Intelligent assimilation construction method and device for frequency modulation broadcast map based on deep learning
By introducing deep learning technology in the construction of FM broadcast maps, using the LSTM model and the Haiou excellence algorithm, the problem of insufficient broadcast map construction accuracy in complex scenarios is solved, and high-precision broadcast propagation model and map construction is realized, which improves the broadcast quality of radio and television broadcasting.
Patent Information
- Application Number
- CN202411964692.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-30
- Publication Date
- 2025-04-29
AI Technical Summary
The prior art is difficult to provide accurate and reliable broadcast maps in complex scenarios, especially in environments where data is sparse or unevenly distributed, making it difficult to achieve high-precision broadcast propagation model construction.
A deep learning-based method is adopted to build a high-precision broadcast propagation model using a long-term and short-term memory neural network (LSTM), and a hyperparameter automatic optimization search is performed by combining the Haiou excellence algorithm, and then a high-precision frequency modulation broadcast map is constructed through the parameter construction method.
It significantly improves the prediction accuracy of the broadcast propagation model in complex propagation scenarios, can achieve high-precision signal field strength prediction in areas with sparse data distribution, supports refined management of broadcast resources, and improves the quality of broadcasting and television broadcasting.
Smart Images
Figure CN119906510A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an application field based on deep learning, and in particular to a method and device for intelligent assimilation construction of FM broadcast maps based on deep learning. Background Art
[0002] As an important public communication medium, radio plays an indispensable role in the transmission of information and social communication. Among them, FM radio plays an important role in the field of broadcasting and communication due to its strong anti-interference ability, high-quality audio transmission effect and wide coverage. Especially in extreme natural environments, disaster emergency communications and information dissemination in remote areas, FM radio can often provide more reliable guarantees than other communication methods.
[0003] Broadcast maps are a form of electromagnetic spectrum maps. Their function is to provide a comprehensive and quantitative description of the broadcast electromagnetic environment from multiple dimensions such as time, frequency, space and field strength, and to visualize the distribution of electromagnetic energy and the utilization of spectrum resources in combination with geographic information. High-precision broadcast maps can not only reflect the distribution of broadcast resources, but also support the refined management of spectrum resources, providing data support for the coverage of radio and television signals and service optimization.
[0004] The current methods for constructing broadcast maps are mainly divided into two types: spatial interpolation construction method and parameter construction method. The spatial interpolation construction method realizes the continuous modeling of the broadcast electromagnetic environment by interpolating and estimating discrete measurement data, but this method relies heavily on the density of data sampling and is difficult to incorporate prior knowledge, so its accuracy is limited. In contrast, the parameter construction method constructs a map by establishing a prediction model based on propagation characteristics and using the prediction results of the model. The construction accuracy of this method mainly depends on the reliability and accuracy of the propagation prediction model.
[0005] During the FM broadcasting process, the signal will be affected by a variety of environmental factors when it propagates within a range of hundreds of kilometers, such as complex terrain conditions, multipath propagation effects, atmospheric attenuation and the Doppler effect, resulting in highly nonlinear and uncertain signal attenuation during the propagation process. This poses great technical challenges to the high-precision distribution modeling and map construction of FM broadcasting resources. Existing methods are usually unable to provide accurate and reliable broadcast maps in complex scenarios where data is sparse or unevenly distributed. Therefore, how to use advanced deep learning technology to improve the accuracy of broadcast map construction in complex scenarios has become an urgent problem to be solved. Summary of the invention
[0006] The purpose of the present invention is to overcome the deficiencies in the prior art, improve the quality of FM radio and television broadcasting, and provide a method and device for intelligent assimilation construction of FM radio maps based on deep learning. The present invention is oriented to complex scenarios where the measurement data is unevenly distributed and sparse, and a high-precision broadcast propagation model is constructed based on a long short-term memory neural network (LSTM), and then a high-precision FM radio map is constructed based on a parameter construction method. The present invention can realize the construction of broadcast maps in extreme environments, and is a key enabling technology for accurately obtaining the distribution status of broadcast resources, supporting the refined management of broadcast resources, and thus improving the quality of radio and television broadcasting.
[0007] The objective of the present invention is achieved through the following technical solutions:
[0008] A method for intelligent assimilation construction of FM broadcast map based on deep learning, comprising:
[0009] S1, data processing and hyperparameter setting;
[0010] Collect the propagation characteristic data of broadcasts within 120km from the transmitter, and divide the propagation characteristic data into training set, test set and validation set to build a broadcast propagation model based on LSTM;
[0011] Set the initial values of the hyperparameters of the broadcast propagation model for iterative training of the broadcast propagation model;
[0012] The propagation characteristic data include the distance between transceivers, terrain clearance angle, transmitting antenna altitude, receiving antenna altitude, terrain roughness, frequency and receiving field strength;
[0013] S2, iteratively train the broadcast propagation model and optimize the hyperparameters;
[0014] The collected propagation characteristic data and the predicted values of the ITU-R P.1546 and 2001 models are used as the input of the broadcast propagation model and iterative training is carried out. During the training process, the hyperparameters of the broadcast propagation model are adjusted using the Seagull optimization algorithm. The minimum root mean square error between the measured receiving field strength and the output result of the broadcast propagation model is used as the evaluation index. After completing the iterative training and hyperparameter optimization process, a trained broadcast propagation model is obtained.
[0015] S3. Constructing a broadcast map; using a parameter construction method and a trained broadcast propagation model to construct a broadcast map.
[0016] Furthermore, step S1 specifically includes:
[0017] S101, data collection and processing;
[0018] The distance between transceivers, terrain clearance angle, transmitting antenna altitude, receiving antenna altitude, terrain roughness and frequency are used as basic input parameters, and the predicted values of ITU-R P.1546 and 2001 models are used as assimilated input parameters. The basic input parameters and assimilated input parameters are input into the broadcast propagation model for training.
[0019] S102, data division;
[0020] The propagation characteristic data are randomly divided into a training set, a test set, and a validation set in a ratio of 6:2:2, which are used for training the broadcast propagation model, optimizing the hyperparameters of the broadcast propagation model, and verifying the final performance of the broadcast propagation model.
[0021] S103, setting the initial values of the hyperparameters of the broadcast propagation model;
[0022] The initial maximum number of iterations is set to 20, the initial learning rate is set to 0.005, the initial number of hidden layer neurons is set to 10, the initial number of hidden layers is set to two, and the initial learning rate descent algorithm is set to stochastic gradient descent with momentum;
[0023] S104, setting a hyperparameter optimization range for the broadcast propagation model;
[0024] Further, step S2 includes:
[0025] S201, applying the training set to the broadcast propagation model for training;
[0026] S202, input the test set into the broadcast propagation model, and compare and analyze the output results with the measured values to verify the prediction accuracy;
[0027] S203, determining whether the accuracy of the broadcast propagation model meets the expected requirements or whether the number of iterations of the broadcast propagation model meets the preset requirements;
[0028] S204, if the accuracy requirement or the number of iterations requirement is not met, the Seagull optimization algorithm is used to optimize the optimal hyperparameters within a preset range and the broadcast propagation model is retrained using the training set data;
[0029] S205: If the requirements are met, the verification set data is input into the trained broadcast propagation model to verify the final performance and save it.
[0030] Further, step S3 includes:
[0031] S301, determine the resolution of the broadcast map; according to the actual construction scenario, input the resolution required for broadcast map construction. The higher the resolution, the longer the construction time is, and the finer the constructed broadcast map is;
[0032] S302, broadcast propagation prediction: according to the resolution determined in step S301, the broadcast propagation model trained in step S2 is used to perform broadcast propagation prediction and calculate the predicted field strength information;
[0033] S303, constructing a broadcast map; based on the predicted field strength calculated in step S302, the broadcast map is constructed using a parameter construction method.
[0034] The present invention also provides an intelligent assimilation construction device for FM broadcast maps based on deep learning, comprising:
[0035] The data collection unit is used to collect the propagation characteristic data of the broadcast within 120km from the transmitter, and divide the propagation characteristic data into a training set, a test set and a validation set; the propagation characteristic data includes the distance between the transceivers, the terrain clearance angle, the altitude of the transmitting antenna, the altitude of the receiving antenna, the terrain roughness, the frequency and the receiving field strength;
[0036] A model building unit is used to build a broadcast propagation model based on the LSTM algorithm and set the initial values of the hyperparameters of the broadcast propagation model for iterative training of the broadcast propagation model;
[0037] A training and hyperparameter optimization unit, used for iteratively training and hyperparameter optimization of the broadcast propagation model to obtain a trained broadcast propagation model;
[0038] The broadcast map construction unit is used to construct a broadcast map using a parameter construction method and a trained broadcast propagation model.
[0039] The data partitioning unit is used to randomly divide the propagation characteristic data into a training set, a test set, and a validation set in a ratio of 6:2:2:
[0040] A hyperparameter initial value setting unit, used to set the initial values of the hyperparameters of the broadcast propagation model;
[0041] The judgment unit is used to judge whether the accuracy of the broadcast propagation model meets the expected requirements or whether the number of iterations of the broadcast propagation model meets the preset requirements.
[0042] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method for intelligent assimilation construction of an FM radio map based on deep learning are implemented.
[0043] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for intelligent assimilation construction of FM broadcast maps based on deep learning.
[0044] Compared with the traditional broadcast map construction method, the present invention successfully realizes the construction of a high-precision broadcast propagation model by introducing LSTM and taking advantage of its advantages in processing time series data and complex nonlinear relationships. At the same time, the Seagull optimization algorithm is combined to automatically optimize hyperparameters, which significantly improves the prediction accuracy of the model. The following are the beneficial effects brought about by the technical solution of the present invention:
[0045] 1. High-precision prediction capability: By assimilating the prediction results of the LSTM network with the traditional propagation prediction model (such as the ITU-R P.1546 model), the present invention effectively combines the characteristics of prior knowledge and deep learning, and significantly improves the prediction accuracy of the broadcast propagation model in complex propagation scenarios. Even in areas with sparse data distribution, high-precision signal strength prediction can be achieved.
[0046] 2. Improved data utilization efficiency: The present invention makes full use of existing measurement data and output information of traditional prediction models through data assimilation, reduces dependence on additional measurement data, and thus reduces the cost and time of data collection.
[0047] 3. Strong scenario adaptability: Whether in cities, suburbs or rural areas with complex terrain, the model of the present invention can accurately model the broadcast propagation characteristics and provide reliable support for broadcast resource management in different scenarios.
[0048] 4. Support refined management: The high-resolution broadcast map constructed can intuitively display the distribution of broadcast electromagnetic energy, help the refined management and optimization of broadcast resources, and provide technical support for improving the quality of radio and television broadcasts.
[0049] 5. Broad application prospects: The broadcast map constructed by the present invention can provide support for multiple fields such as radio and television signal coverage optimization, emergency communication guarantee, electromagnetic environment monitoring, etc., and has high application value and promotion significance.
[0050] In summary, the present invention effectively solves the problems of poor adaptability to complex scenarios and insufficient accuracy in the prior art, and provides important technical support for the development of broadcast communication technology and efficient use of spectrum resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 It is a schematic flow diagram of the method of the present invention.
[0052] Figure 2 yes Figure 1 Schematic diagram of the specific process of data processing and hyperparameter setting.
[0053] Figure 3 yes Figure 1 Schematic diagram of the specific process of iterative training and hyperparameter optimization.
[0054] Figure 4 yes Figure 1 Schematic diagram of the specific process of building a broadcast map. DETAILED DESCRIPTION
[0055] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0056] See Figure 1 This embodiment provides a method for intelligent assimilation construction of FM broadcast map based on deep learning, including:
[0057] S1, data processing and hyperparameter setting;
[0058] Data collection: Targeted collection of broadcast band propagation characteristic data, including carrier frequency (f), transmit and receive distance (d), receiver altitude (hr), transmitter altitude (ht), relative clearance (c) and receiving field strength (E (dbμV / m)).
[0059] In addition, the modeling input parameters also need to use the predicted values of the ITU-R P.1546 and 2001 models, which are the methods described in the ITU-R P.1546 and 2001 recommendations respectively. The ITU-R P.1546 model is applicable to the prediction of ground point-to-surface wireless point propagation in the frequency range of 30MHz-4000MHz. The model interpolates / extrapolates the empirically derived field strength curve by inputting the above-collected d, hr, ht, f and setting the time percentage to 50% to obtain the predicted field strength. The ITU-RP.2001 model is a universal ground propagation model widely applicable to the frequency range of 30MHz-50GHz. The model calculates the predicted loss by inputting f, the transmit and receive antenna gain, the propagation path terrain height interpolation and the time percentage (50% is selected in this embodiment) (the predicted field strength required in this embodiment is converted based on the predicted loss by the method described in the ITU-R P.525 recommendation).
[0060] Dataset division: After data collection is completed, all data in the dataset are divided into training set, test set and validation set in a ratio of 6:2:2. The training set data is used to train the neural network model; the test set data is used to test the performance of the trained model and optimize the model hyperparameters based on the test results. The validation set data does not participate in the training process and is used to verify the final performance of the model.
[0061] S2, iteratively train the broadcast propagation model and optimize the hyperparameters;
[0062] The collected propagation characteristic data and the predicted values of the ITU-R P.1546 and 2001 models are used as the input of the broadcast propagation model and iterative training is carried out. During the training process, the hyperparameters of the broadcast propagation model are adjusted using the Seagull optimization algorithm. The minimum root mean square error between the measured receiving field strength and the output result of the broadcast propagation model is used as the evaluation index. After completing the iterative training and hyperparameter optimization process, a trained broadcast propagation model is obtained.
[0063] 3. Construct broadcast maps; use parameter construction method and trained broadcast propagation model to construct broadcast maps.
[0064] See Figure 2 , step 1 specifically includes:
[0065] S101, data collection and processing;
[0066] The distance between the transceivers (d), the terrain clearance angle (c), the height of the transmitting antenna above sea level (h t ), receiving antenna altitude (h r ), terrain roughness (Δh) and frequency (f) are used as basic input parameters. In this embodiment, the predicted values of ITU-R P.1546 and 2001 models are used as assimilation input parameters, and the basic input parameters and assimilation input parameters are input into the broadcast propagation model for training;
[0067] S102, data division;
[0068] The collected propagation characteristic data are randomly divided into training set, test set and validation set in the ratio of 6:2:2, which are used for training the broadcast propagation model, optimizing the hyperparameters of the broadcast propagation model and verifying the final performance of the broadcast propagation model respectively;
[0069] S103, setting the initial values of the hyperparameters of the broadcast propagation model;
[0070] The initial maximum number of iterations is set to 20, the initial learning rate is set to 0.005, the initial number of hidden layer neurons is set to 10, the initial number of hidden layers is set to two, and the initial learning rate descent algorithm is set to stochastic gradient descent with momentum;
[0071] S104, set the hyperparameter optimization range of the broadcast propagation model (as shown in Table 1); in order to achieve the best training effect of the broadcast propagation model, this embodiment uses the Seagull optimization algorithm to automatically find the optimal value of the hyperparameter of the broadcast propagation model. The initial value setting is shown in Table 2;
[0072] The hyperparameter optimization range setting of the broadcast propagation model is shown in Table 2. Based on the reference optimization range shown in Table 2, the Seagull optimization algorithm is used to automatically find the optimal values of the hyperparameters of the broadcast propagation model.
[0073] Table 2 Hyperparameter optimization range of broadcast propagation model
[0074]
[0075] Further, step S2 includes:
[0076] S201, applying the training set to the broadcast propagation model for training;
[0077] S202, input the test set into the broadcast propagation model, and compare and analyze the output results with the measured values to verify the prediction accuracy;
[0078] S203, determining whether the accuracy of the broadcast propagation model meets the expected requirements or whether the number of iterations of the broadcast propagation model meets the preset requirements;
[0079] S204, if the accuracy requirement or the number of iterations requirement is not met, the Seagull optimization algorithm is used to optimize the optimal hyperparameters within a preset range and the broadcast propagation model is retrained using the training set data;
[0080] S205: If the requirements are met, the verification set data is input into the trained broadcast propagation model to verify the final performance and save it.
[0081] Further, step S3 includes:
[0082] S301. Determine the broadcast map resolution. According to the actual construction scenario, input the resolution required for broadcast map construction. The broadcast map resolution is the main factor affecting the broadcast map construction time. The higher the selected resolution, the longer the construction time is, and the more detailed the constructed broadcast map is.
[0083] S302, broadcast propagation prediction: according to the resolution determined in step S301, the broadcast propagation model trained in step S2 is used to perform broadcast propagation prediction and calculate the predicted field strength information;
[0084] S303, constructing a broadcast map; based on the predicted field strength calculated in step S302, the broadcast map is constructed using a parameter construction method.
[0085] Alternatively, the present invention further provides an intelligent assimilation construction device for FM broadcast maps based on deep learning, comprising:
[0086] The data collection unit is used to collect the propagation characteristic data of the broadcast within 120km from the transmitter, and divide the propagation characteristic data into a training set, a test set and a validation set; the propagation characteristic data includes the distance between the transceivers, the terrain clearance angle, the altitude of the transmitting antenna, the altitude of the receiving antenna, the terrain roughness, the frequency and the receiving field strength;
[0087] A model building unit is used to build a broadcast propagation model based on the LSTM algorithm and set the initial values of the hyperparameters of the broadcast propagation model for iterative training of the broadcast propagation model;
[0088] A training and hyperparameter optimization unit, used for iteratively training and hyperparameter optimization of the broadcast propagation model to obtain a trained broadcast propagation model;
[0089] The broadcast map construction unit is used to construct a broadcast map using a parameter construction method and a trained broadcast propagation model.
[0090] The data partitioning unit is used to randomly divide the propagation characteristic data into a training set, a test set, and a validation set in a ratio of 6:2:2:
[0091] A hyperparameter initial value setting unit, used to set the initial values of the hyperparameters of the broadcast propagation model;
[0092] The judgment unit is used to judge whether the accuracy of the broadcast propagation model meets the expected requirements or whether the number of iterations of the broadcast propagation model meets the preset requirements.
[0093] Preferably, the embodiment of the present application also provides a specific implementation of an electronic device capable of implementing all steps in the method for intelligent assimilation construction of FM broadcast maps based on deep learning in the above embodiment, and the electronic device specifically includes the following contents:
[0094] Processor, memory, communications interface and bus;
[0095] Among them, the processor, memory, and communication interface communicate with each other through the bus; the communication interface is used to realize information transmission between related devices such as server-side devices, metering devices, and user-side devices.
[0096] The processor is used to call the computer program in the memory, and when the processor executes the computer program, it implements all the steps in the intelligent assimilation construction method of the FM radio map based on deep learning in the above embodiment.
[0097] An embodiment of the present application also provides a computer-readable storage medium capable of implementing all the steps of the method for intelligent assimilation and construction of an FM radio map based on deep learning in the above-mentioned embodiment. A computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, all the steps of the method for intelligent assimilation and construction of an FM radio map based on deep learning in the above-mentioned embodiment are implemented.
[0098] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the hardware + program embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0099] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0100] Although the present application provides method operation steps such as embodiments or flow charts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment).
[0101] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0102] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0104] The present invention is not limited to the embodiments described above. The above description of the specific embodiments is intended to describe and illustrate the technical solution of the present invention. The above specific embodiments are merely illustrative and not restrictive. Without departing from the scope of the present invention and the scope of protection of the claims, a person of ordinary skill in the art can also make many forms of specific changes under the guidance of the present invention, which all fall within the scope of protection of the present invention.
Claims
1. A method for intelligent assimilation construction of FM broadcast map based on deep learning, characterized in that: include: S1, data processing and hyperparameter setting; Collect the propagation characteristic data of broadcasts within 120km from the transmitter, and divide the propagation characteristic data into training set, test set and validation set to build a broadcast propagation model based on long short-term memory neural network (LSTM); Set the initial values of the hyperparameters of the broadcast propagation model for iterative training of the broadcast propagation model; The propagation characteristic data includes the distance between transceivers, terrain clearance angle, transmitting antenna altitude, receiving antenna altitude, terrain roughness, frequency and receiving field strength; S2, iteratively train the broadcast propagation model and optimize the hyperparameters; The collected propagation characteristic data and the predicted values of the ITU-R P.1546 and 2001 models are used as the input of the broadcast propagation model and iterative training is carried out. During the training process, the hyperparameters of the broadcast propagation model are adjusted using the Seagull optimization algorithm. The minimum root mean square error between the measured receiving field strength and the output result of the broadcast propagation model is used as the evaluation index. After completing the iterative training and hyperparameter optimization process, a trained broadcast propagation model is obtained. S3, building a broadcast map; The broadcast map is constructed using the parameter construction method and the trained broadcast propagation model.
2. According to claim 1, a method for intelligent assimilation construction of FM broadcast map based on deep learning, characterized in that: Step 1 specifically includes: S101, data collection and processing; The distance between transceivers, terrain clearance angle, transmitting antenna altitude, receiving antenna altitude, terrain roughness and frequency are used as basic input parameters, and the predicted values of ITU-R P.1546 and 2001 models are used as assimilated input parameters. The basic input parameters and assimilated input parameters are input into the broadcast propagation model for training. S102, data division; The propagation characteristic data are randomly divided into a training set, a test set, and a validation set in a ratio of 6:2:2, which are used for training the broadcast propagation model, optimizing the hyperparameters of the broadcast propagation model, and verifying the final performance of the broadcast propagation model. S103, setting the initial values of the hyperparameters of the broadcast propagation model; The initial maximum number of iterations is set to 20, the initial learning rate is set to 0.005, the initial number of hidden layer neurons is set to 10, the initial number of hidden layers is set to two, and the initial learning rate descent algorithm is set to stochastic gradient descent with momentum; S104: Setting a hyperparameter optimization range for the broadcast propagation model.
3. According to claim 1, a method for intelligent assimilation construction of FM broadcast map based on deep learning, characterized in that: Step S2 includes: S201, applying the training set to the broadcast propagation model for training; S202, input the test set into the broadcast propagation model, and compare and analyze the output results with the measured values to verify the prediction accuracy; S203, determining whether the accuracy of the broadcast propagation model meets the expected requirements or whether the number of iterations of the broadcast propagation model meets the preset requirements; S204, if the accuracy requirement or the number of iterations requirement is not met, the Seagull optimization algorithm is used to optimize the optimal hyperparameters within a preset range and the broadcast propagation model is retrained using the training set data; S205. If the requirements are met, the verification set data is input into the trained broadcast propagation model to verify the final performance and save it.
4. According to claim 1, a method for intelligent assimilation construction of FM broadcast map based on deep learning, characterized in that: Step S3 includes: S301, determining the resolution of the broadcast map; according to the actual construction scenario, inputting the resolution required for broadcast map construction, the higher the resolution, the longer the construction time, and the finer the constructed broadcast map; S302, broadcast propagation prediction: according to the resolution determined in step S301, the broadcast propagation model trained in step S2 is used to perform broadcast propagation prediction and calculate the predicted field strength information; S303, constructing a broadcast map; based on the predicted field strength calculated in step S302, the broadcast map is constructed using a parameter construction method.
5. An intelligent assimilation construction device for FM broadcasting maps based on deep learning, characterized in that: include: The data collection unit is used to collect the propagation characteristic data of the broadcast within 120km from the transmitter, and divide the propagation characteristic data into a training set, a test set and a validation set; the propagation characteristic data includes the distance between the transceivers, the terrain clearance angle, the altitude of the transmitting antenna, the altitude of the receiving antenna, the terrain roughness, the frequency and the receiving field strength; A model building unit is used to build a broadcast propagation model based on the LSTM algorithm and set the initial values of the hyperparameters of the broadcast propagation model for iterative training of the broadcast propagation model; A training and hyperparameter optimization unit, used for iteratively training and hyperparameter optimization of the broadcast propagation model to obtain a trained broadcast propagation model; The broadcast map construction unit is used to construct a broadcast map using a parameter construction method and a trained broadcast propagation model.
6. The intelligent assimilation construction device of FM broadcast map based on deep learning according to claim 5, characterized in that: The data partitioning unit is used to randomly divide the propagation characteristic data into a training set, a test set, and a validation set in a ratio of 6:2:2: A hyperparameter initial value setting unit, used to set the initial values of the hyperparameters of the broadcast propagation model; The judgment unit is used to judge whether the accuracy of the broadcast propagation model meets the expected requirements or whether the number of iterations of the broadcast propagation model meets the preset requirements.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the intelligent assimilation construction method of the FM broadcast map based on deep learning as described in any one of claims 1 to 4 are implemented.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for intelligent assimilation and construction of FM broadcast maps based on deep learning as described in any one of claims 1 to 4 are implemented.