Radiation source positioning method, computer equipment and computer storage medium

Abstract

The embodiment of the invention discloses a radiation source positioning method, computer equipment and a computer storage medium. The invention provides a universal radiation source positioning technology based on an efficient neural network, and for a lightweight electromagnetic spectrum sensing terminal, the energy load is limited, the computing power is limited, operators capable of being calculated by a carried computing chip are limited, and the real-time requirement of a positioning algorithm needs to be carried out, so that the universal target positioning technology research based on the efficient neural network needs to be carried out. According to the embodiment of the invention, the sensing field covering the whole graph is realized by optimally selecting the network parameters, so that under the condition that the positioning precision is not excessively reduced, the calculation overhead is reduced, the output time delay is reduced, the calculation efficiency of the neural network model is improved, a higher-degree general operator is adopted, the algorithm can be deployed to a wider calculation platform, and the calculation efficiency is improved. And possibility is provided for actual deployment.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of image processing, and specifically to a radiation source positioning method, a computer device, and a computer storage medium. Background Art

[0002] The traditional target location positioning method needs to first estimate the angle of the radiation source, the time difference of arrival at each platform, the signal strength and other information, and then solve the nonlinear equation, so it is also called two-step positioning.

[0003] However, in urban environments, due to the widespread presence of building obstructions, the radiation source is often not within the line of sight of the electromagnetic sensing device, reducing the efficiency of locating the target. In addition, the traditional two-step positioning method accumulates and amplifies positioning errors, and usually requires solving nonlinear equations, which will lead to the amplification of positioning errors, resulting in low positioning accuracy and low positioning efficiency. Summary of the invention

[0004] The embodiments of the present application provide a radiation source positioning method, a computer device, and a computer storage medium. By adjusting the deep network hyperparameters, a perception field covering the entire map is achieved, thereby obtaining a higher passive positioning accuracy.

[0005] A first aspect of an embodiment of the present application provides a radiation source positioning method, the method comprising:

[0006] Acquire a simulation data set of an urban geographical environment, the simulation data set comprising a global path loss map of a simulated radiation source, a simulated raster image of the urban geographical environment, and preset spatial coordinates of the simulated radiation source;

[0007] Acquire an initial neural network, and use the initial neural network to perform spatial random sampling on the global path loss map to obtain an electromagnetic information image of the urban geographical environment;

[0008] Inputting the simulated grid image and the electromagnetic information image into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment according to the simulated grid image and the electromagnetic information image, and obtaining the predicted position coordinates of the radiation source output by the initial neural network;

[0009] The model parameters of the initial neural network are adjusted according to the predicted position coordinates and the preset spatial coordinates, and the model training is stopped until the model training of the initial neural network reaches a convergence condition to obtain a target neural network, which is used to predict the position of the radiation source in the urban geographical environment.

[0010] A second aspect of an embodiment of the present application provides a computer device, the computer device comprising:

[0011] A first acquisition unit is used to acquire a simulation data set of an urban geographical environment, wherein the simulation data set includes a global path loss map of a simulated radiation source, a simulated raster image of the urban geographical environment, and preset spatial coordinates of the simulated radiation source;

[0012] A second acquisition unit is used to acquire an initial neural network, and use the initial neural network to perform spatial random sampling on the global path loss map to obtain an electromagnetic information image of the urban geographical environment;

[0013] A model training unit, used for inputting the simulated grid image and the electromagnetic information image into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment according to the simulated grid image and the electromagnetic information image, and obtains the predicted position coordinates of the radiation source output by the initial neural network;

[0014] The model training unit is also used to adjust the model parameters of the initial neural network according to the predicted position coordinates and the preset spatial coordinates, and stop the model training until the model training of the initial neural network reaches the convergence condition to obtain the target neural network, which is used to predict the position of the radiation source in the urban geographical environment.

[0015] A third aspect of an embodiment of the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the method of the first aspect when executing the computer program.

[0016] A fourth aspect of an embodiment of the present application provides a computer storage medium, in which instructions are stored. When the instructions are executed on a computer, the computer executes the method of the first aspect.

[0017] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0018] For lightweight electromagnetic spectrum sensing terminals, their energy load is limited, computing power is limited, the number of operators that can be calculated by the computing chips they carry is limited, and the real-time requirements of the positioning algorithm require research on general target positioning technology based on efficient neural networks. This embodiment achieves a perception field covering the entire map by optimally selecting network parameters, thereby reducing computing overhead and output latency without excessively reducing positioning accuracy, improving the computing efficiency of the neural network model, and using more general operators to enable the algorithm to be deployed on a wider range of computing platforms, providing possibilities for actual deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1A schematic diagram of a flow chart of a radiation source positioning method in an embodiment of the present application;

[0020] Figure 2 An exemplary simulation diagram of an urban geographical environment obtained by using simulation software in an embodiment of the present application;

[0021] Figure 3 Based on Figure 2 The simulation example diagram shown is an exemplary schematic diagram of obtaining a simulated raster image of an urban geographical environment;

[0022] Figure 4 Based on Figure 2 The simulation example shown in the figure obtains the global path loss of the simulated radiation source in the urban geographical environment Figure 1 An exemplary schematic diagram;

[0023] Figure 5 An exemplary structural diagram of the backbone structure of the initial neural network in the embodiment of the present application;

[0024] Figure 6 This is an exemplary schematic diagram of the positioning effect of the target neural network based on the radiation source output by the test set in the embodiment of the present application;

[0025] Figure 7 This is a schematic diagram of the structure of a computer device in an embodiment of the present application;

[0026] Figure 8 This is another structural diagram of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The embodiments of the present application provide a radiation source positioning method, a computer device, and a computer storage medium. By adjusting the deep network hyperparameters, a perception field covering the entire map is achieved, thereby obtaining a higher passive positioning accuracy.

[0028] The radiation source positioning method of the embodiment of the present application uses the propagation characteristics of electromagnetic signals to carry out research on the direct estimation technology of the target position. Although the urban environment is complex and changeable, this real-time mapping method can often provide rough prior information of the environment, which can be used for position inference. In the embodiment of the present application, based on the encoder-decoder network structure, a general radiation source positioning technology based on an efficient neural network is proposed. This technology achieves a perceptual field covering the entire map by adjusting the deep network hyperparameters, thereby obtaining a higher passive positioning accuracy.

[0029] The following is a description of the radiation source positioning method in the embodiment of the present application:

[0030] See also Figure 1 In the embodiment of the present application, one embodiment of the radiation source positioning method includes:

[0031] 101. Acquire a simulation data set of an urban geographical environment, wherein the simulation data set includes a global path loss map of a simulated radiation source, a simulated raster image of the urban geographical environment, and preset spatial coordinates of the simulated radiation source;

[0032] The method of this embodiment can be applied to a computer device, which can be a device such as a terminal, a server, etc. that has certain computing power and data processing capabilities. The computer device can use a simulation software program to simulate an urban geographical environment, and the urban geographical environment can be any regional location. In the process of simulating the urban geographical environment using the simulation software program, a simulation data set of the urban geographical environment generated by the urban geographical environment simulation can be obtained, and the simulation data set includes a global path loss map of the simulated radiation source, a simulated raster image of the urban geographical environment, and the preset spatial coordinates of the simulated radiation source.

[0033] The radiation source to be located in this embodiment refers to an object that can generate electromagnetic radiation, including but not limited to communication base stations, black radio stations, etc.

[0034] For example, to build a large-scale simulation data set, we can first collect diverse urban geographic information from large-scale open source world maps (such as OpenStreetMap and other map software). OpenStreetMap contains crowdsourced annotated shape, geometry and height information of urban buildings, which has a key impact on the electromagnetic spectrum propagation of the city. Specifically, the simulation process is as follows:

[0035] 1) Importing urban geographic information into WallMan of Feko-Winprop high-precision electromagnetic simulation software, the urban geographic information includes the height and shape of buildings in the urban geographical environment, etc., which is used to describe the impact of buildings on electromagnetic spectrum propagation;

[0036] 2) Set the electromagnetic property parameters of urban buildings in WallMan;

[0037] 3) Set the antenna parameters and position of the radiation source in PropMan of Feko-Winprop, as well as the parameters of the simulated receiving antenna, and simulate the signal propagation of the radiation source at a fixed height. Then perform spatial and grayscale quantization, and store the resulting rasterized image as a global path loss map of the radiation source ~X i ; Among them, the antenna parameters may include frequency response, height, gain and other parameters related to the antenna;

[0038] 4) Save the geometric shape of the city building represented by the rasterized grayscale image from a bird's-eye view ~I iAs urban geographic information, non-zero grayscale value areas are selected to represent the shape and height of urban buildings, and zero grayscale value areas represent open areas without buildings in the city;

[0039] 5) Save this round of simulation and the corresponding spatial coordinates of the radiation source

[0040] For example, by using the Feko-Winprop simulation software, the following Figure 2 The simulation example diagram shown in FIG. 1 is used to obtain the following Figure 3 Simulated raster image shown And get Figure 4 Global path loss plot for the simulated radiator shown

[0041] It should be noted that, in this embodiment, OpenStreetMap and Winprop are used as examples in the simulation data set generation step. In fact, maps and electromagnetic field distributions generated by other simulation software may also be used.

[0042] 102. Obtain an initial neural network, and use the initial neural network to perform spatial random sampling on the global path loss map to obtain an electromagnetic information image of the urban geographical environment;

[0043] The computer device can further deploy an initial neural network, and use the initial neural network to perform spatial random sampling on a global path loss map of a simulated radiation source to obtain an electromagnetic information image of the urban geographical environment.

[0044] For example, taking the training process of the simulation data set as an example, the overall framework of the initial neural network adopts an encoder-decoder structure, and the first layer input is a simulated raster image of the urban geographical environment. And the global path loss graph The electromagnetic information image of the urban geographical environment is obtained by random sampling in space

[0045] 103. Input the simulated grid image and the electromagnetic information image into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment according to the simulated grid image and the electromagnetic information image, and obtains the predicted position coordinates of the radiation source output by the initial neural network;

[0046] This embodiment uses signal strength data with spatial coordinates and urban geographic environment data as deep learning training input to directly estimate the spatial coordinates of the radiation source in an end-to-end manner.

[0047] Therefore, after obtaining the electromagnetic information image of the urban geographical environment, the simulated raster image and the electromagnetic information image can be input into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment based on the simulated raster image and the electromagnetic information image, and obtains the predicted position coordinates of the radiation source output by the initial neural network.

[0048] It should be noted that the neural network perception field, which is determined by the convolution kernel size, expansion coefficient and sampling scaling ratio, has a great influence on the positioning accuracy. Therefore, in this embodiment, a generative model is actually used to solve the radiation source position. The goal is to obtain the simulated raster image of the input urban geographical environment. And the electromagnetic information image obtained by spatial sampling The predicted position coordinates of the radiation source output by the model are obtained In this way, the real radiation source position η is estimated i .

[0049] Among them, when obtaining the predicted position coordinates of the radiation source output by the initial neural network, an optional method is to obtain the conditional distribution of the predicted position coordinates of the radiation source output by the initial neural network, which conditional distribution is used to represent each predicted position coordinate and its corresponding confidence probability.

[0050] Specifically, since the model training task of this embodiment is inherently random, that is, there is a conditional distribution The idea of ​​generative modeling is to use neural networks to represent the conditional distribution For this method, the real radiation source position η i The information is contained in the simulated raster image represented graphically. and electromagnetic information images Therefore, the key to the design of the neural network structure for improving the positioning accuracy of the model training task in this embodiment is to design a neural network that can better represent the With η i After using raster images to represent the urban geographical environment and electromagnetic environment, their overall information is contained in and The convolutional neural network needs to model the relationship between the overall grid points so that it can effectively represent The receptive field of a neural network determines the spatial range in which the neural network can characterize the relationship between modeling grid points in a two-dimensional grid space. Experimental results show that a receptive field covering the entire image is crucial to the positioning accuracy of the model training task of this embodiment.

[0051] 104. Adjust the model parameters of the initial neural network according to the predicted position coordinates and the preset spatial coordinates, stop the model training until the model training of the initial neural network reaches a convergence condition, and obtain a target neural network, wherein the target neural network is used to predict the position of the radiation source in the urban geographical environment;

[0052] When the model outputs the predicted position coordinates of the radiation source in the urban geographical environment, a loss function can be constructed according to the predicted position coordinates and the preset spatial coordinates of the simulated radiation source set in the simulation process of the urban geographical environment, and the model parameters of the initial neural network are adjusted according to the loss function. When the loss function meets the convergence condition, it is determined that the model training of the initial neural network has reached the convergence condition. At this time, the model training can be stopped to obtain the target neural network obtained by the initial neural network after model training. The target neural network can be used to predict the position of the radiation source in the urban geographical environment.

[0053] For example, based on the current parameters and considering the computer power constraints, a neural network backbone structure can be selected, whose structure is as follows Figure 5 As shown in the table. As shown in the figure, the 10th layer 2D-Softmax normalizes the spatial dimension index of the feature, and the 11th layer, which is the last layer, uses the 2D-CenterMass layer to output the predicted location coordinates of the radiation source. The 11th layer 2D-CenterMass linearly combines the results of the 10th layer with the grid coordinates to obtain the location coordinates of the radiation source and output the predicted position coordinates of the radiation source. In the network, the spatial downsampling features are calculated by the strided convolution, the receptive field is increased by the dilated convolution, and the spatial upsampling features are calculated by the combination module of the nearest neighbor interpolation and convolution. The convolution kernel size, dilation coefficient and sampling scaling ratio are shown in the table.

[0054] In terms of model training, the mean square error between the predicted position coordinates and the actual position coordinates of the radiation source is used as the loss function, the optimizer is Adam, the parameters β1 = 0.9, β2 = 0.999, and the learning rate is set to 1e -3 .

[0055] In an optional implementation of the present embodiment, after the target neural network is obtained through model training, a pre-collected spatial spectrum data set of the urban geographical environment and a real-collected raster image of the urban geographical environment can be obtained, and the spatial spectrum data set includes multiple spatial spectrum data samples, each of which includes the spatial coordinates of the sample of the urban geographical environment and the electromagnetic spectrum signal reception intensity of the urban geographical environment.

[0056] Afterwards, the spatial data and intensity data of the spatial spectrum dataset and the real-sampled raster image can be quantified respectively to obtain a rasterized spatial spectrum dataset corresponding to the spatial spectrum dataset and a quantitative map of the urban geographic environment corresponding to the real-sampled raster image.

[0057] The rasterized spatial spectrum dataset and the urban geographic environment quantization map are input into the target neural network, so that the target neural network outputs the predicted radiation source location according to the rasterized spatial spectrum dataset and the urban geographic environment quantization map.

[0058] An optional implementation method of quantizing spatial data of a spatial spectrum dataset is to select a top view of a first preset size in the spatial spectrum dataset, and divide the top view into multiple grids with a second preset size as a unit size in the top view to obtain a rasterized spatial spectrum dataset represented by a rasterized image, wherein the second preset size is smaller than the first preset size.

[0059] An optional implementation of quantizing the intensity data of the spatial spectrum dataset is to regard the grayscale data of the image as the intensity data of the image, quantize the grayscale data of the spatial spectrum dataset by a preset bit, and obtain a rasterized spatial spectrum dataset.

[0060] For example, suppose that the spatial-spectral data of the urban geographical environment is collected at a fixed height, and a single spatial-spectral data sample is set to x ij =[x ij ,y ij ,a ij ], labelled as η i . Where x ij and ij They represent the spatial coordinates of the jth sample collected in urban geographical environment i, a ij represents the corresponding signal receiving strength collected by the electromagnetic spectrum payload, η i represents the real spatial coordinates of the radiation source corresponding to the space-spectrum data set. Assuming that the data collected from the same radiation source location in the same urban geographical environment contains L samples, the corresponding space-spectrum data can be expressed as X i =[x i1 ,x i2 ,...,x iL ]; the associated urban geographical environment can be represented by a raster image as i .

[0061] Afterwards, in order to i and I i All the images are processed in a unified way as raster images, which can be used for X i The spatial data and intensity data are quantized and preprocessed respectively, and the rasterized spatial spectrum data represented by the rasterized image is further obtained. When quantizing spatial data, the urban geographical environment in a 256m×256m square area of ​​the overhead view can be selected, and 1m×1m can be used as the unit distance to generate 256×256 spatial resolution space-spectrum data. In the intensity (grayscale) quantization part, 8-bit quantization can be used. Similarly, the real-sampled raster image I of the urban geographical environment data can also be i Similarly, quantitative processing of spatial data and intensity data can be performed to obtain a quantitative map of the urban geographical environment corresponding to the actual raster image.

[0062] Among them, the multiple intensity data located at the same grid point after spatial quantization can be averaged. Therefore, after the above preliminary preprocessing, the actual space-spectrum data set represented in the form of a raster image can be obtained. Where N is the number of spatial-spectral data pairs in the dataset.

[0063] In addition, when storing, in order to standardize the data set, image data with a spatial resolution of 256×256 in the range of 256m×256m corresponding to the real map can be intercepted and stored.

[0064] For example, under the current parameter settings mentioned above, using the actual spatial-spectral data set obtained above As a test set for model testing, the data set is input into the target neural network that has completed model training, and the positioning effect of the target neural network output for the radiation source can be obtained, such as Figure 6 As shown in the figure, the black area represents the urban building, the red mark center represents the actual radiation source position, and the blue mark center represents the positioning position predicted by the model based on the green marked spatial sampling point. Experiments on the test set show that the positioning accuracy of the method in this embodiment is 4.03 pixels on average, that is, the simulation positioning error of this method is 4.03 meters within the urban geographical area of ​​256m×256m.

[0065] Therefore, this embodiment proposes a general radiation source positioning technology based on an efficient neural network: for lightweight electromagnetic spectrum sensing terminals, their energy load is limited, computing power is limited, the number of operators that can be calculated by the computing chip they carry is limited, and the real-time requirements of the positioning algorithm require research on general target positioning technology based on efficient neural networks. This embodiment achieves a perception field covering the entire map by optimally selecting network parameters, thereby reducing computing overhead and output latency without excessively reducing positioning accuracy, improving the computing efficiency of the neural network model, and using a higher degree of general operators, so that the algorithm can be deployed on a wider range of computing platforms, providing possibilities for actual deployment.

[0066] The radiation source positioning method in the embodiment of the present application is described above. The computer device in the embodiment of the present application is described below. Figure 7 In the embodiment of the present application, one embodiment of the computer device includes:

[0067] A first acquisition unit is used to acquire a simulation data set of an urban geographical environment, wherein the simulation data set includes a global path loss map of a simulated radiation source, a simulated raster image of the urban geographical environment, and preset spatial coordinates of the simulated radiation source;

[0068] A second acquisition unit is used to acquire an initial neural network, and use the initial neural network to perform spatial random sampling on the global path loss map to obtain an electromagnetic information image of the urban geographical environment;

[0069] A model training unit, used for inputting the simulated grid image and the electromagnetic information image into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment according to the simulated grid image and the electromagnetic information image, and obtains the predicted position coordinates of the radiation source output by the initial neural network;

[0070] The model training unit is also used to adjust the model parameters of the initial neural network according to the predicted position coordinates and the preset spatial coordinates, and stop the model training until the model training of the initial neural network reaches the convergence condition to obtain the target neural network, which is used to predict the position of the radiation source in the urban geographical environment.

[0071] In a preferred implementation manner of this embodiment, the method further includes:

[0072] A third acquisition unit is used to acquire a pre-collected spatial spectrum data set of the urban geographical environment, and to acquire a real-collected raster image of the urban geographical environment, wherein the spatial spectrum data set includes a plurality of spatial spectrum data samples, and each of the spatial spectrum data samples includes the spatial coordinates of the sample of the urban geographical environment and the electromagnetic spectrum signal receiving intensity of the urban geographical environment;

[0073] A quantization unit, used to quantize the spatial data and intensity data of the spatial spectrum data set and the real-sampled raster image, respectively, to obtain a rasterized spatial spectrum data set corresponding to the spatial spectrum data set and a quantized urban geographic environment map corresponding to the real-sampled raster image;

[0074] The testing unit is used to input the rasterized spatial spectrum data set and the urban geographic environment quantization map into the target neural network, so that the target neural network outputs the predicted radiation source position according to the rasterized spatial spectrum data set and the urban geographic environment quantization map.

[0075] In a preferred implementation of this embodiment, the quantization unit is specifically used for:

[0076] Selecting a top view of a first preset size in the spatial spectrum dataset, and dividing the top view into a plurality of grids with a second preset size as a unit size in the top view, so as to obtain a rasterized spatial spectrum dataset represented by a rasterized image;

[0077] The second preset size is smaller than the first preset size.

[0078] In a preferred implementation of this embodiment, the quantization unit is specifically used for:

[0079] The grayscale data of the image is regarded as the intensity data of the image, and the grayscale data of the spatial spectrum data set is quantized by a preset bit to obtain the rasterized spatial spectrum data set.

[0080] In a preferred implementation of this embodiment, the model training unit is specifically used for:

[0081] The conditional distribution of the predicted position coordinates of the radiation source output by the initial neural network is obtained, and the conditional distribution is used to represent each of the predicted position coordinates and its corresponding confidence probability.

[0082] In this embodiment, the operations performed by each unit in the computer device are the same as those described above. Figure 1 The description in the illustrated embodiment is similar and will not be repeated here.

[0083] The computer device in the embodiment of the present application is described below. Figure 8 In the embodiment of the present application, one embodiment of the computer device includes:

[0084] The computer device 800 may include one or more central processing units (CPU) 801 and a memory 805 , wherein the memory 805 stores one or more application programs or data.

[0085] The memory 805 may be a volatile storage or a persistent storage. The program stored in the memory 805 may include one or more modules, each of which may include a series of instruction operations in the computer device. Furthermore, the central processing unit 801 may be configured to communicate with the memory 805 and execute a series of instruction operations in the memory 805 on the computer device 800.

[0086] The computer device 800 may also include one or more power supplies 802, one or more wired or wireless network interfaces 803, one or more input and output interfaces 804, and / or one or more operating systems, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0087] The CPU 801 can execute the aforementioned Figure 1 The operations performed by the computer device in the illustrated embodiment will not be described in detail here.

[0088] The present application also provides a computer storage medium, wherein one embodiment includes: the computer storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the aforementioned Figure 1 The operations performed by the computer device in the illustrated embodiment.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0090] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0091] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0092] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0093] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

Claims

1. A radiation source positioning method, characterized in that: The method comprises: Acquire a simulation data set of an urban geographical environment, the simulation data set comprising a global path loss map of a simulated radiation source, a simulated raster image of the urban geographical environment, and preset spatial coordinates of the simulated radiation source; Acquire an initial neural network, and use the initial neural network to perform spatial random sampling on the global path loss map to obtain an electromagnetic information image of the urban geographical environment; Inputting the simulated grid image and the electromagnetic information image into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment according to the simulated grid image and the electromagnetic information image, and obtaining the predicted position coordinates of the radiation source output by the initial neural network; The model parameters of the initial neural network are adjusted according to the predicted position coordinates and the preset spatial coordinates, and the model training is stopped until the model training of the initial neural network reaches a convergence condition to obtain a target neural network, which is used to predict the position of the radiation source in the urban geographical environment.

2. The method according to claim 1, characterized in that The method further comprises: Acquire a pre-collected spatial spectrum data set of the urban geographical environment, and acquire a real-collected raster image of the urban geographical environment, wherein the spatial spectrum data set includes a plurality of spatial spectrum data samples, and each of the spatial spectrum data samples includes the spatial coordinates of the sample of the urban geographical environment and the electromagnetic spectrum signal receiving intensity of the urban geographical environment; Quantifying the spatial data and intensity data of the spatial spectrum dataset and the real-collected raster image respectively, to obtain a rasterized spatial spectrum dataset corresponding to the spatial spectrum dataset and a quantitative map of the urban geographic environment corresponding to the real-collected raster image; The rasterized spatial spectrum data set and the urban geographic environment quantization map are input into the target neural network, so that the target neural network outputs the predicted radiation source position according to the rasterized spatial spectrum data set and the urban geographic environment quantization map.

3. The method according to claim 2, characterized in that quantifying spatial data of the spatial spectrum data set, including: Selecting a top view of a first preset size in the spatial spectrum dataset, and dividing the top view into a plurality of grids with a second preset size as a unit size in the top view, so as to obtain a rasterized spatial spectrum dataset represented by a rasterized image; The second preset size is smaller than the first preset size.

4. The method according to claim 2, characterized in that: quantifying the intensity data of the spatial spectrum data set, including: The grayscale data of the image is regarded as the intensity data of the image, and the grayscale data of the spatial spectrum data set is quantized by a preset bit to obtain the rasterized spatial spectrum data set.

5. The method according to claim 1, characterized in that The obtaining of the predicted position coordinates of the radiation source output by the initial neural network comprises: The conditional distribution of the predicted position coordinates of the radiation source output by the initial neural network is obtained, and the conditional distribution is used to represent each of the predicted position coordinates and its corresponding confidence probability.

6. A computer device, characterized in that: The computer device comprises: A first acquisition unit is used to acquire a simulation data set of an urban geographical environment, wherein the simulation data set includes a global path loss map of a simulated radiation source, a simulated raster image of the urban geographical environment, and preset spatial coordinates of the simulated radiation source; A second acquisition unit is used to acquire an initial neural network, and use the initial neural network to perform spatial random sampling on the global path loss map to obtain an electromagnetic information image of the urban geographical environment; A model training unit, used for inputting the simulated grid image and the electromagnetic information image into the initial neural network, so that the initial neural network predicts the position of the radiation source of the urban geographical environment according to the simulated grid image and the electromagnetic information image, and obtains the predicted position coordinates of the radiation source output by the initial neural network; The model training unit is also used to adjust the model parameters of the initial neural network according to the predicted position coordinates and the preset spatial coordinates, and stop the model training until the model training of the initial neural network reaches the convergence condition to obtain the target neural network, which is used to predict the position of the radiation source in the urban geographical environment.

7. The computer device according to claim 6, characterized in that The method further comprises: A third acquisition unit is used to acquire a pre-collected spatial spectrum data set of the urban geographical environment, and to acquire a real-collected raster image of the urban geographical environment, wherein the spatial spectrum data set includes a plurality of spatial spectrum data samples, and each of the spatial spectrum data samples includes the spatial coordinates of the sample of the urban geographical environment and the electromagnetic spectrum signal receiving intensity of the urban geographical environment; A quantization unit, used to quantize the spatial data and intensity data of the spatial spectrum data set and the real-sampled raster image, respectively, to obtain a rasterized spatial spectrum data set corresponding to the spatial spectrum data set and a quantized urban geographic environment map corresponding to the real-sampled raster image; The testing unit is used to input the rasterized spatial spectrum data set and the urban geographic environment quantization map into the target neural network, so that the target neural network outputs the predicted radiation source position according to the rasterized spatial spectrum data set and the urban geographic environment quantization map.

8. The computer device according to claim 6, characterized in that The model training unit is specifically used for: The conditional distribution of the predicted position coordinates of the radiation source output by the initial neural network is obtained, and the conditional distribution is used to represent each of the predicted position coordinates and its corresponding confidence probability.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

10. A computer storage medium, characterized in that: The computer storage medium stores instructions, and when the instructions are executed on a computer, the computer is caused to execute the method according to any one of claims 1 to 5.

Images