Electromagnetic map generation method
Through the electromagnetic map generation method based on UNet++ model, combined with building distribution information and sampling point position information, the problem of dependence on prior information and data in the existing technology of electromagnetic map generation method is solved, and the problem of data in complex environments is realized High-precision electromagnetic map generation.
Patent Information
- Application Number
- CN202411125319.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-12-13
AI Technical Summary
The existing electromagnetic map generation method has the problem of inaccurate interpolation results when dealing with discrete and inaccurate spectrum-aware data, and it relies heavily on the acquisition of prior information, which limits its application scope and accuracy, especially in complex environments and variable The modeling accuracy decreases when signal propagating characteristics.
The electromagnetic map generation method based on UNet++ model is adopted to obtain the building distribution information and the location information of sampling points, generate electromagnetic sampling maps and mask maps to fusion, and use multi-scale fusion deep supervision network to generate electromagnetic maps, avoiding prior test Information dependence and step loss function improve the model's delineation accuracy of the radiation source region.
It realizes the generation of accurate electromagnetic maps in low-frequency spectrum-aware data and complex environments, reduces the requirements for data distribution uniformity and sampling point density, improves the accuracy of electromagnetic maps, especially in the fine delineation ability of radiation source areas.
Smart Images
Figure CN119131273A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of electromagnetic spectrum map construction, and in particular to an electromagnetic map generation method. Background Art
[0002] Electromagnetic maps can intuitively present information such as electromagnetic energy distribution and spectrum resource distribution. Ideally, the distribution of electromagnetic signal strength can accurately map the number of radiation sources and their locations, thereby effectively assisting in the precise positioning of radiation sources. The main principles and shortcomings of existing electromagnetic map generation are as follows:
[0003] 1. Spatial interpolation method
[0004] Main principle: By utilizing the monitoring data of known discrete points, various interpolation algorithms are used to estimate the electromagnetic environment information of unknown points without relying on other prior knowledge.
[0005] Disadvantages: When processing discrete and inaccurate spectrum sensing data, the interpolation results may be inaccurate due to uneven data distribution or sparse sampling points.
[0006] 2. Parameter Construction Method
[0007] Main principle: It relies on prior information such as transmitter information and radio wave propagation environment parameters, and is modeled through forward modeling or position estimation methods. It has high accuracy but is limited by the acquisition of prior information.
[0008] Disadvantages: It is highly dependent on the acquisition of prior information, which is often difficult to obtain or incomplete in actual scenarios, thus limiting its scope of application and accuracy.
[0009] 3. Hybrid Construction
[0010] Main principle: It combines the advantages of spatial interpolation and parameter construction methods to generate electromagnetic maps through preliminary construction and subsequent correction.
[0011] Disadvantages: Although it combines the advantages of the first two, it may still face the problem of reduced modeling accuracy when dealing with complex environments and variable signal propagation characteristics.
[0012] Therefore, a method for generating an electromagnetic map is needed that can solve the problems existing in the prior art. Summary of the invention
[0013] The technical problem to be solved by the present invention is to provide an electromagnetic map generation method, which can effectively improve the accuracy and precision of the generated electromagnetic map.
[0014] The technical solution adopted by the present invention to solve the technical problem is: to provide an electromagnetic map generation method, comprising the following steps:
[0015] Get a mask map containing building distribution information in the target area;
[0016] Draw a grayscale image including the location information of each sampling point in the target area and the received signal strength thereof as an electromagnetic sampling map;
[0017] fusing the mask map and the electromagnetic sampling map to obtain an input image sample;
[0018] Generate an electromagnetic map based on the input image sample using an electromagnetic map generation model; the electromagnetic map generation model is constructed based on a UNet++ model, specifically including:
[0019] An encoder and a decoder, wherein the encoder and the decoder form a nested connection of UNet sub-encoder-decoder networks of different depths, each of the UNet sub-encoder-decoder networks generates a feature map of the same size as the input image sample;
[0020] A deep supervision module is used to predict and obtain an electromagnetic map based on the feature map.
[0021] Furthermore, the drawing of a grayscale image including the location information of each sampling point in the target area and the received signal strength thereof as an electromagnetic sampling map includes:
[0022] Convert the received signal strength into a grayscale value between 0 and 1;
[0023] The grayscale value of the pixel corresponding to the sampling point is set to the received signal strength at the corresponding position after conversion, and the grayscale value of the pixel in the remaining area is set to 0, so as to obtain an electromagnetic sampling map.
[0024] Furthermore, in the mask map, the grayscale value of the pixels in the building area is 1, and the grayscale value of the pixels in the non-building area is 0.
[0025] Furthermore, when the electromagnetic map generation model is trained, training labels are generated by the following method:
[0026] Draw a grayscale image of received signal strengths at each sampling point in the target area as a reference electromagnetic map;
[0027] The grayscale values of pixels in a circular area with the radiation source coordinates as the center and a set value as the radius in the reference electromagnetic map are retained, and the grayscale values of pixels in the remaining area are set to 1, so as to obtain a new reference electromagnetic map as the training label.
[0028] Furthermore, in the reference electromagnetic map, the grayscale value of the pixels in the building area is 1, and the grayscale value of the pixels in the non-building area is between 0 and 1.
[0029] Furthermore, the predicting and obtaining the electromagnetic map according to the characteristic map includes:
[0030] The feature maps generated by the UNet sub-encoder / decoder networks of different depths are convolved respectively and then spliced into a fused feature map;
[0031] The fused feature map is sequentially passed through an attention module and a convolution operation to output the electromagnetic map.
[0032] Furthermore, the loss function of the electromagnetic map generation model is
[0033]
[0034] Where W is the number of input image samples, InputMap i and GroundTruth i are the i-th input image sample and the i-th training label respectively, ε(pixel) is the normalized grayscale value of the pixel in the training label, pixel∈GroundTruth,F Map (·) and Θ denote the model and model parameters, respectively.
[0035] Furthermore, the UNet sub-codec networks of different depths include UNet sub-codec networks with depths of 2, 3, 4, and 5, respectively.
[0036] Beneficial Effects
[0037] Due to the adoption of the above technical scheme, the present invention has the following advantages and positive effects compared with the prior art: the present invention only needs to obtain the building distribution information, the location information of the sampling points and the received signal strength of the corresponding positions, and can obtain the final electromagnetic map without relying on the acquisition of prior information, and has a wide range of application scenarios; the present invention uses the location information of the sampling points and the received signal strength of the corresponding positions to generate an electromagnetic sampling map, and superimposes it with the mask map containing the building distribution information. The generated input image samples are analyzed and processed by the electromagnetic map generation model based on UNet++ to obtain the final electromagnetic map. The requirements for spectrum sensing data are low, and it is possible to process discrete and inaccurate data. The generated electromagnetic map will not be distorted due to uneven data distribution or sparse sampling points, and a relatively accurate electromagnetic map can still be obtained when processing complex environments and variable signals; the present invention uses a step loss function to promote the model training to focus on the fine depiction of the radiation source and its adjacent areas, and accurately capture the electromagnetic characteristics of these key areas to ensure that the generated electromagnetic map shows a high degree of accuracy in the key areas of the radiation source. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 is a flow chart of an embodiment of the present invention;
[0039] Figure 2 It is a structural diagram of an electromagnetic map generation model according to an embodiment of the present invention. DETAILED DESCRIPTION
[0040] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.
[0041] The embodiments of the present invention relate to a method for generating an electromagnetic map based on deep learning, such as Figure 1 As shown, the following steps are included:
[0042] Get a mask map containing building distribution information in the target area;
[0043] Draw a grayscale image containing the location information of each sampling point in the target area and its received signal strength as an electromagnetic sampling map;
[0044] The input image samples are obtained by fusing the mask map and the electromagnetic sampling map;
[0045] An electromagnetic map generation model is used to generate an electromagnetic map based on an input image sample.
[0046] Among them, the electromagnetic map generation model is a multi-scale fusion deep supervision network built based on the UNet++ model, such as Figure 2 As shown, including:
[0047] The encoder part is used to extract and compress features from the input data;
[0048] The decoder part is used to decode the features extracted by the encoder and gradually restore the spatial resolution of the feature map to generate a result corresponding to the input image;
[0049] The nested and dense skip connections connect different levels in the encoder and decoder, bridging the semantic gap between the encoder and decoder, allowing the model to better utilize feature information at different levels;
[0050] The deep supervision part is used to introduce the output results of the intermediate layer to jointly supervise the training of the model and improve the convergence performance and generalization ability of the model.
[0051] The encoder part includes five VGG Blocks, each of which contains two encoding blocks. Each encoding block contains a 3×3 convolution layer, a batch normalization layer, and a LeakyReLU activation function. VGG Blocks are downsampled by average pooling to compress features.
[0052] The decoder part contains five VGG Blocks, each of which contains two decoding blocks. Each decoding block contains a 3×3 convolution layer, a batch normalization layer, and a LeakyReLU activation function. The sub-VGG Blocks are upsampled by nearest neighbor interpolation to gradually restore the original size of the feature map.
[0053] The nested and dense skip connections part nests UNet sub-encoder-decoder networks with depths of 2, 3, 4, and 5 in the overall encoder-decoder network. Specifically, the deeper sub-network contains the shallower sub-network as part of it, forming a hierarchical structure. The VGG Block in the decoding structure receives the outputs from the previous VGG Blocks of the same layer and the upsampled outputs of the lower layer VGG Blocks.
[0054] Deep supervision concatenates the outputs of sub-encoder / decoder networks of different depths through a 1×1 convolutional layer, inputs the channel attention module and the spatial attention module, and finally passes through a 1×1 convolutional layer to obtain the final electromagnetic map generation result.
[0055] The training set Input the model and optimize the model by minimizing the loss function. The loss function is expressed as:
[0056]
[0057] Among them, W is the number of training samples, InputMap i and GroundTruth i are the i-th sample and the i-th label respectively, pixel∈GroundTruth,F Map (·) and Θ represent the model and model parameters respectively. The trained model is the multi-scale fusion deep supervision electromagnetic map generation model.
[0058] This embodiment uses the following method to establish an image sample data set and conduct a simulation experiment:
[0059] Make size N x ×N yThe building mask map BuildingMask is used to set the radiation source parameters, thereby simulating the electromagnetic map based on the received signal strength (RSS) in the urban environment. Normalize BuildingMask and RadioMap, where the pixels corresponding to buildings in BuildingMask are 1 and the pixels corresponding to non-building areas are 0, and the pixels corresponding to buildings in RadioMap are 1 and the pixels corresponding to non-building areas are between (0,1). Based on BuildingMask, simulate the process of sparse sampling around the building by the measurement terminal to obtain the mask SampledMask representing the RSS sampling position, where the pixel corresponding to the sampling position is 1 and the remaining pixels are 0. Based on SampledMask, keep the pixels at the sampling position on RadioMap and set the remaining pixels to 0 to obtain the sampled map SampledMap;
[0060] Keep the pixels in the circle with the radiation source coordinates as the center and r as the radius in RadioMap, and set the remaining positions to 1 to obtain the label GroundTruth;
[0061] Add SampledMap and BuildingMask to get InputMap;
[0062] The dataset Divided into training set and test set, where i is the sample number and M is the total number of samples.
[0063] In practical applications, the location information of the sampling points and the received signal strength of the corresponding positions can be obtained by deploying a sampling data acquisition module in the target area. Then, the sampling data in the target area obtained by the sampling data acquisition module is represented as a grayscale image to obtain an electromagnetic sampling map. In the figure, a pixel point represents a sampling position or a position to be estimated. The pixel value at the sampling position is represented by different grayscale levels according to the strength of the received signal, and the pixel values of the remaining positions are set to 0. A building mask map is generated based on the environmental image data of the target area. The corresponding pixels of the buildings in the figure are 1, and the corresponding pixels of the non-building areas are 0. The image obtained by superimposing the electromagnetic sampling map and the building mask map is input into the above-mentioned trained model to generate the final electromagnetic map.
Claims
1. A method for generating an electromagnetic map, characterized in that: The following steps are involved: Get a mask map containing building distribution information in the target area; Draw a grayscale image including the location information of each sampling point in the target area and the received signal strength thereof as an electromagnetic sampling map; fusing the mask map and the electromagnetic sampling map to obtain an input image sample; generating an electromagnetic map based on the input image sample using an electromagnetic map generation model; The electromagnetic map generation model is built based on the UNet++ model, specifically including: An encoder and a decoder, wherein the encoder and the decoder form a nested connection of UNet sub-encoder-decoder networks of different depths, each of the UNet sub-encoder-decoder networks generates a feature map of the same size as the input image sample; A deep supervision module is used to predict and obtain an electromagnetic map based on the feature map.
2. The method according to claim 1, characterized in that The drawing of a grayscale image including the location information of each sampling point in the target area and the received signal strength thereof as an electromagnetic sampling map comprises: Convert the received signal strength into a grayscale value between 0 and 1; The grayscale value of the pixel corresponding to the sampling point is set to the received signal strength at the corresponding position after conversion, and the grayscale value of the pixel in the remaining area is set to 0, so as to obtain an electromagnetic sampling map.
3. The method according to claim 1, characterized in that: The grayscale value of the pixels in the building area in the mask map is 1, and the grayscale value of the pixels in the non-building area is 0.
4. The method according to claim 1, characterized in that: When the electromagnetic map generation model is trained, training labels are generated by the following method: Draw a grayscale image of received signal strengths at each sampling point in the target area as a reference electromagnetic map; The grayscale values of pixels in a circular area with the radiation source coordinates as the center and a set value as the radius in the reference electromagnetic map are retained, and the grayscale values of pixels in the remaining area are set to 1, so as to obtain a new reference electromagnetic map as the training label.
5. The method according to claim 4, characterized in that The grayscale value of the pixels in the building area in the reference electromagnetic map is 1, and the grayscale value of the pixels in the non-building area is between 0 and 1.
6. The method according to claim 1, characterized in that The step of predicting and obtaining an electromagnetic map according to the characteristic map comprises: The feature maps generated by the UNet sub-encoder / decoder networks of different depths are respectively subjected to 1×1 convolution operations and then spliced into a fused feature map; The fused feature map is sequentially passed through an attention module and a 1×1 convolution operation to output the electromagnetic map.
7. The method according to claim 1, characterized in that The loss function of the electromagnetic map generation model is Where W is the number of input image samples, InputMap i and GroundTruth i are the i-th input image sample and the i-th training label respectively, ε(pixel) is the normalized grayscale value of the pixel in the training label, pixel∈GroundTruth,F Map (·) and Θ denote the model and model parameters, respectively.
8. The method according to claim 1, characterized in that The UNet sub-codec networks of different depths include UNet sub-codec networks with depths of 2, 3, 4, and 5 respectively.