Unmanned aerial vehicle monitoring method and system in complex electromagnetic environment

Abstract

The invention discloses an unmanned aerial vehicle monitoring method and system in a complex electromagnetic environment, and the method comprises the steps: extracting a signal feature from a spectrum signal collected in real time, and obtaining a first feature vector; adopting a pre-constructed random forest model to classify the first feature vectors and match the first feature vectors with known unmanned aerial vehicle information to obtain a first model and a first serial number of the unmanned aerial vehicle; according to the coordinates of the at least three measurement stations, calculating the three-dimensional coordinates of the unmanned aerial vehicle by adopting a time difference of arrival positioning technology and an angle of arrival measurement technology; calculating the radial speed of the unmanned aerial vehicle relative to each measurement station by adopting a Doppler effect, and calculating a three-dimensional speed vector of the unmanned aerial vehicle according to the radial speed; and according to the three-dimensional coordinates and the three-dimensional speed vector of the unmanned aerial vehicle, predicting the change trend of the three-dimensional coordinates and the three-dimensional speed vector of the unmanned aerial vehicle in a preset time window by adopting a particle filtering algorithm. By adopting the embodiment of the invention, the identification accuracy, the positioning precision and the anti-interference capability of unmanned aerial vehicle monitoring in a complex electromagnetic environment can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) monitoring, and particularly to a method and system for monitoring UAVs in a complex electromagnetic environment. Background Art

[0002] With the rapid development and wide application of UAV technology, its safety supervision work is facing huge challenges. Especially in a complex electromagnetic environment, how to accurately identify and monitor UAVs has become an urgent problem to be solved. Currently, the mainstream UAV monitoring technologies mainly include methods such as radar detection, optoelectronic detection, acoustic detection, and radio frequency signal detection.

[0003] However, the performance of these existing technologies in a complex electromagnetic environment still has significant deficiencies: traditional radar detection technology is not only easily affected by electromagnetic interference, but also has limited detection ability for small low-altitude flying targets; optoelectronic detection technology is restricted by environmental factors such as weather and light, and has poor performance at night or under adverse weather conditions; acoustic detection technology also faces the problem of environmental noise interference, and its detection range is relatively limited. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for monitoring UAVs in a complex electromagnetic environment, which can improve the recognition accuracy, positioning accuracy, and anti-interference ability of UAV monitoring in a complex electromagnetic environment.

[0005] An embodiment of the present invention provides a method for monitoring UAVs in a complex electromagnetic environment, including:

[0006] Extracting signal features from the spectrum signals collected in real time to obtain a first feature vector; the frequency of the spectrum signals is within a preset frequency band;

[0007] Using a pre-constructed random forest model to classify the first feature vector and match it with known UAV information to obtain the first model and the first serial number of the UAV;

[0008] According to the coordinates of at least three measurement stations, using the time difference of arrival positioning technology and the angle of arrival measurement technology to calculate the three-dimensional coordinates of the UAV;

[0009] Calculating the radial velocity of the UAV relative to each of the measurement stations using the Doppler effect, and calculating the three-dimensional velocity vector of the UAV according to the radial velocity;

[0010] According to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV, using the particle filter algorithm to predict the change trend of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window.

[0011] As an improvement to the above solution, extracting signal features from the spectrumsignal collected in real time to obtain a first feature vector includes:

[0012] Collect the spectrum signal in real time according to a preset frequency band;

[0013] Extract the key features of the spectrum signal, and calculate the mean and standard deviation of the key features to obtain an initial feature vector; the key features include at least one of signal strength, bandwidth, center frequency, signal duration, peak power spectral density, spectral kurtosis, and skewness;

[0014] Perform normalization processing on the initial feature vector to obtain a first feature vector.

[0015] As an improvement to the above solution, the steps of pre - constructing the random forest model include:

[0016] Collect the characteristic data of several unmanned aerial vehicles (UAVs) under different environmental conditions and working states; the characteristic data includes the second feature vector, the second model number, and the second serial number of the UAV;

[0017] Construct n decision trees using the random forest algorithm, where the node splitting criterion of the decision tree is the Gini coefficient, and when each node splits, it randomly selects from the second feature vector;

[0018] Use the second feature vector as the input, and the second signal and the second serial number as the output to train the classification ability of each decision tree;

[0019] Construct a random forest model based on the trained n decision trees.

[0020] As an improvement to the above solution, using the pre - constructed random forest model to classify the first feature vector and match it with the known UAV information to obtain the first model number and the first serial number of the UAV includes:

[0021] Input the first feature vector into each decision tree in the pre - constructed random forest model, and each decision tree classifies independently to obtain a number of predicted model numbers and predicted serial numbers;

[0022] According to the number of predicted model numbers and the predicted serial numbers, use the majority voting method to obtain the first model number and the first serial number of the UAV.

[0023] As an improvement to the above solution, calculating the three - dimensional coordinates of the UAV according to the coordinates of at least three measurement stations by using the time - difference - of - arrival positioning technology and the angle - of - arrival measurement technology includes:

[0024] Obtain the coordinates of at least three measurement stations, and use the time difference of arrival positioning technology to measure the time difference of the UAV signal arriving at the at least three measurement stations;

[0025] Adopt the angle of arrival measurement technology to measure the angle information of the UAV relative to the at least three measurement stations;

[0026] According to the time difference and the angle information, establish a positioning equation set of the UAV;

[0027] According to the signal strength received by the at least three measurement stations, constrain the positioning equation set, solve the positioning equation set, and obtain the three-dimensional coordinates of the UAV.

[0028] As an improvement of the above solution, the angle information includes azimuth angle and pitch angle; the establishing a positioning equation set of the UAV according to the time difference and the angle information includes:

[0029] Establish a first positioning equation according to the time difference

[0030] Establish a second positioning equation (y - y i ) = (x - x i ) · tan(α i );

[0031] Establish a third positioning equation according to the pitch angle

[0032] Obtain the positioning equation set of the UAV according to the first positioning equation, the second positioning equation and the third positioning equation;

[0033] Among them, (x, y, z) are the three-dimensional coordinates of the UAV to be calculated; (x i , y i , z i ) are the coordinates of the i-th measurement station; c is the electromagnetic wave propagation speed; τ i1 is the time difference between the UAV signal arriving at the i-th measurement station and the first measurement station; α i is the azimuth angle of the UAV relative to the i-th measurement station; β i is the pitch angle of the UAV relative to the i-th measurement station.

[0034] As an improvement of the above solution, the constraining the positioning equation set according to the signal strength received by the at least three measurement stations, solving the positioning equation set, and obtaining the three-dimensional coordinates of the UAV includes:

[0035] Establish an RSSI-based distance estimation constraint according to the signal strength received by the at least three measurement stations Use the described distance estimation constraint to constrain the system of equations;

[0036] Solve the positioning system of equations to obtain the three-dimensional coordinates of the UAV;

[0037] where d i is the actual distance from the UAV to the i-th measurement station; d0 is the reference distance; P0 is the signal power received at the reference distance d0; P i is the signal power received at the i-th measurement station; η is a preset path loss exponent.

[0038] As an improvement to the above solution, calculating the radial velocity of the UAV relative to each of the measurement stations using the Doppler effect, and calculating the three-dimensional velocity vector of the UAV according to the radial velocity, includes:

[0039] Using the Doppler effect to obtain the frequency offset;

[0040] According to the frequency offset, through calculate the radial velocity v r of the UAV relative to each of the measurement stations to obtain the radial velocity measurement results of each of the measurement stations; where c is the electromagnetic wave propagation speed; Δf d is the frequency offset; f t is the original frequency of the signal emitted by the UAV;

[0041] Combine the radial velocity measurement results of each of the measurement stations to calculate the three-dimensional velocity vector of the UAV.

[0042] As an improvement to the above solution, predicting the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window using the particle filter algorithm according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV, includes:

[0043] Initialize a particle swarm in the state space and assign a preset initial weight to each particle in the particle swarm; the particle swarm represents the position and velocity states of the UAV;

[0044] According to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV, obtain the real-time observation vector;

[0045] Adjust the weights of the particles based on the uniform motion model according to the change of the real-time observation vector;

[0046] Predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window according to the weights of the particles adjusted in real time.

[0047] An embodiment of the present invention also provides a UAV monitoring system under a complex electromagnetic environment, including:

[0048] A feature extraction module, configured to extract signal features from the spectrum signals collected in real time to obtain a first feature vector; the frequency of the spectrum signals is within a preset frequency band;

[0049] A UAV information matching module, configured to use a pre-constructed random forest model to classify the first feature vector and match it with known UAV information to obtain a first model number and a first serial number of the UAV;

[0050] A coordinate calculation module, configured to calculate the three-dimensional coordinates of the UAV according to the coordinates of at least three measurement stations by using the time difference of arrival positioning technology and the angle of arrival measurement technology;

[0051] A velocity vector calculation module, configured to calculate the radial velocity of the UAV relative to each of the measurement stations by using the Doppler effect, and calculate the three-dimensional velocity vector of the UAV according to the radial velocity;

[0052] A trend prediction module, configured to predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window by using a particle filter algorithm according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV.

[0053] Compared with the prior art, a UAV monitoring method and system under a complex electromagnetic environment disclosed by the present invention extract signal features from the spectrum signals collected in real time to obtain a first feature vector; the frequency of the spectrum signals is within a preset frequency band; use a pre-constructed random forest model to classify the first feature vector and match it with known UAV information to obtain a first model number and a first serial number of the UAV; calculate the three-dimensional coordinates of the UAV according to the coordinates of at least three measurement stations by using the time difference of arrival positioning technology and the angle of arrival measurement technology; calculate the radial velocity of the UAV relative to each of the measurement stations by using the Doppler effect, and calculate the three-dimensional velocity vector of the UAV according to the radial velocity; predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window by using a particle filter algorithm according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV. By using the embodiment of the present invention, the recognition accuracy, positioning accuracy and anti-interference ability of UAV monitoring under a complex electromagnetic environment can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a schematic flowchart of the steps of the UAV monitoring method under a complex electromagnetic environment provided by an embodiment of the present invention;

[0055] Figure 2 is a schematic structural diagram of the UAV monitoring system under a complex electromagnetic environment provided by an embodiment of the present invention. Detailed implementation manners

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0057] In the description of the specification and claims, it should be understood that the terms first, second, etc. in the specification and claims are only used for the purpose of distinguishing the description of the same technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features, nor necessarily describing the order or time sequence. The terms can be interchanged under appropriate circumstances. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature.

[0058] The embodiment of the present invention provides a method for monitoring unmanned aerial vehicles (UAVs) in a complex electromagnetic environment. Please refer to Figure 1 , in this embodiment, the method for monitoring UAVs in the complex electromagnetic environment is specifically executed through steps S1 to S5:

[0059] S1. Extract signal features from the spectrum signals collected in real time to obtain a first feature vector; the frequency of the spectrum signals is within a preset frequency band.

[0060] It should be noted that the signals emitted by different UAVs have characteristic differences. In some cases, the differences in UAV signals are reflected in the signal intensity, while in other cases, they are reflected in characteristics such as spectral kurtosis. By extracting the first feature vector, the differentiation and identification of UAVs can be effectively achieved.

[0061] S2. Use a pre-constructed random forest model to classify the first feature vector and match it with known UAV information to obtain a first model number and a first serial number of the UAV.

[0062] It should be noted that the decision tree in the random forest model can quickly and accurately match the first feature vector with the known UAV information. It should be noted that the random forest model needs to cover the UAV information of all UAVs that may be used for monitoring in this complex electromagnetic environment. If the first feature vector obtained by monitoring cannot be matched with the known UAV information, it may indicate that it is an external UAV signal and a warning needs to be given internally.

[0063] S3. According to the coordinates of at least three measurement stations, use the time difference of arrival positioning technology and the angle of arrival measurement technology to calculate the three-dimensional coordinates of the UAV.

[0064] In an embodiment of the present invention, the drone is measured by multiple measurement stations, and by combining the Time Difference of Arrival (TDOA) positioning technology and the Angle of Arrival (AOA) measurement technology, the three-dimensional positioning accuracy of the drone is improved, and the average positioning error is reduced.

[0065] S4. Calculate the radial velocity of the drone relative to each of the measurement stations using the Doppler effect, and calculate the three-dimensional velocity vector of the drone based on the radial velocity.

[0066] It should be noted that calculating the three-dimensional velocity vector of the drone can, on the one hand, predict the three-dimensional coordinates of the drone in the future time, and on the other hand, verify the accuracy of the three-dimensional coordinate calculation.

[0067] S5. According to the three-dimensional coordinates and the three-dimensional velocity vector of the drone, use the particle filter algorithm to predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the drone within a preset time window.

[0068] In an embodiment of the present invention, the particle filter algorithm is used for data processing, realizing real-time smooth estimation and prediction of the position and velocity of the drone, improving the continuity and stability of monitoring, being able to reduce the tracking interruption time and shorten the response time to sudden state changes in case of short-term signal occlusion or loss.

[0069] In the above solution, accurate identification of multiple drones can be achieved through a pre-constructed random forest model. Combining the TDOA and AOA technologies can improve the three-dimensional positioning accuracy of the drone. In addition, the three-dimensional coordinates and three-dimensional velocity vector of the drone are predicted through the particle filter algorithm, improving the continuity and stability of monitoring. Using the embodiment of the present invention can effectively improve the recognition accuracy, positioning accuracy and anti-interference ability of drone monitoring in a complex electromagnetic environment.

[0070] As a preferred embodiment, step S1: Extract signal features from the real-time collected spectrum signal to obtain a first feature vector, including:

[0071] Collect the spectrum signal in real time according to a preset frequency band;

[0072] Extract the key features of the spectrum signal, and calculate the mean and standard deviation of the key features to obtain an initial feature vector; the key features include at least one of signal strength, bandwidth, center frequency, signal duration, peak power spectral density, spectral kurtosis and skewness;

[0073] Perform normalization processing on the initial feature vector to obtain a first feature vector.

[0074] It should be noted that the preset frequency band is determined by the frequency of the signal transmitted by the drone in actual applications. In the embodiments of the present invention, monitoring of multiple frequency bands can be realized, and the frequencies used by most commercial drones can be covered.

[0075] Exemplarily, the signal strength feature is obtained by measuring the signal power level (unit: dBm), calculating the average power, and recording the power fluctuation range; the bandwidth feature is obtained by measuring the 3dB bandwidth, calculating the occupied bandwidth, and analyzing the spectrum expansion characteristics; the center frequency feature includes determining the main frequency point position, analyzing the frequency stability, and measuring the frequency offset; the signal duration feature is obtained by recording the signal appearance duration, counting the duty cycle, and analyzing the time series characteristics; the peak power spectral density is obtained by calculating the power spectral density distribution and extracting the peak position information; the spectral kurtosis is obtained by calculating the kurtosis value of the spectral distribution and analyzing the spectral concentration degree; the spectral skewness is obtained by calculating the skewness value of the spectral distribution and analyzing the spectral symmetry.

[0076] In one embodiment, the spectrum signals in the range of 430 MHz - 5.8 GHz are collected, among which, the signals in the frequency bands of 430 - 440 MHz, 840 - 845 MHz, 902 - 928 MHz, 1080 - 1280 MHz, 1430 - 1444 MHz, 2400 - 2500 MHz, and 5725 - 5875 MHz are monitored with emphasis.

[0077] Then, the signal strength, bandwidth, center frequency, signal duration, peak power spectral density, spectral kurtosis, and skewness of the spectrum signal are extracted, and the mean and standard deviation are calculated for each key feature to form a 14 - dimensional initial feature vector. Then, the initial feature vector is normalized by Min - Max scaling to obtain the first feature vector. This normalization process maps all features into the interval [0, 1] to eliminate the influence of dimensions.

[0078] As a preferred implementation manner, the steps of pre - constructing the random forest model include:

[0079] Under different environmental conditions and working states, the characteristic data of several drones are collected; the characteristic data includes the second feature vector, the second model, and the second serial number of the drone;

[0080] n decision trees are constructed using the random forest algorithm, and the node splitting criterion of the decision tree is the Gini coefficient. When each node splits, it is randomly selected from the second feature vector;

[0081] Using the second feature vector as the input and the second signal and the second serial number as the output, the classification ability of each decision tree is trained;

[0082] Based on the trained n decision trees, a random forest model is constructed.

[0083] Corresponding to the above embodiment, the second feature vector has the same dimension as the first feature vector.

[0084] Exemplarily, for each type of drone, no less than 1000 groups of valid samples are collected to ensure the sufficiency of the training data. The random forest algorithm is used to construct 100 decision trees, the Gini coefficient is used as the node splitting criterion, and 4 features (about the square root of the total number of features) are randomly selected from the second feature vector to participate in the calculation when each node splits. The maximum tree depth is set to 20 layers, and the minimum number of samples in the leaf nodes is 5. During the training process, the bootstrap sampling method is used to generate the training set. For each decision tree, 80% of the samples are randomly selected for training, and the remaining 20% are used for verification to construct a random forest model.

[0085] As a preferred implementation manner, in step S2, using the pre-constructed random forest model, classifying the first feature vector and matching it with the known drone information to obtain the first model number and the first serial number of the drone, including:

[0086] Input the first feature vector into each decision tree in the pre-constructed random forest model, and each decision tree classifies independently to obtain a number of predicted model numbers and predicted serial numbers;

[0087] According to the number of predicted model numbers and the predicted serial numbers, use the majority voting method to obtain the first model number and the first serial number of the drone.

[0088] In some preferred implementation manners, after inputting the first feature vector into each decision tree in the pre-constructed random forest model, and each decision tree classifies independently to obtain a number of predicted model numbers and predicted serial numbers, step S2 further includes:

[0089] According to the number of predicted model numbers and the predicted serial numbers, obtain the confidence level of the prediction result;

[0090] When the confidence level is not greater than the preset confidence level threshold, continue to collect the spectrum signal, and re-predict in combination with the newly collected spectrum signal until the confidence level of the prediction result is greater than the preset confidence level threshold.

[0091] Exemplarily, input the first feature vector into the pre-constructed random forest model. During the recognition process, first check the first feature vector to ensure that it meets the model input requirements. Then input the first feature vector into each decision tree in the random forest model, and layer by layer down through feature comparison until reaching the leaf node to obtain the classification result of a single tree. Each decision tree independently completes the classification process and outputs the predicted drone model number and serial number.

[0092] The classification result is determined by the majority voting method, and the category predicted by the most decision trees in the random forest model is used as the final recognition result. During the voting process, the prediction results of 100 decision trees are counted, and the model number and serial number with the highest frequency are selected as the final recognition result. At the same time, the confidence level of the prediction result is calculated, that is, the proportion of the number of decision trees predicting this category to the total number of trees. When the confidence level corresponding to the highest number of votes is lower than 60%, this recognition result is marked as a low-confidence result. To improve the recognition reliability, for the recognition results with low confidence, the system will continuously collect data for re-recognition within the next 3 seconds until a high-confidence recognition result is obtained.

[0093] As a preferred implementation manner, in step S3, according to the coordinates of at least three measurement stations, the time difference of arrival positioning technology and the angle of arrival measurement technology are used to calculate the three-dimensional coordinates of the UAV, which is specifically executed through steps S31 - S34:

[0094] S31. Obtain the coordinates of at least three measurement stations, and use the time difference of arrival positioning technology to measure the time difference of the UAV signal arriving at the at least three measurement stations;

[0095] S32. Use the angle of arrival measurement technology to measure the angle information of the UAV relative to the at least three measurement stations;

[0096] S33. According to the time difference and the angle information, establish a positioning equation set of the UAV;

[0097] S34. According to the signal strength received by the at least three measurement stations, constrain the positioning equation set, solve the positioning equation set, and obtain the three-dimensional coordinates of the UAV.

[0098] In the above solution, by combining the time difference of arrival positioning technology and the angle of arrival measurement technology, and at the same time using the received signal strength for assisted positioning, the three-dimensional positioning accuracy of the UAV is improved, and the average positioning error is reduced.

[0099] In some preferred embodiments, step S31, obtaining the coordinates of at least three measurement stations and using the time difference of arrival positioning technology to measure the time difference of the UAV signal arriving at the at least three measurement stations, includes:

[0100] Obtain the coordinates of N measurement stations, where the coordinates of the i-th measurement station are (x i , y i , z i ); N≥3, i = 1, 2,..., N;

[0101] Calculate the time difference of the UAV signal arriving at the i-th measurement station through τ i1 =(d i -d1) / c; where, d iis the distance from the drone to the i-th measurement station, (x, y, z) are the three-dimensional coordinates of the drone to be calculated; (x i , y i , z i ) are the coordinates of the i-th measurement station; c is the electromagnetic wave propagation speed.

[0102] In some preferred embodiments, in step S32, the angle measurement technique is adopted to measure the angle information of the drone relative to the at least three measurement stations, including:

[0103] By measuring the angle information of the drone relative to the i-th measurement station;

[0104] where α i is the azimuth angle information of the drone relative to the i-th measurement station; w α is the measurement noise; y - y i and x - x i are respectively the relative positions of the target relative to the i-th measurement station on the y-axis and the x-axis; β i is the pitch angle information of the drone relative to the i-th measurement station; z - z i is the height difference of the target relative to the measurement station.

[0105] It can be understood that gives the angle between the horizontal direction pointing from the measurement station to the target and the positive direction of the x-axis, is the distance from the target to the measurement station on the horizontal plane, gives the angle between the line connecting the measurement station and the target and the horizontal plane.

[0106] Furthermore, preferably, in step S33, according to the time difference and the angle information, a positioning equation set of the drone is established, including:

[0107] According to the time difference, a first positioning equation is established

[0108] According to the azimuth angle, a second positioning equation (y - y i ) = (x - x i ) · tan(α i );

[0109] According to the pitch angle, a third positioning equation is established

[0110] According to the first positioning equation, the second positioning equation and the third positioning equation, the positioning equation set of the drone is obtained;

[0111] Among them, (x, y, z) are the three-dimensional coordinates of the UAV to be calculated; (x i , y i , z i ) are the coordinates of the i-th measurement station; c is the electromagnetic wave propagation speed; τ i1 is the time difference between the UAV signal arriving at the i-th measurement station and the 1st measurement station; α i is the azimuth angle of the UAV relative to the i-th measurement station; β i is the pitch angle of the UAV relative to the i-th measurement station.

[0112] Preferably, step S34, constraining the positioning equations according to the signal strengths received by the at least three measurement stations, and solving the positioning equations to obtain the three-dimensional coordinates of the UAV, includes:

[0113] Based on the signal strengths received by the at least three measurement stations, establish an RSSI-based distance estimation constraint Use the distance estimation constraint to constrain the equations;

[0114] Solve the positioning equations to obtain the three-dimensional coordinates of the UAV;

[0115] Among them, d i is the actual distance from the UAV to the i-th measurement station; d0 is the reference distance; P0 is the signal power received at the reference distance d0; P i is the signal power received by the i-th measurement station; η is a preset path loss exponent.

[0116] It should be noted that according to the signal strengths received by each measurement station, a log-distance path loss model can be established Among them, P i is the signal power (dBm) received by the i-th measurement station, P0 is the received power at the reference distance d0, where d0 is usually measured and calibrated at 1 meter; η is the path loss exponent, and its value varies with the environment. When propagating in free space, η≈2, in an urban environment η≈3.5 - 4, and when there is occlusion, η can reach 4 - 6; w P is the measurement noise; d i is the actual distance from the UAV to the i-th measurement station, that is

[0117] By transforming the log-distance path loss model, the distance estimation constraint described in the embodiments of the present invention can be obtained

[0118] The distance estimation constraint converts the received signal strength into a distance estimate, and as a supplementary constraint condition for the positioning equations, the constraint on the possible position range of the UAV can be expressed as where ε RSSI is the tolerance range determined according to the RSSI measurement accuracy.

[0119] It should be noted that the value of the tolerance range takes into account the signal strength measurement accuracy of the receiver, the influence degree of the environmental multipath effect, the gain change brought by the antenna directivity, the influence of the atmospheric conditions on the signal propagation, etc. In practical applications, the tolerance range usually takes 5 - 10 meters and can be dynamically adjusted according to the on-site test results.

[0120] By introducing the distance estimation constraint, the positioning accuracy of the UAV can be effectively improved, especially in the case where the time difference positioning and the angle of arrival measurement accuracy are not high.

[0121] As a preferred implementation manner, in step S4, the Doppler effect is used to calculate the radial velocity of the UAV relative to each of the measurement stations, and the three-dimensional velocity vector of the UAV is calculated according to the radial velocity, including:

[0122] Using the Doppler effect to obtain the frequency offset;

[0123] According to the frequency offset, through calculate the radial velocity v of the UAV relative to each of the measurement stations r , and obtain the radial velocity measurement results of each of the measurement stations; where c is the electromagnetic wave propagation speed; Δf d is the frequency offset; f t is the original frequency of the signal emitted by the UAV;

[0124] Combining the radial velocity measurement results of each of the measurement stations, calculate the three-dimensional velocity vector of the UAV.

[0125] It should be noted that based on the Doppler effect principle, the relationship between the received frequency and the transmitted frequency can be expressed as where Δf d is the Doppler frequency shift, f r is the received frequency, f t is the transmitted frequency, v r is the radial velocity (a positive value indicates that the target is moving away from the receiving station, and a negative value indicates that the target is approaching the receiving station), and c is the electromagnetic wave propagation speed. Further, the radial velocity can be obtained. It can be seen that by performing spectral analysis on the UAV signals received by each receiving station and measuring the frequency offset Δf d , the radial velocity of the UAV relative to each receiving station can be calculated.

[0126] Fusing the radial velocity information measured by multiple receiving stations, calculate the velocity vector of the UAV in three-dimensional space. Assuming there are N receiving stations, for the i-th receiving station, the measured radial velocity v r,iThe relationship with the three-dimensional velocity components (v x , v y , v z ) of the UAV can be expressed as v r,i = v x cos(α i )cos(β i ) + v y sin(α i )cos(β i ) + v z sin(β i ); where α i is the azimuth angle and β i is the pitch angle.

[0127] Combining the equations of N measurement stations, the matrix V r = HV can be obtained; where V r is the radial velocity vector measured by N receiving stations, that is, V r = [v r,1 , v r,2 , …, v r,N T , V is the three-dimensional velocity vector to be solved, that is, V = [v x , v y , v z T , H is the observation matrix composed of the geometric position relationship of the measurement stations, where h i1 = cos(α i )cos(β i ), h i2 = sin(α i )cos(β i ), h i3 = sin(β i ).

[0128] Solving this overdetermined system of equations using the least squares method gives From this, the three-dimensional velocity vector (v x , v y , v z ) of the UAV can be obtained.

[0129] As a preferred implementation manner, step S5, according to the three-dimensional coordinates of the UAV and the three-dimensional velocity vector, using the particle filter algorithm to predict the change trend of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window, includes:

[0130] ​​Initialize a particle swarm in the state space and assign a preset initial weight to each particle in the particle swarm; the particle swarm represents the position and velocity states of the UAV.

[0131] Obtain a real-time observation vector according to the three-dimensional coordinates of the UAV and the three-dimensional velocity vector.

[0132] Adjust the weight of the particle based on the uniform motion model according to the change of the real-time observation vector.

[0133] Predict the change trend of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window according to the weight of the particle adjusted in real time.

[0134] In a preferred embodiment, about 1000 particles are initialized in the state space. Each particle contains position and velocity information and is assigned the same initial weight of 1 / 1000. The particle state is dynamically evolved through the prediction update iteration process, where the prediction step calculates the state transition based on the uniform motion model, and its state transition equation is where Δt is the sampling time interval, taken as 0.1 second, (x k , y k , z k ) are the three-dimensional coordinates of the UAV, v k is the velocity of the UAV obtained according to the three-dimensional velocity vector of the UAV; w k is Gaussian white noise with a mean of 0 and a covariance matrix of Q; is the process noise variance of the x position component, is the process noise variance of the y position component, is the process noise variance of the z position component, is the process noise variance of the velocity component.

[0135] The update step adjusts the particle weight according to the observation data, and its observation equation is where d TDOA is the distance difference obtained by TDOA measurement, α AOA and β AOA are the azimuth angle and elevation angle obtained by AOA measurement respectively, v Doppler is the radial velocity obtained by Doppler measurement, d RSSI is the distance estimated by RSSI measurement, h(·) is the observation function, X k is the state transition vector, ∈ k is the observation noise vector, which follows a multi-dimensional Gaussian distribution with a mean of 0 and a covariance matrix of R. is the variance of TDOA measurement, is the variance of azimuth angle measurement, is the variance of pitch angle measurement, is the variance of Doppler velocity measurement, is the variance of RSSI measurement.

[0136] The particle weight update uses Gaussian likelihood to calculate, where, is the weight of the i-th particle at the k-th moment, is the weight of this particle at the previous moment, Z k is the actual observation value vector, is the predicted observation value corresponding to the i-th particle, R is the observation noise covariance matrix, represents the error between the actual observation value and the predicted observation value.

[0137] Finally, based on the state vectors and weights of all particles, the weighted average is calculated as the current state estimate of the UAV. Combining the historical state sequence and the kinematic model, the evolution trend of the position coordinates and three-dimensional velocity components of the UAV within the t + 3s time window is predicted.

[0138] A UAV monitoring method in a complex electromagnetic environment provided by an embodiment of the present invention can accurately identify multiple UAVs through a pre-constructed random forest model, and can improve the three-dimensional positioning accuracy of the UAV by combining TDOA and AOA technologies. In addition, the three-dimensional coordinates and three-dimensional velocity vectors of the UAV are predicted through a particle filter algorithm, improving the continuity and stability of monitoring. Using the embodiment of the present invention can effectively improve the recognition accuracy, positioning accuracy and anti-interference ability of UAV monitoring in a complex electromagnetic environment.

[0139] An embodiment of the present invention provides a UAV monitoring system in a complex electromagnetic environment. Please refer to Figure 2 , the UAV monitoring system in the complex electromagnetic environment includes a feature extraction module 11, a UAV information matching module 12, a coordinate calculation module 13, a velocity vector calculation module 14 and a trend prediction module 15, where:

[0140] The feature extraction module 11 is used to extract signal features from the spectrum signal collected in real time to obtain a first feature vector; the frequency of the spectrum signal is within a preset frequency band;

[0141] The UAV information matching module 12 is used to classify the first feature vector by using a pre-constructed random forest model and match it with known UAV information to obtain the first model and the first serial number of the UAV;

[0142] The coordinate calculation module 13 is used to calculate the three-dimensional coordinates of the UAV according to the coordinates of at least three measurement stations by using the time difference of arrival positioning technology and the angle of arrival measurement technology;

[0143] A velocity vector calculation module 14, which is used to calculate the radial velocity of the UAV relative to each of the measurement stations by using the Doppler effect, and calculate the three-dimensional velocity vector of the UAV according to the radial velocity;

[0144] A trend prediction module 15, which is used to predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window by using a particle filter algorithm according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV.

[0145] As a preferred implementation manner, the feature extraction module 11 is specifically used for:

[0146] Collect spectral signals in real time according to a preset frequency band;

[0147] Extract key features of the spectral signals, and calculate the mean and standard deviation of the key features to obtain an initial feature vector; the key features include at least one of signal intensity, bandwidth, center frequency, signal duration, peak power spectral density, spectral kurtosis, and skewness;

[0148] Perform normalization processing on the initial feature vector to obtain a first feature vector.

[0149] As a preferred implementation manner, the steps of pre-constructing the random forest model include:

[0150] Collect characteristic data of a plurality of UAVs under different environmental conditions and working states; the characteristic data includes the second feature vector, the second model, and the second serial number of the UAV;

[0151] Construct n decision trees by using a random forest algorithm, where the node splitting criterion of the decision tree is the Gini coefficient, and when each node splits, it is randomly selected from the second feature vector;

[0152] Use the second feature vector as the input, and use the second signal and the second serial number as the output to train the classification ability of each of the decision trees;

[0153] Construct a random forest model according to the trained n decision trees.

[0154] As a preferred implementation manner, the UAV information matching module 12 is specifically used for:

[0155] Input the first feature vector into each decision tree in the pre-constructed random forest model, and each decision tree performs independent classification to obtain a plurality of predicted models and predicted serial numbers;

[0156] According to the plurality of predicted models and the predicted serial numbers, use the majority voting method to obtain the first model and the first serial number of the UAV.

[0157] As a preferred embodiment, the coordinate calculation module 13 includes:

[0158] A time difference calculation unit, configured to obtain the coordinates of at least three measurement stations, and measure the time difference between the arrival of the UAV signal at the at least three measurement stations by using the time difference of arrival positioning technology;

[0159] An angle information measurement unit, configured to measure the angle information of the UAV relative to the at least three measurement stations by using the angle of arrival measurement technology;

[0160] An equation set establishment unit, configured to establish a positioning equation set of the UAV according to the time difference and the angle information;

[0161] An equation set solving unit, configured to constrain the positioning equation set according to the signal strength received by the at least three measurement stations, solve the positioning equation set, and obtain the three-dimensional coordinates of the UAV.

[0162] Further preferably, the angle information includes an azimuth angle and a pitch angle; the equation set establishment unit is specifically configured to:

[0163] Establish a first positioning equation according to the time difference

[0164] Establish a second positioning equation (y - y i ) = (x - x i ) · tan(α i ) according to the azimuth angle;

[0165] Establish a third positioning equation according to the pitch angle

[0166] Obtain the positioning equation set of the UAV according to the first positioning equation, the second positioning equation, and the third positioning equation;

[0167] Wherein, (x, y, z) are the three-dimensional coordinates of the UAV to be calculated; (x i , y i , z i ) are the coordinates of the i-th measurement station; c is the electromagnetic wave propagation speed; τ i1 is the time difference between the arrival of the UAV signal at the i-th measurement station and the first measurement station; α i is the azimuth angle of the UAV relative to the i-th measurement station; β i is the pitch angle of the UAV relative to the i-th measurement station.

[0168] Preferably, the equation set solving unit is specifically configured to:

[0169] Based on the signal strengths received by the at least three measurement stations, establish RSSI-based distance estimation constraints Use the distance estimation constraints to constrain the system of equations;

[0170] Solve the positioning system of equations to obtain the three-dimensional coordinates of the UAV;

[0171] where d i is the actual distance from the UAV to the i-th measurement station; d0 is the reference distance; P0 is the signal power received at the reference distance d0; P i is the signal power received at the i-th measurement station; η is a preset path loss exponent.

[0172] As a preferred implementation, the velocity vector calculation module 14 is specifically configured to:

[0173] Use the Doppler effect to obtain the frequency offset;

[0174] According to the frequency offset, through calculate the radial velocity v of the UAV relative to each of the measurement stations r , to obtain the radial velocity measurement results of each of the measurement stations; where c is the electromagnetic wave propagation speed; Δf d is the frequency offset; f t is the original frequency of the signal emitted by the UAV;

[0175] Combine the radial velocity measurement results of each of the measurement stations to calculate the three-dimensional velocity vector of the UAV.

[0176] As a preferred implementation, the trend prediction module 15 is specifically configured to:

[0177] Initialize a particle swarm in the state space and assign a preset initial weight to each particle in the particle swarm; the particle swarm represents the UAV position and velocity states;

[0178] According to the three-dimensional coordinates of the UAV and the three-dimensional velocity vector, obtain the real-time observation vector;

[0179] According to the change of the real-time observation vector, adjust the weight of the particle based on the uniform motion model;

[0180] According to the weight of the particle adjusted in real time, predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window.

[0181] An unmanned aerial vehicle (UAV) monitoring system in a complex electromagnetic environment provided by an embodiment of the present invention can accurately identify various UAVs through a pre-constructed random forest model. Combining the TDOA and AOA technologies can improve the three-dimensional positioning accuracy of UAVs. In addition, the three-dimensional coordinates and three-dimensional velocity vectors of UAVs are predicted through a particle filter algorithm, improving the continuity and stability of monitoring. Using the embodiment of the present invention can effectively improve the recognition accuracy, positioning accuracy and anti-interference ability of UAV monitoring in a complex electromagnetic environment.

[0182] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), or the like.

[0183] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art in the technical field, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A method for monitoring unmanned aerial vehicles (UAVs) in a complex electromagnetic environment, characterized in that, Including: Extract signal features from the spectrum signal collected in real time to obtain a first feature vector; the frequency of the spectrum signal is within a preset frequency band; Use a pre-constructed random forest model to classify the first feature vector and match it with known drone information to obtain the first model and the first serial number of the drone; According to the coordinates of at least three measurement stations, use the time difference of arrival positioning technology and the angle of arrival measurement technology to calculate the three-dimensional coordinates of the drone; Use the Doppler effect to calculate the radial velocity of the drone relative to each of the measurement stations, and calculate the three-dimensional velocity vector of the drone according to the radial velocity; According to the three-dimensional coordinates and the three-dimensional velocity vector of the drone, use the particle filter algorithm to predict the change trend of the three-dimensional coordinates and the three-dimensional velocity vector of the drone within a preset time window.

2. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 1, wherein, The extracting signal features from the spectrum signal collected in real time to obtain a first feature vector includes: Collect spectrum signals in real time according to the preset frequency band; Extract the key features of the spectrum signal, and calculate the mean and standard deviation of the key features to obtain an initial feature vector; the key features include at least one of signal strength, bandwidth, center frequency, signal duration, peak power spectral density, spectral kurtosis, and skewness; Perform normalization processing on the initial feature vector to obtain a first feature vector.

3. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 1, wherein The steps of pre-constructing the random forest model include: Collect the feature data of several drones under different environmental conditions and working states; the feature data includes the second feature vector, the second model, and the second serial number of the drone; Use the random forest algorithm to construct n decision trees, the node splitting criterion of the decision tree is the Gini coefficient, and when each node splits, it is randomly selected from the second feature vector; Use the second feature vector as the input, and use the second signal and the second serial number as the output to train the classification ability of each decision tree; According to the trained n decision trees, construct a random forest model.

4. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 1, wherein, The using a pre-constructed random forest model to classify the first feature vector and match it with known drone information to obtain the first model and the first serial number of the drone includes: Input the first feature vector into each decision tree in the pre-constructed random forest model, and each decision tree classifies independently to obtain a number of predicted models and predicted serial numbers; According to the number of predicted models and the predicted serial numbers, use the majority voting method to obtain the first model and the first serial number of the drone.

5. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 1, wherein The according to the coordinates of at least three measurement stations, using the time difference of arrival positioning technology and the angle of arrival measurement technology to calculate the three-dimensional coordinates of the drone includes: Obtain the coordinates of at least three measurement stations, and use the time difference of arrival positioning technology to measure the time difference of the drone signal arriving at the at least three measurement stations; Use the angle of arrival measurement technology to measure the angle information of the drone relative to the at least three measurement stations; Establish a positioning equation set of the drone according to the time difference and the angle information; Constraining the positioning equation set according to the signal strengths received by the at least three measurement stations, solving the positioning equation set, and obtaining the three-dimensional coordinates of the UAV.

6. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 5, wherein The angle information includes azimuth angle and elevation angle; establishing the positioning equation set of the UAV according to the time difference and the angle information includes: Establish a first positioning equation according to the time difference According to the azimuth, establish a second positioning equation (y - y i ) = (x - x i )·tan(α i ); Establish a third positioning equation according to the pitch angle Obtaining the positioning equation set of the UAV according to the first positioning equation, the second positioning equation, and the third positioning equation; Among them, (x, y, z) are the three-dimensional coordinates of the UAV to be calculated; (x i , y i , z i ) are the coordinates of the i-th measurement station; c is the electromagnetic wave propagation speed; τ i1 is the time difference between the UAV signal reaching the i-th measurement station and the first measurement station; α i is the azimuth angle of the UAV relative to the i-th measurement station; β i is the pitch angle of the UAV relative to the i-th measurement station.

7. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 5, characterized in that, Constraining the positioning equation set according to the signal strengths received by the at least three measurement stations, solving the positioning equation set, and obtaining the three-dimensional coordinates of the UAV, including: Based on the signal strengths received by the at least three measurement stations, establish RSSI-based distance estimation constraints Use the distance estimation constraints to constrain the system of equations; Solving the positioning equation set to obtain the three-dimensional coordinates of the UAV; Among them, d i is the actual distance from the UAV to the i-th measurement station; d0 is the reference distance; P0 is the signal power received at the reference distance d0; P i is the signal power received at the i-th measurement station; η is the preset path loss exponent.

8. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 1, wherein, Calculating the radial velocity of the UAV relative to each of the measurement stations by using the Doppler effect, and calculating the three-dimensional velocity vector of the UAV according to the radial velocity, including: Obtaining the frequency offset by using the Doppler effect; According to the frequency offset, by calculating the radial velocity v of the UAV relative to each of the measurement stations r , obtaining the radial velocity measurement results of each of the measurement stations; where c is the electromagnetic wave propagation speed; Δf d is the frequency offset; f t is the original frequency of the signal transmitted by the UAV; Calculating and obtaining the three-dimensional velocity vector of the UAV by combining the radial velocity measurement results of each of the measurement stations.

9. The method for monitoring an unmanned aerial vehicle in a complex electromagnetic environment according to claim 1, characterized in that, Predicting the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window by using the particle filter algorithm according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV, including: Initializing a particle swarm in the state space and assigning a preset initial weight to each particle in the particle swarm; the particle swarm represents the position and velocity states of the UAV; Obtaining a real-time observation vector according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV; Adjusting the weights of the particles based on a uniform motion model according to the change of the real-time observation vector; Predicting the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window according to the weights of the particles adjusted in real time.

10. An unmanned aerial vehicle monitoring system under a complex electromagnetic environment, characterized in that, Including: A feature extraction module, configured to extract signal features from the spectrumsignal collected in real time to obtain a first feature vector; the frequency of the spectrum signal is within a preset frequency band; A UAV information matching module, configured to classify the first feature vector by using a pre-constructed random forest model and match it with known UAV information to obtain the first model and the first serial number of the UAV; A coordinate calculation module, configured to calculate the three-dimensional coordinates of the UAV by using the time difference of arrival positioning technology and the angle of arrival measurement technology according to the coordinates of at least three measurement stations; A velocity vector calculation module, configured to calculate the radial velocity of the UAV relative to each of the measurement stations by using the Doppler effect and calculate the three-dimensional velocity vector of the UAV according to the radial velocity; A trend prediction module, configured to predict the change trends of the three-dimensional coordinates and the three-dimensional velocity vector of the UAV within a preset time window by using the particle filter algorithm according to the three-dimensional coordinates and the three-dimensional velocity vector of the UAV.

Images