A radio spectrum monitoring method and system for a low altitude defense system

Abstract

The present application relates to the field of data processing, in particular to a radio spectrum monitoring method and system of low altitude defense system, comprising: performing short-time Fourier transform on signal time sequence data collected by any monitoring point to obtain a time-frequency matrix, taking the time-frequency matrix of any monitoring point as a target matrix, calculating the frequency domain reliability of the target matrix according to the energy proportion of each frequency point and the correlation degree of each frequency point, combining with information entropy theory, calculating the periodicity degree of the target matrix, obtaining the monitoring reliability of the monitoring point corresponding to the target matrix according to the frequency domain reliability and the periodicity degree, inputting the time-frequency matrix of each monitoring point into a preset prediction model of the probability of existence of unmanned aerial vehicle to obtain the prediction probability of each monitoring point, taking the monitoring reliability of each monitoring point as a weight, and weighting and summing the prediction probability of all monitoring points to obtain the total probability of existence of unmanned aerial vehicle. The present application can improve the accuracy of unmanned aerial vehicle monitoring.

Description

Technical Field

[0001] This invention relates to the field of data processing. More specifically, this invention relates to a radio spectrum monitoring method and system for a low-altitude defense system. Background Technology

[0002] With the rapid popularization of drone technology, the problem of illegal drone flights in low-altitude areas is becoming increasingly prominent, potentially posing a serious threat to public safety, airspace management, and the protection of critical facilities. Since drone communication, navigation, and motor control all generate radio frequency signals in specific frequency bands, drone monitoring can be achieved by capturing and analyzing these signals. This is the principle behind drone monitoring based on radio spectrum monitoring technology.

[0003] Existing radio spectrum monitoring systems typically use distributed monitoring points for radio spectrum acquisition. However, different monitoring points are subject to varying degrees of environmental interference due to their different geographical locations. For example, monitoring points in different locations may experience different levels of unknown interference, or they may have varying degrees of sensitivity to UAV signals. This results in inconsistent monitoring data quality from different monitoring points, making it easy for false alarms or missed alarms from UAVs to occur when the monitoring results from all monitoring points are fused for decision-making. Summary of the Invention

[0004] The main objective of this application is to propose a radio spectrum monitoring method and system for low-altitude defense systems, aiming to improve the accuracy of UAV monitoring.

[0005] To achieve the above objectives, an embodiment of the first aspect of this application proposes a radio spectrum monitoring method for a low-altitude defense system. The method includes: performing a short-time Fourier transform on the signal time-series data collected at any monitoring point to obtain a time-frequency matrix, wherein each row of the time-frequency matrix corresponds to a frequency point, each column of the time-frequency matrix corresponds to a time window, and the matrix elements of the time-frequency matrix are signal energies; using the time-frequency matrix of any monitoring point as a target matrix, calculating the energy proportion of each frequency point in the target matrix, obtaining the correlation degree between each frequency point and the probability of the existence of a UAV, calculating the frequency domain credibility of the target matrix based on the energy proportion and the correlation degree of each frequency point, and combining information entropy theory, calculating the periodicity of the target matrix, weighted summing the frequency domain credibility and the periodicity to obtain the monitoring credibility of the monitoring point corresponding to the target matrix, and iterating to obtain the monitoring credibility of each monitoring point; inputting the time-frequency matrix of each monitoring point into a preset prediction model of the probability of the existence of a UAV to obtain the predicted probability of each monitoring point, using the monitoring credibility of each monitoring point as a weight, weighted summing the predicted probabilities of all monitoring points to obtain the total probability of the existence of a UAV, thus completing the monitoring.

[0006] In some embodiments, calculating the energy percentage of each frequency point in the target matrix includes: summing all signal energies of the target matrix as the total energy of the target matrix; calculating the cumulative value of signal energy of each frequency point in all time windows of the target matrix to obtain the first cumulative energy of each frequency point; and using the ratio of the first cumulative energy of each frequency point to the total energy as the energy percentage of each frequency point.

[0007] In some embodiments, obtaining the correlation between each frequency point and the probability of the existence of a drone includes: acquiring multiple samples, each sample being a time-frequency matrix from historical monitoring points, wherein a sample is labeled 1 to indicate the presence of a drone signal and a sample is labeled 0 to indicate the absence of a drone signal; using 1 and 0 as set elements to construct a label set; taking any frequency point as a first target point, calculating the average energy of the first target point in the time dimension of each sample, wherein one sample corresponds to one average energy of the first target point; using all average energies as set elements to construct an energy set of the first target point; calculating the mutual information between the label set and the energy set; iterating through and acquiring the mutual information of each frequency point; and using the ratio of the mutual information of the first target point to the maximum value of all mutual information as the correlation between the first target point and the probability of the existence of a drone; iterating through and acquiring the correlation between each frequency point and the probability of the existence of a drone.

[0008] In some embodiments, calculating the average energy of the first target point in the time dimension of each sample includes: taking any sample as the target sample, accumulating the signal energy of the first target point in all time windows of the target sample to obtain a second cumulative energy, and taking the ratio of the second cumulative energy to the number of time windows as the average energy of the first target point in the target sample; and iterating to obtain the average energy of the first target point in the time dimension of each sample.

[0009] In some embodiments, calculating the frequency domain credibility of the target matrix using information entropy theory includes: taking any frequency point as a second target point, constructing an uncertainty function to obtain the uncertainty value of the energy proportion of the second target point, taking the correlation degree of the second target point, the product of the uncertainty value and the proportion of the second target point as the contribution of the second target point, iterating through to obtain the contribution of each frequency point, taking the negative of the sum of the contributions of all frequency points as the entropy of the target matrix, iterating through to obtain the entropy of the time-frequency matrix of each monitoring point, and using an exponential function to perform a negative correlation mapping on all entropies to obtain the frequency domain credibility of the target matrix.

[0010] In some embodiments, calculating the periodicity of the target matrix includes: taking any time window of the target matrix as a target window, taking any frequency point as a third target point, taking the correlation degree of the third target point as a weight to weight the signal energy of the third target point in the target window to obtain weighted energy, traversing to obtain the weighted energy of each frequency point in the target window, taking the sum of all the weighted energies as the third cumulative energy of the target window, traversing to obtain the third cumulative energy of each time window to construct an energy sequence; obtaining a value interval, taking any value in the value interval as a target quantity, obtaining the advance sequence of the energy sequence, the time length of the advance sequence differing from the energy sequence by the target quantity of time windows, calculating the similarity between the energy sequence and the advance sequence, traversing to obtain the similarity corresponding to each value in the value interval, and taking the maximum value of the similarity as the periodicity of the target matrix.

[0011] In some embodiments, obtaining the value range includes: obtaining the number of time windows in the target matrix, calculating the upper limit of the value based on the number, taking 1 as the lower limit of the value, and constructing the value range based on the upper limit of the value and the lower limit of the value.

[0012] In some embodiments, the construction of the prediction model includes: constructing a training set based on a plurality of samples, wherein a sample labeled 1 indicates the presence of a drone signal and a sample labeled 0 indicates the absence of a drone signal; training a neural network based on the training set using a cross-entropy loss function; and obtaining the prediction model after training is completed.

[0013] An embodiment of the second aspect of this application provides a radio spectrum monitoring system for a low-altitude defense system. The system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-described radio spectrum monitoring method for a low-altitude defense system is implemented.

[0014] The beneficial effects of this invention are:

[0015] This invention analyzes the monitoring data (i.e., the time-frequency matrix) of each monitoring point from both the frequency and time domains to obtain the monitoring reliability of each monitoring point. In the final fusion decision, the monitoring reliability of each monitoring point is used as a weight, and the predicted probabilities of all monitoring points are weighted and summed to obtain the total probability of the presence of the UAV. Compared with the prior art, this invention evaluates the data reliability of each monitoring point to amplify the contribution of the monitoring results corresponding to the monitoring points with higher monitoring reliability and reduce the contribution of the monitoring results corresponding to the monitoring points with lower monitoring reliability during the fusion decision, thereby improving the accuracy of UAV monitoring. Attached Figure Description

[0016] Figure 1 This is a flowchart of steps S1-S3 in a radio spectrum monitoring method for a low-altitude defense system according to an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0018] The specific scenarios targeted by this invention are: mainly low-altitude airspace with complex electromagnetic environments, such as cities, border areas, and areas surrounding important infrastructure, where timely and accurate detection of micro and small drones that illegally intrude or fly in violation of regulations is required.

[0019] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] Reference Figure 1 A radio spectrum monitoring method for a low-altitude defense system includes steps S1-S3, as detailed below:

[0021] Step S1: Perform a short-time Fourier transform on the time-series data of the signal collected at any monitoring point to obtain a time-frequency matrix. Each row of the time-frequency matrix corresponds to a frequency point, each column of the time-frequency matrix corresponds to a time window, and the matrix elements of the time-frequency matrix are the signal energy.

[0022] It should be further noted that the radio spectrum monitoring system of the present invention includes multiple monitoring points, which are distributed throughout the monitoring area. Each monitoring point ( For monitoring point index, The system receives signals from all radio frequency bands within the monitoring area (total number of monitoring points) via an ultra-wideband integrated antenna. These signals form timing data, which is then sequentially amplified by a low-noise amplifier and passed through an ultra-wideband bandpass filter. Finally, a high-speed analog-to-digital converter converts the data into a digital IQ (In-phase / Quadrature) signal stream. This digital IQ signal stream is then converted into a time-frequency matrix using a short-time Fourier transform. A time-frequency diagram can be obtained by visualizing the time-frequency matrix.

[0023] The principle of the Short-Time Fourier Transform (SFT) is as follows: A sliding window is generated on the digital IQ signal stream. The digital IQ signal stream is divided by the sliding window, which is also the time window. Each time the window slides, a Fourier calculation is performed on the data in the window. Each Fourier calculation outputs... Signal energy at each frequency point ( Related to the number of Fourier points (which can be set according to the acquisition frequency of the monitoring points, the frequency resolution required by the scene, etc.), the length of the time window can be set according to the acquisition frequency of the monitoring points, the time resolution required by the scene, etc. Each row of the time-frequency matrix stores the signal energy of a frequency point across all time windows, reflecting the time-domain characteristics of a frequency point; each column of the time-frequency matrix stores the signal energy of all frequency points within a time window, reflecting the frequency-domain characteristics of a time window.

[0024] It should be noted that the number of rows and columns of the time-frequency matrix are the same for each monitoring point.

[0025] It should be noted that, for ease of calculation, the time-frequency matrix of each monitoring point can be preprocessed. Specifically, a bandpass filter is used to remove residual frequency components outside the compliant frequency bands from the original time-frequency matrix, retaining the pure compliant frequency band signals. Based on A local statistics algorithm is used to identify impulse interference transition points, and a moving average of length 5 is employed to smooth these transition points. All matrix elements (signal energy) are normalized to eliminate differences in hardware gain between different nodes.

[0026] It should be noted that the time-frequency matrix uses... It indicates. Among them, For row indexes, corresponding to the frequency dimension. , The number of frequency points; This is a column index, corresponding to the time dimension. , The number of time windows. Matrix elements. Indicates the first In the time-frequency matrix of the monitoring point, the first... The frequency point at the th frequency point Signal energy within a time window.

[0027] Step S2: Take the time-frequency matrix of any monitoring point as the target matrix, calculate the energy proportion of each frequency point in the target matrix, obtain the correlation degree between each frequency point and the UAV, calculate the frequency domain credibility of the target matrix based on the energy proportion and correlation degree of each frequency point, and combine the information entropy theory, calculate the periodicity of the target matrix, and calculate the periodicity of the target matrix. The monitoring credibility of the monitoring point corresponding to the target matrix is ​​obtained by weighted summation of the frequency domain credibility and periodicity. The monitoring credibility of each monitoring point is obtained by iterative acquisition.

[0028] In some embodiments, calculating the energy percentage of each frequency point in the target matrix includes: summing the signal energies of all signals in the target matrix as the total energy of the target matrix; calculating the cumulative value of the signal energy of each frequency point in all time windows of the target matrix to obtain the first cumulative energy of each frequency point; and using the ratio of the first cumulative energy of each frequency point to the total energy as the energy percentage of each frequency point.

[0029] For example, the formula for calculating the energy percentage at each frequency point in the target matrix is ​​as follows:

[0030]

[0031]

[0032] In the formula and formula middle, For the first The frequency point at the th frequency point The cumulative value of signal energy in all time windows of the time-frequency matrix of each monitoring point. For the index of the time window, The number of time windows. Indicates the first In the time-frequency matrix of the monitoring point, the first... The frequency point at the th frequency point Signal energy within a time window. For the first The frequency point at the th frequency point The energy percentage of the time-frequency matrix at each monitoring point. For the index of the frequency point, The number of frequency points.

[0033] For the above formula and formula It should be added that, The larger the value, the higher the value. The frequency point at the th frequency point The larger the energy proportion of the time-frequency matrix of the monitoring point, the greater the proportion of the energy of the monitoring point. The first monitoring point received the first The larger the proportion of the signal corresponding to the frequency point (the first frequency point), the greater the proportion of the signal (the second frequency point). The more concentrated the signals received by the monitoring points are on the first... (Number of frequency points).

[0034] It should be further explained that by calculating the energy proportion of each frequency point in the target matrix, the energy contribution of each frequency point in the overall signal is quantified, thereby clarifying the main frequency range of signal energy distribution. This method allows for the rapid identification of high-energy frequency components, providing data support for subsequent frequency domain reliability analysis.

[0035] In some embodiments, obtaining the correlation between each frequency point and the probability of the existence of a drone includes: acquiring multiple samples, each sample being a time-frequency matrix from historical monitoring points, where a sample is labeled 1 to indicate the presence of a drone signal and a sample is labeled 0 to indicate the absence of a drone signal; using 1 and 0 as set elements to construct a label set; taking any frequency point as a first target point, calculating the average energy of the first target point in the time dimension of each sample, where one sample corresponds to one average energy of the first target point; using all average energies as set elements to construct an energy set of the first target point; calculating the mutual information between the label set and the energy set; iterating through and obtaining the mutual information of each frequency point; using the ratio of the mutual information of the first target point to the maximum value of all mutual information as the correlation between the first target point and the probability of the existence of a drone; and iterating through and obtaining the correlation between each frequency point and the probability of the existence of a drone.

[0036] It should be further explained that the historical time-frequency matrix is ​​obtained by performing a short-time Fourier transform on the historical time-series signal data collected from the monitoring points. The historical time-frequency matrix is ​​then labeled, resulting in a label set. , This indicates that the time-frequency matrix corresponding to the sample contains drone signals, meaning the probability of a drone's presence is 100%. This indicates that the time-frequency matrix corresponding to the sample does not contain drone signals, meaning the probability of a drone's existence is 0.

[0037] In some embodiments, calculating the average energy of the first target point in the time dimension of each sample includes: taking any sample as the target sample, accumulating the signal energy of the first target point in all time windows of the target sample to obtain a second cumulative energy, and using the ratio of the second cumulative energy to the number of time windows as the average energy of the first target point in the target sample; and iterating through the samples to obtain the average energy of the first target point in the time dimension of each sample.

[0038] For example, the formula for calculating the average energy of the first target point over the time dimension of each sample is as follows:

[0039]

[0040] In the formula middle, For the first The frequency point at the th frequency point The average energy of each sample over the time dimension The number of time windows. For the index of the time window, Indicates the first The first sample The frequency point at the th frequency point Signal energy within a time window.

[0041] For the above formula It should be added that the first The energy set at each frequency point is ,in, The number of samples.

[0042] It should be further explained that by accumulating the energy of a frequency point across all time windows of the sample and taking the average, the interference caused by energy fluctuations in a single time window is eliminated, thus more objectively reflecting the energy characteristics of that frequency point throughout the entire sample period. This calculation method can smooth out the influence of temporal noise, making the obtained average energy more representative and ensuring the accuracy of subsequent mutual information calculations.

[0043] For example, the first The formula for calculating the mutual information at each frequency point is as follows:

[0044]

[0045] In the formula middle, For the first Mutual information at each frequency point For joint probability, For the first The average energy at each frequency point is The marginal probability, The marginal probability of the label.

[0046] For the above formula It should be added that, for example, For the first The average energy at each frequency point is The proportion of the number of samples to the total number of samples, for each frequency point Equal and all , For the tag The proportion of the number of samples to the total number of samples, for example, =0.5.

[0047] It should be further explained that, in the samples selected in this invention, the mutual information between the average energy of each frequency point and the sample tag is calculated, thereby quantifying the degree of correlation between each frequency point and the potential existence of the UAV. Indicates the first The degree of correlation between each frequency point and the drone. . The closer it is to 1, the better. The stronger the correlation between a frequency point and the UAV signal, the closer it is to 0, indicating that the frequency point is... The weaker the correlation between a frequency point and the drone signal, the more likely the signal at that frequency point is to come from noise or interference from other non-drone signals.

[0048] It should be further explained that calculating the mutual information based on the label set of historical samples and the energy set of frequency points can uncover whether the energy changes of frequency points in different samples are related to the sample labels (whether the drone signal is present or not). The larger the mutual information, the stronger the indicative significance of that frequency point for drone detection. By normalizing the maximum value of the mutual information with all frequency points, the correlation between each frequency point is unified to the same order of magnitude, which facilitates horizontal comparison of the importance of different frequency points, provides standardized parameters for subsequent weighted calculations, and improves the consistency and stability of the overall algorithm.

[0049] In some embodiments, calculating the frequency domain credibility of the target matrix using information entropy theory includes: taking any frequency point as the second target point, constructing an uncertainty function to obtain the uncertainty value of the energy proportion of the second target point, taking the product of the correlation degree of the second target point, the uncertainty value, and the proportion of the second target point as the contribution of the second target point, iterating through to obtain the contribution of each frequency point, taking the negative of the sum of the contributions of all frequency points as the entropy of the target matrix, iterating through to obtain the entropy of the time-frequency matrix of each monitoring point, and using an exponential function to perform a negative correlation mapping on all entropies to obtain the frequency domain credibility of the target matrix.

[0050] For example, the formula for calculating the entropy of the target matrix is ​​as follows:

[0051]

[0052] In the formula middle, For the first The entropy of the time-frequency matrix of each monitoring point, For the index of the frequency point, The number of frequency points, For the first The degree of correlation between each frequency point and the drone. For the first The frequency point at the th frequency point The energy percentage of the time-frequency matrix at each monitoring point For uncertain functions, For the first The frequency point at the th frequency point The uncertainty value corresponding to the energy proportion of the time-frequency matrix of each monitoring point For the first The contribution of each frequency point.

[0053] For the above formula It should be added that, The smaller (the first) The frequency point at the th frequency point The smaller the energy percentage of each time-frequency matrix, the better. The larger the absolute value, the stronger the excluding the th . The more unknown information other frequencies provide besides a given frequency point, the greater the potential interference, meaning the more likely the spectrum will be chaotic.

[0054] For the above formula It should be added that the first Frequency domain reliability of the time-frequency matrix of each monitoring point . Used for control The effect expended in calculating frequency domain reliability is as follows: for frequency points with a stronger correlation to UAV signals, the effect expended in calculating frequency domain reliability is increased; for frequency points with a weaker correlation to UAV signals, the effect expended in calculating frequency domain reliability is decreased, in order to reduce the impact of interference.

[0055] In some embodiments, calculating the periodicity of the target matrix includes: taking any time window of the target matrix as the target window, taking any frequency point as the third target point, taking the correlation degree of the third target point as the weight, weighting the signal energy of the third target point in the target window to obtain the weighted energy, iterating to obtain the weighted energy of each frequency point in the target window, taking the sum of all weighted energies as the third cumulative energy of the target window, iterating to obtain the third cumulative energy of each time window to construct an energy sequence; obtaining a value range, taking any value in the value range as the target quantity, obtaining the advance sequence of the energy sequence, taking the time length of the advance sequence and the energy sequence as the target quantity of time windows, calculating the similarity between the energy sequence and the advance sequence corresponding to each value, iterating to obtain the similarity corresponding to each value in the value range, and taking the maximum value of the similarity as the periodicity of the target matrix.

[0056] For example, the formula for calculating the third cumulative energy of the target matrix for each time window is as follows:

[0057]

[0058] In the formula middle, For the first The time-frequency matrix of the monitoring point is the first The third accumulated energy within a time window, For the index of the frequency point, The number of frequency points, For the first The degree of correlation between each frequency point and the drone. Indicates the first In the time-frequency matrix of the monitoring point, the first... The frequency point at the th frequency point Signal energy within a time window.

[0059] For the above formula It should be noted that the energy sequence of the target matrix is... It can reflect the key frequency ( The energy sequence shows a relatively large (and potentially highly correlated with drones) power variation trend over time, while suppressing the influence of low-correlation frequencies. The advance sequence of this energy sequence is based on the first... Historical data acquisition from each monitoring point, for example, the last sequence element of the energy sequence is... Then the last sequence element of the advance sequence is , For the target quantity, the energy sequence and the advance sequence have the same sequence length.

[0060] The similarity between the energy sequence and the preceding sequence is calculated using the Pearson correlation coefficient. The closer the similarity is to 1, the stronger the similarity (waveform repeatability) between the energy sequence and the preceding sequence, and the more significant the periodicity of the energy sequence. Since UAV signals usually have a stable period, while interference signals are non-periodic, the stronger the periodicity, the higher the probability that the signal originates from a UAV, and the less affected by interference. Therefore, the time-domain reliability of the data from the monitoring point is also higher.

[0061] In some embodiments, obtaining the value range includes: obtaining the number of time windows in the target matrix, calculating the upper limit of the value based on the number, taking 1 as the lower limit of the value, and constructing the value range based on the upper limit and the lower limit of the value.

[0062] It should be noted that the upper limit of the value is half the number of time windows. The number of time windows is affected by the length of the time windows, which can be set according to the acquisition frequency of the monitoring points and the time resolution required by the scene. Obtaining the similarity corresponding to each value in the range is to explore the periodicity of the energy sequence. The preceding sequence corresponding to the maximum similarity is closest to the previous periodic sequence of the energy sequence. Therefore, the maximum similarity is taken as the degree of periodicity of the target matrix.

[0063] For example, the formula for calculating the credibility of monitoring is as follows:

[0064]

[0065] In the formula middle, For the first The reliability of monitoring at each monitoring point These are the weighting coefficients for frequency domain credibility. For the first Frequency domain reliability of each monitoring point The weighting coefficients represent the degree of periodicity. For the first The degree of periodic regularity of each monitoring point.

[0066] For the above formula It should be added that it is generally set to It can be adjusted according to the actual situation. The larger, the more The more reliable the monitoring data from each monitoring point, the more important it will be in the final fusion decision. The greater the contribution of each monitoring point.

[0067] Step S3: Input the time-frequency matrix of each monitoring point into the preset prediction model of the probability of the existence of drones to obtain the predicted probability of each monitoring point. Use the monitoring credibility of each monitoring point as a weight, and sum the predicted probabilities of all monitoring points to obtain the total probability of the existence of drones, thus completing the monitoring.

[0068] In some embodiments, the construction of the prediction model includes: constructing a training set based on multiple samples, wherein a sample labeled 1 indicates the presence of a drone signal, and a sample labeled 0 indicates the absence of a drone signal; training a neural network using a cross-entropy loss function based on the training set; and obtaining the prediction model after training. For example, the formula for calculating the total probability is as follows:

[0069]

[0070] In the formula middle, The total probability of the existence of drones. For monitoring point index, The total number of monitoring points. For the first The reliability of monitoring at each monitoring point For the first The predicted probability of each monitoring point. This is the normalization formula.

[0071] For the above formula It should be added that, The larger the value, the greater the overall probability of the presence of drones within the monitored area. A decision threshold is set; if... If the threshold is exceeded, and the presence of a drone within the monitored area is determined, processes such as drone location and trajectory tracking can be triggered, thereby providing target dynamic information for low-altitude defense response; if If the decision threshold is not exceeded, then there are no drones.

[0072] The present invention also provides a radio spectrum monitoring system for a low-altitude defense system. The system includes a processor and a memory, the memory storing computer program instructions. When the computer program instructions are executed by the processor, they implement a radio spectrum monitoring method for a low-altitude defense system according to the first aspect of the present invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface; their configuration and functions are known in the art and will not be described further here.

[0073] It should be noted that the preferred embodiments of this application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of this application. For those skilled in the art, various modifications and improvements can be made without departing from the concept of the invention, and these all fall within the protection scope of the invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method of radio spectrum monitoring for a low altitude defense system, characterized by, The method comprises the following steps: performing short-time Fourier transform on the time sequence data of the signals collected at any monitoring point to obtain a time-frequency matrix, each row of the time-frequency matrix corresponding to a frequency point, each column of the time-frequency matrix corresponding to a time window, and the matrix elements of the time-frequency matrix being signal energy; taking the time-frequency matrix of any monitoring point as a target matrix, calculating the energy proportion of each frequency point in the target matrix, obtaining the correlation degree of each frequency point with the possibility of the existence of the unmanned aerial vehicle, calculating the frequency domain reliability of the target matrix according to the energy proportion of each frequency point and the correlation degree of each frequency point and combining the information entropy theory, calculating the periodicity degree of the target matrix, and obtaining the monitoring reliability of the monitoring point corresponding to the target matrix by weighted summation of the frequency domain reliability and the periodicity degree, and obtaining the monitoring reliability of each monitoring point by iteration; inputting the time-frequency matrix of each monitoring point into a preset prediction model of the probability of the existence of the unmanned aerial vehicle to obtain the prediction probability of each monitoring point, taking the monitoring reliability of each monitoring point as a weight, and obtaining the total probability of the existence of the unmanned aerial vehicle by weighted summation of the prediction probabilities of all monitoring points, thereby completing the monitoring; wherein; the calculation of the energy proportion of each frequency point in the target matrix comprises: taking the sum of all signal energies of the target matrix as the total energy of the target matrix, calculating the accumulated value of the signal energy of each frequency point in all time windows of the target matrix to obtain the first cumulative energy of each frequency point, and taking the ratio of the first cumulative energy of each frequency point to the total energy as the energy proportion of each frequency point; the obtaining of the correlation degree of each frequency point with the possibility of the existence of the unmanned aerial vehicle comprises: obtaining a plurality of samples, each sample being a time-frequency matrix in a historical monitoring point, the label of the sample being 1 indicating the existence of the unmanned aerial vehicle signal, and the label of the sample being 0 indicating the non-existence of the unmanned aerial vehicle signal, taking 1 and 0 as set elements to construct a label set, taking any frequency point as a first target point and any sample as a target sample, accumulating the signal energy of the first target point in all time windows of the target sample to obtain a second cumulative energy, taking the ratio of the second cumulative energy to the number of time windows as the average energy of the first target point in the target sample, obtaining the average energy of the first target point in the time dimension of each sample by iteration, wherein one sample corresponds to the average energy of one first target point, taking all average energies as set elements to construct an energy set of the first target point, calculating the mutual information of the label set and the energy set, obtaining the mutual information of each frequency point by iteration, and taking the ratio of the mutual information of the first target point to the maximum value of all mutual informations as the correlation degree of the first target point with the possibility of the existence of the unmanned aerial vehicle, thereby obtaining the correlation degree of each frequency point with the possibility of the existence of the unmanned aerial vehicle by iteration. The calculating the periodic regularity degree of the target matrix comprises: taking any time window of the target matrix as a target window, taking any frequency point as a third target point, taking the correlation degree of the third target point as a weight to weight the signal energy of the third target point in the target window to obtain a weighted energy, traversing to obtain the weighted energy of each frequency point in the target window, taking the sum of all the weighted energies as a third cumulative energy of the target window, traversing to obtain the third cumulative energy of each time window to construct an energy sequence; obtaining a value interval, taking any value in the value interval as a target number, obtaining an advanced sequence of the energy sequence, the advanced sequence and the energy sequence being different by a time length of target number of time windows, calculating the similarity between the energy sequence and the advanced sequence, traversing to obtain the similarity corresponding to each value in the value interval, taking the maximum value of the similarity as the periodic regularity degree of the target matrix; The calculation formula of the frequency domain reliability of the target matrix is: is a frequency domain confidence of a target matrix, is an exponential function, is a number of frequency points, is a correlation degree of the th frequency point and the possibility of the existence of the UAV, is an index of a monitoring point to which the target matrix belongs, is a correlation degree of the th frequency point and the possibility of the existence of the UAV, is an uncertainty function, is an uncertainty value corresponding to the energy proportion of the th frequency point in the target matrix, is a contribution of the th frequency point, is a sum of contributions of all frequency points, is an entropy of the target matrix.

2. The method of claim 1, wherein, The obtaining of the value interval comprises: Obtaining the number of time windows in the target matrix, calculating an upper limit value according to the number, taking 1 as a lower limit value, and constructing a value interval according to the upper limit value and the lower limit value.

3. The method of claim 1, wherein, The constructing of the prediction model comprises: constructing a training set according to a plurality of samples, wherein the label of the sample is 1, indicating that there is a drone signal, the label of the sample is 0, indicating that there is no drone signal, training a neural network based on the training set using a cross-entropy loss function, and obtaining a prediction model after training.

4. A radio spectrum monitoring system for a low altitude defense system, characterized by, comprise: a processor and a memory, the memory storing computer program instructions, when the computer program instructions are executed by the processor, realizing the radio spectrum monitoring method of the low-altitude defense system according to any one of claims 1-3.

Images