "a" or "the" are intended to include the plural forms unless the context clearly dictates otherwise.

[0079] It should also be understood that the term "and / or" as used in this document means and covers any and all possible combinations of the listed items. The term "and / or" is a term of art used to describe a relationship between associated objects, indicating the existence of three relationships, for example, A and / or B, which means: A alone, both A and B together, or B alone. Additionally, the character " / " in this application generally indicates an "or" relationship between the associated objects before and after it.

[0080] It should also be understood that in the various embodiments of this application, the numbering of the processes does not imply the order of execution. The order of execution of the processes should be determined by their function and inherent logic, and should not limit the implementation process of the embodiments of this application.

[0081] It should also be understood that determining B based on A does not mean that B is determined solely based on A, but also based on A and / or other information. It should also be understood that the term "comprising" (also referred to as "includes," "including," "Comprises," and / or "Comprising") when used in this specification indicates the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or their groupings.

[0082] It should also be understood that the tenn "if can be interpreted to mean "when" ("when" or "upon") or "in response to determining" or "in response to detecting." Similarly, depending on the context, the phrases "if it is determined that..." or "if [the stated condition or event] is detected" can be interpreted to mean "when... is determined" or "in response to determining..." or "when [the stated condition or event] is detected" or "in response to detecting [the stated condition or event]."

[0083] The following reference drawings are described to assist in the comprehensive understanding of the various embodiments of this application as defined by the claims and their equivalents. This description includes various specific details to aid in understanding but should be considered exemplary only. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described here without departing from the scope and spirit of this application. Additionally, for clarity and brevity, the description of well-known functions and structures may be omitted.

[0084] The tenninology and wording used in the following description and claims are not limited to their dictionary meanings but are used by the inventor solely to enable a clear and consistent understanding of this application. Therefore, it should be apparent to those skilled in the art that the descriptions provided below of the various embodiments of this application are for illustrative purposes only and are not intended to limit the scope of this application as defined by the appended claims and their equivalents.

[0085] It should be understood that the singular forms "one," "a," and "the" may also include plural references unless the context clearly indicates otherwise. For example, the tenn "component surface" includes one or more such surfaces. When we refer to an element being "connected" or "coupled" to another element, it can mean that one element is directly connected or coupled to another element or that one element is connected or coupled to another element through an intermediary element. Additionally, the terms "connect" or "couple" used here can include wireless connections or wireless coupling.

[0086] The term "comprising" or "can comprise" indicates the presence of the corresponding disclosed functions, operations, or components in the various embodiments of this application, without limiting the presence of one or more additional functions, operations, or features. Furthermore, the term "comprising" or "having" can be interpreted to represent certain characteristics, numbers, steps, operations, components, or their combinations, but should not be interpreted as excluding the possibility of the existence of one or more other characteristics, numbers, steps, operations, components, or their combinations. The term "or" as used in the various embodiments of this application includes any listed term and all combinations thereof. For example, "A or B" can include A, can include B, or can include both A and B. When describing multiple (two or more) items, if the relationship between the multiple items is not explicitly defined, it can refer to one, multiple, or all of the multiple items, for example, the description "parameter A includes Al, A2, A3" can be implemented as parameter A including Al or A2 or A3, and can also be implemented as parameter A including at least two of the three items Al, A2, A3.

[0087] Unless differently defined, all terms (including technical or scientific terms) used in this application have the same meaning understood by those skilled in the art in the context of this application. Common terms defined in dictionaries are interpreted to have meanings consistent with the context in the relevant technical field and should not be idealized or overly formalized unless explicitly defined as such in this application.

[0088] At least some of the functions of the devices or electronic devices provided in the embodiments of this application can be implemented through Al models, such as at least one module of the device or electronic device being implemented through Al models. Functions associated with AT can be executed through non-volatile memory volatile memory, and processors.

[0089] The processor may7include one or more processors. At this time, the one or more processors can be general processors, such as a Central Processing Unit (CPU), an Application Processor (AP), etc., or purely graphic processing units, such as a Graphics Processing Unit (GPU), a Vision Processing Unit (VPU), and / or Al-specific processors, such as a Neural Processing Unit (NPU).

[0090] The one or more processors control the processing of input data based on predefined operational rules or artificial intelligence (Al) models stored in non-volatile memory and volatile memory. Predefined operational rules or Al models are provided through training or learning.

[0091] Here, providing through learning means obtaining predefined operational rules or Al models with desired characteristics by applying learning algorithms to multiple learning data. The learning can be executed in the Al device or electronic device itself according to the embodiment, and / or can be implemented through a separate server / system.

[0092] Al models can include multiple neural network layers. Each layer has multiple weight values, and each layer performs neural network calculations through calculations between the input data (such as the results of the previous layer's calculations and / or the input data of the Al model) and the multiple weight values of the current layer. Examples of neural networks include but arc not limited to Convolutional Neural Networks (CNN), Deep Neural Networks (DNN), Recurrent Neural Networks (RNN), Restricted Boltzmann Machines (RBM), Deep Belief Networks (DBN), Bidirectional Recurrent Deep Neural Networks (BRDNN), Generative Adversarial Networks (GAN), and Deep Q Networks. Learning algorithms are methods that use multiple learning data to train a predetermined target device (e.g., a robot) to enable, allow, or control the target device to make determinations or predictions. Examples of learning algorithms include but are not limited to supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

[0093] According to this application, at least one step in the methods performed by electronic devices, such as recognizing architectural elements, classifying planar map areas, and other steps, can be implemented using artificial intelligence models. The processor of the electronic device can perform pre-processing operations on data to convert it into aform suitable for input to an artificial intelligence model. Artificial intelligence models can be obtained through training. Here, "obtained through training" means obtaining predefined operational rules or artificial intelligence models configured to perform desired features (or purposes) by training an underlying artificial intelligence model with multiple training data.

[0094] Below is an explanation of the terminology involved in this application.

[0095] Artificial Intelligence (Al) involves the use of digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use knowledge to achieve the best results. In other words, Al is an interdisciplinary technology in computer science that attempts to understand the essence of intelligence and to produce a new type of intelligent machine that can respond in a manner similar to human intelligence. Al is also the study of the design principles and implementation methods of various intelligent machines, enabling machines to have the capabilities of perception, reasoning, and decision-making.

[0096] Al technology is an interdisciplinary field, involving a wide range of areas, including both hardware and software technologies. Basic Al technologies generally include sensors, dedicated Al chips, cloud computing, distributed storage, big data processing technologies, operating / interaction sy stems, mechatronics, and more. Al software technologies mainly include computer vision technology, speech processing technology, natural language processing technology, and machine leaming / deep learning.

[0097] Computer Vision (CV) is the science of making machines "see." More specifically, it involves using cameras and computers to replace the human eye for machine vision tasks such as identification, tracking, and measurement, and further for image processing to make the computer processing more suitable for the human eye to observe or transmit to instruments for detection. As a scientific discipline, computer vision studies the relevant theories and technologies, attempting to establish an Al system capable of extracting information from images or multidimensional data. Computer vision technology typically includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, 3D object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, and more. It also includes common biometric recognition technologies such as facial and fingerprint recognition.

[0098] Key technologies in Speech Technology include Automatic Speech Recognition (ASR), Text-to-Speech (ITS), and voiceprint recognition technology. Enabling computers to listen, see, speak, and feel represents the future direction of human-computer interaction, with speech being one of tire most promising modes of human-computer interaction in the future.

[0099] Natural Language Processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. NLP is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language, the language people use daily, and thus is closely related to linguistic research. Natural language processing technologies typically include text processing, semantic understanding, machine translation, robot Q&A, knowledge graphs, and more.

[0100] Machine Learning (ML) is an interdisciplinary subject involving probability theory, statistics, approximation theory convex analysis, algorithm complexity' theory, and more. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and is the fundamental way to make computers intelligent, with its applications throughout the field of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, Bayesian networks, reinforcement learning, transfer learning, inductive learning, and formula teaching.

[0101] With the research and advancement of Al technology, Al has been studied and applied in various fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, autonomous driving, unmanned driving, drones, robots, smart healthcare, and smart customer service. It is believed that with the development of technology, Al will be applied in more fields and play an increasingly important value.

[0102] The solution provided by the embodiments of this application involves technologies such as Al's computer vision, speech technology, natural language processing, and machine learning, which are specifically illustrated through the following examples.

[0103] With the widespread deployment of 4G and 5G cellular wireless technologies, their applications are gradually expanding into the core areas of various industries. More and more industries are introducing wireless private networks to support business operations, thereby imposing higher demands on network design and planning. However, current wireless network planning typically relies on design schemes from professional teams or outputs from professional simulation tools. For small and medium-sized enterprises, this approach is not only costly but also has a relatively high technical barrier. The rapid development of artificial intelligence technology offers new possibilities for solving these problems. In particular, large Al models can better perform multimodal generation tasks and demonstrate strong data processing and analytical capabilities. By training these models on a vast amount of target scenano data, it is possible to extract key scenario information during the inference phase, thereby predicting the coverage and capacity of wireless networks. This method is expected to significantly reduce tire barriers to obtaining high-performance wireless private network planning and design.

[0104] In recent years, with the rapid development of big data and Al technology, using machine learning, especially deep learning models, for indoor signal strength prediction has become an emerging and promising method. Al-based models have several notable advantages: First, Al models have powerful data processing and analytical capabilities. By training these models on a large amount of target scenario data, it is possible to extract key scenario information during the inference phase, thereby predicting the coverage and capacity of wireless networks. Second, Al models have strong generalization capabilities and are easy to integrate and extend. When the actual environment changes in real-time, the model can promptly update the estimation results. Moreover, with the continuous emergence of new technologies, the model can be easily expanded to support more input features and data types.

[0105] In summary, the method of using Al models for indoor signal strength generation can provide a flexible, accurate, and efficient solution, helping to promote the popularization and development of private network construction. However, to achieve this goal, careful design is needed during the model design phase. Existing machine learning models still face many challenges in processing raw data and transforming input features, especially when dealing with indoor environment data that contains a variety of different physical meanings How to effectively extract and utilize useful information from these data to generate indoor signal strength coverage maps of buildings has become an urgent problem to be solved. Therefore, the embodiments of this application have developed a new method that can extract key features from data containing indoor spatial features and generate corresponding indoor signal strength coverage maps.

[0106] Below is an explanation of the implementation environment of this application. Figure 1 is a schematic diagram of the implementation environment of a wireless private network signal coverage map generation method provided in an embodiment of this application. The implementation environment includes terminal 101, or the implementation environment includes terminal 101 and wireless private network signal coverage map generation platform 102. Terminal 101 is connected to the wireless private network signal coverage map generation platform 102 via a wireless or wired network. Terminal 101 is at least one of a smartphone, gaming console, desktop computer, tablet computer, e-book reader, MP3 (Moving Picture Experts Group Audio Layer III, dynamic image expert compression standard audio level 3) player, or MP4 (Moving Picture Experts Group Audio Layer IV, dynamic image expert compression standard audio level 4) player, laptop portable computer. Terminal 101 installs and runs an application that supports the generation of wireless private network signal coverage maps. Exemplarily, terminal 101 has data acquisition and data processing functions. It collects user-input data, preprocesses and extracts features from the collected data, and then generates a wireless private network signal coverage map inside the building based on an image generation model. It achieves wireless private network planning and design, providing feedback to the user. Terminal 101 can independently complete this task, or it can collect input data and transmit it to the wireless private network signal coverage map generation platform 102 for data processing sendees. Tins application embodiment does not limit this arrangement. The wireless private network signal coverage map generation platform 102 includes at least one of a sender, multiple senders, a cloud computing platform, and a virtualization center. The wireless private network signal coverage map generation platfonn 102 is used to provide backend sendees for applications that support the generation of wireless private network signal coverage maps. Optionally, the wireless private network signal coverage map generation platform 102 undertakes the main processing tasks, while terminal 101 undertakes secondary processing tasks; or vice versa; or the wireless private network signal coverage map generation platform 102 or terminal 101 can each undertake processing tasks independently. Alternatively, the wireless private network signal coverage map generation platform 102 and terminal 101 can use a distributed computing architecture for collaborative computation. Optionally, the wireless private network signal coverage map generation platform 102 includes at least one server 1021 and a database 1022. The database 1022 is used for storing data. In this application embodiment, the database 1022 stores sample input data and sample annotation data. It can also store user input data, providing data sendees for at least one server 1021. A server is a standalone physical server, a sener cluster or distributed s stem composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud sendees, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security senices, CDN, and big data and Al platforms. Terminals are smartphones, tablet computers, laptops, desktop computers, smart speakers, smartwatches, etc., but are not limited to these. Those skilled in the art understand that the number of tenninals 101 and seners 1021 can be more or less. For example, the aforementioned terminal 101 and sener 1021 can be just one, or there can be dozens or hundreds of terminals 101 and seners 1021, or even more. This application embodiment does not limit the quantity and type of terminals or seners.

[0107] Figure 2 is a flowchart of a wireless private network signal coverage map generation method provided in an embodiment of this application, which is applied in an electronic device, the electronic device being a terminal or a sener. See Figure 2, the method includes the following steps. 201. The electronic device obtains input data, which includes target building data and / or user requirement data to be planned for a wireless private network, and the input data includes one or more modalities.

[0108] In this embodiment of the application, users can input data on the electronic device, enabling the device to plan a wireless private network for the user based on the input data, and output a wireless private network signal coverage map for the interior of a building. The input data can include data containing building information, user requirement-related data, or a combination of both types of data.

[0109] The user can be an individual or a corporate entity. For example, small and medium-sized enterprises that need to deploy a wireless private network can input their target building data and / or user requirement data through the electronic device.

[0110] If the electronic device is a terminal, the user can input data on that electronic device, allowing the device to collect the user's input data. If the electronic device is a server, the user can input data on a terminal device, which can then send the collected input data to the server. The server receives this input data, thus obtaining it. This embodiment of the application does not limit the specific type of electronic device.

[0111] In some embodiments, the input data can be unimodal or multimodal inputs containing building information and / or user requirements obtained through specific methods (i .e., the way the input data is acquired). The input data includes multiple modalities, which can include one or more of photographic image modality, planar image modality, graphic file modality, and video modality.

[0112] In some embodiments, tire input data can also include one or more of LiDAR modality, infrared sensing modality, text modality, and voice modality.

[0113] In some embodiments, the input data can include at least one of architectural images, architectural videos, two-dimensional structural diagrams, or three-dimensional structural diagrams of the interior layout of a building. That is, in a specific example, the system's input consists of images, videos, and / or drawings containing building information obtained through specific methods, as well as data containing indoor signal environment characteristics. The types of input forthe obtained images, videos, and / or drawings containing building information include but are not limited to the following types or combinations of types: a. Architectural images: These images can be obtained by the user themselves by shooting or by other means: b. Architectural videos: These videos can be obtained by the user themselves by shooting or by other means; c. Two-dimensional planar diagrams / three-dimensional structural diagrams containing the interior layout of the building.

[0114] In a specific possible embodiment, the input data may consist of architectural images captured by users in the photographic image modality; architectural videos recorded by users in the video modality, floor plans of the building's interior layout in tire planar image modality, architectural sketches drawn by users in the planar image modality, or graphic files output by professional drawing software in the graphic file modality, either individually or in any combination. That is, the input data can include, but is not limited to, data from the aforementioned modalities or any combination thereof.

[0115] In another specific possible embodiment, the input data may also consist of LiDAR data in the LiDAR modality, infrared sensing data in the infrared sensing modality; textual information from users in the text modality, or voice information from users in the voice modality, either individually or in any combination.

[0116] It should be noted that the aforementioned input data represents only some ty pes of data set by relevant technical personnel according to requirements. In actual applications, technical personnel may also set or add other types of input data based on requirements or experience. In other embodiments, the input data may also consist of other information obtained from users that includes building information and / or user requirements. This application example does not limit this and, therefore, the input data may include, but is not limited to, the aforementioned ty pes of data or combinations of data ty pes.

[0117] In some embodiments, the input data includes multiple modalities, correspondingly, the input data includes at least two of the following: architectural images captured by users in the photographic image modality, architectural videos recorded by users in the video modality; floor plans of the building's interior layout in the planar image modality; architectural sketches drawn by users in the planar image modality; and graphic files output by professional drawing software in the graphic file modality.

[0118] In other embodiments, the input data may also include at least one of the following: LiDAR data in the LiDAR modality; infrared sensing data in the infrared sensing modality; textual information from users in the text modality, and voice information from users in the voice modality.

[0119] Regarding the specific methods of obtaining the input data (i.e., the acquisition methods), these include, but are not limited to, the following methods or any combination thereof:

[0120] Acquisition method one: Through user uploads on website pages, user online drawing, etc.

[0121] Acquisition method two: Through user uploads, user creation, etc., in mobile phone, computer, tablet (Pad) applications.

[0122] Acquisition method three: Outputs from other building, drawing professional software, etc.

[0123] Of course, the methods of obtaining the input data may also include other ways, such as users entering a URL from which the electronic device accesses and downloads the data. This application example does not limit this. 202. The electronic device preprocesses the input data to obtain standardized data in the target format.

[0124] Considering that data of different modalities have different forms of presentation, data types, and feature distributions, after the electronic device obtains the input data, it can first preprocess the input data, converting it into standardized data and adjusting it to a preset unified format. This ensures the standardization of the processed data, facilitating subsequent processing and integration of features of different modalities within a unified framework.

[0125] Also, because data of different modalities have different forms of presentation, datatypes, and feature distributions, the preprocessing methods used by the electronic device for input data of different modalities vary.

[0126] The target format can be set by relevant technical personnel according to needs or experience, and this embodiment of the application does not make specific limitations on this. Accordingly, for data of different modalities, the target format can be the same or different.

[0127] In some embodiments, the electronic device can preprocess the input data according to the modality of the input data and the preprocessing method corresponding to that modality, converting the input data into standardized data in the format corresponding to that modality.

[0128] In some embodiments, a correspondence between modalities and preprocessing methods can be preset, and the electronic device can identify the modality of the input data. Based on this correspondence, it can determine the preprocessing method for the input data and then perform preprocessing.

[0129] In some embodiments, a configuration file storing the correspondence between modalities and preprocessing methods can be preset in the electronic device. When input data needs to be obtained, after identifying the modality of the input data, the configuration file can be called or read. The preprocessing method corresponding to the modality is matched from this configuration file, and then the execution instructions of tire preprocessing method are executed to preprocess the input data accordingly, con erting it into standardized data.

[0130] 203. The electronic device performs feature extraction and processing on the standardized data in the target format to obtain a three-dimensional structural diagram and target feature vectors of the interior of the target building.

[0131] In the aforementioned Step 202, each modality of the input data was standardized to a unified fonnat. allowing for the convenient merging of multimodal features after extraction, meaning that data processing can be carried out within the same framework.

[0132] This step 203 requires obtaining two types of information: a three-dimensional structural diagram of the interior of the taiget building and target feature vectors. The method for obtaining the three-dimensional structural diagram of the interior of tire target building is described below. The three-dimensional structural diagram of the interior of the target building can be obtained in various ways. From the aforementioned Step 201, it is known that the input data may include a three-dimensional structural diagram of the interior of the target building. In this case, the electronic device can extract the three-dimensional structural diagram of the interior of the target building from the input data. In another possible scenario, the input data may not include a three-dimensional structural diagram of tire interior of the target building, and the electronic device can process the standardized data in the target format to obtain a three- dimensional structural diagram of the interior of the target building. Specifically, the electronic device can perform feature extraction on the standardized data in the target format to obtain building space information and input this building space information into an image generation model. The image generation model generates a three-dimensional structural diagram of the interior of the target building based on this building space information. Building space information can refer to building features and spatial information.

[0133] In some embodiments, the electronic device can input the building space information into the image generation model, which generates a predicted two-dimensional plan design diagram of the interior of the target building based on the building space information. A generative neural network generates a three-dimensional structural diagram of the interior of the target building based on this two-dimensional plan design diagram, dividing the three-dimensional structural diagram into different grid point areas according to the target distance, with each grid point area corresponding to a spatial area with physical significance in the space of the target building. The image generation model can be a pre-trained artificial intelligence model, and similarly, the generative neural network can also be pre-trained. This application example does not elaborate on tire training process of tire two.

[0134] In some embodiments, the input data includes at least one of architectural images, architectural videos, two-dimensional structural diagrams, or three-dimensional structural diagrams of the interior layout of a building. Accordingly, in the aforementioned Steps 202 and 203, the electronic device can preprocess the input data to obtain images in the target format. The electronic device converts the images in the target format into a unified dot matrix format and resolution, performs edge detection on the converted images to obtain edge feature information, performs image segmentation and linear projection mapping on the converted images to obtain spatial vectors corresponding to the image block sequence of each image, and uses a convolutional neural network to identify and encode features of the converted images to obtain embedded vectors containing building feature information for each image. It records the positional information and / or relative positional information of the interior space of the target building to obtain positional encoding of the image block sequence, and adds the embedded vector and positional encoding of each image to the spatial vector corresponding to the image block sequence to obtain a sequence containing building space information.

[0135] In a specific embodiment, corresponding preprocessing steps are required for specific input types to transform the corresponding inputs into a unified format of a three-dimensional structural diagram of the interior of a building. The specific preprocessing process is as follows : a. Architectural images: Adjust the image to a unified preset resolution to ensure the consistency of subsequent processing steps and the quality of the output data; apply image processing techniques to grayscale the processed dot matrix, reducing the computational complexity of subsequent operations. b. Architectural videos: Decode the video file and convert it into a frame sequence; generate a corresponding color histogram for each frame of the image, and calculate the histogram similarity between the color histograms of consecutive frames; when the histogram similarity is below a certain threshold, determine that a scene change has occurred and extract the frame as a candidate key frame; perform similarity matching on the extracted candidate key frames, remove redundant frames with high similarity from the candidate key frames, and determine the remaining frames as the final key frames; perform the same preprocessing operations as described above for each final key frame as for architectural images. c. Two-dimensional plan diagrams of the interior layout of the building: Adjust the resolution, for example, adjust the image resolution to a unified preset resolution, to ensure the consistency of subsequent processing steps and the quality of the output data.

[0136] After preprocessing, the electronic device can convert the prcproccsscd input (i.c., the standardized data in the taiget format) into a unified dot matrix format, and uniformly adjust the resolution of the dot matrix during the conversion process. Use an edge detection algorithm to identity’ edge feature information in the image, which usually corresponds to the internal structures of the building such as walls and pillars. Segment the image with edge feature information into multiple image blocks, thereby transforming the entire image into an image block sequence, and flatten each image block into a vector. Then, through a learnable linear projection mapping, map the image into a high-dimensional space vector. Then, use a convolutional neural network to identify specific features in the dot matrix, such as the location of doors and windows, room segmentation, etc., map the information of specific features according to a preset dictionary into specific encoding, and add it as an embedded vector containing building feature information and the image block sequence. At the same time, record the positional information or relative positional information of the interior space of the building as spatial information, add it as the position encoding of the image block to the image block sequence, thereby obtaining a sequence containing building feature information and spatial information, which is also building space information.

[0137] The generation of a signal strength coverage map, in addition to the three-dimensional structural diagram of the interior of the building obtained in step 1), also requires a feature vector that includes indoor spatial features. The following describes the method of obtaining the taiget feature vector. In some embodiments, the target feature vector includes, but is not limited to: a. Location features: Tire x, y, z coordinate values of each grid point in the three- dimensional structure diagram within the three-dimensional spatial coordinate system. b. Material features: The material of the structure contained at each grid point, as well as the corresponding reflection coefficients and absorption coefficients of these materials. This feature can affect the distribution of wireless signals to varying degrees. For example, glass and wooden lightweight partitions have less obstruction to signals, while metal and concrete walls can severely obstruct signals; energy-saving glass with a metal coating may block more wireless frequency signals, etc. c. Signal spatial features: Whether the grid point is located in an area (Focus Zone) with specific requirements for targeted network coverage. The method of marking this signal spatial feature can be: if the point is in the aforementioned area, the corresponding feature value is recorded as 1 ; if the point is not in the aforementioned area, the corresponding feature value is recorded as 0. The physical significance of this signal spatial feature is that network users may designate areas with specific requirements for targeted network coverage according to different practical needs, such as areas where video conference rooms that require high-speed data transmission are located, areas where industrial automation production lines that require immediate response arc located, and areas that require reliability such as key infrastructure areas and emergency sendee areas. d. Base station features: Whether a base station is installed at that grid point. If the point is a base station installation point, the corresponding features of the base station need to be attributed to that grid point. Base station features can be divided into numerical features, such as the maximum throughput of the device, coverage range, power consumption, etc.; categorical features: the frequency range supported by the device, specific functions supported, power consumption level, etc. e. Propagation path features: Propagation path features include the direct path length and phase angle between that point and the base station installation point, the length and phase angle of the first-order reflection path, the length and phase angle of the second-order or higher- order reflection path, the length and phase angle of the diffraction path, the length of the transmission path, the number of times the transmission is penetrated, the phase angle of the transmission path, and the loss coefficient of the transmission penetration, etc.

[0138] In some embodiments, the multiple modalities include at least two types of modality types, including image modalities, planar diagram modalities, and language modalities, where the image modalities include photographic image modalities, graphic file modalities, video modalities, LiDAR modalities, and infrared induction modalities. The planar diagram modalities include planar image modalities; the language modalities include text modalities and voice modalities. That is, the input data includes data of multiple modality' types.

[0139] Accordingly, in the aforementioned Step 202, the electronic device can perform modality' alignment on standardized data of the same modality type to obtain the target standardized data for each modality. Then, in the aforementioned Step 203, the feature extraction step can be executed on the modality-aligned target standardized data to obtain the features of the target standardized data. Subsequent Step 204 can then meige the features of the target standardized data and proceed with the subsequent image generation process.

[0140] Data that has been standardized, if it is multimodal data, can undergo modality alignment to ensure the consistency of features extracted from different modalities. Modality alignment refers to the merging of features from different modalities during the feature extraction phase. This approach is suitable for modalities with small differences in feature scale and feature properties, facilitating better model learning of the relationships between modalities. Through modality alignment, different modalities can be divided into the following three typical types of modalities, which are also divided into the three types of modality types described below.

[0141] In a specific possible embodiment, the electronic device may have predefined preprocessing instructions corresponding to different modalities. The electronic device can recognize the modality of the input data, determine the corresponding preprocessing instructions based on the recognized modality, and execute these preprocessing instructions on the input data. For different modalities in the input data, the electronic device performs different preprocessing operations, which is to say, it executes different preprocessing instructions.

[0142] For example, modalities can include photographic image modalities, planar image modalities, graphic file modalities, video modalities, LiDAR modalities, infrared induction modalities, text modalities, voice modalities, etc., and relevant technical personnel can freely add more as needed; this application example does not limit this.

[0143] In some embodiments, the input data includes multiple modalities, including at least two of photographic image modalities, planar image modalities, graphic file modalities, video modalities, LiDAR modalities, infrared induction modalities, text modalities, and voice modalities. The input data includes at least two of the following: architectural images captured by users in the photographic image modality, architectural videos recorded by users in the video modality, floor plans of the building's interior layout in the planar image modality, architectural sketches drawn by users in the planar image modality, graphic files output by professional drawing software in the graphic file modality, LiDAR data in the LiDAR modality, infrared sensing data in the infrared sensing modality, textual information from users in the text modality, and voice information from users in the voice modality.

[0144] Accordingly, Step 202 may include any combination of at least two of the following situations. Of course, the multiple modalities are not limited to the aforementioned ones, and the input data is also not limited to the aforementioned types. Correspondingly, this step may also include other situations or any combination of other situations with the following ones.

[0145] Situation One: In response to the input data including architectural images captured by the user, the electronic device adjusts the resolution of the architectural images to a target resolution, processes the dot matrix of the architectural images to grayscale to obtain a first image, performs edge detection on this first image to obtain the contour features of the architectural structure within the first image, and uses this first image and the contour features as the standardized data for the architectural image.

[0146] The target resolution is a preset unified resolution, which can be set by relevant technical personnel based on requirements or experience. This application example does not limit the specific value of this target resolution. By adjusting the image resolution, it ensures the consistency of subsequent processing steps and the quality of the output data.

[0147] The grayscale processing uses image processing techniques to reduce memory occupancy and increase processing speed. By reducing color information, it visually enhances contrast and highlights the target area. In some embodiments, the process of grayscale processing of the dot matrix of the architectural image can be implemented in various ways, such as the maximum value method, average value method, weighted average method, etc. This application example does not limit the specific method.

[0148] The edge detection process uses image edge detection technology to extract the contour features of the architectural structure in the image, facilitating the enhancement of the visibil ity of key architectural information in the image.

[0149] Situation Two: In response to the input data including architectural videos recorded by the user in video modality, the electronic device decodes the architectural video, converts the architectural video into a frame sequence, and performs key frame extraction on the frame sequence to obtain the target key frames. The electronic device performs resolution adjustment, grayscale processing, and edge detection on these taiget key frames to obtain the standardized data for the target key frames.

[0150] The resolution adjustment, grayscale processing, and edge detection are the same as in Situation One and will not be further elaborated here.

[0151] In Situation Two, the process of extracting the target key frames from tire frame sequence is called key frame extraction. This key frame extraction process can be implemented in various ways. For example, a method based on shot segmentation involves analyzing the video's shot boundaries and selecting the first or last frame of each shot as a key frame. Another example is a method based on color features, which calculates the color differences between different frames in the frame sequence to identify key frames. Additionally, there is a deep learning method that applies a neural network model to extract key points and local features from video frames and obtains key frames by comparing feature changes between consecutive frames. There is also a method based on changes in the viewing angle, which detects changes in the scene viewing angle in the video to extract key frames. Of course, the key frame extraction process can also use other methods. A specific possible implementation method is provided below, and this application example does not limit the specific implementation method.

[0152] In a specific possible implementation example, for the first and second frames ofthe frame sequence, the electronic device generates color histograms for the first and second frames, respectively, calculates the histogram similarity between the color histograms of the first and second frames, and extracts the second frame as a candidate key frame in response to the histogram similarity being lower than the first threshold. The first and second frames are any two adjacent frames in the frame sequence, with the first frame preceding the second frame. For all candidate key frames, similarity matching is performed, and one of the two candidate key frames with a similarity higher than the second threshold is removed to obtain the target key frame. Then, the aforementioned operations of adjusting resolution, grayscale processing, and edge detection processing can be performed on each target key frame.

[0153] That is to say, for each frame of the image, a corresponding color histogram is generated, and the histogram similarity between the color histograms of two consecutive frames is calculated; when the histogram similarity is below a certain threshold, it is detennined that a scene change has occurred and the frame is extracted as a candidate key frame. The first threshold and the second threshold can both be set by relevant technical personnel according to requirements or experience, and the implementation example of this application does not limit their values.

[0154] Situation Three: In response to the input data including floor plans and / or architectural sketches drawn by the user, the electronic device performs resolution adjustment and edge detection processing on the floor plan and / or architectural sketch to obtain standardized data of the floor plan.

[0155] In this situation, the floor plans and / or architectural sketches drawn by the user arc generally black and white, thus there is no need for grayscale processing. It only requires resolution adjustment and edge detection processing to standardize the data into a unified format as in the first two situations. The resolution adjustment and edge detection processing are the same as in Situation One, and will not be further elaborated here.

[0156] Situation Four: In response to the input data including graphic files output by professional drawing software, the electronic device calls upon a specialized conversion tool to convert the graphic file into a second image in the target image format. The electronic device then adjusts the resolution of this second image to the target resolution to obtain standardized data of the graphic file.

[0157] Since graphic files are output by professional drawing software and their formats are not easily processed, they can first be converted into a more manageable image format, the target image format, which can be set by relevant technical personnel based on experience. This application example does not limit the specific format. The target image format can be a single format or multiple formats, set according to requirements.

[0158] Situation Five: In response to the input data including textual information entered by the user, tire electronic device removes stop words and irrelevant symbols from the text. Tire device also converts abbreviations, synonyms, and variant words in the text information, transforming the processed text into a taiget text format to obtain standardized data of the text information.

[0159] In this situation, stop words and irrelevant symbols in the text information are not helpful for semantic analysis and are considered redundant information. Therefore, this redundant information can be removed first. As for abbreviations, synonyms, and variant words in the text information, they can be converted into a standard expression, the target text format, to facilitate the uniformity' of subsequent data processing.

[0160] The target text format can be set by relevant technical personnel based on requirements or experience; this application example does not limit the specific settings.

[0161] Situation Six: Tn response to the input data, which includes voice information input by' the user, the electronic device performs noise reduction processing on the voice information. The electronic device then performs speech recognition on the noisc-rcduccd voice information to obtain the first text information corresponding to the voice information. The electronic device performs preprocessing on the first text information, including removing stop words, irrelevant symbols, abbreviations, synonyms, and variant word conversions, as well as text format conversion, to obtain the standardized data of the voice information.

[0162] Noise reduction processing can apply audio processing techniques to remove background noise from voice information, thereby improving voice clarity. This noise reduction process can be implemented in various ways. For example, voice information can be processed through amplifiers, filters, etc., to optimize signal transmission and processing. Additionally, automatic gain control (AGC), high-pass filters, and other methods can be used to remove low-frequency noise and enhance signal quality. Advanced noise reduction algorithms can also be employed, such as adaptive filters (e.g., LMS (Least Mean Squares), NLMS (Normalized Least Mean Square) algorithms), spectral subtraction in tire frequency domain, and deep learning algorithms (e.g., noise reduction based on neural networks). Of course, the noise reduction process can also be implemented in other ways, and this example does not limit the scope of the application.

[0163] Automatic Speech Recognition (ASR) technology aims to convert the vocabulary content in human speech into computer-readable input, such as keystrokes, binary codes, or character sequences. Unlike speaker recognition and speaker verification, which attempt to identify or confirm the speaker rather than the vocabulary content, speech recognition focuses on the words contained within the speech. After converting voice information to text information, the same preprocessing operations as in Situation Six can be performed.

[0164] Situation Seven: Tn response to the input data, which includes LiDAR data, the electronic device performs filtering on the LiDAR data. The electronic device then transforms the filtered LiDAR data from the device coordinate system to the global coordinate system, obtaining the standardized data of the LiDAR data. hr Situation Seven, filtering algorithms are used to remove invalid or noisy points contained in the LiDAR data to improve the quality and clarity of the LiDAR data. By transforming the LiDAR data into the global coordinate system, the data is unified under the same coordinate system, making the connections between the data more closely related and accurate. The filtering algorithm can be any algorithm, and this example does not limit the scope of the application.

[0165] Situation Eight: In response to the input data, which includes infrared induction data, the electronic device performs filtering on the infrared induction data. The filtered infrared induction data is then calibrated for temperature, and the calibrated infrared induction data is enhanced to obtain the standardized data of the infrared induction data.

[0166] Tn Situation Eight, the filtering process is similar to that in Situation Seven, removing noise points from the infrared induction data without further elaboration.

[0167] Temperature calibration of infrared induction data is performed to ensure the accuracy of the data. This temperature calibration process can be implemented in various ways. Of course, the temperature calibration process can also be implemented in other ways, and this example does not limit the scope of the application.

[0168] The image enhancement process involves adding some information to the original image or transforming data to selectively highlight features ofinterest in the image or suppress (mask) certain undesired features, making the image match the visual response characteristics.

[0169] The image enhancement process can be implemented using frequency domain methods or spatial domain methods. The frequency domain method treats the image as a two-dimensional signal and enhances it based on two-dimensional Fourier transform. Low-pass filtering (allowing only low-frequency signals to pass) can remove noise from the image, while high- pass filtering can enhance high-frequency signals such as edges, making blurry images clear. In the spatial domain method, representative algorithms include local averaging and median filtering (taking the middle pixel value in a local neighborhood), which can be used to remove or reduce noise. The image enhancement process can adopt any image enhancement method, and this example does not limit the scope of the application.

[0170] In some embodiments, the multimodality includes multiple types of modalities, one type of modality7includes one or more modalities, and one type of modality corresponds to a unified format. Accordingly, in step 202, for input data of multiple modalities of the same type, the electronic device can separately preprocess the input data of the multiple modalities through the corresponding preprocessing methods of the multiple modalities, adjusting the input data of the multiple modalities to standardized data in the unified format of that type of modality. For example, in a specific example, the input data is a combination of situation one and situation two, which includes images of buildings taken by the user and videos of buildings recorded by the user. The image of the building taken by the user is a photography modality, and the video of the building recorded by the user is a video modality. The electronic device can separately preprocess the building image and the building video. During the preprocessing process, the building image is preprocessed according to the method shown in situation one, and the building video is preprocessed according to the method shown in situation two. In both situations, the building image underwent processes such as resolution adjustment, grayscale processing, and edge detection. After processing the building video to obtain a sequence of frames and extracting the target key frames, the same preprocessing operations as the building image were performed on the target key frames. As a result, the input data of the two modalities will ultimately be adjusted to standardized data in a unified format forthe image modality type.

[0171] Image modalities: Images of buildings taken by users, key frames extracted from videos of buildings recorded by users, processed infrared induction images, LiDAR data, graphic files output by professional drawing software, etc. For image modalities, after the first step of preprocessing, the input data is converted into images and / or image sets with a unified format and resolution, thus allowing direct modality alignment.

[0172] That is to say, in response to the standardized data being image modality data, the electronic device can perform modality alignment on the standardized data of the image modality to obtain the target standardized data of the image modality.

[0173] Planar diagram modalities: Floor plans of building interiors, sketches of buildings drawn by users, etc. For planar diagram modalities, since there are significant differences between user-drawn sketches and standard architectural floor plans in terms of precision, standardization, and detail richness, it is necessary to input the sketches into a pre-trained artificial intelligence model (such as a generative adversarial network) to generate corresponding architectural floor plans with a standard format before performing modality alignment.

[0174] That is to say, in response to the standardized data being planar diagram modality data, the electronic device inputs the standardized data of the planar diagram modality into a drawing generation model. The drawing generation model generates corresponding architectural floor plans with a standard format based on the standardized data. Then, modality alignment is perfonned on the architectural floor plans with a standard format to obtain the target standardized data of the planar diagram modality.

[0175] Language modalities: Voice input from users, text input from users, etc. For language modalities, after the first step of preprocessing, text data in the same format can be obtained, which can be directly aligned modality -wise.

[0176] That is to say, in response to the standardized data being a language modality, the electronic device can perform modality alignment on the standardized data of the language modality to obtain the target standardized data of the language modality.

[0177] After modality alignment, the three types of modalities still have significant differences in terms of feature scale, data type, and feature distribution. Therefore, different artificial intelligence models can be used to extract features from different modalities, and then the feature results output by different models can be aligned and fused to form tire final output

[0178] In some embodiments, the electronic device can have pre-trained artificial intelligence models. The electronic device can input the standardized data or target standardized data of different modalities into the artificial intelligence model of that modality, and tire artificial intelligence model can extract features from the standardized data or target standardized data to obtain the features of that modality.

[0179] It should be noted that the input data obtained in step 201 includes target building data and / or user requirement data for wireless private network planning. Therefore, in this embodiment of the application, the extracted features can include building-related features and / or user requirement features.

[0180] Tn some embodiments, building-related features can include at least one of building structure information, building material information, and building spatial information.

[0181] In some embodiments, building structure information and corresponding building matcnal information include but are not limited to at least one of beam, column, staircase, wall, partition, window, and other building structure information, as well as the corresponding material information used in these structures, such as concrete, metal, glass, etc. Building structure information and building material information can have varying degrees of impact on the distribution of wireless signals. For example, glass and wooden lightweight partitions have less obstruction to wireless signal transmission, while metal and concrete walls can severely hinder wireless signal transmission. Energy-saving glass with a metal coating may block more wireless signal transmission. In addition, building material information can also include the reflection coefficient and absorption coefficient corresponding to each material.

[0182] In some embodiments, building spatial information includes but is not limited to building length and width information, floor height information, floor thickness, number of floors, etc. Among them, building length and width information and floor height information are crucial for determining the efficiency of indoor space utilization and indoor network planning. Floor thickness affects the signal penetration ability between floors, and the number of floors affects the final base station configuration. For example, high-rise buildings may require the installation of micro base stations or repeaters to ensure co erage.

[0183] In some embodiments, in the network planning and design of wireless private networks (such as 5th Generation Mobile Communication Technology (5G) private networks), the demands that users can propose are usually related to their expectations for network performance, usage scenarios, and specific needs of individuals or organizations. These features are very important for customizing solutions to meet specific network requirements.

[0184] User requirement features can generally be divided into two types. One type is related to the network implementation method and can be called the first requirement feature, including but not limited to the expected location of base station installation, the model of network equipment used, the frequency information of the network equipment expected to be used, etc. These network devices include but are not limited to core network devices, wireless access network devices, transmission network devices, gateway devices, etc. These network devices all have features related to network performance indicators, which can be divided into numerical features and categorical features according to the type of feature . Numerical features such as the maximum throughput of the device, coverage range, power consumption, etc. Categorical features such as the frequency range supported by the device, specific functions supported, power consumption level, etc. Fornumerical features, they can be directly processed into standardized features by normalization. For categorical features, they are first converted into integer-type features by feature encoding, and then normalized.

[0185] The other type of requirement feature is related to the output target, that is, the capacity and coverage of the network (for example, 5G), and can be called the second requirement feature. For example, users may specify specific areas that they hope the network will cover according to different actual needs, such as video conference rooms that require high-speed data transmission, industrial automation production lines that require instant response, key infrastructure areas and emergency sendee areas that require reliability', etc. For these types of features, there can be different processing methods. In some embodiments, they can also be preprocessed and feature extracted, thereby included in the aforementioned steps 201 to step 203. Tn other embodiments, they can be detected and processed separately without the aforementioned processing, thereby assisting the image generation process in step 205. For more details, please refer to the corresponding content of the input data processing for the target type in step 205.

[0186] In some embodiments, different feature extraction steps can be performed according to the modality type of the standardized data or target standardized data to extract different features. The following is a description of feature extraction for different modality types of data. The feature extraction is only an example, and related technical personnel can also add other feature extraction methods and types of extracted features according to needs. This embodiment of the application does not limit this.

[0187] For image modality data, in response to the standardized data or the target standardized data being image modality data, the electronic device can perform at least one of the following steps A to step C on the standardized data or target standardized data:

[0188] Step A: The electronic device can use an image recognition model to perform object recognition on the standardized data or target standardized data, output the building elements in the image, analyze the texture of local areas in the image using a classifier, identify the building material information in the image, and based on the building material information, obtain the corresponding reflection coefficient and / or absorption coefficient.

[0189] In step A, a pre-trained image recognition model is used to identify building elements in the image; a pre-trained classifier is used to analyze the texture of local areas in the image, identify different building materials, and based on the identified building materials, query or calculate the corresponding reflection coefficient and absorption coefficient.

[0190] Step B: The electronic device performs image segmentation on the standardized data or target standardized data based on a semantic segmentation model to obtain visual cues in the image, and based on these visual cues, obtain the spatial layout and the connection relationships between spaces in the image.

[0191] In step B, a pre-trained semantic segmentation model is used to analyze visual cues in the image, and thereby infer the spatial layout and the connection relationships between spaces, such as floor height.

[0192] Step C: The electronic device uses monocular depth estimation technology to extract three-dimensional spatial information from two-dimensional images.

[0193] Three-dimensional spatial information can include various types, such as floor thickness, and this embodiment of the application does not limit this.

[0194] Step D: If the standardized data is LiDAR data, the electronic device uses ground segmentation technology to separate ground points from non-ground points in the LiDAR data, clusters the point data to obtain objects and scene elements in the LiDAR data, and performs feature extraction on these objects and scene elements to obtain spatial layout features and / or distance features.

[0195] Ground segmentation technology using height thresholds is used to separate ground points from non-ground points in the LiDAR data, and clustering algorithms are applied to cluster the point data, thereby distinguishing different objects and scene elements. Key features arc extracted from the distinguished data for extracting spatial layout and distance information.

[0196] The aforementioned steps B to step D are for the process of extracting building spatial information and arc only an illustrative example. Relevant technical personnel can set which features to extract based on what technology according to needs or experience. This embodiment of the application does not limit this.

[0197] For planar diagram modality data, in response to the standardized data or the target standardized data being planar diagram modality data, the electronic device can perform at least one of the following steps E to step H on the standardized data or target standardized data:

[0198] Step E: The electronic device uses Optical Character Recognition (OCR) technology to extract building matenal information and / or building spatial information from the planar diagram, and based on the building material information, obtain the corresponding reflection coefficient and absorption coefficient.

[0199] In step E, Optical Character Recognition technology is used to extract key building information from the planar diagram, such as floor height, material markings, etc., and based on the identified materials, the corresponding reflection coefficient and absorption coefficient are queried or calculated.

[0200] Step F: The electronic device segments the planar diagram based on a semantic segmentation model to obtain different building areas and building elements, classifies these building areas and target building elements, and extracts building spatial information from tire classification results. In step F, a pre -trained semantic segmentation model is used to segment the planar diagram areas, and based on the building areas and specific building elements (beams, columns, windows, stairs, etc.), the floor boundaries, wall thickness, and other building spatial information are extracted from the features of different classifications.

[0201] Step G: The electronic device identifies the target area from the planar diagram, detects the planar diagram based on image recognition technology, and obtains tire base station installation points marked in the planar diagram, with the target area being the area in the planar diagram that requires targeted network coverage.

[0202] Tn step G, specific areas that require targeted network coverage are identified from the drawings, and the base station installation points marked in the drawings are detected and located using image recognition technology.

[0203] Step H: The electronic device records the position point information and / or reflection point location in the planar diagram, and based on this position point information and / or reflection point location, calculates at least one of the direct path, reflection path, direct path phase angle, and reflection path phase angle.

[0204] Tn step H, the position point information and reflection point location in the image are recorded for subsequent calculations of direct paths and reflection paths, direct path phase angles, and reflection path phase angles.

[0205] For language modality data, in response to the standardized data or the target standardized data being language modality data, the electronic device can perform at least one of the following steps I to step J on the standardized data or target standardized data:

[0206] Step I: The electronic device uses a Named Entity Recognition (NER) model to extract keywords from the text to obtain at least one of tire following: building structure information, building material information, and building spatial information.

[0207] Tn step 1, a trained Named Entity Recognition model is used to extract keywords from the text, such as information about building structures, materials, and spaces.

[0208] Step J: The electronic device uses a relationship extraction model to determine the relationships between entities in the text.

[0209] In step J, a relationship extraction model is used to determine the relationships between entities in the text, such as which materials are used for specific building structures.

[0210] It should be noted that the aforementioned steps A to step J are only an illustrative example, and only building -related features are used as an example. In actual applications, the electronic device can also extract user requirement features from standardized data of different modalities. User requirements can include first requirement features and can also include second requirement features. For details, please refer to the aforementioned content, which is not further elaborated here. Of course, relevant technical personnel can set which features to extract based on what technology according to needs or experience. This embodiment of the application does not limit this. After preprocessing the input data of multiple modalities and then extracting features to obtain their respective features, the extracted features can be merged into a feature vector, that is, the features of multiple modalities are aligned and fused to obtain a comprehensive feature vector. In this way, the artificial intelligence model can process the feature vector to generate a wireless private network signal coverage map.

[0211] In some embodiments, the features of the standardized data include building structure information, building material information, building spatial information, and user requirement information.

[0212] In some embodiments, the target feature vector includes multiple dimensions, and the target feature classification is used to indicate the dimensions in which different classified features arc located in the target feature vector. Accordingly, step 204 can be: the electronic device determines the dimension in the target feature vector corresponding to each extracted feature according to the target feature classification, and writes each extracted feature into the corresponding dimension in the target feature vector.

[0213] This process means that relevant technical personnel can preset the dimensions of the target feature vector according to needs or experience in advance, and then after extracting the features of the standardized data, they can fill them into the corresponding dimensions of the target feature vector. For example, if the target feature vector sets the floor height in the third dimension, the extracted feature of floor height can be filled into the third dimension of the target feature vector. In some embodiments, for features not extracted from the input data, such as user requirement-related information, building material information, building structure information, or building spatial information, preset default values are used instead.

[0214] 204. The electronic device matches the target feature vector with the three-dimensional structural diagram, and estimates the signal strength level for each grid point corresponding to the base station in the three-dimensional structural diagram based on the matching results and the feature values of each dimension in the target feature vector.

[0215] After obtaining the target feature vector and the three-dimensional structural diagram, the electronic device can match their positions. Specifically, the position features of the target feature vector can be matched with the positions in the three-dimensional structural diagram. In this way, each grid point in the three-dimensional structural diagram will have a matched feature component, which is also the feature value of each dimension in the target feature vector. Thus, the signal strength level can be estimated based on the matched feature components of each grid point.

[0216] In some embodiments, different feature components may require different signal strength estimation methods. The electronic device can process different feature components of each grid point separately to estimate the signal strength level. That is, the position features of the target feature vector are matched with the positions in the three-dimensional structural diagram of the building interior, and then the feature components on each grid point after matching are combined, i.e., material feature components, signal spatial feature components, base station feature components, and propagation path feature components. Then, according to the characteristics of each feature component, they are processed separately, and finally, the signal strength level for a certain grid point corresponding to a certain base station is obtained. The specific steps include step one, step two, and step three.

[0217] Step one: The electronic device matches the target feature vector with the three- dimensional structural diagram and combines the different feature components matched for each grid point in the matching results.

[0218] Step two: The electronic device processes the values of different feature components matched for each grid point separately according to the correspondence between the types of feature components and the methods of signal strength estimation.

[0219] In some embodiments, the different feature components include at least two of material features, signal spatial features, base station features, and propagation path features. For each type of feature component, the electronic device can adopt different processing methods. Several processing methods for feature components are provided below.

[0220] For material feature components, the electronic device determines the electromagnetic parameters and rcflcction / absorption coefficients of the matcnal indicated by the matcnal feature component based on the electromagnetic parameters and reflection / absorption coefficients of each material. Based on these electromagnetic parameters and rcflcction / absorption coefficients, the signal strength coefficient of the material under different conditions is queried.

[0221] Hie material feature ultimately affects the strength of the signal. Therefore, it is necessary to convert it into a signal strength coefficient according to the attributes of this feature component. After the signal strength of each grid point is finally estimated, this signal strength coefficient is used to weight the signal strength to obtain the final signal strength. The specific steps are: i. Obtain the electromagnetic parameters of the material according to the material information, including but not limited to dielectric constant, conductivity, permeability; etc.: ii. Estimate the impact of the material on signal strength based on the obtained electromagnetic parameters and the absorption coefficient, reflection coefficient, etc., of the material, through pre-calculated tables or pre-trained artificial intelligence models.

[0222] For signal spatial feature components, the electronic device extracts the signal spatial feature component from the target feature vector, uses interpolation technology to fit the signal spatial feature component into a continuous numerical vector, and adds this numerical vector to the material feature component, base station feature component, or propagation path feature component for signal strength estimation.

[0223] Tliis feature reflects the specific area that needs targeted signal coverage. For such features, an embedded encoding method is adopted to convert these discrete feature information into continuous numerical vectors, which are then superimposed on other feature components. This allows regions of the grid map with similar requirements to be close to each other in vector space, and regions with different requirements to be in different vector spaces even if they are highly similar in other features. This achieves the customization of network design and planning schemes according to user requirements.

[0224] For base station feature components, for grid points of base station installation points, the electronic device normalizes the numerical features in the base station feature component, converts the categorical features in the base station feature component into integer encoding, normalizes this integer encoding, and merges the normalized numerical features and categorical features into a first feature vector.

[0225] For numerical features, they can be directly processed into standardized features by normalization. For categorical features, they are first converted into integer-type features by feature encoding, and then normalized; if the point is not a base station installation point, it is not processed, and only the propagation path feature of this grid point is processed.

[0226] For propagation path feature components, the electronic device converts the phase angle in the propagation path feature component into sine and / or cosine values, standardizes the sine and / or cosine values, path length, and loss coefficient, and merges the standardized sine and / or cosine values, path length, and loss coefficient into a second feature vector.

[0227] First, all feature components containing phase angles are transformed into sine and / or cosine values, and then they are standardized along with features such as path length and loss coefficient to have the same scale.

[0228] Step three: The electronic device synthesizes the processing results of different feature components matched for each grid point to estimate the signal strength level for each grid point, and obtains the signal strength level for each grid point corresponding to the base station in the three-dimensional structural diagram.

[0229] In some embodiments, the different feature components include at least two of material features, signal spatial features, base station features, and propagation path features.

[0230] For the first grid point of the base station installation point, the electronic device estimates the signal strength level of the first grid point based on the first feature vector of the first grid point and the signal strength coefficient of the material under different conditions.

[0231] For the second grid point that is not a base station installation point, the electronic device iterates through each grid point that serves as a receiving point, determines the signal strength component of the second grid point through different propagation paths based on the signal strength coefficient of the material under different conditions for each grid point and the propagation path feature component of the second grid point, and determines the signal strength level of the second grid point corresponding to the base station based on this signal strength component.

[0232] For any grid point that is not a base station installation point, the following iterative estimation is required to obtain the final signal strength: iterate through each grid point that serves as a receiving point, obtain the signal strength through different propagation paths on this grid point, for example, for each grid point, input the path length and path phase angle corresponding to direct radiation, first-order reflection, second-order reflection, third-order reflection, and transmission separately five times, combine the penetration loss coefficient, use the trained artificial intelligence model, and weight tire output based on the weight obtained through the material feature to obtain the signal strength of direct radiation, first-order reflection, second-order reflection, third-order reflection, and transmission for these grid points. The final estimated signal strength of this grid point is determined by the signal strength components obtained on different propagation paths.

[0233] 205. The electronic device maps the signal strength level corresponding to tire base station for each grid point in the three-dimensional structural diagram into the diagram, generating a wireless private network signal coverage map within the target building.

[0234] Once the electronic device obtains the signal strength level corresponding to the base station for each grid point, it can display the signal strength differences of different grid points in the three-dimensional structural diagram using a specific method, thus obtaining the indoor signal level distribution.

[0235] Tn some embodiments, by mapping the signal strength of each grid point into the three- dimensional structural diagram of the building interior, the indoor signal level distribution of the building can be obtained. A color-coding method (such as red for weak signal and green for strong signal) is used to generate an indoor signal strength map for the entire building, which is the wireless private network signal coverage map within the target building.

[0236] In some embodiments, the wireless private network signal coverage map within the target building can be generated using a pre-trained artificial intelligence model based on the fused features.

[0237] In some embodiments, the wireless private network signal coverage map can be in the form of a heat map; of course, the wireless private network signal coverage map can also take other forms, and this embodiment of the application does not limit this.

[0238] Tn some embodiments, estimating the signal strength for each grid point in the three- dimensional structural diagram is implemented based on an artificial intelligence model. The electronic device can also use the measured signal strength within the target building after the base station deployment as the label for the target feature vector and the three-dimensional structural diagram, train the artificial intelligence model based on the target feature vector, the three-dimensional structural diagram, and the label, and update the weights of the artificial intelligence model.

[0239] Regarding the aforementioned steps 201 to 205 and their various implementation methods, two specific examples are provided below to detail the aforementioned process.

[0240] In a specific possible example, the wireless private network signal coverage map generation method is applied to a computer system, executed by a node, and the method may include the following steps one to four:

[0241] This embodiment provides a method executed by a node in a computer system, which includes:

[0242] A large office building with many floors and a complex internal structure, including multiple meeting rooms, offices, corridors, rest areas, etc. With the deployment of wireless networks, it is necessary to accurately predict the indoor signal level of the office building to ensure comprehensive and efficient network coverage. This embodiment will describe in detail the entire process from image input to the final signal strength map generation, including model fine-tuning steps based on actual testing.

[0243] Step 1 : Input Preprocessing

[0244] Input data collection: Obtain multiple high-definition building images of the office building, including external facades, corridors of each floor, and main rooms, and ensure that detailed room distribution, door and window locations, and other information are included; obtain indoor spatial feature vector data, which should be organized in a predetermined format and include all necessary parameter information.

[0245] Adjust all building images to a uniform preset resolution (e g., 1920x1080 pixels). Apply image processing techniques, such as grayscale, to reduce computational complexity. Use edge detection algorithms to identify edge information such as walls and pillars, and perform image block segmentation and feature vector mapping.

[0246] Segment the extracted edge feature images into image blocks, and the entire image is transformed into an image block sequence with a length of. Flatten each image block into a vector, and then map the image to a vector with a dimension of D through a learnable linear projection. Then, use a convolutional neural network to identify specific features in the grid diagram, such as the location of doors and windows, room segmentation, etc., map the information of specific features to specific encoded values through a preset dictionary, and add it to the embedded vector and image block projection sequence to get the sequence. At the same time, record the position information of the interior space of the building and add it to the position encoding of the image block to get the sequence.

[0247] Input the sequence containing building features and spatial information into the trained generative neural network to convert the two-dimensional design plan into a three-dimensional structure diagram of the building interior. Rasterize the three-dimensional structure diagram, with each grid point corresponding to a spatial area in the actual building.

[0248] Step 2: Feature Vector Matching

[0249] For the input feature vector, first check for missing values, outliers, or error entries in the data and perform corresponding processing (such as filling, deleting, or correcting). Ensure that all feature values are within a reasonable physical range, such as path lengths not being negative. Since different features may have different dimensions and value ranges, directly inputting these features into the neural network may lead to unstable training or slow convergence. Therefore, it is necessary to standardize or normalize the features to have the same scale. Common methods include min-max normalization (scaling features to the [0,1] interval) or Z-score normalization (adjusting features to a distribution with a mean of 0 and a standard deviation of 1). For categorical features such as material information, use label encoding; for periodic features such as phase angles, use position encoding to transform them into corresponding sine and cosine values. Finally, match the position features of the feature vector with the positions in the three-dimensional structure diagram. Obtain the combination of feature components for each grid point, including: material feature components, signal spatial feature components, base station feature components, and propagation path feature components.

[0250] Step 3: Signal Strength Estimation

[0251] Process the feature components separately:

[0252] For material feature components, collect or obtain the electromagnetic parameters (such as dielectric constant, magnetic permeability, etc.) and reflection / absorption coefficients of each material. Then, based on the electromagnetic parameters and reflection / absorption coefficients, use a pre-calculated table to query' the signal strength coefficients of the material under different frequencies, incident angles, and other conditions, and store these results in a table. Finally, apply the found signal strength coefficients to the corresponding stages of signal processing, such as adjusting signal gain or attenuation.

[0253] For signal spatial feature components, first, extract the discrete spatial feature information of the signal, such as base station location, beam direction, etc., from the feature vector, and use interpolation technology to fit these discrete feature information into continuous numerical vectors. Finally, superimpose these continuous numerical vectors on other feature components.

[0254] For base station feature components, first, for numerical features (such as base station transmission power, antenna gam, etc.), use normalization methods (such as min-max normalization, Z-score normalization) to scale them to the same scale. For categorical features (such as base station ty pe, frequency band, etc ), first convert them to integer encoding, and then use similar normalization methods to scale the integer values to a certain range. Combine the normalized numerical and categorical features into a feature vector for subsequent processing or analysis.

[0255] For propagation path feature components, first convert the phase angle of the propagation path into sine and cosine values to eliminate the periodic impact of the phase angle and facilitate subsequent processing. Then standardize the path length to a relative value based on a certain reference or reference length. Also standardize the loss coefficient to eliminate the impact of different magnitudes. Finally, combine the sine values, cosine values, standardized path length, and loss coefficient into a feature vector to describe the characteristics of the propagation path.

[0256] For each grid point, based on the superimposed feature components, use tire framed artificial intelligence model to calculate the signal strength corresponding to direct radiation, first-order reflection, second-order reflection, third-order reflection, and transmission. Then, weight the output of each component's artificial intelligence model according to the weight obtained from tire material feature to get the final signal strength.

[0257] Finally, map the signal strength of each grid point to the three-dimensional structure diagram and generate a signal strength map using color encoding.

[0258] After deploying the base station, collect actual signal strength data. Combine the signal strength reported by users as anew label with the corresponding grid point feature vector. Finetune the original model with this data to optimize the weights and improve the accuracy of signal strength estimation.

[0259] Result display and application: Display the indoor signal strength map of the entire office building to ensure the efficiency and comprehensiveness of network planning. According to the results after model fine-tuning, make targeted optimizations for w eak signal areas, such as adding base stations or adjusting antenna directions

[0260] Through the above embodiment, it is possible to accurately predict and optimize the indoor signal coverage strength of large office buildings, providing strong support for private network deployment.

[0261] The advantages of the present invention include, but are not limited to, the following aspects: supporting a variety’ of input methods, including architectural images captured by users, architectural videos recorded by users, two-dimensional floor plans or three-dimensional structural diagrams of building interiors, etc. This multimodal input method greatly enhances the flexibility and applicability of the sy stem, allowing users to more conveniently provide input data according to their needs; the system can generate customized network design and planning solutions according to user requirements. By using embedded encoding methods, the signal spatial features specified by users are transformed into continuous numerical vectors, reflecting the differences in user requirements in vector space, thereby meeting specific needs in different scenarios; by comprehensively considering multi-dimensional information such as location features, material features, signal spatial features, base station features, and propagation path features, the system can accurately predict the signal strength of each grid point. This multi-dimensional feature analysis method improves the accuracy and reliability of signal strength estimation; the present invention has abroad application prospect in fields such as 5G netw ork or other wireless network planning, and indoor signal coverage optimization. By improving the precision and reliability of indoor signal coverage, it helps to enhance user experience and meet communication needs in various complex scenarios.

[0262] Hie data processing method of the present invention provides a complete pre-processing and feature extraction method for multimodal data input by users, transforming multimodal 1 data containing complex information into standardized data suitable for artificial intelligence model learning and reasoning, and outputting feature vectors containing key information required for generating 5G signal coverage maps through pre-trained artificial intelligence models. At the same time, the multimodal data processing and feature extraction processes provided are all automated processes that do not require additional user operations. In addition, users only need to input their expected requirements, and the user requirement feature extraction module can extract the expected requirement information of users, making the feature vectors output by the artificial intelligence model more in line with the actual needs of users. Some user requirements can also be used to guide the model to focus on covering specific areas when generating signal coverage heat maps. These operations reduce the threshold for obtaining network planning and design solutions, making it easier for small and medium-sized enterprises to deploy 5G private networks. Of course, it is not limited to 5G private networks.

[0263] In some embodiments, the implementation of the wireless private network signal coverage map generation method provided in this application example helps to generate high-quality training data and annotated data, and improves the performance and reliability of large models in wireless private network (such as 5G network) planning and design.

[0264] The artificial intelligence model used in this application example can flexibly, accurately, and efficiently solve the problems that wireless network planning usually relics on professional team design solutions or professional simulation tool outputs, and helps to promote the popularization and development of private network construction: the present application example can obtain a three-dimensional structural diagram of the building interior through input data processing, and determine the signal strength corresponding to the base station for each grid point in the three-dimensional structural diagram through the target feature vector obtained from the input data, to generate a wireless private network signal coverage map inside the building. The entire process is automated, and users only need to input building data or requirement data to automatically implement wireless private network planning, reducing the threshold for network planning and design solutions, and improving the efficiency and universality of wireless private network planning, and reducing its cost.

[0265] All the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be repeated one by one.

[0266] Figure 3 is a structural schematic diagram of a wireless private network signal coverage map generation device provided by an embodiment of the present application. See Figure 3, the device includes:

[0267] A preprocessing module 301, which is used to preprocess the input data to obtain standardized data in the target format. The input data includes target building data and / or user requirement data for wireless private network planning, and the input data includes one or more modalities;

[0268] A processing module 302, which is used to perform feature extraction and processing on the standardized data in the target format to obtain a three-dimensional structural diagram and a target feature vector of the interior of the target building;

[0269] An estimation module 303, which is used to match the target feature vector with the three- dimensional structural diagram, and based on the matching result and the feature values of each dimension in the target feature vector, estimate the signal strength level corresponding to the base station for each grid point in the three-dimensional structural diagram;

[0270] A generation module 304, which is used to map the signal strength level corresponding to the base station for each grid point in the three-dimensional structural diagram into the diagram to generate a wireless private network signal coverage map inside the target building.

[0271] In some embodiments, the processing module 302 is used to perform any of the following:

[0272] • Extract a three-dimensional structural diagram of the interior of the target building from the input data;

[0273] • Perform feature extraction on the standardized data in the target format to obtain building spatial information, and input the building spatial information into an image generation model, which generates a three-dimensional structural diagram of the interior of the target building based on the building spatial information.

[0274] In some embodiments, the processing module 302 is used to:

[0275] • Input the building spatial information into the image generation model, which generates a predicted two-dimensional floor plan of the interior of the target building based on the building spatial information; generate a three-dimensional structural diagram of the interior of the target building based on the two-dimensional floor plan using a generative neural network;

[0276] • Divide the three-dimensional structural diagram into different grid point areas according to the target distance, with each grid point area corresponding to a spatial area with physical significance in the target building space.

[0277] In some embodiments, the input data includes at least one of architectural images, architectural videos, two-dimensional structural diagrams, or three-dimensional structural diagrams of the interior of the building;

[0278] The preprocessing module 301 and the processing module 302 are used to:

[0279] • Preprocess the input data to obtain an image in the target format;

[0280] • Convert the image in the target format into a unified raster format and resolution, perform edge detection on the converted image to obtain edge feature information;

[0281] • Perform image segmentation and linear projection mapping on the converted image to obtain a spatial vector corresponding to the sequence of image blocks for each image;

[0282] • Use a convolutional neural network to perform feature recognition and encoding on the converted image to obtain an embedded vector containing building feature information for each image;

[0283] • Record the position information and / or relative position information of the interior space of the target building to obtain position encoding for the sequence of image blocks; • Add the embedded vector and position encoding for each image to the spatial vector corresponding to the sequence of image blocks for that image to obtain a sequence containing building spatial information.

[0284] In some embodiments, the estimation module 303 is used to:

[0285] • Match the target feature vector with the three-dimensional structural diagram, and combine the different feature components matched for each grid point in the matching result;

[0286] • Process the values of different feature components matched for each grid point separately according to the correspondence between the types of feature components and the methods of signal strength estimation;

[0287] • Synthesize the processing results of different feature components matched for each grid point to estimate the signal strength for each grid point, and obtain the signal strength level corresponding to the base station for each grid point in tire three-dimensional structural diagram.

[0288] In some embodiments, the different feature components include at least two of material features, signal spatial features, base station features, and propagation path features;

[0289] The estimation module 303 is used to perform at least two of the following:

[0290] • For material feature components, determine the electromagnetic parameters and reflection / absorption coefficients of the material indicated by the material feature component based on the electromagnetic parameters and reflection / absorption coefficients of each material; query the signal strength coefficient of the material under different conditions based on the electromagnetic parameters and reflection / absorption coefficients;

[0291] • For signal spatial feature components, extract the signal spatial feature component from the target feature vector, use interpolation technology to fit the signal spatial feature component into a continuous numerical vector, and add the numerical vector to the material feature component, base station feature component, or propagation path feature component for signal strength estimation;

[0292] • For base station feature components, for grid points that are base station installation points, normalize the numerical features in the base station feature component, convert the categorical features in the base station feature component into integer encoding, normalize the integer encoding, and merge the normalized numerical features and categorical features into a first feature vector;

[0293] • For propagation path feature components, convert the phase angle in the propagation path feature component into sine and / or cosine values, standardize the sine and / or cosine values, path length, and loss coefficient, and merge the standardized sine and / or cosine values, path length, and loss coefficient into a second feature vector.

[0294] In some embodiments, the estimation module 303 is used to:

[0295] • For the first grid point that is a base station installation point, estimate the signal strength of the first grid point based on the first feature vector of the first grid point and the signal strength coefficient of the material under different conditions; • Forthe second grid point that is not a base station installation point, iterate through each grid point that serves as a receiving point, determine the signal strength component of the second grid point through different propagation paths based on the signal strength coefficient of the material under different conditions for each grid point and the propagation path feature component of the second grid point, and determine the signal strength level corresponding to the base station for the second grid point based on this signal strength component.

[0296] In some embodiments, the estimation of signal strength for each grid point in the three- dimensional structural diagram is implemented based on an artificial intelligence model: the device also includes a training module, which is used to:

[0297] • Use the signal strength measured within the target building after the deployment of the base station as the label for the target feature vector and the three-dimensional structural diagram, train the artificial intelligence model based on the target feature vector, the three- dimensional structural diagram, and the label, and update the weights of the artificial intelligence model.

[0298] The device provided by the present application example uses an artificial intelligence model to flexibly, accurately, and efficiently solve the problems that wireless network planning usually relies on professional team design solutions or professional simulation tool outputs, and helps to promote the popularization and development of private network construction; the present application example can obtain a three-dimensional structural diagram of the interior of the building through input data processing, and determine the signal strength corresponding to the base station for each grid point in the three-dimensional structural diagram through the target feature vector obtained from the input data, to generate a wireless private network signal coverage map inside the building. The entire process is automated, and users only need to input building data or requirement data to automatically implement wireless private network planning, reducing the threshold for network planning and design solutions, and improving the efficiency and universality of wireless private network planning, and reducing its cost.

[0299] It should be noted that: the wireless private network signal coverage map generation device provided by the above embodiment is only exemplified by the division of the above functional modules when generating the wireless private network signal coverage map. In actual application, the above functions can be allocated to different functional modules as needed, that is, the internal structure of the wireless private network signal coverage map generation device is divided into different functional modules to complete all or part of the functions described above. In addition, the wireless private network signal coverage map generation device provided by the above embodiment belongs to the same conception as the wireless private network signal coverage map generation method embodiment, and the specific implementation process is detailed in the method embodiment, which is not repeated here.

[0300] Figure 4 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. The electronic device 400 can vary significantly due to different configurations or performance, including one or more processors (Central Processing Units, CPUs) 401 and one or more memories 402. The memory 402 stores at least one computer program, which is loaded and executed by the processor 401 to implement the wireless private network signal coverage map generation method provided by the various method embodiments described above. The electronic device also includes other components for implementing the device's functions. For example, the electronic device also has wired or wireless network interfaces and input / output interfaces, etc., for input and output. This embodiment of the application does not provide further elaboration.

[0301] The electronic device in the above method embodiments is implemented as a terminal. For example, Figure 5 is a structural block diagram of a terminal provided by an embodiment of the present application. The terminal 500 can be a portable mobile terminal, such as a smartphone, tablet computer, MP3 (Moving Picture Experts Group Audio Layer III, dynamic image expert compression standard audio level 3) player, MP4 (Moving Picture Experts Group Audio Layer IV, dynamic image expert compression standard audio level 4) player, laptop computer, or desktop computer. The terminal 500 may also be referred to as a user device, portable terminal, laptop terminal, desktop terminal, or other names.

[0302] Typically, the terminal 500 includes: a processor 501 and a memory 502.

[0303] The processor 501 may include one or more processing cores, such as a 4-corc processor, an 8-core processor, etc. The processor 501 may be implemented using at least one hardware form of DSP (Digital Signal Processing, digital signal processing), FPGA (Field- Programmable Gate Array, field-programmable gate array), PLA (Programmable Logic Array, programmable logic array). The processor 501 may also include a main processor and a coprocessor, where the main processor is a processor used for processing data in the wake state, also known as the CPU (Central Processing Unit, central processing unit); the coprocessor is a low-power processor used for processing data in the standby state. In some embodiments, the processor 501 may integrate a GPU (Graphics Processing Unit, graphics processor), which is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, the processor 501 may also include an Al (Artificial Intelligence, artificial intelligence) processor, which is used to process computational operations related to machine learning.

[0304] The memory' 502 may include one or more computer-readable storage media, which may be non-volatile. The memory 502 may also include high-speed random access memory, as well as non-volatile memory, such as one or more disk storage devices, flash memory' storage devices. In some embodiments, the non-volatile computer-readable storage medium in the memory' 502 is used to store at least one instruction, which is executed by the processor 501 to implement the wireless private network signal coverage map generation method provided by the method embodiments in this application.

[0305] In some embodiments, the terminal 500 may also optionally' include: a peripheral device interface 503 and at least one peripheral device. The processor 501, memory’ 502, and peripheral device interface 503 may be connected by a bus or signal line. Each peripheral device may be connected to the peripheral device interface 503 via a bus, signal line, or circuit board. Specifically, the peripheral devices include at least one of: RF circuit 504, display 505, camera module 506, audio circuit 507, positioning component 508, and power supply 509.

[0306] Tire peripheral device interface 503 may be used to connect at least one peripheral device related to I / O (Input / Output, input / output) to the processor 501 and memory' 502. In some embodiments, the processor 501, memory’ 502, and peripheral device interface 503 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 501, memory 502, and peripheral device interface 503 may be implemented on separate chips or circuit boards, and this embodiment docs not limit this.

[0307] The RF circuit 504 is used to receive and transmit RF (Radio Frequency, radio frequency’) signals, also known as electromagnetic signals. The RF circuit 504 communicates with communication networks and other communication devices through electromagnetic signals. The RF circuit 504 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the RF circuit 504 includes: an antenna sy stem, an RF transceiver, one or more amplifiers, tuners, oscillators, digital signal processors, codec chipsets, Universal Integrated Circuit Cards (UICCs), etc. The RF circuit 504 can communicate with other terminals through at least one wireless communication protocol. This wireless communication protocol includes but is not limited to: the World Wide Web, metropolitan area networks, intranets, generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks, and / or WiFi (Wireless Fidelity, wireless fidelity) networks, hi some embodiments, the RF circuit 504 may’ also include circuits related to NFC (Near Field Communication, near-field wireless communication), and this application does not limit this.

[0308] The display 505 is used to display the UI (User Interface, user interface). The UI may’ include graphics, text, icons, videos, and any combination thereof. When the display 505 is a touch screen, the display 505 also has the ability to capture touch signals on the surface or above the surface of the display 505. The touch signal can be input as a control signal to the processor 501 for processing. At this time, the display 505 can also be used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards. In some embodiments, the display 505 may’ be one, set on the front panel of the terminal 500: in other embodiments, the display 505 may’ be at least two, set on different surfaces of the terminal 500 or in a folding design; in other embodiments, the display 505 may be a flexible display, set on the curved surface or folding surface of the terminal 500. Even the display 505 can also be set to an irregular shape that is not a rectangle, that is, an irregular screen. The display 505 can be made of materials such as LCD (Liquid Crystal Display, liquid cry stal display) and OLED (Organic Light-Emitting Diode, organic light-emitting diode). The camera module 506 is used to capture images or videos. Optionally, the camera module 506 includes a front camera and a rear camera. Typically, the front camera is set on the front panel of the terminal, and the rear camera is set on the back of the terminal. In some embodiments, the rear camera is at least two, which are any one of the main camera, depth of field camera, wide-angle camera, and telephoto camera, to achieve background blurring functions by fusing the main camera and the depth of field camera, panoramic shooting and VR (Virtual Reality, virtual reality) shooting functions or other fusion shooting functions by fusing the main camera and the wide-angle camera. In some embodiments, the camera module

[0309] 506 may also include a flash. The flash can be a monochromatic flash or a dual-color temperature flash. A dual-color temperature flash refers to the combination of warm light flash and cold light flash, which can be used for light compensation under different color temperatures.

[0310] The audio circuit 507 may include a microphone and a speaker. The microphone is used to capture user and environmental sound waves and convert the sound waves into electrical signals input to the processor 501 for processing, or input to the RF circuit 504 to achieve voice communication. For the purpose of stereo sound capture or noise reduction, there may be multiple microphones, respectively set in different parts of the terminal 500. The microphone can also be an array microphone or an omnidirectional capture-type microphone. The speaker is used to convert the electrical signals from the processor 501 or the RF circuit 504 into sound waves. The speaker can be a traditional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert electrical signals into sound waves that humans can hear but also convert electrical signals into sound waves that humans cannot hear for ranging and other purposes, hi some embodiments, the audio circuit

[0311] 507 may also include a headphone jack.

[0312] The positioning component 508 is used to locate the current geographical location of the terminal 500 to achieve navigation or LBS (Location Based Service, location-based service). The positioning component 508 can be a positioning component based on the American GPS (Global Positioning System, global positioning system), the Chinese Beidou system, or the Russian GLONASS system.

[0313] The power supply 509 is used to power various components in the terminal 500. The power supply 509 can be AC, DC, disposable batteries, or rechargeable batteries. When the power supply 509 includes a rechargeable battery, the rechargeable battery can be a wired battery or a wireless battery. The wired battery is a battery' that is charged through wired lines, and the wireless battery is a battery that is charged through wireless coils. The rechargeable battery can also be used to support fast charging technology.

[0314] In some embodiments, the terminal 500 may also include one or more sensors 510. The one or more sensors 510 include but are not limited to: an acceleration sensor 511, a gyroscope sensor 512, a pressure sensor 513, a fingerprint sensor 514, an optical sensor 515, and a proximity sensor 516.

[0315] The acceleration sensor 511 can detect the acceleration magnitude on the three coordinate axes of the coordinate system established by the terminal 500. For example, the acceleration sensor 511 can be used to detect the components of gravitational acceleration on the three coordinate axes. The processor 501 can control the display 505 to display the user interface in landscape or portrait view based on the gravitational acceleration signal collected by the acceleration sensor 511. The acceleration sensor 511 can also be used for games or user motion data collection.

[0316] The gyroscope sensor 512 can detect the orientation and rotation angle of the terminal 500, and the gyroscope sensor 512 can work in conjunction with the acceleration sensor 511 to capture the user's 3D movements with respect to the terminal 500. The processor 501, based on the data collected by the gyroscope sensor 512, can implement functions such as: motion sensing (e.g., changing the UI based on the user's tilting operation), image stabilization during photography, game control, and inertial navigation.

[0317] The pressure sensor 513 can be set on the side frame and / or under the display 505 of the terminal 500. When the pressure sensor 513 is set on the side frame of the terminal 500, it can detect the user's grip signal on the terminal 500, and the processor 501 can recognize left and right hand use or perform quick operations based on the grip signal collected by the pressure sensor 513. When the pressure sensor 513 is set under the display 505, the processor 501 can control operable controls on the UI interface based on the pressure operations of the user on the display 505. The operable controls include at least one of button controls, scroll bar controls, icon controls, and menu controls.

[0318] Hie fingerprint sensor 514 is used to collect the user's fingerprints, and the processor 501 identifies the user's identity based on the fingerprints collected by the fingerprint sensor 514, or the fingerprint sensor 514 identifies the user's identity based on the collected fingerprints. When tire user's identity is recognized as a trusted identity, the processor 501 authorizes the user to perform related sensitive operations, including unlocking the screen, viewing encrypted information, downloading software, making payments, and changing settings. The fingerprint sensor 514 can be set on the front, back, or side of the terminal 500. When the terminal 500 has physical buttons or a manufacturer's logo, the fingerprint sensor 514 can be integrated with the physical buttons or the manufacturer's logo.

[0319] The optical sensor 515 is used to collect ambient light intensity. In one embodiment, the processor 501 can control the display brightness of the display 505 based on the ambient light intensity collected by the optical sensor 515. Specifically , when the ambient light intensity is high, the display brightness of the display 505 is increased; when the ambient light intensity is low, the display brightness of the display 505 is decreased. In another embodiment, the processor 501 can also dynamically adjust the shooting parameters of the camera module 506 based on the ambient light intensity collected by the optical sensor 515. The proximity sensor 516, also known as the distance sensor, is typically set on the front panel of tire terminal 500. The proximity' sensor 516 is used to collect the distance between the user and the front of the terminal 500. In one embodiment, when the proximity sensor 516 detects that the distance between the user and the front of the terminal 500 is gradually7decreasing, the processor 501 controls the display 505 to switch from the bright screen state to the screen-off state; when the proximity sensor 516 detects that the distance between the user and the front of the terminal 500 is gradually increasing, the processor 501 controls the display7505 to switch from the screen-off state to the bright screen state.

[0320] Persons skilled in the art can understand that the structure shown in Figure 5 does not limit the terminal 500 and can include more or fewer components than shown, combine certain components, or adopt different component layouts.

[0321] The electronic device in tire above method embodiments is implemented as a server. For example. Figure 6 is a structural schematic diagram of a server provided by an embodiment of the present application. The server 600 can vary significantly due to different configurations or performance, including one or more processors (Central Processing Units, CPUs) 601 and one or more memories 602, wherein the memory7602 stores at least one computer program, which is loaded and executed by the processor 601 to implement the wireless private network signal coverage map generation method provided by the various method embodiments described above. Of course, the server also has wired or w'ireless network interfaces as well as input / output interfaces for input and output, and the server includes other components for implementing the device’s functions, which arc not further described here.

[0322] In an exemplary implementation, a computer-readable storage medium is also provided, such as a memory' including at least one computer program, which can be executed by a processor to complete the wireless private network signal coverage map generation method described in the above embodiments. For example, the computer-readable storage medium is Read-Only Memory7(ROM), Random Access Memory (RAM), Compact Disc Read-Only7Memory (CD-ROM), magnetic tape, floppy disk, and optical data storage devices.

[0323] In an exemplary implementation, a computer program product or computer program is also provided, which includes one or more program codes stored on a computer-readable storage medium. One or more processors of an electronic device read the one or more program codes from the computer-readable storage medium, and the one or more processors execute the one or more program codes, causing the electronic device to perform the wireless private network signal coverage map generation method described above.

[0324] In some embodiments, the computer program involved in the present application embodiment may be deployed to be executed on a computer device, or executed on multiple computer devices located at a single site, or executed on multiple computer devices distributed across multiple sites and interconnected by a communication network, with the multiple computer devices distributed across multiple sites and interconnected by a communication network forming a blockchain system.

[0325] The ordinary technician in the field understands that all or part of the steps of the above embodiments are completed by hardware, and also by a program instructing the relevant hardware to complete, with the program stored on a computer-readable storage medium, the storage medium mentioned above being Read-Only Memory, disk, or disc, etc.

[0326] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present application, should be included within the protection scope of the present application.

Claims

CLAIMSWhat is claimed is:

1. A method for generating a wireless private network signal coverage map, the method comprising:Preprocessing the input data to obtain standardized data in the target format, said input data including target building data and / or user demand data for wireless private network planning, said input data including one or more modalities;Feature extraction and processing on the standardized data in the taiget format to obtain a three-dimensional structure diagram and taiget leal lire vectors inside the target building;Matching the target feature vectors with the three-dimensional structure diagram, and estimating the signal strength corresponding to the base station for each grid point in the three- dimensional structure diagram according to the matching result and the feature values of each dimension in the target feature vectors;Mapping the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram to the three-dimensional structure diagram to generate a wireless private network signal coverage map inside the target building.

2. The method of claim 1 , wherein the feature extraction and processing on the standardized data in the target format to obtain a three-dimensional structure diagram inside the target building, comprises any one of the following:Extracting a three-dimensional structure diagram of the interior of the target building from the input data;Feature extraction on the standardized data in the target format to obtain building spatial information, inputting the building spatial information into an image generation model, and generating a three-dimensional structure diagram of the interior of the taiget building based on the building spatial information by the image generation model.

3. The method of claim 2, wherein inputting the building spatial information into the image generation model and generating a three-dimensional structure diagram of the interior of the target building based on the building spatial information by the image generation model, comprises:Inputting the building spatial information into the image generation model to generate a predicted two-dimensional design drawing of the interior of the target building based on the building spatial information; generating a three-dimensional structure diagram of the interior of the target building based on the two-dimensional design drawing by a generative neural network;Dividing the three-dimensional structure diagram into different grid point areas accordingto the taiget distance, each grid point area corresponding to a spatial area with physical significance in the space of the target building.

4. The method of claim 2. wherein the input data includes at least one of building images, building videos, two-dimensional structure diagrams, or three-dimensional structure diagrams of the interior of the building;The preprocessing of the input data to obtain standardized data in the target format; the feature extraction on the standardized data in the target format to obtain building spatial information, comprises:Preprocessing the input data to obtain standardized images in the target format;Converting the standardized images into a unified raster format and resolution, performing edge detection on the converted images to obtain edge feature information;Segmenting the converted images and performing linear projection mapping to obtain spatial vectors corresponding to the image block sequence of each image;Feature recognition and encoding of the converted images based on a convolutional neural network to obtain an embedded vector containing building feature infonnation for each image;Recording position information and / or relative position information of the interior space of the target building to obtain position encoding of the image block sequence;Adding the embedded vector and position encoding of each image to the spatial vector corresponding to the image block sequence to obtain a sequence containing building spatial information.

5. Tire method of claim 1, wherein matching the target feature vectors with the three- dimensional structure diagram, and estimating the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram according to the matching result and the feature values of each dimension in the target feature vectors, comprises:Matching the target feature vectors with the three-dimensional structure diagram, combining different feature components matched for each grid point in the matching result of the taiget feature vectors;Processing the values of different feature components matched for each grid point according to the corresponding relationship between the type of feature component and the signal strength estimation method;Synthesizing the processing results of different feature components matched for each grid point to estimate the signal strength for each grid point to obtain the signal strength corresponding to the base station for each grid point in the three-dimensional structure diagram.

6. A method according to claim 5, characterized in that, the different feature componentsinclude at least two of material characteristics, signal spatial characteristics, base station characteristics, and propagation path characteristics, and processing the values of different feature components matched at each grid point according to the corresponding relationship between the type of feature component and the signal strength estimation method, including at least two of the following:For material characteristic components, detennine the electromagnetic parameters and reflection / absorption coefficients of the material indicated by the material characteristic component: based on the electromagnetic parameters and reflection / absorption coefficients, query the signal strength coefficient of the material under different conditions; For signal spatial characteristic components, extract the signal spatial characteristic component from the target feature vector, use interpolation technology to fit the signal spatial characteristic component into a continuous numerical vector, and add tire numerical vector to the material characteristic component, base station characteristic component, or propagation path characteristic component for signal strength estimation;For base station characteristic components, for grid points where the base station is installed, normalize the numerical features in the base station characteristic component, convert the categorical features in the base station characteristic component into integer encoding, normalize the integer encoding, and merge the normalized numerical features and categorical features into a first feature vector;For propagation path characteristic components, convert the phase angles in the propagation path characteristic components into sine and / or cosine values, standardize the sine and / or cosine values, path length, and loss coefficients, and merge the standardized sine and / or cosine values, path length, and loss coefficients into a second feature vector.

7. A method according to claim 6, characterized in that, synthesizing the processing results of different feature components matched at each grid point to estimate tire signal strength at each grid point to obtain the signal strength corresponding to the base station at each grid point in the three-dimensional structure diagram, including:For the first grid point where the base station is installed, estimate the signal strength of the first grid point based on the first feature vector of the first grid point and the signal strength coefficient of the material under different conditions;For the second grid point where the base station is not installed, iterate through each grid point that serves as a receiving point, detennine the signal strength components of the second grid point through different propagation paths based on the signal strength coefficient of the material at different conditions and the propagation path characteristic components of the second grid point, and determine the signal strength corresponding to the base station at the second grid point based on the signal strength components.

8. A device for generating a wireless private network signal coverage map, characterized in that, the device includes:A preprocessing module for preprocessing input data to obtain standardized data in the target format, said input data including target building data and / or user requirement data for wireless private network planning, said input data including one or more modalities;A processing module for extracting features from the standardized data in the target format and processing to obtain a three-dimensional structure diagram and target feature vectors of the interior of the target building;An estimation module for matching the target feature vectors with the three- dimensional structure diagram, and estimating the signal strength corresponding to the base station at each grid point in the three-dimensional structure diagram based on the matching results and the feature values of each dimension in the target feature vectors;A generation module for mapping the signal strength corresponding to the base station at each grid point in the three-dimensional structure diagram to the diagram to generate a wireless private network signal coverage map inside the target building.

9. An electronic device, characterized in that, the electronic device includes one or more processors and one or more memories, said one or more memories storing at least one computer program, said at least one computer program being loaded and executed by said one or more processors to implement any of the wireless private network signal coverage map generation methods described in claims 1 to 7.

10. A computer-readable storage medium, characterized in that, tire storage medium stores at least one computer program, said at least one computer program being loaded and executed by a processor to implement any of the wireless private network signal coverage map generation methods described in claims 1 to 7.

Images