Multi-source sensor measurement fusion method based on angle correlation

Abstract

The present invention discloses a multi-source sensor measurement fusion method based on angle association, and its implementation steps are: calculating the association angle between each target and each sensor, determining the measurement angle at each moment in each sensor association set, and using the difference in measurement azimuth angle to screen the measurement association combination; combining the least square method to perform angle association on the screened measurements to obtain the measurement location point; taking the first correction of the degree of association of the positioning line and the second correction of the active sensor position measurement on the measurement location point, eliminating most of the wrong positioning points, and using the judgment rule to correct the wrong positioning points that still exist after the second correction, to obtain the final measurement that can be tracked and filtered. The multi-source sensor system of the present invention has low computational complexity in measurement fusion, improves the measurement fusion accuracy, and improves the accuracy of multi-target tracking.

Description

Technical Field

[0001] The present invention belongs to the field of radar technology, and further relates to a multi-source sensor measurement fusion method based on angle correlation in the field of track fusion technology. The present invention can be used to achieve multi-target precise tracking in different moving target environments. Background Art

[0002] The most representative existing track state estimation fusion methods are measurement fusion (MF), simple fusion (SF) and weighted covariance fusion (WCF). The measurement fusion method has a small amount of calculation but low accuracy; the simple fusion method has a fast calculation speed, but its assumptions do not match the actual situation, and the solution obtained by the SF method is not the optimal solution; the weighted covariance fusion method has high accuracy but huge amount of calculation. Due to the measurement data dimension matching problem, the existing multi-target tracking algorithm is mainly based on a multi-sensor system composed of sensors of the same type or sensors with the same dimensional measurement. The homogeneous measurement data is fused and processed through a unified data processing framework to locate and track the target. The advantage of this type of algorithm is that the measurement data structure is unified, there is no need to build a data processing framework separately, and the development is relatively mature. However, this type of algorithm is mainly based on a single type of sensor, either an active sensor or a passive sensor. The information obtained by a single active sensor is mostly three-dimensional or higher data. The amount of calculation is too large when the data is associated, which is not conducive to calculation; a single passive sensor only obtains angle information and lacks position information for accurate positioning. In addition, the passive sensor system needs to reasonably arrange more sensors in space to cross-find the target, which leads to an increase in calculation amount. In addition, due to the influence of complex environment, it is very easy to cause an increase in the number of erroneous measurement combinations, resulting in a decrease in multi-target tracking performance.

[0003] Harbin Engineering University has disclosed a multi-sensor track fusion method that solves the existing problem of imbalance between computational complexity and accuracy in its patent application document "A multi-sensor track fusion method based on distance map and data cleaning" (patent application number: 202110784798.3, authorization announcement number: CN 113532422 B). The main steps of this method are: (1) Using M active sensors to jointly observe T0 targets, for any two sensors s and l, the observed track at time k is obtained, and the distance map of sensors s and l at time k is constructed through the state vector of the track. The tracks of the sensors are associated through the distance map to obtain the track data of each target; (2) The track data of each target is cleaned separately to obtain the valid track data of each target; (3) The valid track data of each target is fused separately to obtain the track fusion result of each target. The disadvantage of this method is that only a single type of active sensor is used to obtain the observed track, and the subsequent track fusion is computationally intensive. If the number of sensors is too large, the combination explosion problem is prone to occur.

[0004] Hangzhou Dianzi University disclosed a multi-sensor track fusion method based on distributed fusion in its patent application "Multi-sensor track fusion method for track management using target visibility" (patent application number: 201811624341.0, authorization announcement number: CN 109657732 B). The main steps of the method are as follows: Step 1, c sensors upload the initial track sets tracked by themselves to the fusion center respectively; each track information in the initial track set contains state estimation, error covariance and target visibility; assign 2 to p; take the first initial track set as the main initial track set τ, and the pth initial track set as the auxiliary initial track set η; Step 2, the fusion center performs track association fusion on each track in the main initial track set τ and each track in the auxiliary initial track set η; Step 3, if p<c, then increase p by 1, copy the fused track set as the main initial track set τ, and the pth initial track set as the auxiliary initial track set η, set the fused track set to an empty set, and repeat Step 2; if p=c, enter Step 4; Step 4, set the target visibility in the fused track set less than t FT The track with a visibility greater than t is regarded as the termination track and is no longer tracked and is directly deleted from the fused track set; the target visibility greater than t is deleted from the fused track set. CT The track of the target is output as the target track; the target visibility in the fusion track set is at t FT ~t CT The tracks between them are tracked as unknown tracks; FT =0.01; t CT=0.99. The disadvantage of this method is that due to the influence of environmental noise or false alarm, the initial track set obtained by the sensor contains erroneous measurements, which are not processed before being uploaded to the fusion center. In the subsequent processing process, due to too many measurement error association combinations, the subsequent fusion accuracy decreases, thereby reducing the multi-target tracking accuracy. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-source sensor measurement fusion method based on angle association in view of the above-mentioned deficiencies in the prior art, aiming to solve the problems in the prior art of excessively high data association calculation amount in multi-sensor systems and decreased tracking accuracy due to an increase in the number of erroneous measurement combinations.

[0006] The idea of ​​achieving the purpose of the present invention is as follows: the present invention extracts the angle measurement values ​​common to the two types of sensors, calculates the associated angle between each target and each sensor, determines the measurement angle at each moment in each sensor associated set, and excludes the measurements that are not in the measurement associated set; the measurements after the initial screening are combined according to different sensors, and the erroneous measurement associated combinations are eliminated by constructing measurement azimuth angle differences, so that the number of processed measurement associated combinations can reach a level that can be quickly calculated; the remaining measurement associated combinations are angle-associated by combining the least squares method to obtain the measurement location point; finally, by combining the active sensor position measurement properties and other means, a primary correction of the degree of association of the positioning line, a secondary correction of the active sensor position measurement, and a judgment criterion are adopted to eliminate erroneous positioning points and obtain the final measurement that can be tracked and filtered.

[0007] To achieve the above object, the steps of the present invention are as follows:

[0008] Step 1: extract the target angle measurement values ​​of at least 4 different motion states of each sensor at the same time, each motion state has at least two or more targets, calculate the associated angle between each target and each sensor respectively, and obtain the associated set of each sensor;

[0009] Step 2, determining the measurement angle at each moment in each sensor association set, calculating the angle measurement error, and forming all measurements within the measurement angle range into the sensor measurement association set;

[0010] Step 3, the measurement values ​​within the measurement association set range of the active and passive sensors in the system are combined into a measurement association combination of the system sensors, and the errors in the measurement association combination are screened out by using the measurement azimuth difference to obtain a screened measurement association combination;

[0011] Step 4, performing angle correlation fusion positioning on the screened measurement correlation combinations to obtain the target positioning point corresponding to each measurement correlation combination;

[0012] Step 5, using the positioning line correlation degree correction method, screen out the wrong measurement sites in the target positioning points corresponding to each measurement correlation combination, and obtain the measurement sites after the first screening;

[0013] Step 6, using the correction method of active sensor position measurement, screen out the wrong positioning points in the measurement sites after the first screening, and obtain the measurement sites after the second screening;

[0014] Step 7, determining whether the number of measurement sites of the same target after the second screening is 1, if so, executing step 9, otherwise, executing step 8;

[0015] Step 8, taking the average of all the measurement sites of the same target, obtaining the fusion positioning point of the target and then executing step 9;

[0016] Step 9: Use the filtered positioning points as the corrected positioning points.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] First, the present invention combines the advantages of accurate positioning of active sensors and small computational complexity of passive sensors, and fuses the distance and angle information of active sensors with the angle information of passive sensors, thereby overcoming the problem that heterogeneous sensor data are difficult to fuse due to different measurement latitudes. Compared with a single active or single passive sensor system, the active and passive sensor system of the present invention has low computational complexity in measurement fusion.

[0019] Second, the present invention reduces the measurement range that needs to be associated between sensors through initial screening of angle measurements before performing angle association; the present invention proposes secondary screening based on statistics, which further reduces the number of measurement association combinations on the basis of eliminating erroneous measurements in the first time, so that the number of measurement association combinations after two screening processes of the present invention has reached the range of numbers that can be quickly calculated, which reduces the calculation complexity in subsequent fusion positioning and improves the fusion efficiency.

[0020] Third, the present invention combines the least squares method with the measurement set to reduce erroneous measurements to perform angle correlation fusion positioning. All correlation combinations used to calculate positioning points are not completely correct measurement correlation combinations. For erroneous measurement correlation combinations within the angle correlation range, the final fusion positioning result is obtained by correcting the degree of correlation of the positioning line, correcting the position measurement, and correcting the closer positioning points. The positioning results at each moment are connected in series on the time axis as the target tracking track, which improves the tracking accuracy of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of the present invention;

[0022] Figure 2 is a diagram showing the relationship between the angular positions of the sensors of the present invention;

[0023] Figure 3 is a correlation set diagram of the present invention for determining sensors for correlation angle and azimuth;

[0024] Figure 4 It is a definition diagram of the angle measured by the sensor of the present invention;

[0025] Figure 5 It is a diagram showing the relationship between the sensor and the measured angular position of the present invention;

[0026] Figure 6 is a flow chart of determining a sensor measurement association set according to the present invention;

[0027] Figure 7 It is a measurement association set diagram of the sensor determined by the present invention;

[0028] Figure 8 It is a schematic diagram of the measurement association combination of the present invention;

[0029] Fig. 9 This is a schematic diagram of measuring azimuth angle difference of the present invention;

[0030] Fig.10 is a probability density function curve diagram of the present invention for measuring azimuth angle difference;

[0031] Fig.11 It is a schematic diagram of the angle association of the present invention;

[0032] Fig.12 It is the real motion trajectory diagram of the CV model target of the present invention;

[0033] Fig.13 It is the tracking error diagram of the CV model of the present invention in the x, y and z directions for a single run;

[0034] Fig.14 This is a simulation result diagram of 100 Monte Carlo experiments of the CV model of the present invention;

[0035] Fig.15 It is the real motion trajectory diagram of the target of the RCT model of the present invention;

[0036] Fig.16 It is the tracking error diagram of the RCT model of the present invention in the x, y and z directions for a single run;

[0037] Fig.17 This is a simulation result diagram of 100 Monte Carlo experiments of the RCT model of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0039] Reference Figure 1 , further describing the implementation steps of the embodiment of the present invention.

[0040] Step 1: extract the target angle measurement values ​​of at least 4 different motion states of each sensor at the same time. There are at least two or more targets in each motion state. Calculate the associated angle between each target and each sensor respectively to obtain the associated set of each sensor.

[0041] The four different motion states include: uniform linear motion, uniformly accelerated linear motion, constant turning rate motion and non-constant turning rate motion.

[0042] The associated angle between each target and each sensor is determined by the following formula:

[0043]

[0044] Among them, θ ij represents the associated opening angle between the i-th sensor and the j-th sensor, arcsin(·) represents the inverse sine calculation symbol, R i With R j represents the farthest detection radius between the i-th sensor and the j-th sensor, d ij represents the distance between the i-th sensor and the j-th sensor, R min It represents the detection radius of the sensor with the smallest detection radius between the i-th sensor and the j-th sensor.

[0045] Reference Figure 2 , further explanation of the sensor position relationship is given.

[0046] Figure 2 (a) indicates That is, the detection ranges of the two sensors intersect, and the endpoints M and N are the intersection points of the circles rather than the tangent points; Figure 2 (b) indicates That is, the detection ranges of the two sensors intersect, and M and N are tangent points of the circles rather than intersection points; Figure 2 (c) means d ij <R j , that is, the i-th sensor is within the detection range of the j-th sensor.

[0047] The association set of each sensor is The range is determined according to the following rules: ij represents the azimuth angle of the i-th sensor with respect to the j-th sensor;

[0048] The first step is to determine whether the distance between the i-th sensor and the j-th sensor is less than the sum of the farthest detection radius of the two sensors. If so, the second step is executed; otherwise, it is determined that there is no associated set between the two sensors;

[0049] The second step is to determine whether the distance between the ith sensor and the jth sensor is less than the farthest detection radius of the jth sensor. If so, the measurement of the ith sensor is determined to constitute the associated set of the two sensors. Otherwise, the measurement of the ith sensor and the jth sensor within the associated azimuth range is determined to constitute the associated set of the two sensors.

[0050] Reference Figure 3 , further explains the formation of sensor association sets at different locations:

[0051] Figure 3 (a) indicates that the distance between the i-th sensor and the j-th sensor is greater than the sum of the farthest detection radii of the two sensors; Figure 3 (b) indicates the situation where the distance between the i-th sensor and the j-th sensor is less than the sum of the farthest detection radii of the two sensors but greater than the farthest detection radii of the i-th sensor and the j-th sensor and the azimuth angle is greater than 0; Figure 3 (c) indicates the situation where the distance between the i-th sensor and the j-th sensor is less than the sum of the farthest detection radii of the two sensors but greater than the farthest detection radii of the i-th sensor and the j-th sensor and the azimuth angle is less than 0; Figure 3 (d) indicates that the distance between the i-th sensor and the j-th sensor is less than the farthest detection radius of the j-th sensor.

[0052] Step 2: determine the measurement angle at each moment in each sensor association set, calculate the angle measurement error, and form the sensor measurement association set with all measurements within the measurement angle range.

[0053] The measured angle at each moment in each sensor association set is determined by the following formula:

[0054]

[0055] in, represents the measured angle between the i-th sensor and the j-th sensor, ∠(·) represents the angle sign, O i The coordinate point representing the position of the i-th sensor in the spatial rectangular coordinate system, O j represents the coordinate point of the jth sensor in the spatial rectangular coordinate system, E represents the angle measurement positioning point of the ith sensor relative to the measurement, F represents the angle measurement positioning point of the ith sensor relative to the measurement, R i With R j represents the farthest detection radius between the i-th sensor and the j-th sensor, dij represents the distance between the i-th sensor and the j-th sensor, l2 represents the line segment O i F is the distance measured by the i-th sensor relative to the measurement.

[0056] Reference Figure 4 , the definition of measuring the opening angle is further explained.

[0057] The definition of the measured opening angle is: i The vector is measured to make a ray that intersects the points E and F on the circle. j Connected with E and F, among all the obtained angles, the minimum angle containing the measurement position is the measurement angle in the associated set of the i-th sensor and the j-th sensor.

[0058] Reference Figure 5 , the calculation of the measured angle for different sensor positions is explained:

[0059] Figure 5 (a) indicates R j <d ij <R i +R j The sensor location situation; Figure 5 (b) means d ij <R j And l2<R i The sensor location situation; Figure 5 (c) means d ij <R j And l2 ≥ R i The sensor location situation.

[0060] The angle measurement error is determined by the following formula:

[0061]

[0062] Among them, v α represents the angle measurement error of each sensor to the target, N(·) represents the normal distribution, σ α Represents the standard deviation of the normal distribution.

[0063] Reference Figure 6 , the judgment process of the measurement association set is further explained.

[0064] The sensor measurement association set is composed of all measurements within the measurement angle range and is determined according to the following rules:

[0065] The first step is to determine d ij Does it satisfy R j <d ij <R i+R j , if yes, execute step 2; otherwise, execute step 9;

[0066] The second step is to determine whether the azimuth angle of the mth measurement is less than the azimuth angle of the line connecting the ith sensor and the jth sensor. If so, execute the third step; otherwise, execute the sixth step.

[0067] The third step is to determine α ij Is it greater than or equal to 0? If so, go to step 4; otherwise, go to step 5.

[0068] The fourth step is to determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0069] The fifth step is to determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0070] Step 6: Determine α ij Is it greater than or equal to 0? If so, go to step 7; otherwise, go to step 8;

[0071] Step 7: Determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0072] Step 8: Determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0073] Step 9: Determine d ij and l2 satisfy d ij <R j And l2<R i , if yes, execute step 10, otherwise execute step 17;

[0074] Step 10, determine whether the mth measurement is within the detection range along the -y axis divided by the line connecting the positions of the ith sensor and the jth sensor. If so, execute step 11, otherwise, execute step 14;

[0075] Step 11: Determine α ij Is it greater than or equal to 0? If so, go to step 12; otherwise, go to step 13;

[0076] Step 12: Determine whether [αji -3σ α ,α ji +∠O i O j F+3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0077] Step 13: Determine whether [α ji -∠O i O j F-3σ α ,α ji +3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0078] Step 14: Determine α ij Is it greater than or equal to 0? If so, go to step 15; otherwise, go to step 16;

[0079] Step 15: Determine whether [α ji -∠O i O j F-3σ α ,α ji +3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0080] Step 16: Determine whether [α ji -3σ α ,α ji +∠O i O j F+3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the ith sensor.

[0081] Step 17: Determine d ij and l2 satisfy d ij <R j And l2 ≥ R i If yes, execute the fourth step; otherwise, consider the calculation error and execute the first step again.

[0082] Reference Figure 7 , further explanation is given on the situation of different position measurements within the sensor detection range.

[0083] Figure 7 (a) represents α ij When it is greater than 0, the measurement is within the detection range along the -y axis divided by the line connecting the i-th sensor and the j-th sensor position, where A represents the detection range along the -y axis divided by the line connecting the i-th sensor and the j-th sensor position, and B represents the detection range along the y axis divided by the line connecting the i-th sensor and the j-th sensor position; Figure 7 (b) represents α ij When it is greater than 0, the measurement is in range B; Figure 7 (c) represents α ij When it is less than 0, the measurement is in range A; Figure 7 (d) represents α ij When it is less than 0, the measurement is in range B.

[0084] Step 3, the measurement values ​​within the measurement association set range of the active and passive sensors in the system are combined into a measurement association combination of the system sensors, and the errors in the measurement association combination are screened out by using the measurement azimuth difference to obtain the screened measurement association combination.

[0085] The measurement values ​​within the measurement association set of active and passive sensors in the system form the measurement association combination of the system sensors, which is determined according to the following rules:

[0086]

[0087] in, and represents the azimuth and elevation angles of the rth measurement in the passive sensor S1, and represents the azimuth and elevation angles measured by the passive sensor S2, and represents the azimuth and elevation angles measured by the active sensor S3, represents the coordinate position of the tth measurement in active sensor S3, {·} T represents the transposition operation, and m represents the sampling time sequence number.

[0088] Reference Figure 8 , a sensor system consisting of one active sensor and two passive sensors is further explained to form the process of the measurement correlation combination of the three sensors at each moment.

[0089] The sensor's measurement at the current moment is the measurement value of all targets within the sensor's associated set. The active sensor obtains the target measurement values ​​of five dimensions: [α, β, x, y, z] T ,in,[·] Trepresents the transposition operation, α represents the azimuth of the measurement point, β represents the pitch angle of the measurement point, and (x, y, z) represents the coordinate position of the measurement point. Each passive sensor obtains a target measurement value with two dimensions [α, β] T .

[0090] Figure 8 At the mth moment, all target measurement values ​​obtained by the active sensor in are as follows: in,{·} T represents the transposition operation, m represents the sequence number of the sampling time, α t represents the azimuth of the tth measurement point, t = 1, 2, ···, T, T represents the total number of active sensor target measurement points, β t represents the pitch angle of the tth measurement point, (x t ,y t ,z t ) represents the coordinate position of the tth measurement point.

[0091] The set of measurements obtained by the first passive sensor at the mth moment is Among them, α r ' represents the azimuth of the rth measurement point, r = 1, 2, ···, R, R represents the total number of target measurement points of the first passive sensor, β r ' represents the pitch angle of the rth measurement point.

[0092] The set of measurements obtained by the second passive sensor at the mth moment is Among them, α s ” represents the azimuth of the sth measurement point, s = 1, 2, ···, S, S represents the total number of target measurement points of the second passive sensor, β s " represents the pitch angle of the sth measurement point; the measurement values ​​of the active sensor and the two passive sensors are arranged and combined to form the measurement association combination at the current moment:

[0093] Reference Fig. 9 , further explanation is given on the measurement of azimuth angle difference.

[0094] The measured azimuth difference is determined by the following formula:

[0095] Δ α =α M -α 3t

[0096] Among them, Δ α is the measured azimuth difference, α M yes Determine the straight line and Determine the azimuth of the intersection point M of the straight lines.

[0097] The error in the measurement association combination is screened out to obtain the screened measurement association combination, which is determined according to the following rules:

[0098] Reference Fig.10 Δ α The probability density function curve further explains the exclusion of measurement-related combinations.

[0099] Set the confidence interval to in Yes α The standard deviation of α If it falls within the confidence interval, then it is considered The three pairs of measurements in the correlation combination may come from the same target and should be retained; otherwise, the correlation combination is judged as an erroneous correlation combination and should be excluded.

[0100] Step 4: Perform angle correlation fusion positioning on the screened measurement correlation combinations to obtain the target positioning point corresponding to each measurement correlation combination.

[0101] The angle correlation fusion positioning is performed on the screened measurement correlation combination to obtain the target positioning point corresponding to each measurement correlation combination, which is determined according to the following rules:

[0102] The first step is to determine the direction of each positioning point relative to each sensor at each moment;

[0103] In the second step, the positioning line of each positioning point relative to each sensor is determined;

[0104] The third step is to determine the distance between each positioning point and the positioning line forming the positioning point;

[0105] The fourth step is to use the least squares method to determine the positioning point of each target.

[0106] Reference Fig.11 , for the sensor system composed of active sensor S3 and passive sensor S1 and passive sensor S2, the positioning line L of the sensor is determined by the measurement association combination at the current moment, and the coordinate position of the point with the shortest sum of the distances of the three positioning lines is taken as the coordinate position of the positioning point.

[0107] The coordinate position of the positioning point is determined according to the following rules:

[0108] The first step is to determine the direction of each positioning point relative to each sensor at each moment according to the following formula:

[0109]

[0110] in, represents the direction of the measurement positioning line of the i-th sensor in the k-th measurement association combination along the x-axis. In the embodiment of the present invention, there is one active sensor and two passive sensors, so i=1,2,3, cos(·) represents the cosine operation, m represents the sequence number of the sampling time, represents the pitch angle measured by the i-th sensor in the k-th measurement association combination at the m-th moment, represents the azimuth angle measured by the i-th sensor in the k-th measurement association combination at the m-th time, represents the direction of the i-th sensor measurement positioning line in the k-th measurement association combination along the y-axis, sin(·) represents the sine operation, It represents the direction of the i-th sensor measurement positioning line in the k-th measurement association combination along the z-axis.

[0111] In the second step, determine the positioning line of each positioning point relative to each sensor according to the following formula:

[0112]

[0113] in, represents the positioning line determined by the measurement of the i-th sensor in the k-th measurement association combination at the m-th time, represents the coordinate position of the positioning point determined by the kth measurement association combination at the mth moment, (x i ,y i ,z i ) represents the coordinate position of the i-th sensor in the spatial rectangular coordinate system.

[0114] The third step is to determine the distance from each positioning point to the positioning line that forms the positioning point according to the following formula:

[0115]

[0116] in, Indicates the distance from the kth positioning point to the positioning line at the current moment The distance Represents the unit vector of the x-axis in the spatial rectangular coordinate system, Represents the unit vector of the y-axis in the spatial rectangular coordinate system, Represents the unit vector of the z-axis in the spatial rectangular coordinate system.

[0117] In the fourth step, the least squares method is used to determine the positioning point of each target as follows:

[0118]

[0119] in,

[0120]

[0121]

[0122]

[0123]

[0124] Step 5: Get the target positioning point according to the following formula:

[0125]

[0126] Step 5: Using the positioning line correlation degree correction method, the erroneous measurement sites in the target positioning points corresponding to each measurement correlation combination are screened out to obtain the measurement sites after the first screening.

[0127] The steps of the positioning line association degree correction method are as follows:

[0128] The first step is to determine the range threshold of the target positioning point according to the following formula:

[0129]

[0130] Among them, γ represents the range threshold of the target positioning point, σ x represents the standard deviation of the distance noise of the active sensor on the x-axis, σ y represents the standard deviation of the distance noise of the active sensor on the y-axis, σ z represents the standard deviation of the distance noise of the active sensor on the z-axis;

[0131] The second step is to determine whether the sum of the distances from each positioning point to the positioning line forming the positioning point is less than γ. If so, the positioning point is retained; otherwise, the positioning point is screened out.

[0132] Step 6, using the correction method of active sensor position measurement, screen out the wrong positioning points in the measurement sites after the first screening, and obtain the measurement sites after the second screening.

[0133] The steps of the active sensor position measurement correction method are as follows:

[0134] The first step is to determine the correction threshold value according to the following formula:

[0135]

[0136] Among them, γ σ,x represents the threshold value of the active sensor on the x-axis, σ x represents the standard deviation of the distance noise of the active sensor on the x-axis, γ σ,y represents the threshold value of the active sensor on the y-axis, σ yrepresents the standard deviation of the distance noise of the active sensor on the y-axis, γ σ,z represents the threshold value of the active sensor on the z-axis, σ z represents the standard deviation of the distance noise of the active sensor on the z-axis,

[0137] The second step is to determine whether the distance between the active sensor and each positioning point is less than three threshold values ​​at the same time. If so, the positioning point is retained; otherwise, the positioning point is screened out.

[0138] Step 7, determining whether the number of measurement sites of the same target after the second screening is 1, if so, executing step 9, otherwise, executing step 8;

[0139] Step 8, taking the average of all the measurement sites of the same target, obtaining the fusion positioning point of the target and then executing step 9;

[0140] Step 9: Use the filtered positioning points as the corrected positioning points.

[0141] The following is a simulation experiment to further illustrate the technical effect of the present invention:

[0142] 1. Conditions of simulation experiment:

[0143] The hardware platform of the simulation experiment of the present invention is: the processor is Intel (R) Core (TM) i7-12700H CPU, the main frequency is 2.7 GHz, and the memory is 16 GB.

[0144] The software platforms for the simulation experiment of the present invention are: Windows 11 operating system and MATLAB R2023a.

[0145] 2. Simulation content and results analysis:

[0146] The simulation experiment of the present invention utilizes the method of the present invention to simulate and verify 6 moving targets under a uniform motion model (CV) and a random uniform turning model (RCT), and analyzes the results. The simulation results are the average results of 100 Monte Carlo experiments.

[0147] Simulation experiment 1:

[0148] The scene in the simulation experiment 1 of the present invention contains 6 targets in total, the motion area is x∈[-5km,5km], y∈[-5km,5km], the altitude is 500m low altitude, and the motion model is uniform linear motion. Targets are born and die, so the number of targets will change over time. The target state vector is x(k)=[x(k),v x (k),y(k),v y (k),z(k),v z (k)]T , which consists of position and velocity. The initial states of the six targets are:

[0149] [2000, -36, 1000, -27, 500, 0] T ,[1500,-18,-2000,28,500,0] T ,[-2000,32,-1000,15,500,0] T ,[1500,-24,-2000,-12,500,0] T ,[-1000,-25,2000,-30,500,0] T ,[-2000,-16,-1000,48,500,0] T ; The birth time of the 6 targets are 1, 1, 10, 20, 30, 40 respectively; The extinction time of the 6 targets are 70, 100, 80, 100, 100, 100 respectively. Fig.12 Represents the real motion trajectory of the target, where "○" represents the starting point of the trajectory and "△" represents the end point of the trajectory. Three sensors are set in the scene, among which S1 and S2 are passive sensors, and their positions are (0,-5km,0) and (5km,0,0) respectively. S3 is an active sensor, and its position is (0,5km,0). The detection range of each sensor is 15km. Measurement vector z(k) = [α(k), β(k), x(k), y(k), z(k)] T It consists of azimuth, elevation and position coordinates. Other parameters in the motion model are set as follows: sampling interval T = 1s, azimuth measurement standard deviation σ α =1rad, standard deviation of pitch angle measurement σ β =1rad, standard deviation of distance in x, y, z directions σ x , σ y , σ z Both are 10m, the target survival probability is 0.98, and the sensor detection probability is 0.98.

[0150] The simulated tracking algorithm selects the PHD filtering algorithm based on random finite sets. In order to reflect the performance improvement of the proposed fusion algorithm, the following simulation results are compared with the PHD tracking results of a single active sensor. The environmental parameters are set exactly the same, that is, a uniformly distributed clutter environment with a clutter rate of λ=60. Fig.13, shows the tracking results of a single simulation in the x, y, and z directions. In the figure, “×” represents the PHD filter estimation value after the fusion of multi-source sensors, that is, active-passive sensors (APS), “·” represents the PHD filter estimation value of a single active sensor (AS), and “×” represents the target measurement or clutter measurement after the fusion of multi-source sensors. Indicates the target measurement or clutter measurement obtained by sensor S3. Fig.13 The simulation results show that the tracking of the target by APS measurement and AS measurement in the x, y, and z directions is close, and there is a small amount of loss in the tracking results of both measurements.

[0151] Fig.14 1 is a simulation result diagram of the simulation experiment 1 of the present invention after 100 Monte Carlo experiments, wherein: Fig.14 (a) Represents the mean of the potential estimate of 100 Monte Carlo experiments, i.e., the estimate of the target number. Fig.14 (a) shows that the results of APS measurement and AS measurement are close to the actual target number.

[0152] Fig.14 (b), Fig.14 (c) and Fig.14 (d) shows the OSPA distance of 100 Monte Carlo experiments (p=1, c=300m). Fig.14 (b) shows that at all sampling moments, the OSPA distance corresponding to the APS measurement is smaller than the OSPA distance corresponding to the AS measurement; Fig.14 (c) shows that at all sampling times, the position error measured by APS is smaller than the position error measured by AS; Fig.14 (d) shows that at most sampling moments, the potential error measured by APS is consistent with the potential error measured by AS.

[0153] Taking into account the position error, potential estimation error and OSPA distance, it is concluded that in the CV model, when the parameters are the same, the tracking performance corresponding to the APS measurement is better than that of the AS measurement, and the tracking accuracy of the target is higher.

[0154] Simulation experiment 2:

[0155] The scene in the simulation experiment 2 of the present invention contains 6 targets in total, the motion area is x∈[-5km,5km], y∈[-5km,5km], the altitude is 500m low altitude, and the motion model is uniform linear motion. Targets are born and die, so the number of targets will change over time. The target state vector is x(k)=[x(k),v x (k),y(k),v y (k),z(k),vz (k),ω] T , which is composed of position, speed and turning angular velocity. The turning angular velocity will change over time. The initial states of the six targets are [-1000, 28, 2000, -20, 500, 0, 3] T ,[1500,-18,-2000,28,500,0,3] T ,[1500,10,-2000,25,500,0,3] T ,[-1000,-25,2000,-30,500,0,3] T ,[-500,50,-1000,-52,500,0,3] T ,[-2000,16,-1000,40,500,0,3] T ; The birth times of the 6 targets are 1, 8, 15, 20, 26, and 30 respectively; The extinction times of the 6 targets are 70, 100, 100, 100, 50, and 100 respectively. Fig.15 The actual motion trajectory of the target is shown, where "○" indicates the starting point of the trajectory and "△" indicates the end point of the trajectory. Three sensors are set in the scene, among which S1 and S2 are passive sensors, and their positions are (0, -5km, 0) and (5km, 0, 0) respectively. S3 is an active sensor, and its position is (0, 5km, 0). The detection range of each sensor is 15km. Measurement vector z(k) = [α(k), β(k), x(k), y(k), z(k)] T It consists of azimuth, pitch and position coordinates. In addition to setting the target survival probability to 0.99 and the sensor detection probability to 0.99, the other parameter settings in the motion model are consistent with the simulation experiment.

[0156] The simulated tracking algorithm selects the PHD filtering algorithm based on random finite sets. In order to reflect the performance improvement of the proposed fusion algorithm, the following simulation results are compared with the PHD tracking results of a single active sensor. The environmental parameters are set exactly the same, that is, a uniformly distributed clutter environment with a clutter rate of λ=60. Fig.13 , shows the tracking results of a single simulation in the x, y, and z directions. In the figure, “×” represents the PHD filter estimation value after the fusion of multi-source sensors, that is, active-passive sensors (APS), “·” represents the PHD filter estimation value of a single active sensor (AS), and “×” represents the target measurement or clutter measurement after the fusion of multi-source sensors. Indicates the target measurement or clutter measurement obtained by sensor S3. Fig.16The simulation results show that the tracking of the target by APS measurement and AS measurement in the x, y, and z directions is close, and there is a small amount of loss in the tracking results of both measurements.

[0157] Fig.17 2 is a simulation result diagram of the simulation experiment 2 of the present invention after 100 Monte Carlo experiments, wherein: Fig.17 (a) Represents the mean of the potential estimate of 100 Monte Carlo experiments, i.e., the estimate of the target number. Fig.17 (a) shows that the results of APS measurement and AS measurement are close to the actual target number.

[0158] Fig.17 (b), Fig.17 (c) and Fig.17 (d) shows the OSPA distance of 100 Monte Carlo experiments (p=1, c=300m). Fig.17 (b) shows that at all sampling moments, the OSPA distance corresponding to the APS measurement is smaller than the OSPA distance corresponding to the AS measurement; Fig.17 (c) shows that at all sampling times, the position error measured by APS is smaller than the position error measured by AS; Fig.17 (d) shows that at most sampling moments, the potential error measured by APS is consistent with the potential error measured by AS.

[0159] Taking into account the position error, potential estimation error and OSPA distance, it is concluded that in the RCT model, under the same parameters, the tracking performance corresponding to APS measurement is better than that of AS measurement, and the tracking accuracy of the target is higher.

[0160] It can be seen from the results of simulation experiments 1 and 2 that the proposed multi-source sensor measurement fusion method based on angle correlation can better handle the measurements of single active sensors and multiple passive sensors, and also significantly improves the tracking accuracy of multiple targets.

Claims

1. A multi-source sensor measurement fusion method based on angle correlation, characterized in that: Calculate the associated angle between each target and each sensor, determine the measurement angle at each moment in each sensor association set, and use the difference in the measured azimuth angle to screen the measurement association combination; the steps of the fusion method include the following: Step 1: extract the target angle measurement values ​​of at least 4 different motion states of each sensor at the same time, each motion state has at least two or more targets, calculate the associated angle between each target and each sensor respectively, and obtain the associated set of each sensor; Step 2, determining the measurement angle at each moment in each sensor association set, calculating the angle measurement error, and forming all measurements within the measurement angle range into the sensor measurement association set; Step 3, the measurement values ​​within the measurement association set range of the active and passive sensors in the system are combined into a measurement association combination of the system sensors, and the errors in the measurement association combination are screened out by using the measurement azimuth difference to obtain a screened measurement association combination; Step 4, performing angle correlation fusion positioning on the screened measurement correlation combinations to obtain the target positioning point corresponding to each measurement correlation combination; Step 5, using the positioning line correlation degree correction method, screen out the wrong measurement sites in the target positioning points corresponding to each measurement correlation combination, and obtain the measurement sites after the first screening; Step 6, using the correction method of active sensor position measurement, screen out the wrong positioning points in the measurement sites after the first screening, and obtain the measurement sites after the second screening; Step 7, determining whether the number of measurement sites of the same target after the second screening is 1, if so, executing step 9, otherwise, executing step 8; Step 8, taking the average of all the measurement sites of the same target, obtaining the fusion positioning point of the target and then executing step 9; Step 9: Use the filtered positioning points as the corrected positioning points.

2. The multi-source sensor measurement fusion method based on angle association according to claim 1 is characterized in that: The four different motion states described in step 1 include: uniform linear motion, uniformly accelerated linear motion, constant turning rate motion and non-constant turning rate motion.

3. The multi-source sensor measurement fusion method based on angle association according to claim 1 is characterized in that: The associated angle between each target and each sensor in step 1 is determined by the following formula: Among them, θ ij represents the associated opening angle between the i-th sensor and the j-th sensor, arcsin(·) represents the inverse sine calculation symbol, R i With R j represents the farthest detection radius between the i-th sensor and the j-th sensor, d ij represents the distance between the i-th sensor and the j-th sensor, R min It represents the detection radius of the sensor with the smallest detection radius between the i-th sensor and the j-th sensor.

4. The multi-source sensor measurement fusion method based on angle association according to claim 3 is characterized in that: The associated set of each sensor in step 1 is The range is determined according to the following rules: ij represents the azimuth angle of the i-th sensor with respect to the j-th sensor; The first step is to determine whether the distance between the i-th sensor and the j-th sensor is less than the sum of the farthest detection radius of the two sensors. If so, the second step is executed; otherwise, it is determined that there is no associated set between the two sensors; The second step is to determine whether the distance between the ith sensor and the jth sensor is less than the farthest detection radius of the jth sensor. If so, the measurement of the ith sensor is determined to constitute the associated set of the two sensors. Otherwise, the measurement of the ith sensor and the jth sensor within the associated azimuth range is determined to constitute the associated set of the two sensors.

5. The multi-source sensor measurement fusion method based on angle association according to claim 1 is characterized in that: The measured angle at each moment in each sensor association set in step 2 is determined by the following formula: in, represents the measured angle between the i-th sensor and the j-th sensor, ∠(·) represents the angle sign, O i The coordinate point representing the position of the i-th sensor in the spatial rectangular coordinate system, O j represents the coordinate point of the jth sensor in the spatial rectangular coordinate system, E represents the angle measurement positioning point of the ith sensor relative to the measurement, F represents the angle measurement positioning point of the ith sensor relative to the measurement, R i With R j represents the farthest detection radius between the i-th sensor and the j-th sensor, d ij represents the distance between the i-th sensor and the j-th sensor, l2 represents the line segment O i F is the distance measured by the i-th sensor relative to the measurement.

6. The multi-source sensor measurement fusion method based on angle association according to claim 1 is characterized in that: The angle measurement error described in step 2 is determined by the following formula: Among them, v α represents the angle measurement error of each sensor to the target, N(·) represents the normal distribution, σ α Represents the standard deviation of the normal distribution.

7. The multi-source sensor measurement fusion method based on angle association according to claim 5 is characterized in that: In step 2, all measurements within the measurement angle range are formed into the sensor measurement association set, which is determined according to the following rules: The first step is to determine d ij Does it satisfy R j <d ij <R i +R j , if yes, execute step 2; otherwise, execute step 9; The second step is to determine whether the azimuth angle of the mth measurement is less than the azimuth angle of the line connecting the ith sensor and the jth sensor. If so, execute the third step; otherwise, execute the sixth step. The third step is to determine α ij Is it greater than or equal to 0? If so, go to step 4; otherwise, go to step 5. The fourth step is to determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; The fifth step is to determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 6: Determine α ij Is it greater than or equal to 0? If so, go to step 7; otherwise, go to step 8; Step 7: Determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 8: Determine The measurements within the range constitute a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 9: Determine d ij and l2 whether they satisfy d ij <R j And l2<R i , if yes, execute step 10, otherwise execute step 17; Step 10, determine whether the mth measurement is within the detection range along the -y axis divided by the line connecting the positions of the ith sensor and the jth sensor. If so, execute step 11, otherwise, execute step 14; Step 11: Determine α ij Is it greater than or equal to 0? If so, go to step 12; otherwise, go to step 13; Step 12: Determine whether [α ji -3σ α ,α ji +∠O i O j F+3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 13: Determine whether [α ji -∠O i O j F-3σ α ,α ji +3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 14: Determine α ij Is it greater than or equal to 0? If so, go to step 15; otherwise, go to step 16; Step 15: Determine whether [α ji -∠O i O j F-3σ α ,α ji +3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 16: Determine whether [α ji -3σ α ,α ji +∠O i O j F+3σ α ] constitutes a measurement association set, where α ji represents the azimuth angle of the jth sensor with respect to the i-th sensor; Step 17: Determine d ij and l2 satisfy d ij <R j And l2 ≥ R i If yes, execute the fourth step; otherwise, consider the calculation error and execute the first step again.

8. The multi-source sensor measurement fusion method based on angle association according to claim 1 is characterized in that: In step 3, the measurement values ​​within the measurement association set of the active and passive sensors in the system are combined to form the measurement association combination of the system sensors as follows: in, and denote the azimuth and elevation angles of the rth measurement in the passive sensor S1, and represents the azimuth and elevation angles measured by the passive sensor S2, and represents the azimuth and elevation angles measured by the active sensor S3, represents the coordinate position of the tth measurement in active sensor S3, {·} T represents the transposition operation, and m represents the sampling time sequence number.

9. The multi-source sensor measurement fusion method based on angle association according to claim 8, characterized in that: The measured azimuth difference described in step 3 is determined by the following formula: D α =a M -a 3t . Among them, Δ α is the measured azimuth difference, α M yes Determine the straight line and Determine the azimuth of the intersection point M of the straight lines.

10. The multi-source sensor measurement fusion method based on angle association according to claim 9, characterized in that: In step 3, the errors in the measurement association combination are screened out, and the measurement association combination after screening is determined according to the following rules: Set the confidence interval to in Yes α The standard deviation of α If it falls within the confidence interval, then it is considered The three pairs of measurements in the associated combination may come from the same target and are retained; Otherwise, the association combination is determined as an erroneous association combination and is excluded.

11. The method of claim 1, wherein the angle correlation is performed on the measurement correlation combination to obtain the measurement location, wherein: The steps of performing angle correlation fusion positioning on the screened measurement correlation combinations described in step 4 to obtain the target positioning point corresponding to each measurement correlation combination are as follows: The first step is to determine the direction of each positioning point relative to each sensor at each moment; In the second step, the positioning line of each positioning point relative to each sensor is determined; The third step is to determine the distance between each positioning point and the positioning line forming the positioning point; The fourth step is to use the least squares method to obtain the target positioning point corresponding to each measurement association combination.

12. The multi-source sensor measurement fusion method based on angle association according to claim 1, characterized in that: The steps of the positioning line correlation degree correction method described in step 5 are as follows: The first step is to determine the range threshold of the target positioning point according to the following formula: Among them, γ represents the range threshold of the target positioning point, σ x represents the standard deviation of the distance noise of the active sensor on the x-axis, σ y represents the standard deviation of the distance noise of the active sensor on the y-axis, σ z represents the standard deviation of the distance noise of the active sensor on the z-axis; The second step is to determine whether the sum of the distances from each positioning point to the positioning line forming the positioning point is less than γ. If so, the positioning point is retained; otherwise, the positioning point is screened out.

13. The multi-source sensor measurement fusion method based on angle association according to claim 1, characterized in that: The steps of the active sensor position measurement correction method described in step 6 are as follows: The first step is to determine the correction threshold value according to the following formula: Among them, γ σ,x represents the threshold value of the active sensor on the x-axis, σ x represents the standard deviation of the distance noise of the active sensor on the x-axis, γ σ,y represents the threshold value of the active sensor on the y-axis, γ σ,z represents the threshold value of the active sensor on the z-axis, σ z represents the standard deviation of the distance noise of the active sensor on the z-axis; The second step is to determine whether the distance between the active sensor and each positioning point is less than three threshold values ​​at the same time. If so, the positioning point is retained; otherwise, the positioning point is screened out.

Images