Multi-station unmanned aerial vehicle track batch processing method in complex background noise environment

Abstract

The invention discloses a multi-station unmanned aerial vehicle track batch processing method in a complex background noise environment. The method is operated based on a radar network comprising a plurality of low-altitude monitoring radars and a data center. Firstly, a Cartesian coordinate system used for measuring a detected airspace is established, three-dimensional coordinates of all echo points detected by radars are determined, then all the echo points detected by the low-altitude detection radars in a period time are superposed into a blank frame according to original positions, and a comprehensive echo frame is generated. Pairing the echo points belonging to the two adjacent comprehensive echo frames in pairs, selecting the echo points which may reflect the real position of the target through a Hough transformation voting mechanism, and building a corresponding observation vector according to the selected echo points; and then the observation vectors are input into an LSTM network for analysis and screening, which are track features and which are noise are judged, the noise is eliminated, the track features are reserved and output, and then a final track result is obtained through FCN.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicle (UAV) track initiation, and in particular relates to a multi-station UAV track initiation batch processing method under a complex background noise environment. Background Art

[0002] With the widespread use of drones in logistics, delivery, aerial photography, and other fields, low-altitude monitoring and control have become increasingly important. The initial track, which reflects the path a drone takes from takeoff to entering its intended route, is a crucial step in low-altitude monitoring. Derivation of the track relies on processing the chronologically ordered target positions. Efficient and accurate track initiation processing is crucial for continuous monitoring of target trajectories. Currently, drone monitoring is mostly performed using traditional single-station radar monitoring. Initial track processing methods primarily include the logical initiation method, which uses logical judgment to select valid target points and establish an initial track; the multi-hypothesis correlation method, which performs unilateral multi-hypothesis correlation on radar point data to form a temporary track and gradually verify its validity; and the Hough transform method, which uses the Hough transform to extract linear / curved features from radar or visual data. However, drones often operate in complex environments during takeoff. Echoes from objects such as buildings and trees, as well as complex terrain, can add noise to radar echoes, significantly impacting continuous radar monitoring of drones. In addition, due to the strong maneuverability of UAVs, when using the above-mentioned traditional track processing methods to perform initial batch processing of UAV tracks, there will be problems such as long track processing cycle and low detection probability.

[0003] With the widespread use of low-altitude monitoring radars and the development of the internet, radar networks consisting of multiple radar stations and data centers can achieve multi-angle monitoring of targets by operating simultaneously at multiple stations and uploading monitoring data to the data center in real time. This effectively avoids incomplete monitoring data from a single station due to distance or azimuth limitations. The potential information contained in multi-station monitoring data can also reduce noise, that is, the impact of ground object echoes on track processing. However, monitoring data measured by multi-station radars scanning the airspace at different times and wavelengths still needs to be processed using an effective data processing method to derive tracks. Only then can radar networks realize these advantages. Currently, the exploration of the field of drone track monitoring is still in its infancy, and research on track initiation batch processing methods based on multi-station radars has not yet achieved significant results. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-station UAV track batch processing method in a complex background noise environment. The method is based on the echo frame data measured by multi-station radar. The echo points in the echo frame that may reflect the true position of the target are screened out through Hough transform, and a target feature dataset is constructed according to the potential track information contained in these echo points. Then, a long-short term memory (LSTM) network is used to screen out the data set that can truly reflect the target track based on the correlation characteristics. Finally, the track fitting is completed through a fully convolutional network (FCN).

[0005] The technical solution adopted in the present invention is as follows:

[0006] A method for batch processing multi-station UAV tracks in a complex background noise environment is described. The method operates based on a radar network, wherein the radar network includes multiple low-altitude monitoring radars and a data center. The multiple low-altitude monitoring radars are located in different locations but have the same monitoring airspace range. All low-altitude monitoring radars are data-connected to the data center. The method specifically includes the following steps:

[0007] Step 1. Establish a Cartesian coordinate system for measuring the entire measured airspace. Multiple low-altitude surveillance radars simultaneously monitor the airspace and transmit the measured echo frame data to the data center in real time. Each radar is assigned a unique number. After receiving the echo frame data, the data center determines the azimuth coordinates of all echo points appearing in the echo frame based on the Cartesian coordinate system and records them. Each echo point is also assigned a unique number.

[0008] Step 2. The data center periodically superimposes the echo points that appear in all echo frames received within a period of time onto a blank frame. The newly generated frame is called a synthetic echo frame. The period of time is called a cycle, and the length of the cycle is equal to the interval between the generation of two adjacent synthetic echo frames.

[0009] Step 3. Use the Hough transform voting mechanism to screen all echo points in two adjacent integrated echo frames, select the echo point that may reflect the target's true location, and derive the potential track information of the selected echo point based on the coordinates corresponding to the selected echo point. Then, construct an observation vector corresponding to the echo point, which includes the position information and potential track information of the corresponding echo point.

[0010] Step 4. Further analyze the observation vectors to select the observation vectors that can reflect the actual track, and use the track decision device to fit the echo points corresponding to the selected observation vectors to obtain the starting track.

[0011] In the prior art, the processing method for track initiation is mainly based on the echo frame data measured by a single-station radar. Since a single-station radar is easily interfered with by complex terrain conditions, it is difficult to achieve reliable and stable monitoring of low-altitude targets such as drones. A radar network including multiple-station radars can detect the same airspace from different directions. By analyzing the echo frames received by the multiple-station radars, the probability of measuring the true position of the target can be greatly improved, while minimizing the impact of buildings or terrain obstructions or terrain echoes on target detection. However, the data center of the radar network needs a set of effective processing methods for the echo frame data of multiple-station radars to achieve its advantages. The radar network based on the method of the present invention can be used with radars that emit different electromagnetic wave wavelengths and radars with different scanning periods. This is because the method of the present invention processes the track through the position of the echo point. Regardless of the type of radar, as long as the echo frame contains an echo point representing the true position of the target, it can play a role in the final processed track.

[0012] Further optimization, the radar network contains a total of m radars, m ≥ 2, and the echo frame received by the i-th radar in Step 1 is recorded as Rad_img i , 1≤i≤m; for echo frame Rad_img i All echo points appearing in the echo frame Rad_img are numbered. i All echo points appearing in the frame are represented as Rad_img i ={dot R i_1 ,…,dot R i_P |t s}, where P is the number of echo points in the echo frame, dot R i_1 and dot R i_P are the 1st and Pth echo points detected by the i-th radar, respectively, t s is the moment when the i-th radar receives the echo frame; the azimuth coordinates of any echo point in the echo frame are expressed as {dot R i_j |t s}={pos x ,pos y ,pos z}, where j is the echo point in the echo frame Rad_img i The sequence number in, 1≤j≤P, pos x ,pos y ,pos zThese are the x-axis, y-axis, and z-axis coordinates of the point in the Cartesian coordinate system. The coordinates of the echo point are listed in a set belonging to the point. Then, the echo points contained in each echo frame are listed in a set belonging to the frame. The echo points in the set are numbered sequentially to facilitate subsequent data processing.

[0013] Further optimization, the superposition method of the integrated echo frame in Step 2 is: first find t s All echo frames in the same cycle, and then add each echo point appearing in these echo frames to a blank frame according to the original coordinates, and the resulting frame is the integrated echo frame; the integrated echo frame is recorded as Fusion_img t The integrated echo frame and all the echo points in it are represented as Fusion_img t ={dot F t_1 ,…,dot F t_N |T}, where T is the cycle length, t is the sequence number corresponding to the cycle, N is the number of echo points in the integrated echo frame, and the echo points in the integrated echo frame are recorded as dot F t_k , and record the source information of the echo point, k is the echo point in the integrated echo frame Fusion_img t The sequence number is 1≤k≤N, and the source information includes the Rad_img where the echo point is located. i The corresponding t s The purpose of this step is to classify the echo points into different batches in preparation for subsequent screening. Recording the source information of the echo points is to prepare for the analysis of the potential track information contained in the subsequent echo points.

[0014] For further optimization, the echo points in Step 3 must first be screened by the distance threshold, and then screened by the Hough transform voting mechanism. The specific steps of distance threshold screening are as follows:

[0015] Step 3.1. Set the distance threshold dis_ threshold , the current integrated echo frame is Fusion_img t , the previous integrated echo frame is Fusion_img t-1 , Fusion_img t The middle echo point is recorded as dot F t_k , Fusion_img t-1 The middle echo point is recorded as dot F t-1_k ;

[0016] Step 3.2. Get Fusion_imgt Zhongyi echo dot F t_k , calculate dot F t_k with Fusion_img t-1 Each echo point F t-1_k The distance between The calculation formula is:

[0017]

[0018] where pos x_t ,pos y_t ,pos z_t dot F t_k The corresponding x-axis, y-axis, and z-axis coordinates in the Cartesian coordinate system, pos x_t-1 ,pos y_t-1 ,pos z_t-1 dot F t-1_k The corresponding x-axis, y-axis, and z-axis coordinates in the Cartesian coordinate system;

[0019] If dot F t_k 、dot F t-1_k The distance between two points is less than the distance threshold, that is, If the condition is met, the dot F t_k 、dot F t-1_k Include point set Collection_Dot t middle;

[0020] When Fusion_img t-1 All echo points in the complete and current dot F t_k After calculating the distance, select another dot F t_k Repeat the above process until Fusion_img t All echo points in the calculation are completed.

[0021] The purpose of screening the distance threshold is to preliminarily select echo points related to the track in the integrated echo frame. The target of the monitoring method of the present invention is a drone. Based on the maximum flight speed, the maximum distance that the drone may move in the integrated echo frame period interval can be obtained. If the distance between an echo point and all echo points in the adjacent echo frames is very far, the possibility that the echo point represents a drone can be ruled out.threshold It is an empirical parameter and is determined by the operator based on the target situation and specific monitoring environment.

[0022] Further optimization, the screening of echo points by the Hough transform voting mechanism in Step 3 includes the following steps:

[0023] Step 3.3. The specific steps of selecting echo points through the Hough transform voting mechanism are as follows:

[0024] Step 3.3.1. Assume that there is a point M in the Cartesian coordinate system. The equation through point M can be obtained as follows:

[0025] ρ=pos x_m cosθcosφ+pos y_m sinθcosφ+pos z_m sinφ (2)

[0026] where pos x_m ,pos y_m ,pos z_m are the x-axis, y-axis, and z-axis coordinates of point M, respectively. The projection of the line OM between point M and the origin on the xy plane is l, the angle between l and the positive direction of the x-axis is θ, the angle between the line OM and the xy plane is φ, and ρ is the length of the line OM. Replace pos in formula (2) with x_m ,pos y_m ,pos z_m By replacing the corresponding independent variables x, y, and z, we can get the equation of the line in the Cartesian coordinate system:

[0027] ρ=xcosθcosφ+ysinθcosφ+zsinφ (3)

[0028] This line also represents a point with coordinates (ρ, θ, φ) in Hough space;

[0029] Step 3.3.2. Vote for the points in the Hough space to determine the line that passes through the largest number of echo points, all of which belong to Collection_Dot t The specific steps for voting are:

[0030] a. Divide the range [-90°, 90°] corresponding to θ into multiple parts through the first type of segmentation points, where the first type of segmentation points are the angle values ​​of θ. Divide the range [0, 180°] corresponding to φ into multiple parts through the second type of segmentation points, where the second type of segmentation points are the angle values ​​of φ. Then, combine the first type of segmentation points with the second type of segmentation points in pairs, and substitute each combination into formula (3) to generate a corresponding straight line formula. Then, use Collection_Dot tSubstitute the coordinates of the mid-echo point into each generated straight line formula one by one to calculate the corresponding ρ value; then count the number of occurrences of all ρ values, find the corresponding ρ value with the most repetitions, and the θ and φ in the straight line formula for the ρ value, and determine the point in Hough space by finding the corresponding ρ, θ, and φ values;

[0031] b. Find multiple continuous point sets Collection_Dot t After the straight line with the largest number of echo points passes through the Hough space, there will be multiple points distributed; determine the range of these point distributions, divide ρ within the range into multiple segments of equal length, and divide θ and φ into multiple angles of equal size. The discretization of ρ, θ, and φ divides the entire space within the range into a grid state; then count the number of points contained in each grid, find the grid with the most points and record it as the peak grid; determine the straight line equations corresponding to the points in the peak grid, and find the echo points on these straight line equations.

[0032] Different from the prior art that directly obtains the track through Hough transform voting, the purpose of Hough transform voting in the method of the present invention is to screen echo points, and the Hough transform voting in the method of the present invention includes two voting processes. After the specific coordinates of a certain number of echo points have been obtained, it is a common track processing method to use the Hough transform voting mechanism to screen out a straight line to obtain the track of the target at this stage. However, the existing track processing method transforms the two-dimensional coordinates of the echo points and cannot reflect the pitch motion of the target. The method of the present invention processes the three-dimensional coordinates and can more realistically reflect the target motion trajectory. The Hough transform voting mechanism is based on statistical principles. Multiple points with determined coordinates within a certain range can determine multiple straight lines. The voting mechanism counts which straight line has the most points. Because the target's motion trajectory is continuous, the trajectory within a short distance can be approximately regarded as a straight line. Therefore, the straight line that passes through more points is closer to the target's true trajectory. For the method of the present invention, each point set Collection_Dot t Each Hough transform vote can produce a straight line representing the track. After obtaining multiple straight lines, further processing is required, and the processing method is to vote and filter again. The screening method is to transform these lines into points in Hough space, then divide the entire Hough space into grids and count the number of points in each grid. The division method is also an empirical method. This screening method is essentially a quantification of the coincidence of lines. If many points are located in the same grid, it means that the corresponding lines of these points have a high degree of coincidence. The points corresponding to the actual track can basically be determined to be within this grid, further narrowing the scope of the actual track.

[0033] Further optimization, the specific process of forming the observation vector in Step 3 is:

[0034] Step 3.4. Use the echo points found in Step 3.3 to form a dot for each echo point. F t_k The corresponding observation vector, the elements of which include dot F t_k In the Cartesian coordinate system, dot F t_k The Hough space coordinates corresponding to the line and the target in dot F t_k The speed, heading angle and pitch angle at the target are UAVs for which initial track information needs to be obtained; the speed, heading angle and pitch angle are calculated by dot F t_k and dot F t_k Dots on the same line F t-1_k The Cartesian coordinates of two points are calculated, dot F t_k The velocity is expressed as The heading angle is expressed as Where ξ is the heading offset, which is determined by the specific heading reference direction and the target heading. The pitch angle is expressed as Echo Dot F t_k The observation vector at is denoted as Input_dot F t_k ={pos x_t ,pos y_t ,pos z_t ,ρ k ,θ k ,φ k ,V k avg ,Yaw k ,Pitch k}, where Δt is dot F t_k Rad_img i Corresponding to t s with dot F t-1_k Rad_img i Corresponding to t s The difference, ρ k ,θ k 、φ k Indicates dot F t_k The Hough space coordinates corresponding to the line equation.

[0035] The purpose of constructing observation vectors is to discover the potential track information contained in these filtered echo points for further screening. The inclusion of multiple elements in the observation vector, including position and motion information, is to provide multiple bases for further screening and increase the accuracy of the final track.

[0036] For further optimization, in Step 4, the LSTM network is used to filter the observation vector that can reflect the real track. The track decision device is FCN. The specific steps for obtaining the starting track are as follows:

[0037] Step 4.1. Input the observation vector into the LSTM network. The LSTM network learns the elements contained in the observation vector to acquire the ability to identify the temporal characteristics of the real track. It then analyzes and filters all input observation vectors, defines the observation vectors that can reflect the real track as track features and retains them, and defines the observation vectors other than track features as noise and eliminates them. The track features are then output as the result, which is recorded as Output_dot. F t_k ;

[0038] Step 4.2. Output_dot F t_k Then input it into FCN, and combine it with Output_dot through the deconvolution function of FCN F t_k The position information and track information contained in Output_dot F t_k Corresponding dot F t_k and dot F t_k Dots on the same line F t-1_k Fit the track line to obtain the track starting batch processing results.

[0039] The LSTM network is a neural network with a gate structure, including a forget gate, a candidate gate, an input gate, and an output gate. Its specific operation process is shown in the following formula:

[0040] f t =min(σ(W f [h t-1 ,x t ]+b f ), alpha)

[0041] O t =σ(W o [h t-1 ,x t ]+b o )

[0042] i t =σ(W i [h t-1 ,x t ]+b i ),

[0043] C` t =tanh(W c [h t-1 ,x t ]+b h )

[0044] C t =f t *C t-1 +i t *C` t

[0045] h t =O t *tanh(C t )

[0046] where f t 、C` t 、i t , O t They are forget gate, candidate gate, input gate, output gate, W f 、W c 、W i 、W o They are respectively the forget gate weight matrix, the candidate gate weight matrix, the input gate weight matrix, and the output gate weight matrix, b f 、b i 、b o They are the forget gate bias, input gate bias, and output gate bias respectively. t is the input value, h t With h t-1 They are the current output value and the previous output value, σ is the sigmoid activation function, alpha is a hyperparameter that limits the output range of the forget gate to avoid extreme gradient fluctuations, tanh is the activation function that limits the range of the output value to (-1,1), and C t is the memory cell state, used to control the output value h t .

[0047] From the above formula we can see that W f 、W c 、W i 、W o The calculation of the four weight matrices involves not only the input value but also the previous output value, so f t 、C` t 、it , O t Each output result of the four gates will be affected by the previous output result. In other words, W f 、W c 、W i 、W o The output results of the four weight matrices are modified each time. This is the process of the LSTM network learning the real track time series characteristics through the observation vector. t It consists of two parts, one of which is through the forget gate f t The data that does not need to be retained in the previous memory cell state is filtered out, and the other part is the currently retained data. The two parts are linearly superimposed to form C t For track processing, the forget gate f t It can be used to filter out the observation vector containing noise, input gate i t It is used to retain the selected track features and continue the cycle. Because the UAV has the characteristic of continuous motion within a certain time range, while other noises introduced by ground echoes do not have motion characteristics, motion characteristics can be used as an important basis for distinguishing noise. In addition, the comparison of position information and motion information contained in multiple observation vectors can also be used as another basis for multiple identification to increase accuracy. Using tanh as the activation function can control the influence of the output value of the previous calculation on the next calculation. t If the value is negative, it means that the output is not allowed and should be suppressed. If the value is positive, it means that the output is reasonable. t If Y is positive, t =h t , change Y t As the final output result after all observation vectors are analyzed and filtered by the LSTM network.

[0048] The deconvolution function of FCN is commonly used in image processing. While preserving the pixel spatial information of the input image, the FCN's deconvolution layer can restore low-resolution feature maps into high-resolution images. The echo points corresponding to the track features filtered by the LSTM network can be viewed as corresponding pixels in the low-resolution feature map, while the position and motion information contained in the track features can be viewed as the spatial information of the pixels. While ensuring that the track features reflect the actual track, the FCN's deconvolution function can be used to fit the echo points corresponding to the track features into a track path that is as close to the actual track as possible, completing the initial batch processing of the tracks.

[0049] The beneficial effects of the method of the present invention are:

[0050] 1. The method of the present invention achieves batch processing of the initial flight paths of small, low-altitude targets such as drones using radar network monitoring data. Compared with existing batch processing methods for initial flight paths, the method of the present invention has better processing effects on the initial flight paths of drones.

[0051] 2. The radar network used in the method of the present invention can include radars that transmit electromagnetic waves of different wavelengths and radars with different scanning periods, which is more flexible in organization and scheduling than traditional radar networks.

[0052] 3. The coordinates of the echo points are all three-dimensional coordinates. The observation vector contains not only the position information of the corresponding echo point, but also the motion information. The position information includes the Cartesian coordinates and the Hough space coordinates corresponding to the line on which it lies, making the echo point screening and track processing results more accurate.

[0053] 4. The combined use of LSTM network and FCN can now achieve better results in track initial batch processing than other data processing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the overall process of the track initialization batch processing method of the present invention. DETAILED DESCRIPTION

[0055] To make the purpose, technical solution, and advantages of the method of the present invention more clear, the technical solution of the present invention will be clearly and completely described below through specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0056] Example 1:

[0057] In this embodiment, the radar network includes two low-altitude monitoring radars and a data center. The two low-altitude monitoring radars are located in different locations but have overlapping monitoring airspaces. All low-altitude monitoring radars are data-connected to the data center. The track initiation batch processing method specifically includes the following steps:

[0058] Step 1. Establish a Cartesian coordinate system for measuring the entire measured airspace, with the z-axis pointing upward, the y-axis pointing westward, and the x-axis pointing northward. The origin of the coordinate system is at sea level, and the coordinate unit is m. Two low-altitude monitoring radars simultaneously monitor the airspace continuously and transmit the measured echo frame data to the data center in real time. The two radar pairs are radar 1 and radar 2. The scanning period of radar 1 is 2 seconds, and the scanning period of radar 2 is 3 seconds. 18 seconds after the radar network is turned on, one echo point appears in the echo frame received by radar 1, and two echo points appear in the echo frame received by radar 2. After determining the coordinates of the three echo points, the echo frame and the corresponding echo point coordinates are labeled Rad_img1 = {dot R 1_1 |18},{dot R 1_1 |18}={19,24,50}; Rad_img2={dot R 2_1 , dot R 2_2 |18},{dot R 2_1 |18}={94,80,7},{dot R 2_2 |18}={30,110,10}, the three numbers in the brackets after the echo point are the x-axis, y-axis, and z-axis coordinates of the point. During the period of 16.5 to 24 seconds, in addition to the above three echo points, another five echo points appear, belonging to the following four echo frames: Rad_img1={dot R 1_1 , dot R 1_2 |20},{dot R 1_1 |20}={13,20,50},{dot R 1_2 |20}={544,20,48}; Rad_img2={dot R 2_1 |21},{dot R 2_1 |21}={100,210,70}; Rad_img1={dot R 1_1 , dot R 1_2 |22},{dot R 1_1 |22}={7,16,50},{dot R 1_2 |22}={20,39,57}.

[0059] Step 2. In the time period of 16.5 to 24 seconds, the echo frame superposition period length T = 2.5 seconds, the first period time period is 16.5 to 19 seconds, the second and third periods are similar. The three integrated echo frames and the echo points contained in the frame are represented as: Fusion_img1 = {dot F 1_1 ,dot F 1_2 ,dot F 1_3 |2.5}, dot F 1_1 ={19,24,50|18}, dot F 1_2 ={94,80,7|18}, dot F 1_3 ={30,110,10|18}; Fusion_img2={dot F 2_1 ,dot F 2_2 ,dot F 2_3 |2.5}, dot F 2_1 ={13,20,50|20}, dot F 2_2 ={544,20,48|20}, dot F 2_3 ={100,210,70|21}; Fusion_img3={dot F 3_1 ,dot F 3_2 |2.5}, dot F 3_1 ={7,16,50|22}, dot F 3_2 ={20,39,57|22}. The coordinates of the point are shown before the vertical line in the curly brackets after the echo point, and the source information of the point is shown after the vertical line. The source information here is only the time when the echo point was detected by the radar.

[0060] Step 3. The specific steps for selecting echo points and forming observation vectors are as follows:

[0061] Step 3.1. Set the distance threshold dis_ threshold =50m.

[0062] Step 3.2. Take Fusion_img2 as the current integrated echo frame and take the echo point dot in Fusion_img2 F2_1 , calculate dot by formula (1) F 2_1 The three echo points in Fusion_img1 are After the calculation is completed, replace it with dot F 2_2 , and then replace it with dot F 2_3 , all calculation results are listed in Table 1.

[0063] Table 1 Distances between echo points in Fusion_img1 and Fusion_img2

[0064]

[0065] From Table 1, we can see that only dot F 2_1 and dot F 1_1 The distance between them is less than the distance threshold dis_ threshold , add the above two points into the point set Collection_Dot2.

[0066] Take Fusion_img3 as the current integrated echo frame and take the echo point dot in Fusion_img3 F 3_1 , calculate dot by formula (1) F 3_1 The three echo points in Fusion_img2 are After the calculation is completed, replace it with dot F 3_2 , all calculation results are listed in Table 2.

[0067] Table 2 Distances between echo points in Fusion_img2 and Fusion_img3

[0068]

[0069] From Table 2 we can see that dot F 2_1 with dot F 3_1 、dot F 3_2 The distance between them is less than the distance threshold dis_ threshold , add the above three points into the point set Collection_Dot3.

[0070] Step 3.3. Determine the straight line passing through the two echo points in Collection_Dot2 by voting on the points in Hough space. In this embodiment, the range [-90°, 90°] corresponding to θ in Hough space is divided into 18,000 equal parts, and the range [0, 180°] corresponding to φ is also divided into 18,000 equal parts. At this time, there are 18,001 values ​​of -90°, -89.99°, -89.98°, ..., 89.99°, and 90° on θ, and 18,001 values ​​of 0, 0.01°, 0.02°, ..., 179.99°, and 180° on φ. Combining the values ​​of θ and φ in pairs can produce a total of 324,036,001 combinations. Substituting them into formula (3) can form 324,036,001 straight line equations. Then, the coordinates of the two echo points in Collection_Dot2 (19, 24, 50) and (13, 20, 50) are substituted into the 324,036,001 straight line equations to calculate the corresponding ρ values. Because this step involves calculating too many numbers, the resulting ρ values ​​are not listed here. Statistics show that ρ = 13.15 appears only twice in all the results. When ρ = 13.15, the corresponding θ and φ are θ = -56.31° and φ = 25.66°, respectively. The resulting Hough space coordinates are (13.15, -56.31°, 25.66°), and the corresponding straight line equation is 30x - 45y + 26z - 789 = 0, indicating that the line passes through the two echo points in Collection_Dot2.

[0071] The same method is then used to determine the lines passing through the two echo points in Collection_Dot3. Although Collection_Dot3 contains three echo points, they are not collinear. Therefore, voting on the points in Collection_Dot3 yields three line equations. The lines passing through (13, 20, 50) and (7, 16, 50) also have Hough space coordinates of (13.15, -56.31°, 25.66°). The lines passing through (13, 20, 50) and (20, 39, 57) have Hough space coordinates of (21.33, -20.22°, 19.07°). The lines passing through (7, 16, 50) and (20, 39, 57) have Hough space coordinates of (3.62, -52.63°, 14.84°).

[0072] At this time, a total of four points appear in the Hough space. The range of Hough space ρ value [3,22] is divided by a length of 1, θ is divided by an angle of 1° in the range of [0,-60°], and φ is also divided by an angle of 1° in the range of [10°,30°]. The statistical results show that there are two points in the grid with ρ value [13,14], θ value [-57°,-56°], and φ value [25°,26°], which is the largest number, so this grid is the peak grid. The coordinates of the two points in the peak grid are both (13.15,-56.31°,25.66°), corresponding to the straight line equation 30x-45y+26z-789=0, and the three echo points with coordinates of (19,24,50), (13,20,50), and (7,16,50), namely dot F 1_1 、dot F 2_1 、dot F 3_1 Located on this straight line.

[0073] Step 3.4. The three selected echo points belong to three different integrated echo frames and can form two observation vectors. F 1_1 with dot F 2_1 The velocity calculated from the coordinates of the two points is V 1 avg ={-3,-2,0}. In this embodiment, the heading reference direction is due north. The target has a velocity in the negative direction of the x-axis, so ξ = 180°. The heading angle is Yaw 1 =213.69°, the pitch angle is 0. dot F 2_1 The corresponding observation vector is established as Input_dot F 2_1 ={13,20,50,13.15,-56.31°,25.66°,-3,-2,0,213.69°,0}. dot F 2_1 with dot F 3_1 The speed, heading angle, and pitch angle calculated from the two-point coordinates are all equal to the above, so the observation vector is established as Input_dot F 3_1 ={7,16,50,13.15,-56.31°,25.66°,-3,-2,0,213.69°,0}.

[0074] Step 4. The specific steps for further analyzing the observation vector and deriving the track are as follows:

[0075] Step 4.1. Input the observation vector into a single-layer LSTM network. The LSTM network forget gate weight matrix contains three weight matrices: bewith, cross, and rel. Among them, bewith is the weight of the point belonging to the track, which is used to measure the probability that the echo point belongs to the track and noise; cross is the cross track weight, which is used to measure whether it is a real cross track or interference caused by noise when the echo point connection line crosses; rel is the weight of the correlation between points, which is used to measure the probability that different echo points belong to the same track. The LSTM network performs time series feature analysis and screening on the two input observation vectors, and sets the number of loop calculations to 15 times. The analysis results show that both observation vectors can reflect the real track. The two observation vectors are defined as track features and retained through the input gate, and then the track features are used as the output result. The output track feature is recorded as Output_dot F 2_1 with Output_dot F 3_1 .

[0076] Step 4.2. Output_dot F 2_1 with Output_dot F 3_1 Then input it into FCN, and combine it with Output_dot through the deconvolution function of FCN F 2_1 with Output_dot F 3_1 The location information and track information contained in dot F 1_1 、dot F 2_1 、dot F 3_1 The three points are fitted into the track line, and the track initial batch processing result is a straight line 30x-45y+26z-789=0. So far, the track initial batch processing performed by the method of the present invention has been completed. The overall process of the method of the present invention is as follows: Figure 1 shown.

[0077] Example 2:

[0078] This embodiment is based on the first embodiment. In this embodiment, the grid division in Step 3.3 divides the Hough space ρ value range [0, 24] into lengths of 12, divides θ in the range [0, -60°] into angles of 60°, and divides φ in the range [0, 30°] into angles of 15°. The statistical results show that the grids with ρ values ​​of [12, 24], θ values ​​of [0, -60°], and φ values ​​of [15°, 30°] have the largest number of three points, making this grid the peak grid. The coordinates of two points in the peak grid are (13.15, -56.31°, 25.66°), and the coordinates of another point are (21.33, -20.22°, 19.07°). The corresponding straight line equation is 88.69x-32.66y+32.67z-2133=0. The coordinates of the two echo points are (13, 20, 50) and (20, 39, 57), that is, dot F 2_1 、dot F 3_2 Located on this line. At this point, three observation vectors can be formed in Step 3.4, namely Input_dot F 2_1 ={13,20,50,13.15,-56.31°,25.66°,-3,-2,0,213.69°,0}、Input_dot F 3_1 ={7,16,50,13.15,-56.31°,25.66°,-3,-2,0,213.69°,0}、Input_dot F 3_2 ={20,39,57,-21.33,-20.22°,19.07°,3.5,8.5,3.5,69.78°,19.07°}.

[0079] Input the above three observation vectors into the LSTM network, and find that the lines between the points intersect at point (13, 20, 50). Through the cross track weight analysis, it is concluded that this is not a cross track; through the weight bewith of the point attributed to the track and the correlation weight rel between the points, it is concluded that dot F 1_1 、dot F 2_1 、dot F 3_1 The three points are located on the same straight line and the time series distribution of the three points corresponds to the track information, and dot F 3_2 with dot F 1_1 、dot F 2_1The two points cannot form a reasonable motion trajectory, and the track information does not correspond. Therefore, Input_dot F 3_2 ={20, 39, 57, -21.33, -20.22°, 19.07°, 3.5, 8.5, 3.5, 69.78°, 19.07°} is defined as noise and eliminated through the forget gate. The final track result is the same as that in Example 1. The rest of this embodiment is the same as that described in Example 1.

Claims

1. A method for batch processing multi-station UAV tracks in a complex background noise environment, characterized by: The method operates based on a radar network, wherein the radar network includes multiple low-altitude monitoring radars and a data center. The multiple low-altitude monitoring radars are located in different locations but have the same monitoring airspace range. All low-altitude monitoring radars are data-connected to the data center. The method specifically includes the following steps: Step 1. Establish a Cartesian coordinate system for measuring the entire measured airspace. Multiple low-altitude surveillance radars simultaneously monitor the airspace and transmit the measured echo frame data to the data center in real time. Each radar is assigned a unique number. After receiving the echo frame data, the data center determines the azimuth coordinates of all echo points appearing in the echo frame based on the Cartesian coordinate system and records them. Each echo point is also assigned a unique number. Step 2. The data center periodically superimposes the echo points that appear in all echo frames received within a period of time onto a blank frame. The newly generated frame is called a synthetic echo frame. The period of time is called a cycle, and the length of the cycle is equal to the interval between the generation of two adjacent synthetic echo frames. Step 3. Use the Hough transform voting mechanism to screen all echo points in two adjacent integrated echo frames, select the echo point that may reflect the target's true location, and derive the potential track information of the selected echo point based on the coordinates corresponding to the selected echo point. Then, construct an observation vector corresponding to the echo point, which includes the position information and potential track information of the corresponding echo point. Step 4. Further analyze the observation vectors to select the observation vectors that can reflect the actual track, and use the track decision device to fit the echo points corresponding to the selected observation vectors to obtain the starting track.

2. The method for batch processing multi-station UAV tracks in a complex background noise environment according to claim 1, characterized in that: The radar network contains m radars, m ≥ 2. The echo frame received by the i-th radar in Step 1 is recorded as Rad_img i , 1≤i≤m; for echo frame Rad_img i All echo points appearing in the echo frame Rad_img are numbered. i All echo points appearing in the frame are represented as Rad_img i ={dot R i_1 ,…,dot R i_P |t s }, where P is the number of echo points in the echo frame, dot R i_1 and dot R i_P are the 1st and Pth echo points detected by the i-th radar, respectively, t s is the moment when the i-th radar receives the echo frame; the azimuth coordinates of any echo point in the echo frame are expressed as {dot R i_j |t s }={pos x ,pos y ,pos z }, where j is the echo point in the echo frame Rad_img i The sequence number in, 1≤j≤P, pos x ,pos y ,pos z are the x-axis, y-axis, and z-axis coordinates of the point in the Cartesian coordinate system.

3. The method for batch processing multi-station UAV tracks in a complex background noise environment as claimed in claim 2, characterized in that: The method for superimposing the integrated echo frames in Step 2 is as follows: first find out t s All echo frames in the same cycle, and then add each echo point appearing in these echo frames to a blank frame according to the original coordinates, and the resulting frame is the integrated echo frame; the integrated echo frame is recorded as Fusion_img t The integrated echo frame and all the echo points in it are represented as Fusion_img t ={dot F t_1 ,…,dot F t_N |T}, where T is the cycle length, t is the sequence number corresponding to the cycle, N is the number of echo points in the integrated echo frame, and the echo points in the integrated echo frame are recorded as dot F t_k , and record the source information of the echo point, k is the echo point in the integrated echo frame Fusion_img t The sequence number is 1≤k≤N, and the source information includes the Rad_img where the echo point is located. i The corresponding t s .

4. A method for batch processing multi-station UAV tracks in a complex background noise environment as described in claim 3, characterized in that: In Step 3, the echo points must first be screened by the distance threshold, and then screened by the Hough transform voting mechanism. The specific steps of distance threshold screening are as follows: Step 3.

1. Set the distance threshold dis_ threshold , the current integrated echo frame is Fusion_img t , the previous integrated echo frame is Fusion_img t-1 , Fusion_img t The middle echo point is recorded as dot F t_k , Fusion_img t-1 The middle echo point is recorded as dot F t-1_k ; Step 3.

2. Get Fusion_img t Zhongyi echo dot F t_k , calculate dot F t_k with Fusion_img t-1 Each echo point F t-1_k The distance between The calculation formula is: where pos x_t ,pos y_t ,pos z_t dot F t_k The corresponding x-axis, y-axis, and z-axis coordinates in the Cartesian coordinate system, pos x_t-1 ,pos y_t-1 ,pos z_t-1 dot F t-1_k The corresponding x-axis, y-axis, and z-axis coordinates in the Cartesian coordinate system; If dot F t_k 、dot F t-1_k The distance between two points is less than the distance threshold, that is, If the condition is met, the dot F t_k 、dot F t-1_k Include point set Collection_Dot t middle; When Fusion_img t-1 All echo points in the complete current dot F t_k After calculating the distance, select another dot F t_k Repeat the above process until Fusion_img t All echo points in the calculation are completed.

5. A method for batch processing multi-station UAV tracks in a complex background noise environment as described in claim 4, characterized in that: The step 3 of screening echo points by using the Hough transform voting mechanism includes the following steps: Step 3.

3. The specific steps of selecting echo points through the Hough transform voting mechanism are as follows: Step 3.3.

1. Assume that there is a point M in the Cartesian coordinate system. The equation through point M can be obtained as follows: p = pos x_m cosθcosφ+pos y_m sinθcosφ+pos z_m sinφ (2) where pos x_m ,pos y_m ,pos z_m are the x-axis, y-axis, and z-axis coordinates of point M, respectively. The projection of the line OM between point M and the origin on the xy plane is l, the angle between l and the positive direction of the x-axis is θ, the angle between the line OM and the xy plane is φ, and ρ is the length of the line OM. Replace pos in formula (2) with x_m ,pos y_m ,pos z_m By replacing the corresponding independent variables x, y, and z, we can get the equation of the line in the Cartesian coordinate system: ρ=xcosθcosφ+ysinθcosφ+zsinφ (3)This line also represents a point with coordinates (ρ, θ, φ) in Hough space; Step 3.3.

2. Vote for the points in the Hough space to determine the line that passes through the largest number of echo points, all of which belong to Collection_Dot t The specific steps for voting are: a. Divide the range [-90°, 90°] corresponding to θ into multiple parts through the first type of segmentation points, where the first type of segmentation points are the angle values ​​of θ. Divide the range [0, 180°] corresponding to φ into multiple parts through the second type of segmentation points, where the second type of segmentation points are the angle values ​​of φ. Then, combine the first type of segmentation points with the second type of segmentation points in pairs, and substitute each combination into formula (3) to generate a corresponding straight line formula. Then, use Collection_Dot t Substitute the coordinates of the mid-echo point into each generated straight line formula one by one to calculate the corresponding ρ value; then count the number of occurrences of all ρ values, find the corresponding ρ value with the most repetitions, and the θ and φ in the straight line formula for the ρ value, and determine the point in Hough space by finding the corresponding ρ, θ, and φ values; b. Find multiple continuous point sets Collection_Dot t After the straight line with the largest number of echo points passes through the Hough space, there will be multiple points distributed; determine the range of these point distributions, divide ρ within the range into multiple segments of equal length, and divide θ and φ into multiple angles of equal size. The discretization of ρ, θ, and φ divides the entire space within the range into a grid state; then count the number of points contained in each grid, find the grid with the most points and record it as the peak grid; determine the straight line equations corresponding to the points in the peak grid, and find the echo points on these straight line equations.

6. A method for batch processing multi-station UAV tracks in a complex background noise environment as described in claim 5, characterized in that: The specific process of forming the observation vector in Step 3 is as follows: Step 3.

4. Use the echo points found in Step 3.3 to form a dot for each echo point. F t_k The corresponding observation vector, the elements of which include dot F t_k In the Cartesian coordinate system, dot F t_k The Hough space coordinates corresponding to the line and the target in dot F t_k The speed, heading angle and pitch angle at the target are UAVs for which initial track information needs to be obtained; the speed, heading angle and pitch angle are calculated by dot F t_k and dot F t_k Dots on the same line F t-1_k The Cartesian coordinates of two points are calculated, dot F t_k The velocity is expressed as The heading angle is expressed as Where ξ is the heading offset, which is determined by the specific heading reference direction and the target heading. The pitch angle is expressed as Echo Dot F t_k The observation vector at is denoted as Input_dot F t_k ={pos x_t ,pos y_t ,pos z_t ,ρ k ,θ k ,φ k ,V k avg ,Yaw k ,Pitch k }, where Δt is dot F t_k Rad_img i Corresponding to t s with dot F t-1_k Rad_img i Corresponding to t s The difference, ρ k ,θ k 、φ k Indicates dot F t_k The Hough space coordinates corresponding to the line equation.

7. A method for batch processing multi-station UAV tracks in a complex background noise environment as described in claim 6, characterized in that: In Step 4, the LSTM network is used to select the observation vector that can reflect the actual track. The track decision device is FCN. The specific steps for obtaining the starting track are as follows: Step 4.

1. Input the observation vector into the LSTM network. The LSTM network learns the elements contained in the observation vector to acquire the ability to identify the temporal characteristics of the real track. It then analyzes and filters all input observation vectors, defines the observation vectors that can reflect the real track as track features and retains them, and defines the observation vectors other than track features as noise and eliminates them. The track features are then output as the result, which is recorded as Output_dot. F t_k ; Step 4.

2. Output_dot F t_k Then input it into FCN, and combine it with Output_dot through the deconvolution function of FCN F t_k The position information and track information contained in Output_dot F t_k Corresponding dot F t_k and dot F t_k Dots on the same line F t-1_k Fit the track line to obtain the track starting batch processing results.

Images