Radio map construction method, device and equipment and readable storage medium

Abstract

The embodiment of the invention provides a radio map construction method, device and equipment and a readable storage medium, and belongs to the technical field of radio communication and radio monitoring. The method comprises the following steps: randomly acquiring position coordinates and receiving field intensities of different positions of a specific area; preprocessing the position coordinates of different positions of the area and the receiving field intensity, and generating a first radio map of the area with pixel missing; generating a corresponding mask map according to the first radio map; and fusing the first radio map, the corresponding mask map and the corresponding building distribution map to generate a second radio map of the region. Through a provided radio map construction scheme, a map with pixel missing and a building distribution map are fitted into a complete radio map by using a deep learning model, so that the precision of visual representation of the propagation condition of electric waves is improved.

Description

Technical Field

[0001] The present application relates to the field of radio communication technologies, and in particular, to a method, apparatus, device, and readable storage medium for constructing a radio map. Background Art

[0002] A radio environment map characterizes the field distribution of radio waves in the electromagnetic space, which can be represented by the spatial electric field distribution, power spectral density (PSD) distribution, path loss (PL), etc. It makes a quantitative description of the propagation characteristics of radio waves in the electromagnetic space from multiple dimensions such as time, frequency spectrum, space, and power. The construction of a radio map is closely related to the planning of wireless communication networks and radio management. At the same time, radio maps can also be applied to wireless positioning, unmanned aerial vehicle path planning, and automotive autonomous driving application scenarios.

[0003] The construction of a radio map in the prior art is carried out by sampling. The prediction accuracy of this method for generating a radio map is directly proportional to the number of samples. A large amount of data collection will incur huge time and cost expenses, while with fewer sampled data, the accuracy of constructing a radio map is low. Summary of the Invention

[0004] To solve the above technical problems, embodiments of the present application provide a method, apparatus, device, and readable storage medium for constructing a radio map.

[0005] In a first aspect, embodiments of the present application provide a method for constructing a radio map, the method including:

[0006] Randomly collect the position coordinates and received field strengths at different positions in a specific area in the area;

[0007] Preprocess the position coordinates and received field strengths at different positions in the area to generate a first radio map of the area with pixel missing;

[0008] Generate a corresponding mask map according to the first radio map;

[0009] Fuse the first radio map, the corresponding mask map, and the corresponding building distribution map to generate a second radio map of the area.

[0010] In an implementation manner, preprocessing the received field strengths at different positions in the area includes:

[0011] Determine the true path loss value according to the received field strength;

[0012] Normalize the true path loss value and compare it with a critical value to obtain a mapping value;

[0013] Divide the mapping value into multiple pixel gray levels, where the multiple pixel gray levels correspond to different pixel values.

[0014] In one embodiment, after mapping the received field strength to a pixel value, the method further includes:

[0015] Set a truncation threshold and obtain a path loss threshold under the set truncation threshold;

[0016] Determine whether the true path loss value is greater than or equal to the path loss threshold;

[0017] If the true path loss value is greater than or equal to the path loss threshold, the pixel value is transformed according to the truncation threshold;

[0018] If the true path loss value is less than the path loss threshold, set the pixel value to 0.

[0019] In one embodiment, generating a corresponding mask map according to the first radio map includes:

[0020] Convert the position coordinates collected in the area to 0 and convert the position coordinates not collected in the area to 1 to generate a mask matrix.

[0021] In one embodiment, the fusion of the first radio map, the corresponding mask map, and the corresponding building distribution map includes:

[0022] Build a semantic enhancement model on the baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map.

[0023] In one embodiment, the semantic enhancement model includes an upper branch and a lower branch. Building a semantic enhancement model on the baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map includes:

[0024] The upper branch respectively performs splitting processing, conversion processing, and aggregation processing on the first radio map, its corresponding mask map, and its corresponding building distribution map to obtain first data, second data, and third data, and fuses the first data, the second data, and the third data to obtain first semantic sub-information;

[0025] The lower branch updates the mask matrix using gated convolution to obtain second semantic sub-information;

[0026] Perform weighted fusion on the first semantic sub-information and the second semantic sub-information to obtain the semantic information.

[0027] In one embodiment, the upper branch's splitting and conversion steps include:

[0028] Decompose the feature maps of the first radio map, the corresponding mask map, and the corresponding building distribution map through the splitting step;

[0029] Use dilated convolution to extract the feature information of the feature maps.

[0030] In a second aspect, an embodiment of the present application provides a radio map construction device, which includes:

[0031] An acquisition module, configured to randomly acquire the position coordinates and received field strengths at different positions in the specific area;

[0032] A preprocessing module, configured to preprocess the position coordinates and received field strengths at different positions in the area to generate a first radio map of the area with pixel missing;

[0033] A first generation module, configured to generate a corresponding mask map according to the first radio map;

[0034] A second generation module, configured to fuse the first radio map, the corresponding mask map, and the corresponding building distribution map to generate a second radio map of the area.

[0035] In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor, where the memory is used to store a computer program, and the computer program executes the radio map construction method provided in the first aspect when running on the processor.

[0036] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program executes the radio map construction method provided in the first aspect when running on a processor.

[0037] For the radio map construction method provided in the above present application, randomly acquire the position coordinates and received field strengths at different positions in the specific area; preprocess the position coordinates and received field strengths at different positions in the area to generate a first radio map of the area with pixel missing; generate a corresponding mask map according to the first radio map; fuse the first radio map, the corresponding mask map, and the corresponding building distribution map to generate a second radio map of the area. Through the provided radio map construction scheme, use a deep learning model to fit the map with pixel missing and the building distribution map into a complete radio map, improving the accuracy of the visual representation of the radio wave propagation situation. Description of the Drawings

[0038] To more clearly illustrate the technical solutions of the present application, the following will briefly introduce the accompanying drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application, and thus should not be regarded as limiting the protection scope of the present application. In each drawing, similar components are numbered similarly.

[0039] Figure 1 Fig. 5 shows a schematic flowchart of a radio map construction method provided by an embodiment of the present application;

[0040] Figure 2 Fig. 9 shows a schematic diagram of an experimental result of a radio map construction method provided by an embodiment of the present application;

[0041] Figure 3 Fig. 13 shows another schematic diagram of an experimental result of a radio map construction method provided by an embodiment of the present application;

[0042] Figure 4 Fig. 17 shows a schematic diagram of a semantic enhancement model provided by an embodiment of the present application;

[0043] Figure 5 Fig. 21 shows a schematic structural diagram of a radio map construction device provided by an embodiment of the present application.

[0044] Icons: 500 - radio map construction device, 501 - acquisition module, 502 - preprocessing module, 503 - first generation module, 504 - second generation module. Detailed Embodiments

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some, rather than all, embodiments of the present application.

[0046] Generally, the components of the embodiments of the present application described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts belong to the protection scope of the present application.

[0047] As used hereinafter, the terms "comprising", "having" and their cognates that may be used in various embodiments of the present application are only intended to indicate specific features, numbers, steps, operations, elements, components, or combinations of the foregoing items, and should not be construed as precluding the existence of one or more other features, numbers, steps, operations, elements, components, or combinations of the foregoing items in the first place or as precluding the possibility of adding one or more features, numbers, steps, operations, elements, components, or combinations of the foregoing items.

[0048] In addition, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0049] Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as that commonly understood by those of ordinary skill in the art to which various embodiments of the present application pertain. The terms (such as those defined in commonly used dictionaries) will be construed to have the same meaning as the contextual meaning in the relevant technical field and will not be construed to have an idealized meaning or an overly formal meaning unless clearly defined in various embodiments of the present application.

[0050] Embodiment 1

[0051] An embodiment of the present application provides a method for constructing a radio map.

[0052] See Figure 1 , the method for constructing a radio map includes:

[0053] S101: Randomly collect the position coordinates and receiving field strength (RSS) at different positions in a specific area.

[0054] In this embodiment, first, it is necessary to determine the position and frequency of the emission source, and generate a corresponding labeled radio map according to the frequency and an open-source dataset. The radio map is used to train the fusion of the radio map with pixel missing and the corresponding building distribution map. Under the area of the labeled radio map, sampling points are randomly selected within a given area through a random coordinate generator, and a sensing device or tool with a simulated real discrete distribution is used to measure the receiving field strength of the sampling points. The geopy library of the software can be used to obtain the geographical location coordinates, and scipy or numpy is used to process the receiving field strength. The position coordinates and receiving field strength data are stored in a database or a file for further analysis.

[0055] S102: Preprocess the position coordinates and receiving field strength at different positions in the area to generate a first radio map of the area with pixel missing.

[0056] In this embodiment, before constructing the radio map, the collected data is preprocessed to make it more suitable for subsequent map construction and analysis. The preprocessing includes preprocessing such as data pixelization, threshold transformation, mask generation, and data tensorization of the collected data, to obtain a radio map with pixel missing and a mask map.

[0057] In one embodiment, preprocessing the received field strength at different positions in the region includes: determining a true path loss value according to the received field strength; comparing the normalized true path loss value with a critical value to obtain a mapping value; dividing the mapping value into multiple pixel gray levels, and the multiple pixel gray levels correspond to different pixel values.

[0058] In this embodiment, when an electromagnetic wave encounters an obstacle during propagation, diffraction, reflection, refraction and other phenomena will occur. These phenomena will cause the propagation path of the electromagnetic wave to extend, the propagation direction to change, and the energy to further attenuate. Path loss refers to the energy attenuation of the electromagnetic wave during propagation due to the influence of obstacles and scatterers. As the distance increases, the received field strength becomes smaller and smaller, indicating that the path loss becomes larger and larger. It is necessary to map the received field strength to pixel values for visualization. If M1 represents the maximum path loss in the received field strength, and P L,trnc represents the minimum path loss in the received field strength, and P L is the true value of the path loss, it is defined that Normalize the true path loss value and compare it with the critical value 0 to obtain a mapping value f. If f = 0, it means that all path loss values are lower than the path loss value of the analytical background noise; if f = 1, it means that the transmitter has the maximum equivalent isotropic radiated power. The path loss values between the critical values 0 and 1 are divided into different pixel gray levels, and different pixel gray levels correspond to different pixel values.

[0059] In one embodiment, after mapping the received field strength to pixel values, the method further includes: setting a truncation threshold, and obtaining a path loss threshold under the set truncation threshold; determining whether the true path loss value is greater than or equal to the path loss threshold; if the true path loss value is greater than or equal to the path loss threshold, the pixel value is transformed according to the truncation threshold; if the true path loss value is less than the path loss threshold, the pixel value is set to 0.

[0060] In this embodiment, the path loss threshold is used as a critical coefficient for discriminating the background noise level of the radio map. Let (δ) dB = 10log 10 (w×N0)+NF represent the background noise, with the unit of decibel (dB), where w, N0, and NF are the bandwidth, thermal noise PSD, and noise figure respectively. The path loss threshold is expressed as: PL thr = -(PTX ) dB +SNR thr +(δ) dB , where (P TX ) dB is the path loss value of the emission source, and SNR thr is the signal-to-noise ratio under the threshold. In the map area where P L <PL thr , the path loss value is truncated, that is, the pixel value is set to 0, and the background noise level is the highest; in the map area where P L >PL thr , the pixel value is transformed by , P L is the true path loss, thr is the threshold transformation coefficient, the larger the threshold transformation coefficient, the larger the background noise level, and the darker the image.

[0061] S103: Generate a corresponding mask map according to the first radio map.

[0062] In one embodiment, generating a corresponding mask map according to the first radio map includes:

[0063] Convert the position coordinates collected in the area to 0, and convert the position coordinates not collected in the area to 1 to generate a mask matrix.

[0064] In this embodiment, in the construction of the passive radio map, the reference signal received power (RSRP) collected in a specific area is sparse and discrete. The reconstruction task can be analogized to an image completion task, but there are also many differences. First, from an ideal perspective, the distribution of RSRP should satisfy a certain propagation law, rather than some random pixel points. Second, the positions of the collected data are random, that is, the positions to be completed are irregular, which is quite different from image completion. And the area to be completed accounts for about 60%-70% of the target map area. Therefore, it is difficult to directly predict the received field strength in the unmeasured area. To solve this problem, in this embodiment, a mask matrix of the same size as the pixelated sampling area is pre-constructed, and then the coordinates of the sampling points in the area to be constructed are converted to 0, and the coordinates of the non-sampling points in the area to be constructed are converted to 1. The mask function can be expressed as:

[0065]

[0066] where δ represents the complete received field strength map of the area to be constructed. The condition δ (i,j) = NULL indicates that there is data missing at the coordinate point (i,j). In actual calculation, the preprocessing of the mask satisfies: G′(x) = (1 - Mask(δ(i,j) ))⊙G(x). Here, G(x) is the intensity distribution of the label receiving field strength, and G′(x) is the intensity distribution of the sampled receiving field strength. The mask matrix is a binary matrix containing only 0 or 1. Adopting this mask strategy enables the model to label the positions to be completed as regions of interest, which is beneficial for model learning.

[0067] S104: Fuse the first radio map, the corresponding mask map, and the corresponding building distribution map to generate the second radio map of the region.

[0068] In this embodiment, the prediction of the path loss law in the signal propagation process is achieved through deep learning, that is, using a deep learning model to fit the incomplete map and the building distribution map into a complete radio map, and finally visually representing the propagation situation of the radio wave.

[0069] In one embodiment, the fusing of the first radio map, the corresponding mask map, and the corresponding building distribution map includes: constructing a semantic enhancement model on the baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map.

[0070] In this embodiment, the baseline model is formed by two identical U-Nets connected in series. A semantic enhancement (SE) module is added to the encoder layer of the baseline model to enhance the model's ability to represent the semantic information of the radio map, thereby improving the construction quality of the radio map. It should be specifically noted that each encoding layer structure is composed of a convolutional layer - max pooling layer - SE module in cascade. When the pooling stride is 2, the model performs downsampling, that is, for each layer of pooling of the feature map, the high dimension H and the width dimension W are reduced by 1 / 2 proportionally, and the SE module is attached after the pooling layer and shares the resolution and the number of channels with the pooling layer.

[0071] Optionally, in the embodiments of the present application, in order to capture the global information and local texture of the radio map, the optimization objective of the present application is to minimize the mean square error (MSE) to ensure that the radio map has high global accuracy. The loss function can be expressed as: where y (i,j) is the true receiving field strength intensity at the coordinate (i,j), is the predicted receiving field strength intensity at the coordinate (i,j).

[0072] Optionally, in the embodiments of the present application, a test set is used to evaluate the performance of the proposed method to obtain the RMSE error index and the NMSE error index. The RMSE and NMSE functions can be defined respectively as: where N and M represent the width and height of the radio map, respectively, and r (i,j) represents the true received field strength value of the (i, j) grid, and represents the predicted RSS value of the grid. represents the variance of the true RSS value in the (i, j) grid. The lower the value of the above indicators, the better the performance.

[0073] Optionally, the present invention verifies the performance of the proposed model on the RadioMapSeer dataset. The RadioMapSeer dataset consists of 701 maps, each map has 80 transmitter positions, and there are a total of 56,080 simulated radio maps. See Figure 2 and Figure 3 , in this embodiment, a total of 5 different models are compared, and the comparison results of the composition accuracy of different models are as Figure 2 and Figure 3 shown. In the embodiment of the present application, high-precision radio map estimation is achieved with fewer measurement data. Experiments are carried out by setting the number of sampling points of Ω = 30, Ω = 50, Ω = 100, and Ω = 150. Where Ω is a single pixel point, that is, a sampling area of 1*1. Compared with the percentage sampling method, the sampling quantity is significantly reduced, which not only reduces the labor and material costs, but also better conforms to the actual situation of data collection in complex environments. Compared with other models, the model proposed by the present invention has significantly improved RMSE performance and NMSE performance on the test set. Especially in the interval where the number of sampling points Ω ranges from 30 to 100, the performance improvement is the most obvious. Compared with the currently optimal RadioUnet model, the RMSE accuracy and NMSE accuracy are respectively increased by 26.4% and 34.8% on average.

[0074] In one embodiment, see Figure 4 , the semantic enhancement model includes an upper branch and a lower branch. The semantic enhancement model is constructed on the baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map, including:

[0075] The upper branch sequentially performs splitting processing, conversion processing, and aggregation processing on the first radio map, its corresponding mask map, and its corresponding building distribution map, respectively obtaining first data, second data, and third data, and fusing the first data, the second data, and the third data to obtain first semantic sub-information; the lower branch updates the mask matrix using gated convolution to obtain second semantic sub-information; and the first semantic sub-information and the second semantic sub-information are weighted and fused to obtain the semantic information.

[0076] In this embodiment, the semantic enhancement model consists of an upper branch and a lower branch, and the feature information propagates forward along the upper and lower branches simultaneously. In the design of the upper branch, the semantic enhancement module extracts feature information by using an operation method with the sequential steps of splitting, transformation, aggregation, and fusion from left to right. Among them, splitting is to divide the feature maps of the first radio map, its corresponding mask map, and its corresponding building distribution map to extract semantic information features more precisely. The divided feature maps are input into the transformation module for feature extraction, and then input into the aggregation module to splice the obtained semantic information. The semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map is fused. The semantic enhancement module incorporates a fusion mechanism. The addition of the fusion mechanism can not only improve the performance of the module but also effectively prevent the grid effect generated during transformation. It should be noted that in this embodiment, ordinary convolution with a kernel size of 3×3 is used to fuse the aggregated features. Compared with the traditional ASPP model without fusion, the verification loss is significantly reduced.

[0077] For the lower branch, the semantic enhancement module uses gated convolution to regulate the update of the mask, and the information of the upper branch and the lower branch is fused by weighting with the formula x1×g + x2×(1 - g), where g is a learnable adaptive hyperparameter.

[0078] In one implementation manner, the steps of splitting and transformation adopted by the upper branch include: decomposing the feature maps of the first radio map, the corresponding mask map, and the corresponding building distribution map through the splitting step; using dilated convolution to extract the feature information of the feature maps.

[0079] In this embodiment, the radio map construction algorithm uses dilated convolution to design the transformation step of the upper branch of the semantic enhancement module. The receptive field calculation function of dilated convolution can be expressed as: k′ = k + (k - 1)×(d - 1), where k′ is the equivalent convolution kernel size, k is the original convolution kernel size, and d is the dilation factor. Taking a standard convolution with a size of 3×3 as an example, when the dilation rate is not set, the equivalent convolution kernel is still 3×3. When the dilation factor is set to d = 3, the equivalent convolution kernel is 7×7. The dilated convolution is used to further extract the semantic information of the feature maps output by the splitting step.

[0080] The radio map construction method provided in this embodiment randomly collects the position coordinates and received field strength at different positions in a specific area; preprocesses the position coordinates and received field strength at different positions in the area to generate a first radio map of the area with pixel missing; generates a corresponding mask map according to the first radio map; fuses the first radio map, the corresponding mask map and the corresponding building distribution map to generate a second radio map of the area. Through the provided radio map construction scheme, using a deep learning model to fit a map with pixel missing and a building distribution map into a complete radio map improves the accuracy of the visual representation of the radio wave propagation situation.

[0081] Embodiment 2

[0082] In addition, an embodiment of the present application provides a radio map construction device, which is applied to an electronic device.

[0083] As Figure 5 shown, the radio map construction device 500 includes:

[0084] An acquisition module 501, configured to randomly collect the position coordinates and received field strength at different positions in a specific area;

[0085] A preprocessing module 502, configured to preprocess the position coordinates and received field strength at different positions in the area to generate a first radio map of the area with pixel missing;

[0086] A first generation module 503, configured to generate a corresponding mask map according to the first radio map;

[0087] A second generation module 504, configured to fuse the first radio map, the corresponding mask map and the corresponding building distribution map to generate a second radio map of the area.

[0088] Optionally, the preprocessing module 502 is further configured to determine a true path loss value according to the received field strength; compare the normalized true path loss value with a critical value to obtain a mapping value; divide the mapping value into multiple pixel gray levels, and the multiple pixel gray levels correspond to different pixel values.

[0089] Optionally, the preprocessing module 502 is further configured to obtain a path loss threshold under the set truncation threshold by setting a truncation threshold;

[0090] Judge whether the true path loss value is greater than or equal to the path loss threshold;

[0091] If the true path loss value is greater than or equal to the path loss threshold, the pixel value is transformed according to the truncation threshold;

[0092] If the real loss value of the path is less than the path loss threshold, set the pixel value to 0.

[0093] Optionally, the first generation module 503 is further configured to generate a corresponding mask map according to the first radio map, including:

[0094] Convert the position coordinates collected in the area to 0, and convert the position coordinates not collected in the area to 1 to generate a mask matrix.

[0095] Optionally, the second generation module 504 is further configured to fuse the first radio map, the corresponding mask map, and the corresponding building distribution map, including:

[0096] Construct a semantic enhancement model on the baseline model to extract the semantic information of the first radio map, the corresponding mask map, and the corresponding building distribution map.

[0097] Optionally, the second generation module 504 is further configured to set the semantic enhancement model to include an upper branch and a lower branch. The construction of the semantic enhancement model on the baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map includes: the upper branch respectively performs splitting processing, conversion processing, and aggregation processing on the first radio map, its corresponding mask map, and its corresponding building distribution map in sequence to obtain first data, second data, and third data, and performs fusion processing on the first data, the second data, and the third data to obtain first semantic sub-information; the lower branch updates the mask matrix using gated convolution to obtain second semantic sub-information; and performs weighted fusion on the first semantic sub-information and the second semantic sub-information to obtain the semantic information.

[0098] Optionally, the second generation module 504 is further configured to decompose the feature maps of the first radio map, the corresponding mask map, and the corresponding building distribution map through the splitting step;

[0099] Use dilated convolution to extract the feature information of the feature maps.

[0100] The radio map construction device 500 provided in this embodiment can implement the radio map construction method provided in Embodiment 1. To avoid repetition, it will not be elaborated here.

[0101] The radio map construction device provided in this embodiment randomly collects the position coordinates and received field strengths at different positions in a specific area; preprocesses the position coordinates and received field strengths at different positions in the area to generate a first radio map of the area with missing pixels; generates a corresponding mask map according to the first radio map; fuses the first radio map, the corresponding mask map, and the corresponding building distribution map to generate a second radio map of the area. Through the provided radio map construction scheme, the deep learning model is used to fit the map with missing pixels and the building distribution map into a complete radio map, improving the accuracy of the visual representation of the radio wave propagation situation.

[0102] Embodiment 3

[0103] In addition, an embodiment of the present application provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the computer program runs on the processor, it executes the radio map construction method provided in Embodiment 1.

[0104] The electronic device provided in the embodiment of the present invention can execute the steps of the radio map construction method provided in the above Method Embodiment 1. To avoid repetition, it will not be elaborated here.

[0105] Embodiment 4

[0106] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the radio map construction method provided in Embodiment 1.

[0107] In this embodiment, the computer-readable storage medium can be a read-only memory (ROM for short), a random access memory (RAM for short), a magnetic disk, or an optical disc, etc.

[0108] The computer-readable storage medium provided in this embodiment can implement the radio map construction method provided in Embodiment 1. To avoid repetition, it will not be elaborated here.

[0109] It should be noted that in this article, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or terminal including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or terminal. Without more limitations, the element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article, or terminal including that element.

[0110] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases, the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0111] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms without departing from the purpose of the present application and the scope protected by the claims, and all of them belong to the protection scope of the present application.

Claims

1. A method for constructing a radio map, characterized in that, The method includes: Randomly collecting the position coordinates and received field strengths at different positions in a specific area in the area; Preprocessing the position coordinates and received field strengths at different positions in the area to generate a first radio map of the area with missing pixels; Generating a corresponding mask map according to the first radio map; Fusing the first radio map, the corresponding mask map, and the corresponding building distribution map to generate a second radio map of the area.

2. The method according to claim 1, wherein Preprocessing the received field strengths at different positions in the area includes: Determining the true path loss value according to the received field strength; Normalizing the true path loss value and comparing it with a critical value to obtain a mapping value; Dividing the mapping value into multiple pixel gray levels, and the multiple pixel gray levels correspond to different pixel values.

3. The method according to claim 2, wherein After mapping the received field strength to a pixel value, the method further includes: Setting a truncation threshold and obtaining a path loss threshold under the set truncation threshold; Judging whether the true path loss value is greater than or equal to the path loss threshold; If the true path loss value is greater than or equal to the path loss threshold, the pixel value is transformed according to the truncation threshold; If the true path loss value is less than the path loss threshold, the pixel value is set to 0.

4. The method according to claim 1, wherein Generating a corresponding mask map according to the first radio map includes: Converting the position coordinates collected in the area to 0, and converting the position coordinates not collected in the area to 1 to generate a mask matrix.

5. The method according to claim 4, wherein Fusing the first radio map, the corresponding mask map, and the corresponding building distribution map includes: Constructing a semantic enhancement model on a baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map.

6. The method according to claim 5, characterized in that, The semantic enhancement model includes an upper branch and a lower branch. Constructing a semantic enhancement model on a baseline model to extract the comprehensive semantic information of the first radio map, its corresponding mask map, and its corresponding building distribution map includes: The upper branch respectively performs splitting processing, conversion processing, and aggregation processing on the first radio map, its corresponding mask map, and its corresponding building distribution map in sequence to obtain first data, second data, and third data, and fuses the first data, the second data, and the third data to obtain first semantic sub-information; The lower branch updates the mask matrix using gated convolution to obtain second semantic sub-information; Weightedly fusing the first semantic sub-information and the second semantic sub-information to obtain the semantic information.

7. The method according to claim 6, wherein The splitting and conversion steps adopted by the upper branch include: Decomposing the feature maps of the first radio map, the corresponding mask map, and the corresponding building distribution map through the splitting step; Using dilated convolution to extract the feature information of the feature maps.

8. A radio map construction device, characterized in that, The device includes: An acquisition module for randomly collecting the position coordinates and received field strengths at different positions in a specific area in the area; A preprocessing module, configured to preprocess the position coordinates and received field strength at different positions in the area, and generate a first radio map of the area with missing pixels; A first generation module, configured to generate a corresponding mask map according to the first radio map; A second generation module, configured to fuse the first radio map, the corresponding mask map and the corresponding building distribution map, and generate a second radio map of the area.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory stores a computer program, and the computer program executes the radio map construction method according to any one of claims 1 to 7 when running on the processor.

10. A computer-readable storage medium, characterized in that, It stores a computer program, and the computer program executes the radio map construction method according to any one of claims 1 to 7 when running on the processor.

Images