Method and device for calculating terrain loss of radio wave propagation, and electronic equipment

Abstract

The embodiment of the present application relates to the technical field of radio wave propagation, and discloses a kind of ground feature loss calculation method of radio wave propagation, the method comprises: obtaining the environment feature in target environment, environment feature includes vegetation feature;According to the environment feature and clustering classification model determines target vegetation environment label;Target vegetation environment label is input into vegetation ground feature loss model, and vegetation loss result is obtained;Vegetation ground feature loss model is obtained in advance according to environment label, sample environment feature and sample loss difference to the training of neural network, and sample loss difference is the difference between measured loss data and the estimated loss calculated by the preset standard ground feature loss model;According to vegetation ground feature loss result and standard ground feature loss result, target ground feature loss result is obtained;Wherein, the standard ground feature loss result is calculated according to the environment feature and the preset standard ground feature loss model.The accuracy and universality of ground feature loss calculation are effectively improved in the embodiment of the present application.

Description

Technical Field

[0001] The present invention relates to the field of radio wave propagation technology, specifically to a method, apparatus, electronic device, and computer-readable storage medium for calculating ground object loss during radio wave propagation. Background Technology

[0002] Currently, 5G, as a leading technology in the global technological revolution, provides reliable information exchange support for vertical industry applications such as mobile internet, industrial internet, vehicle-to-everything (V2X), fintech, smart healthcare, new media, and education, thanks to its characteristics of high bandwidth, massive connectivity, high reliability, and low latency. It is a key new infrastructure supporting the networked, digital, and intelligent transformation of the economy and society. In terms of application scenarios, typical 5G network scenarios include three-dimensional urban areas, suburbs, mountainous areas, and roads and bridges. The increased frequency and variety of environments make 5G wireless network planning more complex than 4G in terms of wireless propagation environment and characteristics. The spacing between base stations is also denser, and the number of base stations is greater, placing higher demands on the accuracy and efficiency of network planning. To achieve accurate, efficient, and routine 5G wireless network planning and optimization, it is necessary to rely on precise wireless channel models and efficient simulation methods.

[0003] However, the accuracy of ray tracing simulations depends not only on a correct multipath propagation model but also on the proper combination of scene geometry and material-related model parameters. Therefore, appropriate selection and use of propagation models and material-related model parameters are crucial for ensuring the accuracy and universality of ray tracing simulations in different scenarios. While the ground model is expected to be accurate for scenarios corresponding to actual measurement activities, additional multi-frequency measurement activities in other urban environments are valuable for further validation / calibration. Furthermore, several aspects remain unresolved in the current model: ground cover loss due to vegetation is not modeled. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention provide a method, apparatus, electronic device and computer-readable storage medium for calculating ground object loss during radio wave propagation, to solve the technical problem of poor accuracy and universality of ray tracing simulation in the prior art.

[0005] According to one aspect of the present invention, a method for calculating ground object loss during radio wave propagation is provided, the method comprising:

[0006] Acquire environmental features in the target environment, including vegetation features;

[0007] The target vegetation environment label is determined based on the environmental characteristics and cluster classification model.

[0008] The target vegetation environment label is input into the vegetation loss model to obtain the vegetation loss result. The vegetation loss model is obtained by training the neural network in advance based on the environment label, sample environment characteristics and sample loss difference. The sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model.

[0009] Based on the vegetation loss results and the standard feature loss results, the target feature loss results are obtained; wherein, the standard feature loss results are calculated based on the environmental characteristics and the preset standard feature loss model.

[0010] In one optional approach, before determining the target vegetation environment label based on the environmental features and the clustering classification model, the method further includes: acquiring sample environmental feature data and the measured loss data corresponding to the sample environmental feature data; clustering the sample environmental feature data with the sample loss difference as the optimization objective to obtain multiple environment labels; and training a preset neural network based on the environment labels and the sample environmental feature data to obtain the clustering classification model.

[0011] In one optional approach, the step of clustering the sample environmental feature data to obtain multiple environmental labels, with the sample loss difference as the optimization objective, further includes: calculating the estimated loss of the samples based on a preset standard ground feature loss model and the sample environmental features; calculating the sample loss difference based on the measured loss data and the estimated loss of the samples; and clustering the sample environmental feature data, grouping the sample environmental feature data whose sample loss difference is less than or equal to a preset threshold into the same environmental label to obtain multiple environmental labels.

[0012] In one optional approach, training a preset neural network based on the environmental labels and the sample environmental feature data to obtain the clustering classification model includes: determining the sample environmental feature dataset corresponding to each environmental label; and training a preset neural network based on each environmental label and the corresponding sample environmental feature dataset to obtain the clustering classification model.

[0013] In one optional approach, before inputting the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result, the method further includes: acquiring the sample environment features and sample loss difference corresponding to each environment label; the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model; training the neural network according to the sample environment features corresponding to each type of environment label and the sample loss difference to obtain the vegetation loss model, so as to establish the mapping relationship between the sample environment features and the sample loss difference.

[0014] In one optional approach, the step of training the neural network based on the sample environmental features corresponding to each environmental label and the sample loss difference to obtain the vegetation loss model, thereby establishing a mapping relationship between the sample environmental features and the sample loss difference, includes: inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference; optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, and continuing to execute the steps of inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference, and optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, until the error between the actual output of the loss difference and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations, thereby obtaining the vegetation loss model.

[0015] According to another aspect of the present invention, a ground feature loss calculation device for radio wave propagation is provided, the device comprising:

[0016] The first acquisition module is used to acquire environmental features in the target environment, including vegetation features.

[0017] The clustering module is used to determine the target vegetation environment label based on the environmental characteristics and the clustering classification model;

[0018] The determination module is used to input the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result; the vegetation loss model is obtained by training the neural network in advance based on the environment label, sample environment characteristics and sample loss difference, and the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model.

[0019] The output module is used to obtain the target feature loss result based on the vegetation feature loss result and the standard feature loss result; wherein the standard feature loss result is calculated based on the environmental characteristics and the preset standard feature loss model.

[0020] In an optional embodiment, the apparatus further includes: a second acquisition module, configured to acquire sample environmental feature data and measured loss data corresponding to the sample environmental feature data; a sample clustering module, configured to cluster the sample environmental feature data with the sample loss difference as the optimization objective to obtain multiple environmental labels; and a training module, configured to train a preset neural network based on the environmental labels and the sample environmental feature data to obtain the clustering classification model.

[0021] According to another aspect of the present invention, an electronic device is provided, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0022] The memory is used to store at least one executable instruction that causes the processor to perform the operation of the method for calculating ground feature loss during radio wave propagation.

[0023] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the storage medium stores at least one executable instruction, which, when executed on an electronic device, causes the electronic device to perform the operation of the method for calculating ground feature loss during radio wave propagation.

[0024] This invention, in its embodiments, acquires environmental features within a target environment, determines target vegetation environmental labels based on these features and a clustering classification model, inputs these labels into a vegetation loss model to obtain vegetation loss results, and then obtains target feature loss results based on these vegetation loss results and standard feature loss results. The standard feature loss results are calculated based on the environmental features and the preset standard feature loss model. The environmental labels are obtained by clustering sample environmental feature data with the sample loss difference as the optimization objective. The vegetation loss model is pre-trained on a neural network based on the environmental labels, sample environmental features, and sample loss differences. This approach allows the model to perform loss calculations based on different categories of environmental features, especially vegetation features, effectively improving the accuracy and universality of feature loss calculations.

[0025] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0026] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0027] Figure 1 A flowchart illustrating the ground object loss calculation method for radio wave propagation provided in an embodiment of the present invention is shown.

[0028] Figure 2 This diagram illustrates the training process of the ground feature loss model in the ground feature loss calculation method for radio wave propagation provided in an embodiment of the present invention.

[0029] Figure 3 A schematic diagram of the structure of the ground object loss calculation device for radio wave propagation provided in an embodiment of the present invention is shown;

[0030] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present invention is shown. Detailed Implementation

[0031] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0032] First, let's further introduce the existing technology:

[0033] The final ground cover loss model in ITU-R Recommendation 2108, for areas greater than approximately 1000m, shows that the ground cover loss L... l Modeling with linear functions:

[0034] L l =23.5+9.6log(f) dB(1)

[0035] Where f is the frequency in GHz, and N(0,6) is a normal distribution with zero mean and 6 dB standard deviation. The model parameters were determined by fitting the model to measurements taken at Aalborg and Gothenburg distances greater than 800 m.

[0036] Since no obvious frequency trend of the standard deviation was observed, a value of 6 dB, located at the high end of the distribution, was chosen as a conservative assumption to simplify the model. Therefore, for a range greater than approximately 1000 m, the ground cover loss L... l Modeling with linear functions:

[0037] L l =23.5+9.6log(f)+N(0,6) dB(2)

[0038] For shorter distances (0.26-1.2 km), ground feature loss L s Modeling with linear functions:

[0039] L s =32.98+23.9log(d)+3log(f) dB(3)

[0040] Where d is the distance in kilometers. To simplify the model combination, the standard deviation is aligned with the long-distance model (6 dB), slightly smaller than the 6.89 dB value in P.1411, using a roof model based on Tokyo measurements, and subtracting free space losses. Therefore, ground cover loss Ls is modeled as a linear function:

[0041] L s =32.98+23.9log(d)+3log(f)+N(0,6) dB(4)

[0042] The final ground feature loss model in ITU-R Recommendation RP.2108, ground-to-ground path L ctt The clutter loss at a location where p% does not exceed the specified value is given by the following formula:

[0043]

[0044] Among them, Q -1 (p / 100) is the inverse complementary normal distribution function.

[0045] To provide a single continuous model for all ranges greater than 260m, the two models are mixed using received power in linear units. The resulting combined model may be suitable for both ends of the link if the total link distance is greater than 1000m, given that clutter is embedded at both ends. However, for shorter distances, the line-of-sight (NLoS) model of ITU-RP.1411 should not be used.

[0046]

[0047] Because the elevation angle of the ground scene is expected to be very small, a zero-degree elevation angle is chosen for the Earth-to-space / air model. Comparison of the outputs confirms that the Earth-to-space / air model is too simplistic to simulate the ground scenario. At the low-loss end of the distribution, the Earth-to-air / air model significantly underestimates the losses. Therefore, for higher frequencies, the effective average loss is underestimated by 20 dB.

[0048] As can be seen from the formula calculations above, although the ground model is accurate for scenarios corresponding to actual measurement activities, additional multi-frequency measurement activities in other urban environments are valuable for further verification / calibration. Furthermore, some aspects are not addressed in the current model: ground cover loss due to vegetation is not modeled. Therefore, this invention introduces the impact of vegetation on ground cover loss to optimize its calculation. As mentioned earlier, the motivation for proposing a separate model for ground scenarios is that ground-to-air / air models may be too simplistic for the complex propagation on numerous rooftops in urban environments. This invention classifies vegetation samples according to vegetation characteristics and uses neural network clustering to optimize the formula. It combines environmental classification with radio wave propagation prediction error indicators, performing vegetation classification-machine learning-joint optimization. This allows the propagation environment classification to be based on radio wave propagation, achieving adaptive ground cover loss calculation. By combining machine learning with traditional formula derivation, a simulation method for vegetation classification calculation and automatic adaptation to propagation environment types is proposed for radio wave propagation, improving the accuracy and universality of radio wave propagation prediction and serving 5G multi-scenario wireless network planning.

[0049] Figure 1 A flowchart illustrating a method for calculating ground object loss during radio wave propagation, provided in an embodiment of the present invention, is shown. This method is executed by an electronic device. The electronic device can be a computer device, communication device, etc., and the embodiments of the present invention do not impose specific limitations. Figure 1 As shown, the method includes the following steps:

[0050] Step 110: Obtain environmental features in the target environment, including vegetation features.

[0051] The environmental characteristics of the target environment include vegetation features and transceiver deployment parameters. These vegetation features include vegetation height, vegetation type classification, and the number of electronic map grids occupied. The target environment refers to the environment of the area between the base station transceivers to be estimated. These environmental characteristics can be obtained through electronic maps or similar means.

[0052] Step 120: Determine the target vegetation environment label based on the environmental characteristics and clustering classification model.

[0053] Here, the target vegetation environment label refers to the environment label of the target environment. This environment label refers to multiple environment labels obtained by clustering the sample environment feature data in advance, with the sample loss difference as the optimization objective.

[0054] In this embodiment of the invention, before determining the target vegetation environment label based on the environmental features and the clustering classification model, the environmental features are pre-classified to obtain the environmental labels corresponding to the environmental features, and the neural network is trained based on the environmental labels and the sample environmental features to obtain the clustering classification model.

[0055] Specifically, it includes the following steps:

[0056] Step 1001: Obtain sample environmental feature data and the measured loss data corresponding to the sample environmental feature data.

[0057] First, a sample environmental feature dataset is acquired. This dataset includes multiple sample environmental feature data points, each containing multiple environmental features. These features include a large number of sample vegetation samples with different heights, areas, and types of vegetation, as well as the number of electronic map grids they occupy. Then, corresponding electronic maps, base station antenna patterns, and associated base station parameters are prepared for the area where the measured loss data is located. Finally, drive tests are conducted to obtain the measured loss data.

[0058] The environmental characteristics of the measured sample points are described in detail below:

[0059] Transceiver deployment: Basic deployment includes transmitter (usually a base station) engineering parameters such as location, antenna height, transmit power, azimuth, downtilt angle, frequency, and line loss; receiver (usually a user terminal) includes antenna height and location. This feature can also be parameters calculated from the basic deployment parameters, such as transceiver distance and equivalent transmit power.

[0060] Vegetation type characteristics: The type of vegetation landform is determined by a two-dimensional map of land cover type labels or by secondary calculations within a specific range between transceivers.

[0061] Vegetation height characteristics: The ground height and elevation of vegetation and landforms within a specific range between transceivers, or statistical values ​​obtained through secondary calculations, such as the mean and standard deviation of elevation of various land features, and the mean and standard deviation of ground height.

[0062] Vegetation area characteristics: The area of ​​vegetation and landforms within a specific range between transceivers, or a statistical value obtained through secondary calculation.

[0063] It can be understood that the specific range between transceivers includes, but is not limited to, the direct connection between transceivers and the first to fourth Fresnel ellipsoidal regions.

[0064] Step 1002: Using the sample loss difference as the optimization objective, cluster the sample environmental feature data to obtain multiple environmental labels.

[0065] The sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard ground feature loss model. Specifically, the estimated loss of the sample is calculated based on the preset standard ground feature loss model and the sample environmental characteristics, and the sample loss difference is calculated based on the measured loss data and the estimated loss of the sample. The preset standard ground feature loss model can be an estimation formula from the ITU-R P.1411 protocol. Initial values ​​for material-related and material-independent parameters in the preset standard ground feature loss model are determined, and raster calculations are performed based on the transceiver deployment parameters in the environmental characteristics and the preset standard ground feature loss model to estimate the estimated loss of the sample. Since the preset standard ground feature loss model does not consider the loss caused by vegetation features, the sample loss difference can be understood as the ground feature loss caused by vegetation features.

[0066] The specific process of clustering sample environmental feature data to obtain multiple environmental labels, with the sample loss difference as the optimization objective, is as follows: The sample environmental feature data are clustered, and those with a sample loss difference less than or equal to a preset threshold are grouped into the same environmental label to obtain multiple environmental labels. Sample environmental feature data with a sample loss difference greater than the preset threshold are considered substandard sample points and are re-merged with other sample environmental feature data to generate a new sample environmental feature dataset. This process is repeated iteratively until all sample environmental feature data meet the standard, or the number of iterations exceeds a specified value N, at which point the classification process terminates, resulting in multiple environmental labels. When the process terminates due to reaching the maximum number of iterations, substandard sampling points are considered singular values ​​and discarded.

[0067] Step 1003: Based on the environmental labels and the sample environmental feature data, train the preset neural network to obtain the clustering classification model.

[0068] In this embodiment of the invention, after obtaining each environmental label, a sample environmental feature dataset corresponding to each environmental label is pre-constructed. This sample environmental feature dataset includes multiple sample environmental feature data; each sample environmental feature data includes sample environmental features such as vegetation type, vegetation height, vegetation area, and the number of electronic map grids occupied by vegetation at the transceiver's location.

[0069] Then, based on each environmental label and its corresponding sample environmental feature dataset, a pre-defined neural network is trained to obtain the clustering classification model. The sample environmental feature dataset is divided into a training sample set and a test sample set. Using the sample environmental feature dataset as input and the corresponding environmental labels as ground truth, the neural network is trained. Specifically, the training sample set is input into the neural network, which outputs corresponding classification labels. These classification labels are compared with the ground truth (the environmental labels corresponding to the training sample set). A loss function is calculated based on the classification labels and ground truth. The parameters of the neural network are adjusted according to the loss function, and the training sample set is input into the adjusted neural network for further training. The corresponding classification labels are output again. When the loss function calculated between the classification labels and the ground truth (i.e., the actual environmental labels corresponding to the training sample set) reaches a preset threshold, or when the preset number of iterations is reached, training ends, and the trained neural network is obtained. The trained neural network is iteratively tested using a test set, and the parameters are continuously adjusted until the training and validation achieve the expected accuracy. The classifier is then complete, and the clustering classification model is obtained. In this embodiment of the invention, the specific structure of the neural network is not specifically limited; it can be a fully connected neural network, a convolutional neural network, or a hybrid architecture of different neural networks.

[0070] In this way, the environment labels corresponding to the locations of each receiver within the simulation area can be obtained through this clustering classification model during application.

[0071] Step 130: Input the target vegetation environment label into the vegetation loss model to obtain the vegetation loss results.

[0072] In this embodiment of the invention, before inputting the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result, the vegetation loss model corresponding to each environment label is pre-trained so as to adjust the parameters of the neural network through the environment label.

[0073] Specifically, the sample environmental characteristics and sample loss difference corresponding to each environmental label are first obtained. This sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard land cover loss model.

[0074] Then, the neural network is trained based on the sample environmental features corresponding to each environmental label and the sample loss difference to obtain the vegetation loss model, thereby establishing a mapping relationship between the sample environmental features and the sample loss difference. During training, the sample environmental features corresponding to each environmental label are input into the neural network to obtain the actual loss difference output; the parameters of the neural network are optimized based on the actual loss difference output and the sample loss difference, and the process of inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual loss difference output, and optimizing the parameters of the neural network based on the actual loss difference output and the sample loss difference continues until the error between the actual loss difference output and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations, thus obtaining the vegetation loss model.

[0075] Specifically, please refer to Figure 2 The main objective is to correct the parameters of the neural network model based on the environmental labels of vegetation and landform, and to use this as a standard parameter template for simulation. Taking a fully connected neural network as an example, the training process includes the following steps:

[0076] (1) Initialize the weights W and the threshold θ, that is, assign connection weights W to the input layer units and the hidden layer units. ij The connection weight V from the hidden layer to the output layer jk The threshold θ of the hidden layer j Output layer unit threshold θ k Randomly assign a small value between (0, l).

[0077] (2) Provide learning sample pairs (input and expected output values), and give the input vector X. i =(x1,x2,…x m ) and the corresponding expected output vector x i The values ​​are input to the output layer nodes and calculated sequentially in a forward direction:

[0078]

[0079] (3) Calculate the output value {y} of the output node. k} and expected value Error {δ k}:

[0080]

[0081] (4) Distribute the error in reverse to the hidden layer nodes, that is, use the connection weights {V} jk Output layer

[0082] (5) Using the generalized error {δ} of the output layer unit k The outputs of each unit in the hidden layer {x′} j} Modify the weights of the output layer {V jk} and threshold {θ k}:

[0083] Output layer and hidden layer weight adjustment:

[0084] v jk (t+1)=v jk (t)+ηδ k x′ j (2.22)

[0085] Output layer threshold correction:

[0086] θ k (t+1)=θ k (t)+ηδ k (2.23)

[0087] (6) Generalize the error {δ} using the hidden layer j}, Input {x} of each unit in the input layer i}, adjust the connection weights {W} ij} and threshold {θ j}:

[0088] Input layer and hidden layer connection weight correction:

[0089] w ij (t+1)=w ij (t)+ηδ j x i (2.24)

[0090] Implicit value threshold correction:

[0091] θ j (t+1)=θ j +ηδ j (2.25)

[0092] Repeat step (2), selecting different training samples, and continuously performing the above iterative process until the requirement is met, so that the error δ k When the value becomes sufficiently small or zero, learning stops, and the vegetation loss model is obtained.

[0093] After obtaining the vegetation loss model, inputting the target vegetation environment label into the vegetation loss model will yield the vegetation loss results related to that vegetation feature.

[0094] Step 140: Based on the vegetation loss results and the standard feature loss results, obtain the target feature loss results.

[0095] The standard feature loss result is calculated based on the environmental characteristics and the preset standard feature loss model. That is, feature loss can be predicted using estimation formulas such as those in ITU-R P.1411 combined with environmental characteristics to obtain the standard feature loss result. Since the standard feature loss result does not consider the impact of vegetation on feature loss, the target feature loss result can be obtained based on the vegetation feature loss result and the standard feature loss result.

[0096] This invention, in its embodiments, acquires environmental features within a target environment, determines target vegetation environmental labels based on these features and a clustering classification model, inputs these labels into a vegetation loss model to obtain vegetation loss results, and then obtains target feature loss results based on these vegetation loss results and standard feature loss results. The standard feature loss results are calculated based on the environmental features and the preset standard feature loss model. The environmental labels are obtained by clustering sample environmental feature data with the sample loss difference as the optimization objective. The vegetation loss model is pre-trained on a neural network based on the environmental labels, sample environmental features, and sample loss differences. This approach allows the model to perform loss calculations based on different categories of environmental features, especially vegetation features, effectively improving the accuracy and universality of feature loss calculations.

[0097] Figure 3 A schematic diagram of the structure of the ground feature loss calculation device for radio wave propagation provided in an embodiment of the present invention is shown. Figure 3 As shown, the device 300 includes:

[0098] The first acquisition module 310 is used to acquire environmental features in the target environment, including vegetation features.

[0099] Clustering module 320 is used to determine the target vegetation environment label based on the environmental characteristics and clustering classification model;

[0100] The determination module 330 is used to input the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result; the vegetation loss model is obtained by training the neural network in advance based on the environment label, sample environment characteristics and sample loss difference, and the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model.

[0101] The output module 340 is used to obtain the target feature loss result based on the vegetation feature loss result and the standard feature loss result; wherein the standard feature loss result is calculated based on the environmental characteristics and the preset standard feature loss model.

[0102] In an optional embodiment, the device 300 further includes: a second acquisition module, configured to acquire sample environmental feature data and measured loss data corresponding to the sample environmental feature data; a sample clustering module, configured to cluster the sample environmental feature data with the sample loss difference as the optimization objective to obtain multiple environmental labels; and a training module, configured to train a preset neural network based on the environmental labels and the sample environmental feature data to obtain the clustering classification model.

[0103] In one optional approach, the step of clustering the sample environmental feature data to obtain multiple environmental labels, with the sample loss difference as the optimization objective, further includes: calculating the estimated loss of the samples based on a preset standard ground feature loss model and the sample environmental features; calculating the sample loss difference based on the measured loss data and the estimated loss of the samples; and clustering the sample environmental feature data, grouping the sample environmental feature data whose sample loss difference is less than or equal to a preset threshold into the same environmental label to obtain multiple environmental labels.

[0104] In one optional approach, training a preset neural network based on the environmental labels and the sample environmental feature data to obtain the clustering classification model includes: determining the sample environmental feature dataset corresponding to each environmental label; and training a preset neural network based on each environmental label and the corresponding sample environmental feature dataset to obtain the clustering classification model.

[0105] In one optional approach, before inputting the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result, the method further includes: acquiring the sample environment features and sample loss difference corresponding to each environment label; the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model; training the neural network according to the sample environment features corresponding to each type of environment label and the sample loss difference to obtain the vegetation loss model, so as to establish the mapping relationship between the sample environment features and the sample loss difference.

[0106] In one optional approach, the step of training the neural network based on the sample environmental features corresponding to each environmental label and the sample loss difference to obtain the vegetation loss model, thereby establishing a mapping relationship between the sample environmental features and the sample loss difference, includes: inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference; optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, and continuing to execute the steps of inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference, and optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, until the error between the actual output of the loss difference and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations, thereby obtaining the vegetation loss model.

[0107] This invention, in its embodiments, acquires environmental features within a target environment, determines target vegetation environmental labels based on these features and a clustering classification model, inputs these labels into a vegetation loss model to obtain vegetation loss results, and then obtains target feature loss results based on these vegetation loss results and standard feature loss results. The standard feature loss results are calculated based on the environmental features and the preset standard feature loss model. The environmental labels are obtained by clustering sample environmental feature data with the sample loss difference as the optimization objective. The vegetation loss model is pre-trained on a neural network based on the environmental labels, sample environmental features, and sample loss differences. This approach allows the model to perform loss calculations based on different categories of environmental features, especially vegetation features, effectively improving the accuracy and universality of feature loss calculations.

[0108] Figure 4 The diagram shows a structural schematic of an electronic device provided in an embodiment of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the electronic device.

[0109] like Figure 4 As shown, the electronic device may include: a processor 402, a communications interface 404, a memory 406, and a communications bus 408.

[0110] The processor 402, communication interface 404, and memory 406 communicate with each other via communication bus 408. Communication interface 404 is used to communicate with other network elements, such as clients or other servers. Processor 402 executes program 410, specifically performing the relevant steps in the above-described embodiment of the method for calculating ground feature loss during radio wave propagation.

[0111] Specifically, program 410 may include program code, which includes computer-executable instructions.

[0112] Processor 402 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The electronic device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs.

[0113] Memory 406 is used to store program 410. Memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0114] Specifically, program 410 can be called by processor 402 to cause the electronic device to perform the following operations:

[0115] Acquire environmental features in the target environment, including vegetation features;

[0116] The target vegetation environment label is determined based on the environmental characteristics and cluster classification model.

[0117] The target vegetation environment label is input into the vegetation loss model to obtain the vegetation loss result. The vegetation loss model is obtained by training the neural network in advance based on the environment label, sample environment characteristics and sample loss difference. The sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model.

[0118] Based on the vegetation loss results and the standard feature loss results, the target feature loss results are obtained; wherein, the standard feature loss results are calculated based on the environmental characteristics and the preset standard feature loss model.

[0119] In one optional approach, before determining the target vegetation environment label based on the environmental features and the clustering classification model, the method further includes: acquiring sample environmental feature data and the measured loss data corresponding to the sample environmental feature data; clustering the sample environmental feature data with the sample loss difference as the optimization objective to obtain multiple environment labels; and training a preset neural network based on the environment labels and the sample environmental feature data to obtain the clustering classification model.

[0120] In one optional approach, the step of clustering the sample environmental feature data to obtain multiple environmental labels, with the sample loss difference as the optimization objective, further includes: calculating the estimated loss of the samples based on a preset standard ground feature loss model and the sample environmental features; calculating the sample loss difference based on the measured loss data and the estimated loss of the samples; and clustering the sample environmental feature data, grouping the sample environmental feature data whose sample loss difference is less than or equal to a preset threshold into the same environmental label to obtain multiple environmental labels.

[0121] In one optional approach, training a preset neural network based on the environmental labels and the sample environmental feature data to obtain the clustering classification model includes: determining the sample environmental feature dataset corresponding to each environmental label; and training a preset neural network based on each environmental label and the corresponding sample environmental feature dataset to obtain the clustering classification model.

[0122] In one optional approach, before inputting the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result, the method further includes: acquiring the sample environment features and sample loss difference corresponding to each environment label; the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model; training the neural network according to the sample environment features corresponding to each type of environment label and the sample loss difference to obtain the vegetation loss model, so as to establish the mapping relationship between the sample environment features and the sample loss difference.

[0123] In one optional approach, the step of training the neural network based on the sample environmental features corresponding to each environmental label and the sample loss difference to obtain the vegetation loss model, thereby establishing a mapping relationship between the sample environmental features and the sample loss difference, includes: inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference; optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, and continuing to execute the steps of inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference, and optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, until the error between the actual output of the loss difference and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations, thereby obtaining the vegetation loss model.

[0124] This invention, in its embodiments, acquires environmental features within a target environment, determines target vegetation environmental labels based on these features and a clustering classification model, inputs these labels into a vegetation loss model to obtain vegetation loss results, and then obtains target feature loss results based on these vegetation loss results and standard feature loss results. The standard feature loss results are calculated based on the environmental features and the preset standard feature loss model. The environmental labels are obtained by clustering sample environmental feature data with the sample loss difference as the optimization objective. The vegetation loss model is pre-trained on a neural network based on the environmental labels, sample environmental features, and sample loss differences. This approach allows the model to perform loss calculations based on different categories of environmental features, especially vegetation features, effectively improving the accuracy and universality of feature loss calculations.

[0125] This invention provides a computer-readable storage medium storing at least one executable instruction that, when executed on an electronic device, causes the electronic device to perform the ground feature loss calculation method for radio wave propagation in any of the above method embodiments.

[0126] Executable instructions can be used to cause an electronic device to perform the following operations:

[0127] Acquire environmental features in the target environment, including vegetation features;

[0128] The target vegetation environment label is determined based on the environmental characteristics and cluster classification model.

[0129] The target vegetation environment label is input into the vegetation loss model to obtain the vegetation loss result. The vegetation loss model is obtained by training the neural network in advance based on the environment label, sample environment characteristics and sample loss difference. The sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model.

[0130] Based on the vegetation loss results and the standard feature loss results, the target feature loss results are obtained; wherein, the standard feature loss results are calculated based on the environmental characteristics and the preset standard feature loss model.

[0131] In one optional approach, before determining the target vegetation environment label based on the environmental features and the clustering classification model, the method further includes: acquiring sample environmental feature data and the measured loss data corresponding to the sample environmental feature data; clustering the sample environmental feature data with the sample loss difference as the optimization objective to obtain multiple environment labels; and training a preset neural network based on the environment labels and the sample environmental feature data to obtain the clustering classification model.

[0132] In one optional approach, the step of clustering the sample environmental feature data to obtain multiple environmental labels, with the sample loss difference as the optimization objective, further includes: calculating the estimated loss of the samples based on a preset standard ground feature loss model and the sample environmental features; calculating the sample loss difference based on the measured loss data and the estimated loss of the samples; and clustering the sample environmental feature data, grouping the sample environmental feature data whose sample loss difference is less than or equal to a preset threshold into the same environmental label to obtain multiple environmental labels.

[0133] In one optional approach, training a preset neural network based on the environmental labels and the sample environmental feature data to obtain the clustering classification model includes: determining the sample environmental feature dataset corresponding to each environmental label; and training a preset neural network based on each environmental label and the corresponding sample environmental feature dataset to obtain the clustering classification model.

[0134] In one optional approach, before inputting the target vegetation environment label into the vegetation loss model to obtain the vegetation loss result, the method further includes: acquiring the sample environment features and sample loss difference corresponding to each environment label; the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model; training the neural network according to the sample environment features corresponding to each type of environment label and the sample loss difference to obtain the vegetation loss model, so as to establish the mapping relationship between the sample environment features and the sample loss difference.

[0135] In one optional approach, the step of training the neural network based on the sample environmental features corresponding to each environmental label and the sample loss difference to obtain the vegetation loss model, thereby establishing a mapping relationship between the sample environmental features and the sample loss difference, includes: inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference; optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, and continuing to execute the steps of inputting the sample environmental features corresponding to each environmental label into the neural network to obtain the actual output of the loss difference, and optimizing the parameters of the neural network based on the actual output of the loss difference and the sample loss difference, until the error between the actual output of the loss difference and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations, thereby obtaining the vegetation loss model.

[0136] This invention, in its embodiments, acquires environmental features within a target environment, determines target vegetation environmental labels based on these features and a clustering classification model, inputs these labels into a vegetation loss model to obtain vegetation loss results, and then obtains target feature loss results based on these vegetation loss results and standard feature loss results. The standard feature loss results are calculated based on the environmental features and the preset standard feature loss model. The environmental labels are obtained by clustering sample environmental feature data with the sample loss difference as the optimization objective. The vegetation loss model is pre-trained on a neural network based on the environmental labels, sample environmental features, and sample loss differences. This approach allows the model to perform loss calculations based on different categories of environmental features, especially vegetation features, effectively improving the accuracy and universality of feature loss calculations.

[0137] This invention provides a ground feature loss calculation device for radio wave propagation, used to execute the above-described ground feature loss calculation method for radio wave propagation.

[0138] This invention provides a computer program that can be called by a processor to enable an electronic device to execute the method for calculating ground object loss during radio wave propagation in any of the above-described method embodiments.

[0139] This invention provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed on a computer, cause the computer to perform the method for calculating ground feature loss during radio wave propagation in any of the above-described method embodiments.

[0140] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the content of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0141] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0142] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.

[0143] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0144] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method for calculating ground object loss during radio wave propagation, characterized in that, The method includes: Acquire sample environmental feature data and the corresponding measured loss data; cluster the sample environmental feature data with the sample loss difference as the optimization objective to obtain multiple environmental labels; train a preset neural network based on the environmental labels and the sample environmental feature data to obtain a classification model; The step of clustering sample environmental feature data to obtain multiple environmental labels with sample loss difference as the optimization objective further includes: calculating the estimated loss of samples based on a preset standard land cover loss model and sample environmental features; calculating the sample loss difference based on the measured loss data and the estimated loss of samples; and clustering the sample environmental feature data to group sample environmental feature data with sample loss difference less than or equal to a preset threshold into the same environmental label to obtain multiple environmental labels. The sample environmental features and sample loss difference corresponding to each environmental label are obtained respectively; the sample environmental feature data corresponding to each environmental label are input into the neural network to obtain the actual output of loss difference; the parameters of the neural network are optimized according to the actual output of loss difference and the sample loss difference, and the above steps are repeated until the error between the actual output of loss difference and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations to obtain the vegetation loss model. Acquire environmental features in the target environment, including vegetation features; The target vegetation environment label is determined based on the environmental characteristics and classification model. The environmental features are input into the vegetation loss model corresponding to the target vegetation environment label to obtain the vegetation loss result. The vegetation loss model is obtained by training the neural network in advance based on the environmental label, sample environmental features and sample loss difference. The sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model. Based on the vegetation loss results and the standard feature loss results, the target feature loss results are obtained; wherein, the standard feature loss results are calculated based on the environmental characteristics and the preset standard feature loss model.

2. The method according to claim 1, characterized in that, The step of training a preset neural network to obtain the classification model based on the environmental labels and the sample environmental feature data includes: Determine the sample environment feature dataset corresponding to each of the aforementioned environment labels; The preset neural network is trained based on each of the environmental labels and the corresponding sample environmental feature datasets to obtain the classification model.

3. A device for calculating ground object loss during radio wave propagation, characterized in that, The device includes: The second acquisition module is used to acquire sample environmental feature data and the measured loss data corresponding to the sample environmental feature data. The sample clustering module is used to cluster sample environmental feature data with sample loss difference as the optimization objective to obtain multiple environmental labels. The sample clustering module is further used to: calculate the estimated loss of samples based on a preset standard land cover loss model and sample environmental features; calculate the sample loss difference based on the measured loss data and the estimated loss of samples; and cluster the sample environmental feature data, classifying the sample environmental feature data whose sample loss difference is less than or equal to a preset threshold into the same environmental label to obtain multiple environmental labels. The training module is used to train a preset neural network based on the environmental labels and the sample environmental feature data to obtain a classification model. The first acquisition module is used to acquire environmental features in the target environment, including vegetation features. The clustering module is used to determine the target vegetation environment label based on the environmental characteristics and the clustering classification model; The determination module is used to input the environmental features into the vegetation loss model corresponding to the target vegetation environment label to obtain the vegetation loss result; the vegetation loss model is obtained by training the neural network in advance based on the environmental label, sample environmental features and sample loss difference, and the sample loss difference is the difference between the measured loss data and the estimated loss calculated by the preset standard vegetation loss model. The output module is used to obtain the target feature loss result based on the vegetation loss result and the standard feature loss result; wherein the standard feature loss result is calculated based on the environmental characteristics and the preset standard feature loss model. The device is further configured to: acquire the sample environmental features and sample loss difference corresponding to each environmental label respectively; input the sample environmental feature data corresponding to each type of environmental label into a neural network to obtain the actual output of the loss difference; optimize the parameters of the neural network according to the actual output of the loss difference and the sample loss difference, repeat the above steps until the error between the actual output of the loss difference and the sample loss difference is less than or equal to a preset threshold or the training reaches a preset number of iterations, thereby obtaining a vegetation loss model.

4. An electronic device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation of the method for calculating ground feature loss during radio wave propagation as described in claim 1 or 2.

5. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on the electronic device, causes the electronic device to perform the operation of the ground feature loss calculation method for radio wave propagation as described in claim 1 or 2.

Images