Photoelectric target association method based on two-dimensional detection line cone confidence model

Abstract

The invention discloses a photoelectric target association method based on a two-dimensional detection line cone confidence model, which comprises the following steps of: judging the stability of a target by adopting a time domain buffer and refresh timeout judgment mechanism, and aligning target data in a time domain by adopting a multivariate function optimization time extrapolation method; a target correlation model is established by adopting a centrum confidence model modeling method based on two-dimensional normal distribution confidence estimation, the correlation probability calculation amount is reduced by adopting a lightweight echelon target correlation judgment method based on target dynamic characteristics, and a target correlation loss function is constructed by adopting the correlation probability. The target association problem is solved based on the greedy thought by taking the minimum loss function as the optimization target, so that the association node associates the target two-dimensional detection information of the plurality of detection nodes under the condition of oblique distance information loss, and complements the three-dimensional information to form the global air situation, thereby achieving the advantages of high detection precision and high concealment of the photoelectric detection equipment. And the information processing and cooperative combat capabilities of multi-source two-dimensional photoelectric detection are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of air situation information processing, and in particular relates to a photoelectric target association method based on a two-dimensional detection line cone confidence model. Background Art

[0002] Optoelectronic information correlation technology can match the search information of multiple optoelectronic search devices and link the search data for the same target on each device. However, existing two-dimensional optoelectronic information correlation technology still faces a series of difficulties. The main ones are that optoelectronic search devices have a large number of targets, many of which are false targets; the detection node's measurement values of the target may contain outliers, which, if not eliminated, will affect the correlation effect; due to the different attitude angles of each detection node and the different startup times of the detection equipment, the detection information of different detection nodes is generally recorded at different times, making direct correlation impossible; the measurement errors of optoelectronic data cause the detection lines of different detection nodes for the same target to be non-uniform in three-dimensional space, making it impossible to determine whether the detection lines point to the same target based on whether the detection lines intersect at a point, and it is difficult to calculate the true intersection point of multiple detection lines pointing to the same target; the projections of the detection lines of multiple detection nodes on the geodetic coordinate system intersect with each other, and most intersection points are false points, making it impossible to solve the correlation problem by simply reducing the dimension.

[0003] In view of the difficulties in the photoelectric information correlation problem, how to judge the authenticity and accuracy of the detected target when there are measurement errors causing the detection line to be out of plane has become a key research direction. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: how to design a photoelectric target association method based on a two-dimensional detection line cone confidence model to overcome the difficulties and shortcomings of current photoelectric information association technology, effectively improve the authenticity and accuracy of target association judgment when there is a state where the detection line is out of plane due to measurement errors of photoelectric sensors, give full play to the advantages of high detection accuracy and strong concealment of photoelectric detection equipment, and enhance the information processing and coordinated command capabilities of multi-source two-dimensional photoelectric detection.

[0005] In order to solve the above method problems, the present invention provides a photoelectric target association method based on a two-dimensional detection line cone confidence model, which includes the following steps:

[0006] 1) A data preprocessing method based on multivariate function optimization first filters out invalid data from detection nodes based on their data characteristics, reducing the amount of computation, lowering the interference with the association of real targets, and improving the accuracy and efficiency of the association. A multivariate function is then constructed using a nonlinear fitting method to extrapolate the target data of each detection node to the current moment before static association.

[0007] 2) The confidence modeling method of the two-dimensional detection line cone based on the two-dimensional normal distribution confidence estimation is to first calculate the target azimuth, high and low angle measurement values (β m ,ε m ) and the corresponding azimuth, elevation, and low-angle true values (β, ε) and the corresponding azimuth and elevation measurement error ranges. After obtaining the azimuth and elevation error ranges for a given confidence level, the azimuth and elevation errors are approximately mapped to the auxiliary coordinate system using coordinate transformation and Taylor expansion. A two-dimensional detection line cone confidence model is established in the auxiliary coordinate system, and the established model is mapped to the geodetic rectangular coordinate system using the coordinate transformation formula.

[0008] 3) A multi-detection node passive target association method based on a two-dimensional detection line cone confidence model first combines the target measurement data time series with the spatial position relationship between the detection nodes that obtain the target measurement data, and then performs an association judgment based on the target's dynamic characteristics. Then, combined with the established two-dimensional detection line cone confidence model, a set of measurement data from any two detection nodes is selected. The intersection area of their cone confidence models is used as the integration area, and the joint probability density function when the two detection node measurement true values intersect at a point within the integration area is used as the integrand. This integral value represents the probability that this set of measurement data points to the same target, and the association probability of the measurement data set that passes the association judgment is calculated. Finally, a greedy algorithm is used to solve the target association set corresponding to the minimum loss function. After completing the passive target association, the approximate intersection point between the detection lines is calculated for each valid association. The intersection point coordinates are obtained by solving the minimum sum of the squared distances between the intersection point and all detection lines in the valid association.

[0009] The beneficial effects of the present invention are embodied in the following aspects:

[0010] 1. The present invention achieves association of two-dimensional detection target information by processing two-dimensional photoelectric detection information and establishing a corresponding two-dimensional detection line cone confidence model after time registration;

[0011] 2. The 2D detection line cone confidence model proposed in this paper effectively solves the problem of detection line spatial non-planarity caused by detection errors during the 2D detection process, providing a reference for subsequent 2D detection correlation;

[0012] 3. After establishing the vertebral confidence model of the target detection line of each detection node, the present invention completes the rapid calculation of the target association probability through two steps of coarse association and fine association, solves the target association measurement data set through the target association algorithm based on the greedy algorithm idea, and then combines the positioning algorithm to calculate the target spatial position information to obtain the target's precise position information. This method reduces the amount of calculation, eliminates outliers, and improves the algorithm's operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is an overall flow chart of a photoelectric target association method based on a two-dimensional detection line cone confidence model of the present invention;

[0014] Figure 2 This is the distributed data preprocessing process of the present invention;

[0015] Figure 3 Schematic diagram of the joint probability density function and the marginal probability density function of the present invention;

[0016] Figure 4 Schematic diagram of the integration area of the joint probability density function of the present invention;

[0017] Figure 5 A schematic diagram of the auxiliary coordinate system establishment process of the present invention;

[0018] Figure 6 It is a schematic diagram of an elliptical cone and a top view in the auxiliary coordinate system of the present invention;

[0019] Figure 7 The implementation process of the target rough association method based on target dynamic characteristics of the present invention;

[0020] Figure 8 This is the target association algorithm flow based on the greedy algorithm idea of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, content and advantages of the present invention more clear, the specific implementation methods of the present invention are further described in detail below with reference to the accompanying drawings and examples.

[0022] like Figure 1 As shown in the figure, a photoelectric target association method based on a two-dimensional detection line cone confidence model is proposed. This method uses a data preprocessing method based on multivariate function optimization to filter the target data sent by the detection node to the association node, eliminating data that clearly does not belong to the real target. The filtered target data is then extrapolated to the same time node for subsequent calculations. A two-dimensional detection line cone confidence model modeling method based on two-dimensional normal distribution confidence estimation is used to establish a confidence model for the target data. A photoelectric target association method based on the two-dimensional detection line cone confidence model is used to first determine the target association based on its dynamic characteristics. The association probability of the targets that pass the association determination is then calculated based on the confidence model. Finally, a loss function is calculated based on the association probability. A greedy algorithm is used to determine the target association set that minimizes the loss function. This combined approach addresses the issues of effectiveness, real-time performance, and stability in the association of two-dimensional photoelectric detection information.

[0023] A photoelectric target association method based on a two-dimensional detection line cone confidence model comprises the following steps:

[0024] 1) Data preprocessing method based on multivariate function optimization

[0025] Step 1, such as Figure 2 As shown, data preprocessing is performed on the detected distributed data. This data preprocessing filters out invalid data from the detection node, reduces the amount of computation, reduces interference with the association of the real target, and improves the accuracy and efficiency of the association. After the detection node obtains the target measurement data, if it has previously received three or more measurement data of the target and the new measurement data can pass the outlier judgment, the data is sent to the associated node. If it has not received three or more measurement data of the target, it determines whether the data is an outlier. If it is an outlier, it is discarded; if not, the data is stored. For the data of the same target, when three data are stored for the first time, the three target data and the corresponding acquisition time must be sent to the associated node at once.

[0026] Step 2: Time alignment based on multivariate function optimization. A multivariate function is constructed by a nonlinear fitting method to ensure that all received target data are extrapolated to the current moment before the detected photoelectric data are statically associated.

[0027] Let the elevation angle ε of the jth target of the i-th detection node be ij The function of time t is L ij (t), the detection data set used to calculate the fitting curve is E ij ={ε ij (t k ),ε ij (t k+1 ),…,ε ij (t k+m-1 )}, where m is the number of probe data used, ε ij (t k ) is t k The momentary elevation angle ε ij Construct a set of basis functions related to time t is an n-order polynomial function about time t, and the basis function contains N elements. To ensure the existence of a solution, N must satisfy N≤m. Assume that a set of parameters A={a0,a1,…,a N}, the expression for constructing the curve to be fitted is

[0028]

[0029] In order to ensure that the constructed polynomial fits the existing detection data, the parameter set A used to construct the expression should satisfy the minimum sum of the squares of the difference between the value at the recording time corresponding to the detection data used for fitting and the detection data, and the result is

[0030]

[0031] Convert the minimum value problem into a multivariate function extreme value problem, let Simplification can get a linear equation system. Due to the uncorrelated characteristics of the elements in the basis function, the coefficient matrix of the equation system must be full rank. The equation system has a unique solution. The solution is to get all the parameters of the fitting curve expression:

[0032]

[0033] Substituting into the fitting curve expression is

[0034]

[0035] This method is used to construct fitting functions of the azimuth and elevation angles of the measured values, and the time variable is substituted to align the measured values to the current moment.

[0036] 2) Modeling method of two-dimensional detection line cone confidence model based on two-dimensional normal distribution confidence estimation

[0037] A two-dimensional detection line cone confidence model is established based on the target measurement data. The two-dimensional detection line cone confidence model can establish a set of detection lines in space based on the target measurement data and measurement error. This set contains the true value corresponding to the target measurement data with a given probability. The obtained model can be used for the precise calculation of the subsequent loss function. The detailed calculation steps are as follows:

[0038] (1) Calculate the target measurement value (β m ,ε m ) and the corresponding range of the true value (β,ε) and the corresponding range of the azimuth and elevation angle measurement errors.

[0039] ①Calculate the given azimuth angle measurement value β m 、High and low angle measurement value ε m The joint probability density function characteristics of the true value of the lower azimuth angle β and the true value of the upper and lower angles ε.

[0040] The measurement errors Δβ and Δε of the detection node in azimuth and elevation angle are respectively subject to standard deviation σ β and σ ε The standard normal distribution of m , ε m The relationship with the true values β and ε is

[0041]

[0042] Therefore, the true azimuth and elevation angles corresponding to the given measurement values are respectively obeyed by Since β and ε are independent of each other, (β,ε) obeys the two-dimensional normal distribution (β,ε)~N(β m ,εm ,Δβ,Δε,0), then the corresponding probability density function f(β,ε) is

[0043]

[0044] For a given probability density 0<C≤(2πσ β σ ε ) -1 , taking the logarithm of both sides of the equation and transforming it, we can get

[0045]

[0046] From the above formula, we can see that for a given measurement value (β m ,ε m ), the joint probability density of its true value (β,ε) is calculated based on the measured value (β m ,ε m ) is the center of the ellipse, and On the ellipse with the major and minor semi-axes, that is, the contour lines of the joint probability density expand outward in an elliptical shape from the center of the ellipse.

[0047] ② Calculate the probability density on the edge contour line of the joint probability density function under a given confidence level P, and solve the value range of the azimuth and elevation angle measurement errors of the elliptical area corresponding to the contour line.

[0048] Measurement value (β m ,ε m The probability that the true value (β,ε) of the equation (β,ε) falls within the given elliptical region Ω is

[0049]

[0050] The integral region Ω' is obtained by translating the center of the original integral region ellipse to the origin of the coordinate system after variable substitution. The joint probability density function and its marginal probability density function about (β,ε) after translation are as follows: Figure 3 shown.

[0051] For ease of calculation, the integrand and the integral domain are converted to the polar coordinate system to obtain the functional relationship between the angle θ and the corresponding radius r, thereby obtaining the probability calculation formula in polar coordinates:

[0052]

[0053] After integrating r and θ, we can get

[0054]

[0055] Measurement value (β m ,ε mThe expression for the probability that the true value (β,ε) of ) falls within the given elliptical area Ω is:

[0056] P=1-2πσ β σ ε C

[0057] From the probability expression, we can see that if the confidence level P is given, the probability density of the integral edge C = (1-P) / 2πσ can be inversely solved β σ ε , and then calculate the values of the major and minor axes of the elliptic integral area under the confidence level, which is the measurement error range under the given confidence level. The top view of the joint probability density function under the given confidence level is as follows: Figure 4 As shown. The entire elliptical area is the range of azimuth and elevation angle measurement errors under a given confidence level.

[0058] (2) After obtaining the range of azimuth error and elevation error under a given confidence level, the azimuth error and elevation error are approximately mapped to the auxiliary coordinate system by combining coordinate transformation and Taylor expansion, and a two-dimensional detection line cone confidence model is established in the auxiliary coordinate system. The established model is mapped to the geodetic rectangular coordinate system by combining the coordinate transformation formula.

[0059] ① Establish an auxiliary coordinate system and a conversion function between the translation horizontal coordinate system and the auxiliary coordinate system.

[0060] like Figure 5 As shown, first translate the detection node to the horizontal coordinate system O ps X ps Y ps Z ps With the Z axis as the rotation axis, rotate clockwise until the X axis and the detection line are at the X axis. ps O ps Y ps The projections on the plane coincide with each other, and then the Y-axis of the rotated coordinate system is used as the rotation axis to rotate clockwise until the Z-axis coincides with the target detection line.

[0061] Detection node translation horizontal coordinate system O ps X ps Y ps Z ps Lower coordinate (x ps ,y ps ,z ps ) and the newly constructed auxiliary coordinate system O f X f Y f Z f Lower coordinate (x f ,y f ,z f ) coordinate transformation formula is

[0062]

[0063] ②Construct a two-dimensional detection line cone confidence model in the auxiliary coordinate system.

[0064] The O with measurement error ps X ps Y ps Z ps The lower coordinates (x, y, z) are in polar coordinate form, that is,

[0065]

[0066] Substituting into the coordinate transformation formula, we can simplify it and get

[0067]

[0068] Based on the current measurement error, an elliptical cone is constructed on the auxiliary coordinate system. f O f Y f The center of the ellipse projected on the plane coincides with the origin, the major semi-axis coincides with the Y axis, the minor semi-axis coincides with the X axis, and the altitude of the elliptical cone coincides with the Z axis. The front view and top view are as follows Figure 6 shown.

[0069] Elliptical cone in X f O f Z f The angle between the tangent line of the plane and the Z axis is ε'. f O f Z f The angle between the tangent line of the plane and the Z axis is β'. Under the conditions of Δε→0 and Δβ→0, the tangent values of ε' and β' are obtained respectively.

[0070]

[0071] For any detection range r within the detection power, under a given confidence level P, let ε m is the measured value of the target elevation angle, and the obtained tangent values of ε' and β' can be easily verified to obtain the equation

[0072]

[0073] Therefore, in the auxiliary coordinate system O f X f Y f Z f Order β' to take the value range The value range of ε' is And satisfy the constraints

[0074]

[0075] The elliptical cone constructed at this time can approximately map the set of true values corresponding to the target measurement values under a given confidence level from the translation horizontal coordinate system to the auxiliary coordinate system. The model expression at this time is easy to give, which is

[0076]

[0077] ③Transform the cone confidence model established in the auxiliary coordinate system into the geodetic rectangular coordinate system.

[0078] Combined with the coordinate transformation formula, the cone confidence model is transformed from the auxiliary coordinate system O f X f Y f Z f Down-convert to the translation horizontal coordinate system O ps X ps Y ps Z ps The model expression is

[0079]

[0080] Combine the system origin position and the vehicle position information to translate the cone confidence model from the horizontal coordinate system O ps X ps Y ps Z ps Converted to the geodetic rectangular coordinate system OXYZ, the model expression is

[0081]

[0082] The expression is recorded as g(x,y,z,β m ,σ β ,ε m ,σ ε ,x O ,y O ,z O ,R,P), where (x,y,z) is the value of the required elliptical cone in the geodetic coordinate system; (β m ,σ β ,ε m ,σ ε ) is the target measurement value and measurement error, and the ray where the measurement value is located coincides with the central axis of the elliptical cone; (x O ,y O ,z O ) is the coordinate of the detection node in the geodetic coordinate system; R is the maximum power of the detection; P is the given confidence level.

[0083] 3) Multi-detection node passive target association method based on two-dimensional detection line cone confidence model

[0084] Step 1: Multi-detection node passive target association based on the two-dimensional detection line cone confidence model. First, target association is determined based on the target's dynamic characteristics. Then, the association probability of the measurement data set that passes the association determination is calculated using the two-dimensional detection line cone confidence model. The resulting association probability is used to construct a loss function for any set of measurement data that is an associated combination. Finally, a greedy algorithm is used to determine the target association set that minimizes the loss function.

[0085] (1) According to the dynamic characteristics of the target, the measurement data of any two detection nodes are correlated and judged, such as Figure 7 As shown, let the position of the detection node i in the geodetic coordinate system be (x Oi ,y Oi ,z Oi ), the detection power is R i , the detection value of target j is (β mij ,ε mij ); the position of the detection node n in the geodetic coordinate system is (x On ,y On ,z On ), the detection power is R n , the detection value of target l is (β mnl ,ε mnl ), the probability that the target j of detection node i and the target l of detection node n point to the same target is The specific steps are:

[0086] ① Project the detection line of detection node i to target j and the detection line of detection node n to target l within the detection power into line segments on the XOY plane, and calculate the intersection of the two line segments; the line segments where the corresponding detection lines within the detection power of the two detection nodes are projected on the XOY plane are

[0087]

[0088] The simultaneous equations calculate the intersection position (x', y') of the straight lines where the two detection lines are projected on the XOY plane.

[0089]

[0090] ② If the intersection value does not meet the constraints of x value, it means that the intersection position exceeds the maximum detection power of the detection line, and the two detection lines must not point to the same target. The judgment ends; if the intersection value satisfies the constraint condition of x value, the height of target j and target l when the intersection point is used as the target projection are calculated respectively.

[0091]

[0092] ③ Given a height threshold Z, if |zmij -z mnl When |>Z, the two detection lines must not point to the same target. Otherwise, it may continue to judge the same target. When y'∈[min(y Oi ,y On ),max(y Oi ,y On )], execute ⑤;

[0093] ④If (β mij (t k )-β mij (t k-1 ))(β mnl (t k )-β mnl (t k-1 ))>0, the detection lines of the two detection nodes may point to the same target, and execute ⑥; otherwise, the two detection nodes must not point to the same target, and let and withdraw from judgment;

[0094] ⑤If (β mij (t k )-β mij (t k-1 ))(β mnl (t k )-β mnl (t k-1 ))<0, the detection lines of the two detection nodes may point to the same target, execute ⑥, otherwise the two detection nodes must not point to the same target, let and withdraw from judgment;

[0095] ⑥If ε mij (t k )-ε mij (t k-1 )>0 and β mnl (t k )-β mnl (t k-1 )>0, the targets pointed by the two detection nodes move in the same direction and move closer to the detection node. The two detection lines may point to the same target, and the association probability It is calculated by combining the two-dimensional detection line cone confidence model. In other cases, And withdraw from judgment.

[0096] (2) Combined with the two-dimensional detection line cone confidence model, the association probability of the measurement data group of any two detection nodes through rough association judgment is calculated From the previous calculation, we can get that under the confidence level P, the corresponding elliptical cone expressions of the target j of the detection node i and the target l of the detection node n in the geodetic coordinate system are g(x ij ,y ij ,z ij ,β mij ,σ βi ,ε mij ,σ εi ,x Oi ,y Oi ,z Oi ,R i ,P i ) and g(x nl ,y nl ,z nl ,β mnl ,σ βn ,ε mnl ,σ εn ,x On ,y On ,z On ,R n ,P n ), abbreviated as g ij and g nl . Association probability The detailed calculation steps are:

[0097] ① Calculate the true value of the azimuth and elevation angle of the detection line of detection node i to target j and detection node n to target l for each point (x, y, z) in the intersection area of the two elliptical cones. The calculation formula is:

[0098]

[0099] ②Calculate the probability density when each point in the intersection area of the two elliptical cones corresponds to a set of true measurement values. The calculation formula is:

[0100]

[0101] ③ Integrate the probability density G(x, y, z) of the intersection area of the two elliptical cones to obtain the association probability of targets j and l under a given confidence level The calculation formula is

[0102]

[0103] (3) Based on the greedy algorithm, the corresponding target association set when the loss function is minimized is set T k ={i1,i2,…i N} is a set of associated combinations of the measurement set, and its loss function is defined as

[0104]

[0105] like Figure 8 As shown, let the number of targets detected by the nth detection node be M n , the measurement data association algorithm steps are:

[0106] ①Algorithm initialization:

[0107] a. For any two detection lines of any two detection nodes, first determine whether there is an intersection between the projections of the two detection lines on the XOY plane within the given detection power. If so, proceed to the next step. If not, set i n =1,…,M n ;i l =1,…,M l ; (The value restrictions in subsequent steps are the same and will not be repeated), it is assumed that the two detection lines do not point to the same target.

[0108] b. Take the intersection of the projections of the two detection lines on the XOY plane as the projection of the target on the XOY plane, and calculate the target height under the corresponding two detection lines. If the absolute value of the target height difference is less than the given threshold, proceed to the next step of judgment, otherwise It is assumed that the two detection lines do not point to the same target.

[0109] c. Combine the measured values of the two detection lines at the previous moment to determine the direction of azimuth change. If the projections of the two detection lines are outside the projections of the two detection nodes, then proceed to the next step when the azimuth changes in the same direction. Otherwise, If the projections of the two detection lines are outside the projections of the two detection nodes, then the next step is judged when the azimuth angle changes in the opposite direction. Otherwise,

[0110] d. Combine the measured values of the two detection lines at the previous moment to determine the direction of the height angle change. If the height angles of the detection lines change in a positive direction, the target is close to the detection node and the next step of calculation is entered. Otherwise, the target is far away from the detection node and the next step of calculation is entered.

[0111] e. Calculate the probability that two detection lines point to the same target at a given confidence level

[0112] f. Repeat steps ae until any two detection lines of any two detection nodes are traversed.

[0113] ②Construct possible association sets and simplify them:

[0114] a. The i-th detection node of the o-th detection node containing the most measurement values o (i o =1) Group measurement value With the i-th detection node of other n (i n =1,…,M n ; n=1,…,N; n≠o) group of measurement values The probability of pointing to the same target Compared with the threshold P, if it is greater than the threshold, it is considered that the i-th detection node of the n-th detection node n Group measurement values belong to the set of possible associations

[0115] b. Take the i-th node from the measurement of the n-th (n=1, n≠o) detection node n (i n =1,…,M n ) group measurement, if Combine it with the other detection nodes in the measurement belonging to the associated set Measurement of k (i k =1,…M k ; k=n+1,…N) the probability of pointing to the same target Compared with the threshold P, if it is less than the threshold, then select Delete i k .

[0116] c. Set n = n + 1 and return to b, repeat the cycle until n = N - 1, and obtain a simplified set of possible association sets.

[0117] ③Construct the final association set:

[0118] a. From the simplified association set Choose one measurement from each detection node measurement (if The measurement of the detection node is included in the Calculate all possible combinations A p The loss function C(A p ), select the association combination with the minimum cost As a reduced association set The final connection.

[0119] ④ Let i1 = i1 + 1, return to ②, repeat the cycle until i1 = M1, and obtain the association set containing all the final association sets

[0120] ⑤ Select a final association set from the association set T with a small loss function and the most measurement values As the first confirmed association T1 * , while deleting the collection middle The associated combination Then, in the remaining sets, according to the same premise, select the association combination with the smallest loss function as the second determined association And so on, finally select Q related combinations {T i * ,i=1,…,Q} is the final effective association about Q targets.

[0121] ⑥ Delete all measurement data of detection node o and all successfully associated measurement data of other detection nodes from the measurement data set. If there are still elements in the measurement data set at this time, return to ②. If the measurement data set is empty, terminate the calculation.

[0122] Step 2: After completing the passive target association, for each valid association, calculate the approximate intersection point between the detection lines. The intersection point coordinates can be obtained by solving the minimum sum of the squares of the distances between the intersection point and all the detection lines in the valid association.

[0123] For each valid association Calculate the approximate intersection between the detection lines and set the effective correlation The coordinates of the intersection point in the geodetic coordinate system are (x, y, z), and the distance between the intersection point and any detection line in the valid association is

[0124]

[0125] Then the sum of the squares of the distances between the intersection point and all detection lines in the valid association is

[0126]

[0127] When the measurement error is zero, it is obvious that G(x,y,z)=0 achieves the minimum value. When the measurement error is not zero, the sum of the squares of the distances between the intersection point and all the detection lines should be minimized. The problem of solving the intersection coordinates (x,y,z) can be converted into a problem of finding the extreme value of a multivariate function. Let Simplifying the system of linear equations, the solution is the intersection point.

Claims

1. A photoelectric target association method based on a two-dimensional detection line cone confidence model, characterized in that: It includes the following steps: 1) Data preprocessing method based on multivariate function optimization The stability of the navigation target of the detection node is judged by the time domain buffer and refresh timeout judgment mechanism. The validity of the navigation target is judged by comparing the high-order difference of the target time series with the real target time domain motion constraint. The target detection data that does not meet the stability or validity requirements is eliminated, and only the real target data with stable navigation is uploaded to the associated node, reducing the invalid calculation load of the associated node. A time alignment method based on multivariate function optimization is used to construct a fitting function of the target measurement data with respect to time. The received historical time series data is used as the data to be fitted to construct a multivariate function of the fitting function parameters. The fitting function should ensure that the multivariate function takes the minimum value. Thus, the problem of solving the fitting function is transformed into the problem of finding the extreme value of the multivariate function. All parameters of the fitting function are solved, and the data are aligned in the time domain using the fitting function of the target measurement data. 2) Modeling method of two-dimensional detection line cone confidence model based on two-dimensional normal distribution confidence estimation The two-dimensional measurement data of the target obeys a two-dimensional normal distribution, and its corresponding true value fills the entire variable space. Under a given confidence level, the corresponding true value set is converged into a finite set. The graph formed by the detection lines corresponding to this set in three-dimensional space is the two-dimensional detection line cone confidence model. The specific steps for establishing the model are as follows: The first step is to derive the relationship between confidence and probability density under two-dimensional normal distribution, and inversely solve the confidence interval based on the confidence; first, the target measurement value is taken as the average point of the two-dimensional normal distribution, and the joint probability density function of the target azimuth angle, high and low angle measurement true value is established. Among them, the true value azimuth angle β and elevation angle ε corresponding to the measured value obey β m is the azimuth angle measurement value, σ β is the corresponding standard deviation, ε m is the azimuth angle measurement value, σ ε is the corresponding standard deviation; then let the joint probability density be C and substitute it into the equation to obtain the joint probability density contour expression It is easy to see from the expression that the contour line of the joint probability density expands outward from the average value point in an elliptical shape; then the relationship between the confidence level and the probability density is established by integrating the joint probability density function according to the elliptical area Ω contained in the given contour line. Finally, given the confidence level, the relationship between confidence level and probability density is used to simplify the contour line expression. From the expression, we can get the confidence interval under a given confidence level as Minor semi-axis The elliptical area of ​​ The second step is to establish an auxiliary coordinate system and derive the mapping relationship between the target polar coordinate data in the translation horizontal coordinate system and the data in the auxiliary coordinate system; first, through two coordinate rotations, the Z axis of the translation horizontal coordinate system is made to coincide with the target detection line, and the auxiliary coordinate system O is established. f X f Y f Z f and the translation horizontal coordinate system O ps X ps Y ps Z ps The mapping formula Then the target measurement data in the translation horizontal coordinate system is converted to the auxiliary coordinate system, and the Z coordinate system is used to calculate the target measurement data in the auxiliary coordinate system. f The axis is the vertical line to construct the elliptical cone; finally, the elliptical cone is calculated in the auxiliary coordinate system Y f O f Z f Projection on the plane and Z f The angle β', X f O f Z f Projection on the plane and Z f The mapping relationship between the angle ε' between the axes and the true value of the azimuth angle β and the true value of the elevation angle ε in the horizontal coordinate system Among them, Δβ and Δε are the measurement errors of azimuth and elevation angle respectively; The third step is to establish a two-dimensional detection line cone confidence model in the auxiliary coordinate system according to the confidence interval, and transform it to the translation horizontal coordinate system through the coordinate system mapping relationship; first, according to the relationship between the confidence and confidence interval obtained in the first step, calculate the confidence interval expression under the given confidence level; then, the elliptical cone constructed in the second step in the auxiliary coordinate system is transformed into the Y coordinate system. f O f Z f Projection on the plane and Z f The angle β', X f O f Z f Projection on the plane and Z f Substituting the mapping relationship between the angle ε' between the axes and the true value of the azimuth angle β and the true value of the elevation angle ε in the horizontal coordinate system into the confidence interval expression, we can obtain the elliptical cone in the X-axis of the auxiliary coordinate system that contains all the true value sets of the target measurement data under a given confidence level. f O f Y f The expression of the projection on the plane, and then the expression of the target two-dimensional detection line cone confidence model in the auxiliary coordinate system is obtained where x f 、y f 、z f The confidence model of the two-dimensional detection line cone is in the auxiliary coordinate system X f Axis, Y f Axis, Z f The value on the Z axis f The value range on the axis is determined by the maximum detection distance of the photoelectric device; finally, the target two-dimensional detection line cone confidence model is converted to the earth rectangular coordinate system OXYZ by combining the coordinate system mapping relationship. The model expression is: The expression is recorded as g(x,y,z,β m ,σ β ,ε m ,σ ε ,x O ,y O ,z O ,R,P), where (x,y,z) is the value of the required elliptical cone in the geodetic coordinate system; (β m ,σ β ,ε m ,σ ε ) is the target measurement value and measurement error, and the ray where the measurement value is located coincides with the central axis of the elliptical cone; (x O ,y O ,z O ) is the coordinate of the detection node in the geodetic coordinate system; R is the maximum power of detection; P is the given confidence level; in the following text, the two-dimensional detection line cone confidence model established by the detection node i for the target j in the geodetic coordinate system is abbreviated as g ij ; 3) Multi-detection node passive target association method based on two-dimensional detection line cone confidence model A lightweight hierarchical target association judgment method based on the dynamic characteristics of the target is used to make a preliminary judgment on the target association. First, the association of the detection nodes is judged based on whether there is an intersection between the projections of the detection lines on the ground. If there is an intersection, the association of the detection nodes is judged based on the spatial relationship of the detection lines at the projection intersection. If the judgment is passed, the variation pattern of the measurement data under different relative position relationships between the target and the detection node is analyzed, and the variation trend of the target measurement value under different detection nodes is dynamically judged to determine whether the detection lines of the detection nodes point to the same target. For target detection lines that are judged as unrelated by the lightweight echelon target association judgment method based on target dynamic characteristics, their association probability is set to 0; for target detection lines that are judged as related by the lightweight echelon target association judgment method based on target dynamic characteristics, a target association judgment method based on a two-dimensional detection line cone confidence model is used to make further association judgments; first, corresponding two-dimensional detection line cone confidence models are established for detection lines of different detection nodes in the geodetic rectangular coordinate system; then, the joint probability density function G(x, y, z) of the true value azimuth and elevation angle of the detection line corresponding to the intersection area of the two confidence models is calculated. where β ij , ε ij , β nl , ε nl They are the true values of the measurement data of detection node i for target j and detection node n for target l at any point in the model intersection area, and the calculation formula is: Where (x, y, z) is any point in the intersection area of the model, (x oi ,y oi ,z oi )、(x on ,y on ,z on ) is the position coordinate of detection nodes i and n in the geodetic coordinate system; finally, in the intersection area Ω(g ij ∩g nl ) to integrate the joint probability density function, the integral formula is The integration result is used to represent the probability that two detection lines point to the same target. For detection lines whose integration result is less than a given threshold, they are judged as uncorrelated and the association probability is set to 0. For detection lines whose integration result is greater than or equal to the given threshold, the association probability is set to the integration result. The target association problem solving method based on the greedy algorithm is adopted to quickly obtain the suboptimal solution of the problem through iterative local optimal solution and pruning. The algorithm is mainly divided into 5 steps. The first step is algorithm initialization. First, a loss function is constructed according to the association probability. The loss function decreases as the sum of the association probabilities between any two detection lines in the selected association set increases. The calculation result of the loss function is used to judge the quality of the current association set. Then, the probability that any two detection lines of any two detection nodes point to the same target is calculated by combining the two target association judgment methods mentioned above as the initial input of the algorithm. The second step is to construct a possible association set. The detection node with the largest number of measurement values is selected, and the measurement values in the node and the measurement values of other nodes with a probability greater than the threshold pointing to the same target are combined. Add the possible association set; the third step is to simplify the possible association set; calculate the association probability of the two detection lines contained in any two association nodes other than the selected detection node in the possible association set, and remove the corresponding measurement values when the association probability is less than the threshold from the possible association set; the fourth step is to construct the final association set; select the association combination with the smallest loss function in the simplified possible association set as the final association set of the target; the fifth step is to simplify the original association set; delete all the detection lines contained in the nodes selected in the second step and the detection lines of other detection nodes contained in the final association set from the original association set. If the original association set is not an empty set, return to the second step and start again; after the algorithm is executed, the final association set of all targets corresponding to the original association set can be obtained; After completing the passive target association, for each valid association, the approximate intersection point between the detection lines is calculated, and the intersection point coordinates can be obtained by solving the minimum sum of the squares of the distances between the intersection point and all the detection lines in the valid association.

Images