CN114858677B - Radio large fog weather monitoring method, system, computer equipment and medium - Google Patents

Abstract

The invention belongs to the technical field of electronic information, and discloses a method, a system, computer equipment and a medium for monitoring the atmosphere of a radio large fog. The radio heavy fog weather monitoring method uses a radio and deep learning method, adopts a deep artificial neural network technology with supervision-learning capability to extract characteristics of low-frequency radio signals passing through environments with different concentration fog areas, overcomes the defects of weak generalization capability and the like of the traditional technology, innovatively designs a deep learning special model for low-frequency fog spectrum characteristic identification, and can identify spectrum characteristics of different concentration fog at a low frequency band so as to monitor heavy fog weather. The invention can directly monitor the heavy fog weather by using common broadcast signals or base station signals and the like without installing additional equipment. The method has the advantages of high time resolution, low cost, no maintenance and the like. And continuous real-time monitoring can be realized, and the defect of insufficient spatial resolution of the traditional monitoring method can be overcome.

Description

A radio foggy weather monitoring method, system, computer equipment and medium

Technical Field

The present invention belongs to the field of electronic information technology, and in particular relates to a radio foggy weather monitoring method, system, computer equipment and medium.

Background Art

At present, the detection of fog in meteorological services is mainly based on visual observation, and equipment such as droplet spectrometers, visibility meters, and microwave radiometers are only used in places such as airports.

With the development of radar technology in the 1960s, the interaction between radio propagation and the environment began to be used as a method to monitor changes in the atmospheric environment. For example, the weather radar using Doppler frequency shift technology can monitor atmospheric environments such as rain, fog, clouds, snow, and ice. It uses radar to emit pulsed electromagnetic waves that propagate in an approximately straight line to the sky. When this electromagnetic wave encounters meteorological targets such as clouds and rain, the meteorological targets will disperse the pulsed electromagnetic waves emitted by the radar, and these dispersed pulsed electromagnetic waves will be reflected back to the weather radar. Researchers achieve the purpose of monitoring the atmosphere by analyzing the echoes. However, Doppler weather radars also have problems such as large investment in construction, high cost after operation, and echo information relying on the level of researchers.

In recent years, radio technology has been widely used in various fields. The advent of the 5G era has made radio technology promote the vigorous development of various emerging intelligent industries. Researchers have found that radio waves in higher frequency bands will be affected by weather such as rain and fog and attenuate. When electromagnetic waves pass through different propagation media, they will be affected by particles in the propagation medium, resulting in scattering, absorption, depolarization and other effects. This leads to the attenuation of electromagnetic wave energy. Moreover, when electromagnetic waves pass through foggy areas, the impact of fog on electromagnetic waves increases with the increase of water content in the foggy areas. Although some studies have proved that radio signals below 10GHZ are basically not affected by rain and fog attenuation.

However, through the analysis of its wireless spectrum, we found that although the attenuation of signals below 10GHZ by rain and fog did not cause a significant decrease in signal amplitude, the influence of the channel medium on the signal always exists and is not eliminated with the change of signal frequency.

Through the above analysis, the problems and defects of the prior art are as follows:

(1) The existing method of using droplet spectrometer to monitor foggy weather has the limitation that it can only measure at a fixed point.

(2) The method of using visibility meters to monitor foggy weather has high installation costs and difficult maintenance.

(3) The existing microwave radiometer method cannot measure fog over a large area and can only measure fog in the area vertically above the instrument.

(4) Existing methods all monitor the weather conditions such as rain and fog by monitoring the decrease in signal amplitude when higher frequency radio signals pass through rain and fog areas. However, although the attenuation of low frequency radio signals by rain and fog does not cause a significant decrease in signal amplitude, the influence of some channel media on the signal always exists and is not eliminated with the change of signal frequency.

Summary of the invention

In view of the problems existing in the prior art, the present invention provides a radio fog weather monitoring method, system, computer equipment and medium.

The present invention is implemented in this way: a radio fog weather monitoring method based on deep learning comprises the following steps:

The collected low-frequency radio signals in foggy areas with different concentrations are converted into spectrum waterfall diagrams. Based on the spectrum waterfall diagrams, data processed by two data preprocessing methods, the spectrum waterfall diagram and the original IQ signal, are used as input data and input into the low-frequency radio signal fog feature recognition model to identify the characteristics of fog in the radio signal.

Furthermore, after collecting the low-frequency radio signals in foggy environments with different concentrations, it is necessary to use radio and deep learning methods, and adopt a deep artificial neural network with supervised learning capabilities to extract the characteristics of low-frequency radio signals passing through foggy environments with different concentrations.

Further, in the spectrum waterfall diagram, the low-frequency radio signal is converted into the spectrum waterfall diagram, the IQ signal obtained by orthogonal digital down conversion and the IQ signal are converted into the frequency domain by FFT to obtain the spectrum waterfall diagram.

Furthermore, the low-frequency radio signal fog feature recognition model is composed of a number of convolutional neural networks and attention networks; the convolutional neural network is used to extract features of input data;

The attention network uses the features extracted by the convolutional neural network and obtains the hot spots in the input data by fusing the attention mechanism, so that the next level network can make judgments on a finer scale.

Furthermore, the radio fog weather monitoring method based on deep learning specifically includes the following steps:

Step 1: using a dedicated radio monitoring receiver to collect low-frequency radio signal data under foggy weather with different concentrations;

Step 2: Use orthogonal digital down-conversion technology to convert the received signal into an IQ signal, then convert it into the frequency domain through FFT and draw a spectrum waterfall diagram; evaluate the performance of various data preprocessing schemes and select the appropriate data preprocessing method;

Step 3: Extract the features of input data through convolutional neural network;

Step 4: Use the features extracted by the convolutional neural network to obtain the hot spots in the input data through the fusion attention mechanism, so that the next level of network can make judgments at a finer scale.

Furthermore, in step one, the low-frequency band radio signal is a received low-frequency band baseband signal of a selected frequency band, and during data collection, the position of the receiver in the fog area is continuously changed to eliminate the influence of different channels on the fog characteristics of the radio signal.

Further, in step 2, the received signal is converted into an IQ signal using orthogonal digital down-conversion technology, and the conversion formula used is as follows:

I＝h _LP (s(t)cos(2πft))

Q＝h _LP (s(t)sin(2πft))

Where s(t) is the received signal, f represents the carrier frequency of the transmitted signal, and h _LP represents the system function of the low-pass filter;

In step 4, the attention network is instructed to track multiple regions of interest and identify the characteristics of radio signals under different concentrations of fog.

Another object of the present invention is to provide a radio fog weather monitoring system based on deep learning, comprising:

A signal receiving module is used to receive low-frequency radio signals passing through foggy environments of different densities by using a radio monitoring receiver and setting a center frequency, bandwidth, etc.;

The data preprocessing module is used to convert the low-frequency band radio signal received by the receiver into a spectrum waterfall diagram for identification by the low-frequency band signal fog feature identification module.

The low-frequency signal fog feature recognition module is used to perform feature recognition on radio signals collected at low frequencies and passing through foggy environments of different concentrations through a low-frequency radio signal fog feature recognition model.

Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the radio fog weather monitoring method based on deep learning.

Another object of the present invention is to provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor executes the radio fog weather monitoring method based on deep learning.

In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

First, in view of the technical problems existing in the above-mentioned prior art and the difficulty of solving the problems, the technical solutions to be protected by the present invention and the results and data during the research and development process are closely combined to analyze in detail and deeply how the technical solutions of the present invention solve the technical problems, and some creative technical effects brought about after solving the problems. The specific description is as follows:

1. The present invention uses commercially available wireless base stations to acquire data on the characteristics of the propagation medium in wireless links. These data have high spatial and temporal resolution and operate in real time. Compared with existing monitoring equipment such as droplet spectrometers, visibility meters, and microwave radiometers, the present invention does not require the installation of additional equipment, nor does it require additional maintenance and operating costs, and can detect fog in a larger range.

2. The traditional method of using the attenuation model of high-frequency radio signals in rain and fog to invert foggy weather requires receiving high-frequency radio signals, which has high requirements for receiving equipment, etc. However, the signal receiving module of the present invention receives the lower frequency radio signals of the commercial wireless base station, which is simple to receive and low in cost.

3. Aiming at the problem of low recognition efficiency of traditional methods, the present invention uses a method combining radio and deep learning, and introduces a deep neural network to extract the features of low-frequency radio signals passing through foggy areas of different concentrations. The collected radio signals are converted into a spectrum waterfall diagram, and the original IQ signal and the spectrum waterfall diagram are used as input data. First, the convolutional neural network is used to extract the features of the two input data respectively, and then input into the attention network. The hot spots in the input data are obtained by integrating the attention mechanism, and a more refined scale judgment is performed on them. Then the classifier is input to classify foggy weather of different concentrations. Finally, the recognition accuracy of the four different concentrations of foggy weather, namely 21% without water and fog, 21% with water and fog, 85% with water and fog, and with water and fog in saturated state, reached 86.18%. And compared with the traditional recognition model, the radio signal fog feature recognition model proposed in the present invention has higher recognition accuracy and lower time complexity. Continuous and real-time monitoring of foggy weather can be achieved in a low-cost manner, making up for the defect of insufficient spatial resolution of traditional monitoring methods. In extreme weather conditions, it can serve as a good supplement to traditional meteorological observation methods, and provide decision-making support for relevant departments to carry out fog weather monitoring and early warning, and disaster prevention and mitigation work.

Second, considering the technical solution as a whole or from the perspective of the product, the technical effects and advantages of the technical solution to be protected by the present invention are described in detail as follows:

The present invention breaks through the traditional technical idea of using rain and fog attenuation model to predict rain and fog attenuation and then monitor rain and fog weather. It uses radio + deep learning method and deep artificial neural network technology with supervised learning ability to extract the characteristics of low-frequency radio signals passing through fog areas with different concentrations, overcomes the shortcomings of weak generalization ability of traditional technology, and innovatively designs a deep learning special model for low-frequency fog spectrum feature recognition, which can identify the spectrum characteristics of fog with different concentrations in the low frequency band, and then monitor foggy weather.

Compared with the traditional method of using Doppler weather radar, microwave radiometer, visibility meter, droplet spectrometer and other professional equipment to monitor foggy weather, the radio foggy weather monitoring method and system designed by the present invention does not need to install additional equipment, and can directly use commonly used broadcast signals or base station signals to monitor foggy weather. It has the advantages of high time resolution, low cost, and no maintenance. And it can realize continuous real-time monitoring, which can make up for the defect of insufficient spatial resolution of traditional monitoring methods. In extreme weather conditions, it can serve as a good supplement to traditional meteorological observation methods, and provide decision-making support for relevant departments to carry out foggy weather monitoring and early warning work, disaster prevention and mitigation work. Moreover, the recognition model of the characteristics of radio signals in foggy areas with different concentrations designed by the present invention can significantly improve the accuracy of fog spectrum feature recognition of low-frequency radio signals compared with traditional neural network models.

Third, as auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

The expected benefits and commercial value of the technical solution of the present invention after transformation are:

After the technical solution of the present invention is converted, it can fill the gap in the coverage of traditional monitoring methods and make up for the problem of insufficient spatial resolution of traditional monitoring methods. Since the cost of the technical solution of the present invention is low after conversion, a device converted from the technical solution of the present invention can be installed at intervals in places where traditional monitoring equipment is not installed on the highway. The device converted from the technical solution of the present invention can also be installed in urban areas, rural areas or complex terrains such as mountainous areas, valleys and forests where it is difficult to deploy traditional monitoring equipment. This is of great significance for relevant departments to provide decision-making support for foggy weather monitoring and early warning work and disaster prevention and mitigation work. Therefore, it has high commercial value.

The technical solution of the present invention fills the technical gap in the industry at home and abroad:

At present, the existing methods of using radio signals to monitor foggy weather at home and abroad are to use the attenuation of high-frequency signals in rain and fog, invert weather phenomena through rain and fog attenuation models, and then monitor weather phenomena. The technical solution of the present invention is to use a deep neural network to extract the characteristics of different propagation media left in the received lower-frequency radio signals to monitor foggy weather. The use of radio + deep learning methods overcomes the shortcomings of traditional technologies such as weak generalization ability, and innovatively designs a deep learning dedicated model for low-frequency fog feature recognition, which can identify the spectral characteristics of fog of different concentrations in the low-frequency band. And then monitor foggy weather.

Whether the technical solution of the present invention overcomes technical prejudice:

Traditional methods believe that fog will only attenuate radio signals in higher frequency bands, causing a decrease in signal amplitude. It has almost no effect on radio signals in lower frequency bands. However, the technical solution of the present invention collects low-frequency radio signals in different environments, uses deep neural networks to extract the characteristics of different environments, and finds that although the attenuation of lower frequency signals by rain and fog does not cause a significant decrease in signal amplitude, the influence of the channel medium on the signal always exists and is not eliminated with changes in signal frequency. Therefore, the changes in the propagation medium on the path can be measured by analyzing the characteristics of the signal received in the wireless link, so as to achieve the purpose of monitoring different weather conditions.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a radio fog weather monitoring method based on deep learning provided by an embodiment of the present invention;

FIG2 is a spectrum waterfall diagram obtained after conversion of a low-frequency band radio signal received by a dedicated radio monitoring receiver (3900A) in a data preprocessing stage provided by an embodiment of the present invention;

Among them, Figure 2 (a) collects radio signals passing through fog areas of different densities and converts them into spectrum waterfall diagrams, spectrum waterfall diagram 1 of signals collected under fog areas of different densities; 2 (b) collects radio signals passing through fog areas of different densities and converts them into spectrum waterfall diagrams, spectrum waterfall diagram 2 of signals collected under fog areas of different densities;

3 is a structural diagram of a dedicated model for feature recognition of low-frequency radio signals in fog environments of different concentrations provided by an embodiment of the present invention;

FIG4 is a diagram of a data collection process provided by an embodiment of the present invention;

FIG5 is a schematic diagram of a radio fog weather monitoring system based on deep learning provided by an embodiment of the present invention;

In the figure: 1. Signal receiving module; 2. Data preprocessing module; 3. Low-frequency band signal fog feature recognition module.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the 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.

1. Explanatory Examples In order to enable those skilled in the art to fully understand how to implement the present invention, this section provides an illustrative example that expands and describes the technical solution of the claims.

The radio fog weather monitoring method based on deep learning provided by the embodiment of the present invention first collects low-frequency radio signals in fog areas of different concentrations, and identifies the characteristics of fog in the radio signals through the low-frequency radio signal fog feature recognition model; in terms of low-frequency radio signal fog feature recognition, the collected radio signals are first converted into a spectrum waterfall diagram, and a variety of data preprocessing schemes are used as input data, and then input into the low-frequency radio signal fog feature recognition model for fog feature recognition. The present invention breaks through the traditional technical idea of using rain and fog attenuation models to predict rain and fog attenuation and then monitor rain and fog weather. It uses the radio + deep learning method and adopts deep artificial neural network technology with supervised-learning capabilities to overcome the shortcomings of weak generalization ability of traditional technologies, and innovatively designs a deep learning dedicated model for low-frequency fog spectrum feature recognition, which can identify the spectrum characteristics of fog of different concentrations in the low frequency band. And then monitor foggy weather.

Example 1

As shown in FIG1 , the radio fog weather monitoring method based on deep learning provided by an embodiment of the present invention includes the following steps:

S101, using a dedicated radio monitoring receiver (3900A) to collect low-frequency radio signal data under different concentrations of fog.

S102, using orthogonal digital down-conversion technology to convert the received signal into an IQ signal, and then convert it into the frequency domain through FFT and draw a spectrum waterfall diagram. Evaluate the performance of various data preprocessing schemes and select the appropriate data preprocessing scheme. And through S103 and S104, identify the characteristics of low-frequency radio signals in fog areas of different concentrations.

S103, extracting features of input data through a convolutional neural network.

S104, using the features extracted by the convolutional neural network, the hot spots in the input data are obtained by integrating the attention mechanism, so that the next level of network can make judgments at a finer scale. At the same time, the attention network is made to track multiple areas of interest to ensure that the characteristics of radio signals in foggy weather with different concentrations can be accurately identified.

In an embodiment of the present invention, a spectrum waterfall diagram is also called a spectrum array diagram, which is a three-dimensional spectrum diagram that describes the power spectrum or amplitude spectrum of a radio signal changing with time, and shows how the energy of each frequency in the radio signal changes with time. Its horizontal axis represents frequency, and its vertical axis represents time. The size of each point in the spectrum waterfall diagram represents the energy of the frequency point at that moment. As shown in FIG2 , an embodiment of the present invention uses a spectrum waterfall diagram obtained by converting a low-frequency radio signal received by a dedicated radio monitoring receiver (3900A) in a data preprocessing stage. Among them, radio signals passing through foggy environments of different concentrations are collected and converted into spectrum waterfall diagrams. It can be seen that the spectrum waterfall diagrams of signals collected under foggy environments of different concentrations are different, and there are other spectrum lines on both sides of the two waterfall diagrams, and the power, position, and number of spectrum lines are different, as shown in FIG2 (a) and FIG2 (b).

When radio signals pass through foggy weather environments of different densities, fog features will remain in the signals, but the fog features may be obliterated in the transmitted signals. This is one of the reasons why traditional methods believe that fog has no effect on the propagation of low-frequency signals. At the same time, when receiving signals, they need to undergo multiple filtering processes and extractions, which may mix some noise into the received signals, making it more difficult to extract the features of fog. Moreover, after converting radio signals collected under foggy weather of different densities into spectrum waterfall diagrams, their core contours are basically the same, with only slight differences. Therefore, traditional methods cannot be used for identification.

In recent years, with the development of artificial intelligence, convolutional neural networks have been widely used in deep learning. They can continuously extract features through convolution kernels. They are often used in the field of image recognition. Therefore, deep neural networks can be used to extract fog features from radio signals collected under different concentrations of foggy weather, and then monitor foggy weather of different concentrations. Therefore, the intelligent model independently developed and verified by the research team in recent years is used to automatically learn the characteristics of low-frequency radio signal fog and effectively identify the characteristics of fog.

Since radio signal data is quite different from image data, a “special model” must be developed for fog feature recognition of low-frequency radio signals, as shown in Figure 3.

As shown in FIG3 , the present invention first converts the received radio signal into an IQ signal using orthogonal digital down-conversion technology, and then converts it into the frequency domain through FFT and draws a spectrum waterfall diagram. In the data preprocessing stage, the data is normalized and the data obtained by the two different data preprocessing methods are respectively input into the discriminant model based on deep learning. In the discriminant model based on deep learning, the features of the input data are first extracted by the convolutional neural network. Then it is input into the attention network, and the attention network uses the features extracted by the convolutional neural network to obtain the hot spot area in the input data by fusing the attention mechanism. At the same time, the attention network is made to track multiple areas of interest to obtain the data of the area of interest for the next level network to judge on a finer scale. Then the extracted refined features are input into the classifier for classification and identification, and then the different environmental conditions of 21% no water and no fog, 21% water and no fog, 85% water and fog, and water and fog under saturation are monitored.

The model is composed of several convolutional neural networks and attention networks. The convolutional neural network is used to extract the features of the input data. The attention network uses the features extracted by the convolutional neural network to obtain the hot spots in the input data by fusing the attention mechanism, so that the next level of network can make judgments on a finer scale. At the same time, the attention network tracks multiple areas of interest to ensure that the characteristics of radio signals in foggy weather with different concentrations can be accurately identified. The input data of the model are the IQ signals obtained by converting the low-frequency radio signals collected in foggy weather with different concentrations through orthogonal digital down-conversion technology and the spectrum waterfall diagram obtained by converting the IQ signals to the frequency domain through FFT.

In capturing low-frequency radio signals, a dedicated radio monitoring receiver (3900A) is used to receive low-frequency baseband signals passing through foggy environments of different concentrations, wherein the data collection process is shown in FIG4 .

As shown in FIG4 , a diagram of the data acquisition process of the present invention is shown. In the present invention, after the radio signal passes through a channel filled with fog of different concentrations, a 3900A radio monitoring receiver is used at the RF receiving end to receive the signal, and then the received IQ data is stored in a storage device, and the IQ data is converted to the frequency domain through FFT to obtain spectrum data.

The radio monitoring receiving equipment is designed to reduce the impact of different hardware on the signal spectrum. A high-performance dedicated receiver is used. During the data collection process, the position of the receiver in the fog area is constantly changed to eliminate the impact of the same channel on the fog characteristics of the radio signal. Finally, the data is received.

The collected radio signals are converted into IQ signals through orthogonal digital down-conversion technology, and then the IQ signals are converted into spectrum waterfall diagrams. The fog features are identified using the low-frequency signal fog feature recognition module.

Example 2

Based on the deep learning-based radio fog weather monitoring method provided in Example 1, as a preferred embodiment, in step S101, the low-frequency band radio signal is a low-frequency band baseband signal of a received selected frequency band. During the data collection process, the position of the receiver in the fog area is continuously changed to eliminate the influence of the same channel on the fog characteristics of the radio signal.

Example 3

Based on the radio fog weather monitoring method based on deep learning provided in Example 1, as a preferred embodiment, in step S102, the received signal is converted into an IQ signal using orthogonal digital down-conversion technology, and the conversion formula used is as follows:

I＝h _LP (s(t)cos(2πft))

Q＝h _LP (s(t)sin(2πft))

Where s(t) is the received signal, f represents the carrier frequency of the transmitted signal, and h _LP represents the system function of the low-pass filter.

Example 4

Based on the radio fog weather monitoring method based on deep learning provided in Example 1, as shown in FIG5 , the embodiment of the present invention also provides a radio fog weather monitoring system based on deep learning, including:

The signal receiving module 1 is used to receive low-frequency radio signals passing through foggy environments of different densities by using a 3900A radio monitoring receiver by setting the center frequency, bandwidth, etc.

The data preprocessing module 2 is used to convert the IQ signal received by the receiver into a spectrum waterfall diagram so as to be identified by the low-frequency signal fog feature recognition module.

The low-frequency signal fog feature recognition module 3 is used to perform feature recognition on radio signals collected at low frequencies and passing through foggy environments of different concentrations through a feature recognition model.

Example 5

Based on the radio fog weather monitoring system based on deep learning provided in Example 4, the feature recognition model in the low-frequency signal fog feature recognition module 3 includes:

Several convolutional neural networks are used to extract the characteristics of the propagation medium left in the low-frequency radio signals passing through fog areas of different concentrations and their spectrum waterfall diagrams.

The attention network is used to extract features from neural networks at all levels, and obtain hot spots in the input data through the fusion attention mechanism, so that the next level of network can make judgments at a finer scale. At the same time, the attention network tracks multiple areas of interest to ensure that the characteristics of radio signals in foggy weather with different concentrations can be accurately identified.

2. Application Examples: In order to prove the creativity and technical value of the technical solution of the present invention, this section provides application examples of the claimed technical solution on specific products or related technologies.

In a specific application scenario, in the signal receiving module, a 3900A radio monitoring receiver is used to receive low-frequency radio signals, such as broadcast signals, in four different fog environments: 21% without water and fog, 21% with water and fog, 85% with water and fog, and with water and fog in a saturated state. In addition, during the acquisition process, the position of the receiver in the fog area is constantly changed to eliminate the influence of the same channel on the fog characteristics of the radio signal.

In the data preprocessing module, the received signal is converted into an IQ signal through orthogonal digital down-conversion technology, and the data is normalized, and then converted into the frequency domain through FFT and a spectrum waterfall diagram is drawn. The two data preprocessing schemes, IQ signal and spectrum waterfall diagram, are used as input data and input into the low-frequency signal fog feature recognition module.

In the low-frequency signal fog feature recognition module, the features of the input data are first extracted through the convolutional neural network. Then it is input into the attention network. The attention network uses the features extracted by the convolutional neural network to obtain the hot spots in the input data by integrating the attention mechanism. At the same time, the attention network tracks multiple areas of interest to obtain the data of the areas of interest for the next level of network to make judgments on a finer scale. The extracted refined features are then input into the classifier for classification and identification, and then different environmental conditions are monitored, including 21% no water and no fog, 21% water and no fog, 85% water and fog, and water and fog under saturated state. The classification effect of the model is evaluated by the accuracy rate. The higher the accuracy rate, the better the classification effect of the model, and the better the monitoring effect of foggy weather with different concentrations. The calculation formula is as follows:

Among them, TP is the total number of positive detections, TN is the number of other positive detections, FP is the total number of false detections, and FN is the number of missed detections.

The radio fog weather monitoring method based on deep learning recorded in the above embodiment can be run in a computer device and applied in related scenarios. The computer device includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the following steps:

The received signal is converted into an IQ signal using orthogonal digital down-conversion technology, and then converted to the frequency domain through FFT and a spectrum waterfall diagram is drawn. The performance of various data preprocessing schemes is evaluated and the appropriate data preprocessing scheme is selected. The signal is then input into the subsequent network to identify the characteristics of low-frequency radio signals in fog areas of different concentrations.

The features of the input data are extracted through convolutional neural networks.

Using the features extracted by the convolutional neural network, the hot spots in the input data are obtained by integrating the attention mechanism, so that the next level of network can make judgments at a finer scale. At the same time, the attention network tracks multiple areas of interest to ensure that the characteristics of radio signals in foggy weather with different concentrations can be accurately identified.

The radio fog weather monitoring method based on deep learning described in the above embodiment can be run in a computer-readable storage medium and applied to related electronic devices. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor executes the following steps:

The received signal is converted into an IQ signal using orthogonal digital down-conversion technology, and then converted to the frequency domain through FFT and a spectrum waterfall diagram is drawn. The performance of various data preprocessing schemes is evaluated and the appropriate data preprocessing scheme is selected. The signal is then input into the subsequent network to identify the characteristics of low-frequency radio signals in fog areas of different concentrations.

The features of the input data are extracted through convolutional neural networks.

Using the features extracted by the convolutional neural network, the hot spots in the input data are obtained by integrating the attention mechanism, so that the next level of network can make judgments at a finer scale. At the same time, the attention network tracks multiple areas of interest to ensure that the characteristics of radio signals in foggy weather with different concentrations can be accurately identified.

3. Evidence of the effects of the embodiments. The embodiments of the present invention have achieved some positive effects during the development or use process, and indeed have great advantages over the prior art. The following content is described in conjunction with the data, charts, etc. of the test process.

The present invention firstly conducted a fog monitoring experiment. The experiment collected broadcast signal data under two categories: a relative humidity of 21% antenna wet environment without fog and a saturated antenna wet environment with fog. The collected data was divided into a training set, a validation set and a test set in a ratio of 8:1:1. Finally, a recognition accuracy rate of 95.28% was obtained. The experiment proved that the scheme adopted by the present invention can effectively monitor the presence of fog.

In order to test the recognition performance of the scheme adopted by the present invention for different concentrations of fog, the experiment collected broadcast signal data under four categories: 21% antenna dry environment without fog, 21% antenna wet environment without fog, 85% antenna wet environment with fog, and antenna wet environment with fog under saturation state. The training set, validation set, and test set are still divided according to the ratio of 8:1:1. Finally, the recognition accuracy rate of 86.18% was obtained. The experiment proves the feasibility and effectiveness of using low-frequency radio signals for foggy weather monitoring, and also provides ideas for further research in this direction.

The deep learning workstation used in the experiment is configured as a Windows system and implemented using the Pytorch framework. The hardware configuration is GPU: NVIDIA GeForce RTX 2080 Ti, CPU: Intel(R) Core(TM) i9-9820X CPU@3.30GHZ. The signal receiving device uses a 3900A radio monitoring receiver.

It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and/or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims (8)

1. A radio foggy weather monitoring method based on deep learning, characterized in that the radio foggy weather monitoring method based on deep learning comprises the following steps: The collected low-frequency radio signals in foggy areas with different concentrations are converted into a spectrum waterfall diagram. Based on the spectrum waterfall diagram, data processed by two data preprocessing methods, the spectrum waterfall diagram and the original IQ signal, are used as input data and input into a low-frequency radio signal fog feature recognition model to recognize the features of fog in the radio signal. After collecting the low-frequency radio signals in foggy areas with different concentrations, it is also necessary to: use radio and deep learning methods, and adopt a deep artificial neural network with supervised learning capabilities to extract the characteristics of the low-frequency radio signals passing through foggy areas with different concentrations; The low-frequency radio signal fog feature recognition model is composed of several convolutional neural networks and attention networks; the convolutional neural network is used to extract the features of input data; The attention network uses the features extracted by the convolutional neural network and obtains the hot spots in the input data by fusing the attention mechanism, so that the next level network can make judgments on a finer scale. 2. The radio fog weather monitoring method based on deep learning as described in claim 1 is characterized in that when the low-frequency radio signal is converted into a spectrum waterfall diagram, the IQ signal obtained by orthogonal digital down-conversion and the IQ signal are converted to the frequency domain through FFT to obtain the spectrum waterfall diagram. 3. The radio fog weather monitoring method based on deep learning according to claim 1, characterized in that the radio fog weather monitoring method based on deep learning specifically comprises the following steps: Step 1: using a dedicated radio monitoring receiver to collect low-frequency radio signal data under foggy weather with different concentrations; Step 2: Use orthogonal digital down-conversion technology to convert the received signal into an IQ signal, then convert it into the frequency domain through FFT and draw a spectrum waterfall diagram; evaluate the performance of various data preprocessing schemes and select the appropriate data preprocessing method; Step 3: Extract the features of input data through convolutional neural network; Step 4: Use the features extracted by the convolutional neural network to obtain the hot spots in the input data through the fusion attention mechanism, so that the next level of network can make judgments at a finer scale. 4. The radio fog weather monitoring method based on deep learning as described in claim 3 is characterized in that, in step 1, the low-frequency band radio signal is a low-frequency band baseband signal of a selected frequency band received, and during the data collection process, the position of the receiver in the fog area is continuously changed to eliminate the influence of the same channel on the fog characteristics of the radio signal. 5. The radio fog weather monitoring method based on deep learning as claimed in claim 3 is characterized in that, in step 2, the received signal is converted into an IQ signal using orthogonal digital down-conversion technology, and the conversion formula used is as follows: I＝h _LP (s(t)cos(2πft)) Q＝h _LP (s(t)sin(2πft)) Where s(t) is the received signal, f represents the carrier frequency of the transmitted signal, and h _LP represents the system function of the low-pass filter; In step 4, the attention network is instructed to track multiple regions of interest and identify the characteristics of radio signals under different concentrations of fog. 6. A radio fog weather monitoring system based on deep learning for implementing the radio fog weather monitoring method based on deep learning according to any one of claims 1 to 5, characterized in that the radio fog weather monitoring system based on deep learning comprises: A signal receiving module is used to receive low-frequency radio signals passing through foggy environments of different densities by using a radio monitoring receiver and setting a center frequency, bandwidth, etc.; A data preprocessing module, used for converting the low-frequency band radio signal received by the receiver into a spectrum waterfall diagram for identification by the low-frequency band signal fog feature identification module; The low-frequency signal fog feature recognition module is used to perform feature recognition on radio signals collected at low frequencies and passing through foggy environments of different concentrations through a low-frequency radio signal fog feature recognition model. 7. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the radio fog weather monitoring method based on deep learning as described in any one of claims 1 to 5. 8. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the radio fog weather monitoring method based on deep learning as described in any one of claims 1 to 5.