Inference method and system for complementing three-dimensional space position of electromagnetic radiation source based on frequency spectrum situation data

Abstract

The invention discloses a reasoning method and system for complementing the three-dimensional space position of an electromagnetic radiation source based on spectrum situation data, and the method comprises the steps: obtaining the GIS data of a target region, constructing a geographic model of the target region, simulating the channel loss between a radiation source and each receiving point in the geographic model through a dominant path propagation model, obtaining the intensity value of an original received signal, and calculating the intensity value of the original received signal. Forming electromagnetic spectrum situation initial data; complementing a received signal strength vacant area in the electromagnetic spectrum situation initial data to form a three-dimensional electromagnetic spectrum situation matrix, and storing the matrix locally according to a radiation source position index; the three-dimensional electromagnetic spectrum situation matrix is input into a pre-trained neural network, a corresponding three-dimensional probability thermodynamic diagram is output, and the position coordinate with the maximum probability value in the thermodynamic diagram is the corresponding radiation source position. The method has good expansibility, and can be widely applied to complex scenes depending on spatial spectrum sensing, such as spectrum resource planning, electromagnetic situation evaluation and network electric countermeasure.

Description

Technical Field

[0001] This invention relates to a method and system for inferring the three-dimensional spatial location of electromagnetic radiation sources based on spectrum situational data, and belongs to the field of wireless communication technology. Background Technology

[0002] With the accelerating pace of social progress, the electromagnetic environment is becoming increasingly complex. The presence of illegal signal sources not only interferes with legitimate communications but also threatens the security and stability of critical wireless systems. Wireless sensor networks (WSNs) have become an important means of monitoring the electromagnetic environment due to their flexible deployment and low power consumption. However, in complex electromagnetic environments, traditional geometric spatial location reasoning methods (such as TOA, TDOA, AOA, and RSS) struggle to obtain high-precision spatial location reasoning results due to factors such as building obstruction and NLOS paths. In recent years, deep learning has been introduced into the field of electromagnetic spatial location reasoning due to its excellent feature extraction capabilities, but it still faces the following key challenges: difficulty in data acquisition (high cost of acquiring spectrum situation data in real-world scenarios, and simulation data often deviating from reality); sparse sensor deployment (limited by cost and practical conditions, sensor deployment density cannot meet the needs of data-driven models); and NLOS propagation uncertainty (models lacking geographic information support struggle to identify and handle errors caused by NLOS paths). Current research attempts to apply deep neural networks to the task of spatial location reasoning for wireless radiation sources. For example, 2D spatial location reasoning is achieved by extracting features from 2D RSS maps using CNNs. However, such methods can only handle two-dimensional planar scenes, fail to depict the height dimension features in actual terrain, and exhibit accuracy degradation under sparse nodes. Summary of the Invention

[0003] This invention provides a method and system for inferring the three-dimensional spatial location of electromagnetic radiation sources based on spectral situation data, which improves the three-dimensional modeling and information completion capabilities, enables precise spatial location reasoning, and solves the problems disclosed in the background art.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for inferring the three-dimensional spatial location of electromagnetic radiation sources based on spectral situation data: Acquire GIS data of the target area, construct a geographic model of the target area, use the dominant path propagation model to simulate the channel loss between the radiation source and each receiving point in the geographic model, obtain the original received signal strength value, and constitute the initial data of the electromagnetic spectrum situation. The missing regions of received signal strength in the initial electromagnetic spectrum situation data are filled in to form a three-dimensional electromagnetic spectrum situation matrix, which is stored locally according to the radiation source location index. The three-dimensional electromagnetic spectrum situation matrix is ​​input into a pre-trained neural network, which outputs a corresponding three-dimensional probability heatmap. The coordinates of the location with the highest probability value in the heatmap are the locations of the corresponding radiation sources.

[0005] Furthermore, the method for filling in the missing regions of received signal strength in the initial electromagnetic spectrum situation data is as follows: for any radiation source s, assuming the target area is X*Y*Z and the number of its receiving sensing nodes is a*b*c, zeros are filled in the positions where there are no sensing nodes in the area.

[0006] Furthermore, the training method for the neural network includes: The three-dimensional electromagnetic spectrum situation matrix is ​​constructed as the input feature matrix, and the label is the probability distribution matrix of the real radiation source in three-dimensional space. The distribution matrix is ​​centered on the signal source position, and the probability of other positions is zero, forming a supervised learning dataset. For any radiation source s, a neural network is trained based on the local three-dimensional electromagnetic spectrum situation matrix and the tag probability distribution matrix.

[0007] Furthermore, the training method for the neural network also includes: The radiation sources deployed in the target area X*Y*Z and the received values ​​of the corresponding sensing nodes in the area are used as samples. n radiation sources are set, the location information of the radiation sources is used as labels, and the received values ​​of the sensing nodes in the corresponding areas are used as the corresponding label information. One label and its corresponding information are counted as one sample. The samples are randomly divided into training set and validation set, with 80% used for training set and 20% used for validation set.

[0008] Furthermore, the neural network is implemented using the PyTorch framework in Python, employing an hourglass-shaped convolutional neural network architecture to balance multi-scale spatial feature extraction and spatial location inference accuracy recovery. It consists of five main parts: a downsampling module, an upsampling module, a semantic embedding block, a heatmap generation layer, and a customized loss function. The downsampling module comprises multiple convolutional layers, a BatchNorm layer, and a ReLU activation function, working in conjunction with a max-pooling layer to achieve multi-layer feature extraction and spatial dimension compression of the input spectral data. The upsampling module uses bilinear interpolation to progressively upsample and recover the spatial resolution of the spatial location inference output.

[0009] A second aspect of the present invention provides a system for inferring the three-dimensional spatial location of an electromagnetic radiation source based on spectral situational data, comprising: The electromagnetic spectrum situation initial data module is used to acquire GIS data of the target area, construct a geographic model of the target area, use the dominant path propagation model to simulate the channel loss between the radiation source and each receiving point in the geographic model, obtain the original received signal strength value, and constitute the electromagnetic spectrum situation initial data. The three-dimensional electromagnetic spectrum situation matrix module is used to fill in the missing areas of received signal strength in the initial electromagnetic spectrum situation data to form a three-dimensional electromagnetic spectrum situation matrix, which is stored locally according to the radiation source location index. The neural network module is used to take a three-dimensional electromagnetic spectrum situation matrix as input and output a corresponding three-dimensional probability heat map. The coordinates of the position with the highest probability value in the heat map are the corresponding radiation source positions.

[0010] A third aspect of the present invention provides a computer-readable storage medium for storing one or more programs, characterized in that: the one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods described above.

[0011] A fourth aspect of the present invention provides a computing device, comprising: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described above.

[0012] This invention is applied to complex electromagnetic environments and sparse sensing node conditions for accurate spatial location reasoning of illegal radiation sources. It is applicable to the deep coupling problem of spectrum situation awareness and source location information in non-line-of-sight (NLOS) propagation scenarios. Based on deep learning modeling and spectrum situation data completion methods, this technical solution has good scalability and can be widely applied to complex scenarios that rely on spatial spectrum awareness, such as spectrum resource planning, electromagnetic situation assessment, and cyber-electronic countermeasures. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the present invention; Figure 2 This is a schematic diagram of the radiation source and corresponding electromagnetic spectrum of the present invention; Figure 3 This is a schematic diagram of the neural network structure according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the actual label processing input neural network in an embodiment of the present invention; Figure 5 This is the label map of the neural network prediction output in an embodiment of the present invention. Detailed Implementation

[0014] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0015] This invention provides a method for inferring the three-dimensional spatial location of electromagnetic radiation sources based on spectral situational data, such as... Figure 2 As shown, the core concept of the invention lies in obtaining the field strength value only by using sparse receiving sensors and adjusting the network parameters using a deep learning model, thereby realizing the identification of the radiation source location.

[0016] like Figure 1 As shown, the specific steps include:

[0017] Step 1: Construct a regional geographic model using GIS data, and use the dominant path propagation model (DPM) to simulate the channel loss between the radiation source and each receiving point to obtain the original received signal strength (RSS) value, thus forming the initial data of the electromagnetic spectrum situation.

[0018] Step 2: Considering the sparse distribution of sensing nodes, data interpolation is used to fill in the missing areas of the received signal strength (RSS) values, forming a three-dimensional electromagnetic spectrum situation matrix, which is stored locally indexed by the radiation source location. Specifically: For any radiation source s, assuming the simulation area is X*Y*Z and the number of receiving sensing nodes is a*b*c, zeros are padded at locations within the area without sensing nodes to form a complete three-dimensional electromagnetic spectrum situation matrix (where spatial XYZ corresponds to power values).

[0019] Step 3: Construct the three-dimensional electromagnetic spectrum situation matrix as the input feature matrix, and the labels are the probability distribution matrices of the real radiation sources in three-dimensional space. This distribution matrix is ​​centered at the signal source location, with the probability of other locations being zero, forming a supervised learning dataset.

[0020] Step 4: For any radiation source *s*, train the neural network based on the local 3D electromagnetic spectrum situation matrix and the label probability distribution matrix. Specifically, the radiation sources deployed in the simulation area X*Y*Z and the received values ​​of the corresponding receiving sensors in that area are used as samples. For example, n radiation sources are set, the radiation source location information is used as labels, and the received values ​​of the sensing nodes corresponding to the radiation sources in that area are used as the label corresponding information. One label and its corresponding information are counted as one sample. The samples are randomly divided into training and validation sets, with 80% used for the training set and 20% for the validation set. The neural network structure model is implemented in Python using the PyTorch framework, employing an hourglass-shaped convolutional neural network architecture to balance multi-scale spatial feature extraction and spatial location inference accuracy recovery capabilities. The model consists of five main parts: a downsampling module, an upsampling module, a semantic embedding block, a heatmap generation layer, and a customized loss function. The downsampling module consists of multiple convolutional layers, a BatchNorm layer, and a ReLU activation function, combined with a max-pooling layer to achieve multi-layer feature extraction and spatial dimension compression of the input spectrum data. The upsampling module uses bilinear interpolation to perform step-by-step upsampling to restore the spatial resolution of the spatial location inference output.

[0021] Step 5: Input the three-dimensional electromagnetic spectrum situation matrix constructed in the test set into the trained neural network model in sequence. The model outputs the corresponding three-dimensional probability heat map. The coordinates of the position with the highest probability value in the heat map are the predicted radiation source positions corresponding to the sample.

[0022] Step 6: Calculate the Euclidean distance between the predicted radiation source location and its corresponding real location in each sample in the test set, and statistically analyze the average spatial location inference error value of all samples to evaluate the overall performance of the model in the three-dimensional spatial location inference task.

[0023] Example:

[0024] In this embodiment, a three-dimensional electromagnetic spectrum situation dataset was constructed based on a certain region (25km*25km*300m). The radiation source frequency was set to 1500MHz, and the transmission power was 20dBm. The radiation sources were deployed in the region at intervals of 300m*300m*50m, while the sensing nodes were deployed at intervals of 500m*500m*50m.

[0025] Table 1 shows the settings for geographic element parameters. Figure 3 The diagram shown is a schematic of a neural network structure. Figure 4 The diagram shown is a schematic of the actual label processing input neural network. Figure 5 The image shows the output label diagram of the neural network.

[0026]

[0027] Table 1

[0028] During the processing of data received from sensing nodes, due to the sparse node deployment, the original matrix formed by the received data is only 50*50*6. Using a 50*50*6 matrix directly without interpolation will amplify errors in spatial location reasoning in the actual physical context. To improve the spatial continuity of the data, an interpolation method is used to complete the data, expanding it into a 250*250*30 three-dimensional matrix.

[0029] The interpolated and completed matrix data is fed in as input. Figure 3 The hourglass-shaped neural network shown is trained. Figure 4 The label matrix structure is shown: only the voxels at the actual radiation source locations have a value of 1, while all other locations have a value of 0. A weighted loss function is introduced during training. This loss function constructs a weight matrix based on a 3D Gaussian distribution centered on the label locations. The weights decrease with increasing spatial distance, thus making the network focus more on the high-power region close to the radiation source. The weight matrix has the same dimensions as the input and output, and its mathematical expression is as follows:

[0030] ;

[0031] in Indicates an illegal signal source The coordinates are also the weighting center. The received data corresponding to the radiation source, after being completed, serves as the information corresponding to the label, forming a sample input to the network for training. Figure 5 This paper presents a heatmap of the model's predictions for a given input data during the validation phase. The predicted results are Gaussian filtered, and the coordinates of the maximum value are extracted as the predicted location of the radiation source. During validation, no label weighting is applied; the positions with a value of 1 in the label matrix are directly used as the true coordinates, and the spatial Euclidean distance is calculated between these coordinates and the predicted results. Experimental results show that the proposed method can still achieve a relative geospatial location inference error of 0.69% even under sparse sensing node deployment conditions. This result validates the accuracy and practicality of the proposed method for three-dimensional spatial location inference of radiation sources in complex electromagnetic environments.

[0032] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

[0033] A computer-readable storage medium storing one or more programs, the programs including instructions that, when executed by a computing device, cause the computing device to perform a method for inferring the three-dimensional spatial location of an electromagnetic radiation source based on spectral situational data.

[0034] A computing device includes one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include methods for performing inference methods for completing the three-dimensional spatial location of electromagnetic radiation sources based on spectral situational data.

[0035] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0036] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0037] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0038] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0039] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for inferring the three-dimensional spatial location of electromagnetic radiation sources based on spectral situation data, characterized in that: Acquire GIS data of the target area, construct a geographic model of the target area, use the dominant path propagation model to simulate the channel loss between the radiation source and each receiving point in the geographic model, obtain the original received signal strength value, and constitute the initial data of the electromagnetic spectrum situation. The missing regions of received signal strength in the initial electromagnetic spectrum situation data are filled in to form a three-dimensional electromagnetic spectrum situation matrix, which is stored locally according to the radiation source location index. The three-dimensional electromagnetic spectrum situation matrix is ​​input into a pre-trained neural network, which outputs a corresponding three-dimensional probability heatmap. The coordinates of the location with the highest probability value in the heatmap are the locations of the corresponding radiation sources.

2. The method for inferring the three-dimensional spatial location of an electromagnetic radiation source based on spectral situational data according to claim 1, characterized in that, The method for filling in the missing regions of received signal strength in the initial electromagnetic spectrum situation data is as follows: For any radiation source s, assuming the target area is X*Y*Z and the number of its receiving sensing nodes is a*b*c, zeros are filled in the positions where there are no sensing nodes in the area.

3. The method for inferring the three-dimensional spatial location of an electromagnetic radiation source based on spectral situational data according to claim 1, characterized in that, The training method for the neural network includes: The three-dimensional electromagnetic spectrum situation matrix is ​​constructed as the input feature matrix, and the label is the probability distribution matrix of the real radiation source in three-dimensional space. The distribution matrix is ​​centered on the signal source position, and the probability of other positions is zero, forming a supervised learning dataset. For any radiation source s, a neural network is trained based on the local three-dimensional electromagnetic spectrum situation matrix and the tag probability distribution matrix.

4. The method for inferring the three-dimensional spatial location of an electromagnetic radiation source based on spectral situation data according to claim 3, characterized in that, The training method for the neural network also includes: The radiation sources deployed in the target area X*Y*Z and the received values ​​of the corresponding sensing nodes in the area are used as samples. n radiation sources are set, the location information of the radiation sources is used as labels, and the received values ​​of the sensing nodes in the corresponding areas are used as the corresponding label information. One label and its corresponding information are counted as one sample. The samples are randomly divided into training set and validation set, with 80% used for training set and 20% used for validation set.

5. The method for inferring the three-dimensional spatial location of an electromagnetic radiation source based on spectral situation data according to claim 1, characterized in that, The neural network is implemented in Python using the PyTorch framework and employs an hourglass-shaped convolutional neural network architecture, balancing multi-scale spatial feature extraction with spatial location inference accuracy recovery. It consists of five main parts: a downsampling module, an upsampling module, a semantic embedding block, a heatmap generation layer, and a customized loss function. The downsampling module comprises multiple convolutional layers, a BatchNorm layer, and a ReLU activation function, working in conjunction with a max-pooling layer to achieve multi-layer feature extraction and spatial dimension compression of the input spectral data. The upsampling module uses bilinear interpolation to perform step-by-step upsampling to restore the spatial resolution of the spatial location inference output.

6. A system for inferring the three-dimensional spatial location of electromagnetic radiation sources based on spectral situation data, characterized in that, include: The electromagnetic spectrum situation initial data module is used to acquire GIS data of the target area, construct a geographic model of the target area, use the dominant path propagation model to simulate the channel loss between the radiation source and each receiving point in the geographic model, obtain the original received signal strength value, and constitute the electromagnetic spectrum situation initial data. The three-dimensional electromagnetic spectrum situation matrix module is used to fill in the missing areas of received signal strength in the initial electromagnetic spectrum situation data to form a three-dimensional electromagnetic spectrum situation matrix, which is stored locally according to the radiation source location index. The neural network module is used to take a three-dimensional electromagnetic spectrum situation matrix as input and output a corresponding three-dimensional probability heat map. The coordinates of the position with the highest probability value in the heat map are the corresponding radiation source positions.

7. A computer-readable storage medium for storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods according to claims 1 to 5.

8. A computing device, characterized in that, include: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods according to claims 1 to 5.

Images