Electromagnetic environment map generation method based on conditional constraints

Abstract

The invention discloses an electromagnetic environment map generation method based on conditional constraints, and the method comprises the steps: S1, collecting electromagnetic environment data of a designated region of a city in a crowdsourcing manner, and enabling the electromagnetic environment data to be used for generating an original collection information data set; s2, integrating data in the original acquisition information data set, and generating an original image data set by adopting a Kriging interpolation method; s3, processing the original image data set, and performing data enhancement and data normalization to form an urban sampling information graph, an urban building distribution graph and an enhanced data set of the original image; s4, constructing a conditional generative adversarial network model, wherein the conditional generative adversarial network model comprises a conditional constraint module, a generator module, a discriminator module and a loss function module; and S5, training a conditional generative adversarial network model by using the three enhanced data sets to generate an electromagnetic environment map. According to the method, the generation efficiency and precision of the electromagnetic environment map can be remarkably improved under the condition of keeping relatively low calculation cost.

Description

Technical Field

[0001] This invention relates to the field of intelligent image generation, and in particular to a method for generating electromagnetic environment maps based on conditional constraints. Background Technology

[0002] In the field of electromagnetic situation generation technology, Radio Environment Map (REM), as an innovative technology, has the core value of integrating multi-dimensional spectrum data such as time, frequency, geographical location, and field strength to present the distribution and utilization of spectrum resources in an intuitive and highly visualized form. This revolutionary presentation method not only profoundly reveals the true face of the radio environment but also opens up new paths for the dynamic sharing and efficient supervision of the radio spectrum. REM technology provides strong support for various key decision-making processes, including but not limited to: improving network performance through refined coverage optimization strategies; achieving intelligent allocation and efficient utilization of spectrum resources; further analyzing and acquiring electromagnetic interference behavior patterns; supporting active interference localization; and directly supporting applications such as electromagnetic situation generation, application deployment, frequency conflict resolution, and electromagnetic spectrum planning.

[0003] The accuracy of electromagnetic environment mapping heavily depends on the amount and accuracy of information collected by sensor nodes. Typically, mapping requires sensor nodes to perceive information across all grid nodes. The grid division also affects map accuracy; smaller grid side lengths generally result in higher map accuracy, but require more sensing information and involve a large computational load. Larger grid side lengths can reduce the required sensing information, but map accuracy decreases accordingly.

[0004] Traditional electromagnetic environment maps require long data collection times, large amounts of data, and have poor real-time performance, which severely limits their application.

[0005] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art. Summary of the Invention

[0006] The purpose of this invention is to provide a method for generating electromagnetic environment maps based on conditional constraints, enabling rapid generation of electromagnetic environment maps in urban environments and solving problems such as long data collection time, large amount of required data, and poor real-time performance.

[0007] To achieve the above objectives, this invention provides a method for generating electromagnetic environment maps based on conditional constraints, comprising: S1. Collect electromagnetic environment data for a designated area of ​​the city using crowdsourcing methods to generate the original data set of collected information for the electromagnetic environment map of that designated area. S2. Integrate the data in the original collected information dataset and use the Kriging interpolation method to generate the original image dataset of the electromagnetic environment map of the specified area. S3. Process the original image dataset, perform data augmentation and data normalization to form three augmented datasets: city sampling information map, city building distribution map and original image; S4. Construct a conditional generative adversarial network model, which includes a conditional constraint module, a generator module, a discriminator module, and a loss function module. S5. Train the conditional generative adversarial network model using the two augmented datasets to generate an electromagnetic environment map.

[0008] Optionally, the geomagnetic environment data includes signal strength, geographic coordinates, and reception time; In the original data collection dataset, multiple data received at the same time are grouped together according to the data reception time. Each group of data includes the signal strength and geographical coordinates at the current time.

[0009] Optionally, step S1 specifically includes the following steps: S1.1 Use several terminal devices to collect electromagnetic environment data, and adjust the terminal devices to receive mode; S1.2 Deploy several terminal devices at preset test locations in a designated area, start the terminal devices to record signal strength data, and synchronize the geographic coordinate data at that moment; collect electromagnetic environment data for several consecutive days; S1.3. The collected electromagnetic environment data is centrally transmitted to the data fusion center to form the original collected information dataset.

[0010] Optionally, step S2 specifically includes the following steps: S2.1. Assign each signal strength in each group of data in the original data collection dataset to multiple geographic coordinates, determine the grid density, divide the specified area into grids, divide the signal strength data into the corresponding grids according to the geographic coordinates, and calculate the average signal strength of the grid by averaging the multiple signal strength data in the same grid. S2.2. The geographic coordinates of each grid center are used as the grid coordinates of the corresponding grid, the average signal strength of each grid is used as the grid strength of the corresponding grid, and Kriging interpolation is performed according to the required map accuracy to calculate the signal strength estimate of each interpolation point, thereby generating the original image of the electromagnetic environment map at a certain time; the original images at multiple different times constitute the original image dataset.

[0011] Optionally, step S2.2 specifically includes: For interpolation points The signal strength is estimated based on N known grid data points around it. The spatial structure of the data is described using a semi-variogram and a covariance function, ensuring that adjacent points are similar. This similarity weakens as the distance between two points increases. The weights of the data points in each grid around the interpolation point are determined by a spatial correlation model. The formula for calculating the signal strength estimate of the interpolation point is as follows: In the formula, The signal strength estimate calculated for the interpolation point. , where is the geographic coordinate of the interpolation point. , where is the grid coordinate of the i-th grid. Let i be the mesh strength of the i-th mesh. The number of grids to be divided, The weight factor for the i-th grid must satisfy the following condition: Furthermore, the calculation results must satisfy the condition of minimizing the estimation error. .

[0012] Optionally, step S3 specifically includes the following steps: S3.1. Use public maps to obtain a two-dimensional city plan and extract the city building distribution map from it; align the original image dataset with the city building distribution map to distinguish between building areas and non-building areas; remove data located in building areas from each original image and retain data located in non-building areas; select the required number of sampling points in non-building areas as needed to obtain the initial dataset of the city sampling information map that retains only the features of the sampling points. S3.2. Data augmentation is achieved by using image rotation and image flipping data preprocessing methods to correspond one-to-one with the images in the initial dataset of the city sampling information map, the city building distribution map and the original image dataset, thereby expanding the scale and diversity of the dataset. S3.3. Normalize the dataset of the enhanced city sampling information map, city building distribution map and original image obtained in step S3.2 to finally obtain the enhanced dataset of the city sampling information map, city building distribution map and original image.

[0013] Optionally, the image rotation refers to rotating the image by a certain angle in a counterclockwise or clockwise direction with a certain point in the image as the origin, specifically: In the formula, These are the coordinates of the data points before the image transformation. These are the coordinates of the data points after image transformation. It is the angle of image rotation; The image folding refers to folding the image with a straight line in the image as the axis of symmetry, specifically: In the formula, and It is the axis of symmetry of the image reflection; The normalization process specifically involves: In the formula, These are the pixel values ​​before image normalization. These are the pixel values ​​after image normalization. This refers to the minimum value of an image sample. This refers to the maximum value of the image sample; for urban sampling infographics, the pixel value includes signal strength.

[0014] Optionally, step S4 includes the following steps: S4.1 Construct a condition constraint module, which includes a five-layer encoder structure for extracting condition features from the input image; The first-layer encoder structure includes a two-dimensional convolutional layer, a batch normalization layer, and an activation function. The last four encoder layers each consist of a max pooling downsampling layer, a two-dimensional convolutional layer, a batch normalization layer, and an activation function. S4.2 Construct a generator module, which includes an encoder part and a decoder part, for outputting a generated image that is as consistent as possible with the original image; The encoder section includes a five-layer encoder structure, which is the same as the condition constraint module. The output of the condition constraint module is convolved and fused with the output of the encoder part of the generator module, and then fed into the decoder part of the generator module. The decoder part includes a five-layer decoder structure, each of which includes a transposed convolutional upsampling layer, a splicing layer, a two-dimensional convolutional layer, a batch normalization layer, and an activation function. S4.3 Construct a discriminator module, which includes a four-layer neural network structure for distinguishing the authenticity of the generated images output by the generator module; The first layer structure includes a two-dimensional convolutional layer, a pooling layer, a batch normalization layer, an activation function, and a random deactivation layer. The latter three layers each include a max pooling downsampling layer, a two-dimensional convolutional layer, a batch normalization layer, an activation function, and a random deactivation layer. S4.4 Construct a loss function module, which includes local pixel error and global image error, which together constitute the conditional constraint penalty term; The global image error is the feature loss error between the generated image and the original image, and the local pixel error is the feature loss error between the generated image and the constraint conditions. The two errors are added together in a certain proportion to obtain the loss value, which is returned to the generator module and the constraint module to guide the parameter optimization process of the model.

[0015] Optionally, step S5 includes the following steps: S5.1 Initialize network weight parameters, set the neural network gradient optimization algorithm, and set the initial learning rate and decay rate of the generator module and discriminator module; S5.2 Perform iterative training loops for the conditional generative adversarial network model to generate the final electromagnetic environment map.

[0016] Optionally, step S5.2 specifically includes: The two augmented datasets, the city sampling information map and the city building distribution map generated in step S3, are input into the conditional constraint module for concatenation, and then input together with random noise into the conditional generative adversarial network model to train the generator module to generate images based on random noise. The discriminator module compares and judges the generated image and the original image, and outputs the judgment result; The loss function module calculates the penalty intensity based on the judgment result, and improves the performance of the generator module and the discriminator module through regularization constraints; Set a threshold for the number of training iterations. When the number of training iterations is less than the threshold, output the loss value for each round. When the number of training iterations reaches the threshold, the loop ends and an electromagnetic environment map that is highly consistent with the actual electromagnetic distribution is generated.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: 1. This invention obtains an electromagnetic environment map of a city area by inputting a building distribution map of a specified area and a small number of sampled information points into a condition constraint module and using a conditional generative adversarial network model to generate a conditionally constrained random noise. This requires no large amount of data, has a short data collection time, and low computational cost.

[0018] 2. When generating electromagnetic environment maps using a condition-constrained generative adversarial network model, this invention can apply conditional constraints based on physical laws and automatically generate images that match the current environment by playing a game with the discriminator. It does not rely on human experience and intuition and can significantly improve the efficiency and accuracy of image construction while maintaining low computational costs. Attached Figure Description

[0019] Figure 1 This is a flowchart of the electromagnetic environment map generation method based on conditional constraints of the present invention; Figure 2 This is a flowchart of the construction conditional generative adversarial network model of the present invention; Figure 3 This is a schematic diagram of urban sampling information points according to the present invention; Figure 4 This is a schematic diagram of the urban building distribution of the present invention; Figure 5 This is a schematic diagram of the electromagnetic environment map generation result in an urban environment according to the present invention. Detailed Implementation

[0020] The following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates the electromagnetic environment map generation method based on conditional constraints proposed in this invention. The advantages and features of this invention will become clearer from the following description. It should be noted that the accompanying drawings are in a very simplified form and use non-precise scales, used only to facilitate and clearly illustrate the embodiments of this invention. Please refer to the accompanying drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, scales, sizes, etc., depicted in the accompanying drawings are only for illustrative purposes to aid those skilled in the art and are not intended to limit the implementation conditions of this invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in scale, or adjustments to the size, without affecting the effects and objectives achieved by this invention, should still fall within the scope of the technical content disclosed in this invention.

[0021] This invention provides a method for generating electromagnetic environment maps based on conditional constraints. For example... Figure 1 As shown, the method includes the following steps: S1. Crowdsourced data collection: Electromagnetic environment data of a designated area in the city is collected through crowdsourcing and transmitted to the data fusion center to generate the original data collection dataset of the electromagnetic environment map of the designated area. Electromagnetic environment data includes signal strength, geographic coordinates, and reception time. Step S1 specifically includes the following steps: S1.1 Use several terminal devices to collect electromagnetic environment data. When using the devices, adjust them to receive mode. The terminal device is a smartphone or an electronic reconnaissance sensor. The smartphone can measure the downlink signal strength of the base station and obtain the current geographical coordinates using a compass. The electronic reconnaissance sensor can receive signal strength of a specific frequency band, thereby measuring the signal strength of the specified frequency band. The electronic reconnaissance sensor is equipped with a navigation device for navigation and positioning to obtain geographical coordinate information. S1.2 Deploy several terminal devices at preset test locations in a designated area, start the terminal devices to record signal strength information, and synchronize the navigation and positioning data at that time, i.e., geographic coordinate information; collect electromagnetic environment data for several consecutive days, and stop collecting after the requirements are met; The terminal equipment can be moved with vehicles or people, or it can be used to perform measurements at a fixed location; the deployment of the terminal equipment should be as dense as possible. Among them, the number of days for collecting electromagnetic environment data must be more than 2 days; S1.3. The electromagnetic environment data collected over a period of time (multiple days) is transmitted to the data fusion center to form the original data collection dataset. In the original data collection dataset, multiple data received at the same time are grouped together according to the data reception time. Each group of data includes at least the signal strength and geographical coordinates at the current time.

[0022] S2. Based on the original collected information dataset, integrate each set of data and use the Kriging interpolation method to generate the original image dataset of the electromagnetic environment map of the specified area. Step S2 specifically includes the following steps: S2.1. Correspond each group of data in the original data acquisition dataset to multiple signal strengths and multiple geographic coordinates, determine the grid density, divide the specified area into grids, divide the signal strength data into the corresponding grids according to the geographic coordinates, and calculate the average signal strength of the grid by averaging multiple signal strength data in the same grid, so as to eliminate the impact of fast fading. S2.2. The geographic coordinates of each grid center are used as the grid coordinates of the corresponding grid, the average signal strength of each grid is used as the grid strength of the corresponding grid, and Kriging interpolation is performed according to the required map accuracy to calculate the signal strength estimate of each interpolation point, thereby generating the original image of the electromagnetic environment map at a certain moment; the original image can be regarded as the real image; multiple original images at different times constitute the original image dataset. Specifically, for each interpolation point on the map It is necessary to estimate its signal strength based on N known grid data points around it. The spatial structure of the data is described using a semi-variogram and a covariance function, ensuring that adjacent points are similar. This similarity weakens as the distance between two points increases. The weights of the data points in each grid around the interpolation point are determined by a spatial correlation model. The formula for calculating the signal strength estimate of the interpolation point is as follows: In the formula, The signal strength estimate calculated for the interpolation point. , where is the geographic coordinate of the interpolation point. , where is the grid coordinate of the i-th grid. Let i be the mesh strength of the i-th mesh. The number of grids to be divided, The weight factor for the i-th grid must satisfy the following condition: Furthermore, the calculation results must satisfy the condition of minimizing the estimation error. .

[0023] S3. Process the original images from multiple different times one by one, and perform data augmentation and data normalization preprocessing to form an urban sampling information map, an urban building distribution map, and an augmented dataset of the original images, which will serve as the training dataset for the neural network; specifically, this includes the following steps: S3.1. Obtain a 2D city plan using a publicly available map and extract the city building distribution map from it. Align the original image dataset with the city building distribution map to distinguish between building areas and non-building areas, achieving spatial decoupling. Further, on each original image, without altering the image structure, remove data located in building areas and retain data located in non-building areas. Based on requirements, select the necessary number of sampling points (e.g., 100 sampling points) in the non-building areas, and obtain a dataset retaining only the features of the sampling points based on their coordinates. This dataset is the initial, unprocessed dataset of the city sampling infographic. Optionally, the publicly available map used is the global open-source map data OpenStreetMap (OSM). S3.2 Data Augmentation: Apply image rotation and image flipping data preprocessing methods to the multiple unprocessed city sampling information maps and city building distribution maps generated in step S3.1 and the multiple original images generated in step S2.2 to achieve data augmentation, expand the scale and diversity of the dataset, and maintain the spatial correspondence between the city sampling information maps, city building distribution maps and original images. Image rotation refers to rotating the image by a certain angle, either clockwise or counterclockwise, with a point in the image as the origin. Specifically: In the formula, These are the coordinates of the data points before the image transformation. These are the coordinates of the data points after image transformation. It is the angle of image rotation, usually rotated counterclockwise around the center point of the image. , , The rotation; Image folding refers to folding an image about a straight line as an axis of symmetry. Specifically: In the formula, and It is the axis of symmetry of the image reflection, usually a straight line passing through the center of the image; S3.3. Normalize the dataset of the enhanced urban sampling information map, urban building distribution map, and original images from step S3.2 to ensure that the pixel values ​​of the images are on the same order of magnitude and to balance the contributions of each feature. Specifically: In the formula, These are the pixel values ​​before image normalization. These are the pixel values ​​after image normalization. This refers to the minimum value of an image sample. This refers to the maximum value of the image sample; for urban sampling infographics, the pixel values ​​include signal strength.

[0024] After the above series of preprocessing operations, we obtain an enhanced dataset of urban sampling information map, urban building distribution map, and original images, which serves as the training dataset for the subsequent neural network. The urban sampling information map and urban building distribution map are shown below. Figure 3 and Figure 4 As shown.

[0025] S4. Construct a conditional generative adversarial network model, which includes a conditional constraint module, a generator module, a discriminator module, and a loss function module. like Figure 2 As shown, step S4 includes the following steps: S4.1 Construct a condition constraint module, which includes a five-layer coded feature extraction structure for extracting condition features from the input image; the city sampling information map and the city building distribution map, which have been preprocessed in step S3, are spliced ​​together and input as the input data of the condition constraint module; The first encoder layer consists of a two-dimensional convolutional layer (Conv2DTranspose), a batch normalization layer, and an activation function; the two-dimensional convolutional layer is used to initially extract shallow features of the image; The last four encoder layers each consist of a max pooling downsampling layer, a two-dimensional convolutional layer (Conv2DTranspose), a batch normalization layer, and an activation function. The max pooling downsampling layer is used to reduce the spatial resolution of the image and expand the receptive field of subsequent convolutional layers, thereby extracting deeper abstract features layer by layer.

[0026] S4.2 Construct a generator module, which includes an encoder part and a decoder part. Its purpose is to output a generated image that is as consistent as possible with the original image based on the input condition information. The encoder part includes a five-layer encoding feature extraction structure, which is the same as the condition constraint module. That is, the first encoder layer consists of a two-dimensional convolutional layer, a batch normalization layer and an activation function, and the last four encoder layers each consist of a max pooling downsampling layer, a two-dimensional convolutional layer, a batch normalization layer and an activation function. The encoder part is used to extract features and downsample the input image. The output of the condition constraint module is convolved and concatenated with the output of the encoder part of the generator module to form a comprehensive feature vector that integrates global conditions and local features. This vector is then fed into the decoder part of the generator module to recover spatial information. The decoder part includes a five-layer decoder structure, which is used to upsample the fused feature vector to gradually restore the spatial resolution and detail information of the image; further, each of the five layers of the decoder structure consists of a transposed convolution upsampling layer, a concatenation layer, a two-dimensional convolutional layer, a batch normalization layer, and a Leaky ReLU activation function.

[0027] S4.3 Construct a discriminator module, which includes a four-layer neural network structure for identifying the authenticity of the generated image output by the generator module, i.e., determining whether the generated image is close to the original image. The first layer consists of a 2D convolutional layer (Conv2D), a pooling layer, a batch normalization layer, a Leaky ReLU activation function, and a dropout layer. The latter three layers each consist of a max pooling downsampling layer, a 2D convolutional layer (Conv2DTranspose), a batch normalization layer, an activation function, and a dropout layer.

[0028] S4.4 Construct a loss function module based on multiple features. The loss function module includes two parts: error based on local pixels and error based on the global image. The two together constitute the conditional constraint penalty term. The global image error is the feature loss error between the generated image and the original image, and the local pixel error is the feature loss error between the generated image and the constraint conditions. The two parts of the error are added together according to a certain ratio to obtain the loss value, which is returned to the generator module and the conditional constraint module to guide the parameter optimization process of the model.

[0029] S5. Train and optimize the conditional generative adversarial network model to generate the final image. Two modules need to be trained simultaneously: one is to train the generator module to simulate the distribution of the original image, and the other is to train the discriminator module to estimate the probability that the generated image is close to the original image. Step S5 includes the following steps: S5.1 Initialize network weight parameters, set the neural network gradient optimization algorithm, and set the initial learning rate and decay rate of the generator module and discriminator module; S5.2 After initialization, perform conditional adversarial generation. The network model undergoes iterative training loops to generate the final image; Specifically, the two enhanced datasets, the city sampling information map and the city building distribution map generated in step S3, are input into the conditional constraint module for splicing, and are also input into the conditional generative adversarial network model along with random noise to train the generator module to generate images based on random noise, thereby realizing the generation of a complete electromagnetic environment map from discrete sampling information points; The discriminator module compares the generated image with the real original image and outputs a judgment result. Based on the judgment result, the loss function module calculates the penalty strength and improves the performance of the generator module and the discriminator module through regularization constraints. A threshold for the number of training iterations is set. When the number of training iterations is less than the threshold, the loss value for each round is output. When the number of training iterations reaches the threshold, the loop ends. The trained network then generates an electromagnetic environment map that is highly consistent with the actual electromagnetic distribution based on the sampled information points and the city building distribution map. A specific example of the electromagnetic environment map is as follows: Figure 5 As shown.

[0030] In summary, the method for generating the city's electromagnetic environment map can quickly generate accurate electromagnetic environment maps at various times using only some sampled information points and building distribution, satisfying the real-time generation and updating of maps, and solving the bottlenecks of large data volume, long time, and low accuracy required for map generation.

[0031] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0032] In the description of this invention, it should be understood that the terms "center," "height," "thickness," "upper," "lower," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0033] In the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0034] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0035] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for generating electromagnetic environment maps based on conditional constraints, characterized in that, include: S1. Collect electromagnetic environment data for a designated area of ​​the city using crowdsourcing methods to generate the original data set of collected information for the electromagnetic environment map of that designated area. S2. Integrate the data in the original collected information dataset and use the Kriging interpolation method to generate the original image dataset of the electromagnetic environment map of the specified area. S3. Process the original image dataset, perform data augmentation and data normalization to form three augmented datasets: city sampling information map, city building distribution map and original image; S4. Construct a conditional generative adversarial network model, which includes a conditional constraint module, a generator module, a discriminator module, and a loss function module. S5. Train the conditional generative adversarial network model using the three augmented datasets to generate an electromagnetic environment map.

2. The electromagnetic environment map generation method based on conditional constraints as described in claim 1, characterized in that, The geomagnetic environment data includes signal strength, geographic coordinates, and reception time; In the original data collection dataset, multiple data received at the same time are grouped together according to the data reception time. Each group of data includes the signal strength and geographical coordinates at the current time.

3. The electromagnetic environment map generation method based on conditional constraints as described in claim 2, characterized in that, Step S1 specifically includes the following steps: S1.1 Use several terminal devices to collect electromagnetic environment data, and adjust the terminal devices to receive mode; S1.2 Deploy several terminal devices at preset test locations in a designated area, start the terminal devices to record signal strength data, and synchronize the geographic coordinate data at that moment; collect electromagnetic environment data for several consecutive days; S1.

3. The collected electromagnetic environment data is centrally transmitted to the data fusion center to form the original collected information dataset.

4. The electromagnetic environment map generation method based on conditional constraints as described in claim 2, characterized in that, Step S2 specifically includes the following steps: S2.

1. Assign each signal strength in each group of data in the original data collection dataset to multiple geographic coordinates, determine the grid density, divide the specified area into grids, divide the signal strength data into the corresponding grids according to the geographic coordinates, and calculate the average signal strength of the grid by averaging the multiple signal strength data in the same grid. S2.

2. The geographic coordinates of each grid center are used as the grid coordinates of the corresponding grid, and the average signal strength of each grid is used as the grid strength of the corresponding grid. Kriging interpolation is performed to calculate the signal strength estimate of each interpolation point, thereby generating the original image of the electromagnetic environment map at a certain time. The original images at multiple different times constitute the original image dataset.

5. The electromagnetic environment map generation method based on conditional constraints as described in claim 4, characterized in that, Step S2.2 specifically includes: For interpolation points The signal strength is estimated based on N known grid data points around it. The spatial structure of the data is described using a semi-variogram and a covariance function, ensuring that adjacent points are similar. This similarity weakens as the distance between two points increases. The weights of the data points in each grid around the interpolation point are determined by a spatial correlation model. The formula for calculating the signal strength estimate of the interpolation point is as follows: In the formula, The signal strength estimate calculated for the interpolation point. , where is the geographic coordinate of the interpolation point. , where is the grid coordinate of the i-th grid. Let i be the mesh strength of the i-th mesh. The number of grids to be divided, The weight factor for the i-th grid must satisfy... Furthermore, the calculation results must satisfy the condition of minimizing the estimation error. .

6. The electromagnetic environment map generation method based on conditional constraints as described in claim 4, characterized in that, Step S3 specifically includes the following steps: S3.

1. Use public maps to obtain a two-dimensional city plan and extract the city building distribution map from it; align the original image dataset with the city building distribution map to distinguish between building areas and non-building areas; remove data located in building areas from each original image and retain data located in non-building areas; select the required number of sampling points in non-building areas as needed to obtain the initial dataset of the city sampling information map that retains only the features of the sampling points. S3.

2. Data augmentation is achieved by using image rotation and image flipping data preprocessing methods to correspond one-to-one with the images in the initial dataset of the city sampling information map, the city building distribution map and the original image dataset, thereby expanding the scale and diversity of the dataset. S3.

3. Normalize the datasets of the enhanced city sampling information map, city building distribution map and original image obtained in step S3.2 to finally obtain three enhanced datasets: city sampling information map, city building distribution map and original image.

7. The electromagnetic environment map generation method based on conditional constraints as described in claim 6, characterized in that, The image rotation refers to rotating the image by a certain angle, either counterclockwise or clockwise, with a point in the image as the origin. Specifically: In the formula, These are the coordinates of the data points before the image transformation. These are the coordinates of the data points after image transformation. It is the angle of image rotation; The image folding refers to folding the image with a straight line in the image as the axis of symmetry, specifically: In the formula, and It is the axis of symmetry of the image reflection; The normalization process specifically involves: In the formula, These are the pixel values ​​before image normalization. These are the pixel values ​​after image normalization. This refers to the minimum value of an image sample. This refers to the maximum value of the image sample; for urban sampling infographics, the pixel value includes signal strength.

8. The electromagnetic environment map generation method based on conditional constraints as described in claim 1, characterized in that, Step S4 includes the following steps: S4.1 Construct a condition constraint module, which includes a five-layer encoder structure for extracting condition features from the input image; The first-layer encoder structure includes a two-dimensional convolutional layer, a batch normalization layer, and an activation function. The last four encoder layers each consist of a max pooling downsampling layer, a two-dimensional convolutional layer, a batch normalization layer, and an activation function. S4.2 Construct a generator module, which includes an encoder part and a decoder part, for outputting a generated image that is as consistent as possible with the original image; The encoder section includes a five-layer encoder structure, which is the same as the condition constraint module. The output of the condition constraint module is convolved and fused with the output of the encoder part of the generator module, and then fed into the decoder part of the generator module. The decoder part includes a five-layer decoder structure, each of which includes a transposed convolutional upsampling layer, a splicing layer, a two-dimensional convolutional layer, a batch normalization layer, and an activation function. S4.3 Construct a discriminator module, which includes a four-layer neural network structure for distinguishing the authenticity of the generated images output by the generator module; The first layer structure includes a two-dimensional convolutional layer, a pooling layer, a batch normalization layer, an activation function, and a random deactivation layer. The latter three layers each include a max pooling downsampling layer, a two-dimensional convolutional layer, a batch normalization layer, an activation function, and a random deactivation layer. S4.4 Construct a loss function module, which includes local pixel error and global image error, which together constitute the conditional constraint penalty term; The global image error is the feature loss error between the generated image and the original image, and the local pixel error is the feature loss error between the generated image and the constraint conditions. The two errors are added together in a certain proportion to obtain the loss value, which is returned to the generator module and the constraint module to guide the parameter optimization process of the model.

9. The electromagnetic environment map generation method based on conditional constraints as described in claim 1, characterized in that, Step S5 includes the following steps: S5.1 Initialize network weight parameters, set the neural network gradient optimization algorithm, and set the initial learning rate and decay rate of the generator module and discriminator module; S5.2 Perform iterative training loops for the conditional generative adversarial network model to generate the final electromagnetic environment map.

10. The electromagnetic environment map generation method based on conditional constraints as described in claim 9, characterized in that, Step S5.2 specifically includes: The two augmented datasets, the city sampling information map and the city building distribution map generated in step S3, are input into the conditional constraint module for concatenation, and then input together with random noise into the conditional generative adversarial network model to train the generator module to generate images based on random noise. The discriminator module compares and judges the generated image and the original image, and outputs the judgment result; The loss function module calculates the penalty intensity based on the judgment result, and improves the performance of the generator module and the discriminator module through regularization constraints; Set a threshold for the number of training iterations. When the number of training iterations is less than the threshold, output the loss value for each round. When the number of training iterations reaches the threshold, the loop ends and an electromagnetic environment map that is highly consistent with the actual electromagnetic distribution is generated.

Images