A method for target identity determination and collaborative positioning based on multiple passive sensors

Abstract

The present invention discloses a fast calculation method suitable for collaborative positioning of multiple moving targets and detection target identity determination by multiple passive / passive sensors. Existing research results are mostly based on the distinction of multiple targets of sensors in imaging features to determine target identity, or the determination method of one-to-one matching of multiple targets of each sensor until convergence, both of which require a large amount of calculation, and have a great impact on online operation and subsequent positioning convergence speed. The present invention proposes a target identity determination method based on geometric epipolar constraints, which can pre-eliminate a large number of erroneous matching results through the epipolar geometric constraints when multiple sensors observe the same target, and establish a likelihood function based on the geometric characteristics of the direction-finding line in space, and achieve the final correct match based on the likelihood value, effectively reducing the amount of calculation for target identity determination, improving the efficiency of online operation, and providing a basis for the rapid convergence of subsequent collaborative cross-positioning of multiple stations.

Description

Technical Field

[0001] The present invention belongs to the field of electronic reconnaissance, and in particular relates to a method for determining target identity and co-locating multiple passive sensors based on epipolar geometry constraints. Background Art

[0002] For passive sensors, they rely on signals emitted by the target, such as electromagnetic waves, visible light, infrared radiation, etc., to perceive the target's information. A single passive sensor lacks depth information of the target observation and is a weak observation. Therefore, in order to achieve complete observation of the target position, multiple sensors need to cooperate to observe the target. Taking a typical optical camera as an example, when multiple optical cameras locate multiple targets, each camera will form multiple imaging points on the imaging plane for multiple targets. Each target is on the line connecting the imaging point and the optical center of the camera. The multiple direction-finding lines of each camera are matched with each other to form cross-positioning points that exceed the actual number of targets, some of which are the positions of real targets, but most of them are false positioning points. This paper proposes a false positioning point elimination method for passive cross-positioning based on epipolar geometry constraints and the establishment of likelihood functions.

[0003] Research results addressing these issues are limited. Existing methods mostly rely on accurately matching targets based on their imaging features on the sensor, or on pairwise matching of targets on each sensor, selecting the final converged match as the correct match. The former approach is ineffective in some scenarios, such as infrared sensor observation of targets, and requires significant computational effort to extract and match target imaging features. The latter approach suffers from high computational effort and slow convergence in scenarios with a large number of targets, and similarly suffers from poor practical performance. Summary of the Invention

[0004] To address some of the shortcomings of existing methods, this paper proposes an identity matching method for passive sensor cross-localization based on epipolar geometry constraints. This method, based on epipolar geometry constraints, first narrows the range of correct matches using the two-dimensional information of image pixels. It then constructs a likelihood function based on the geometric properties of the direction-finding lines in space from the correct matches to complete the matching process, thereby reducing computational complexity and improving matching speed.

[0005] First, simplify the problem: multiple targets maneuver in three-dimensional space. At each moment, each target forms an image in each observation sensor, with the image center serving as the target's mapping point on the imaging plane. The following algorithm matches each target in each sensor, and the matching results serve as input to determine the target's position. This algorithm is suitable for scenarios where multiple fixed or moving observation sensors are used to coordinate the localization of multiple targets.

[0006] The technical solution adopted by the present invention to solve the technical problem includes the following steps:

[0007] Step 1: Select and fix the world coordinate system and calculate the S of all sensors i (where i = 0, 1, ..., N-1, N is the total number of sensors) The coordinates O of the optical center of each sensor in the world coordinate system i and imaging plane I i The equation α in this world coordinate system i x+β i y+κ i z+ω i =0, sensor S0 is selected as the reference sensor, and the reference sensor is referred to as sensor S0 hereinafter.

[0008] Step 2: According to the coordinates p of a target imaging pixel in sensor S0 j (where j = 1, ..., M, M is the number of target points imaged in sensor S0) and the number of sensor S0 and other sensors S i Baseline O0O between optical centers i , calculated by p j ,O0,O i The unique polar plane R determined by the coordinates in the world coordinate system ij The expression a ij x+b ij y+c ij z+d ij =0, calculate coefficient a ij ,b ij ,c ij ,d ij The formula is as follows:

[0009]

[0010]

[0011] in, Polar plane R ij The normal vector of is the three-dimensional coordinate of the imaging pixel in the world coordinate system.

[0012] Step 3: Calculate the epipolar plane R ij And the corresponding sensor S i Imaging plane I i The intersection line l ij , the intersection line is the epipolar line in the epipolar geometry constraint,

[0013] Polar line l ij The equation in the world coordinate system can be expressed as:

[0014]

[0015] Among them, a ij ,b ij ,c ij ,d ij and α i ,β i ,κ i ,ω i Represent the polar plane R ij and imaging plane I i The expression coefficients, and there is a constraint relationship (a ij :b ij :c ij )≠(α i :β i :κ i ).

[0016] Convert it to sensor S i Imaging plane I i The equation is:

[0017] A ij u+B ij v+C ij =0

[0018] Among them A ij ,B ij ,C ij Polar line l ij Coefficients of the expression in the imaging plane.

[0019] Step 4: Calculate Sensor S i All imaging points in (i≠0) and the epipolar line l on the plane ij Due to the existence of imaging and measurement errors, the calculated epipolar lines and target imaging points are not accurate. Therefore, the correct matching point may not be on the calculated epipolar line, or there may be multiple target points on the epipolar line. Therefore, a reasonable distance threshold dis0 is set. Compare the distance between each point and the epipolar line. If it is greater than the threshold dis0, it is considered that the possibility of the point being the correct matching point is very low and it is discarded; if it is less than or equal to the threshold dis0, it is considered that the possibility of the point being the correct match is high, thereby obtaining the possible matching set M. ij ={p|p∈I i ,dis(p)≤dis0}, used for subsequent matching. The distance calculation method is as follows:

[0020] dis(p)=|A ij u p +B ij v p +C ij |

[0021] Among them, up ,v p For sensor S i (i≠0) an image point p in the imaging plane I i The two-dimensional pixel coordinates on .

[0022] Step 5: Select a new point p in sensor S0 j , and repeat steps 2-4 until all imaging points of sensor S0 obtain the corresponding possible matching set M j ={M ij ,i=1,...,n-1}.

[0023] Step 6: Calculate the predetermined position. Take an imaging point p in sensor S0 j , and the possible matching set M corresponding to this point j Each subset M in ij Take one point from each of the three sensors S0, S1, and S2 as a possible match, and use this set of points to solve the predetermined location. To make the description easier to understand, the following steps of the invention are all based on the three sensors S0, S1, and S2 as an example, and assume that a set of matching points is q0, q1, and q2. Combined with the camera poses q0, q1, and q2 on the corresponding imaging plane I i The two-dimensional coordinates u and v on the target are used to solve the pitch angle and azimuth of the target in the world coordinate system, and the predetermined positions are solved pairwise.

[0024] Take S1 and S2 as an example to explain how to solve the predetermined position. Assume that the azimuth angle measured by S1 is The azimuth measured by S2 is R1, R2 represent the distance between each sensor and the foot point of the direction-finding line, and m is the distance vector between the two direction-finding lines, that is, the line vector between the foot points of the two direction-finding lines. The geometric vector relationship is as follows:

[0025]

[0026] Written in matrix form:

[0027] HR-D=M

[0028] in

[0029]

[0030]

[0031] For S i The coordinate point of .

[0032] Due to the existence of measurement error or mismatch, m is not zero. Take its minimum value, that is, m is the common perpendicular of the two direction-finding lines, then there is a relationship R1⊥m, R2⊥m, that is, HT M=0, then:

[0033]

[0034] The m obtained at this time is the minimum value, so the foot point X i The coordinates of (i=1,2) can be calculated as:

[0035]

[0036] Take the average of the two perpendicular points as the coordinates of the predetermined site of the group of matching

[0037] Step 7: Use the method in step 6 to calculate the predetermined site X of S0S1, S1S2, and S0S2 respectively 01 ,X 12 ,X 02 , then the positioning coordinates of the three sensors are:

[0038] Step 8: Predetermine the target site based on the calculated possible target Solving for the pitch angle and azimuth The calculation formula is as follows:

[0039]

[0040] Step 9: Based on the posterior angle error of the positioning point and the length L of the common perpendicular between the direction-finding lines ij And the distance L between multiple perpendicular points on the same direction-finding line i Construct a likelihood function and calculate the likelihood values ​​of different matches:

[0041]

[0042] Among them Like nmp express The likelihood value of this set of points, k is the weight coefficient for balancing angle error and distance error.

[0043] Step 10: Based on the calculated likelihood values, the Hungarian algorithm is used to select the target matching results to maximize the sum of the likelihood values ​​between each match. The matching principle is that one target generates only one imaging point on one sensor, and one imaging point corresponds to only one target. Its mathematical expression is:

[0044]

[0045] Among them, M i Indicates S i The total number of targets in , ρ nmp express The probability of this set of matches, ρ nmp ∈{0,1}.

[0046] Step 11: Based on the above matching calculation results, use the vector relationship in steps 6-7 to calculate the coordinates of each target in the three-dimensional space at that moment.

[0047] In the above steps, steps 1-5 describe in detail the algorithm steps for reducing false matches during target matching, thus avoiding one-to-one matching of targets observed by multiple sensors. Steps 6-10 describe the detailed algorithm for completing the final match after reducing the number of matching options. Step 6-7 describes the detailed algorithm for collaborative passive localization of the target after obtaining a correct match. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to make the purpose, technical solutions and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0049] Figure 1 Schematic diagram of epipolar geometry constraints for matching observation points of two sensors in the present invention;

[0050] Figure 2 Flowchart of the algorithm in the present invention;

[0051] Figure 3 Schematic diagram of a scenario in embodiment 1 of the present invention;

[0052] Figure 4 Schematic diagram of the epipolar geometry constraint matching effect in the present invention;

[0053] Figure 5 Schematic diagram of the collaborative positioning vector relationship in the present invention;

[0054] Figure 6 Schematic diagram of the passive positioning sensor and target situation in Example 2 of the present invention;

[0055] Figure 7 Schematic diagram of the passive positioning effect in Example 2 of the present invention;

[0056] Figure 8 Passive positioning error result diagram of Example 2 of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described below with reference to the accompanying drawings and examples.

[0058] The purpose of the present invention is to provide a fast calculation method for determining the identity of observed targets during cooperative cross-positioning of multiple passive sensors. This method addresses the problems of unclear imaging features of targets on sensors and high computational complexity and low efficiency of one-to-one target matching. A method for eliminating mismatches based on epipolar geometry constraints is provided. The constraint diagram is shown in FIG. Figure 1 This method relies solely on the geometric relationship between the positions of individual targets in three-dimensional space and the geometric principles of sensor imaging to determine target identity and effectively reduce the amount of matching calculations, thereby accelerating the convergence of passive cross-location.

[0059] like Figure 2 As shown in the algorithm flow chart of the invention, the method for determining target identity by collaborative positioning and recognition of multiple passive sensors based on epipolar geometry constraints proposed in the invention includes the following steps:

[0060] Example 1:

[0061] This embodiment takes the optical camera model as an example, which consists of an optical center and a rectangular imaging plane. Assume that at a certain time k, there are three optical cameras observing five targets in three-dimensional space, and each passive sensor generates five imaging points on the imaging plane, as shown in the schematic diagram. Figure 3 .

[0062] Assume that sensor S i Optical center O i and the jth target P in three-dimensional space j The connection line of sensor S i Imaging plane I i The intersection point p on ij For sensor S i For target P j The imaging point has a Gaussian error with a mean of zero. The algorithm calculation process starts with this as the initial condition.

[0063] Take the optical center O1 of sensor S1 as the center of the world coordinate system, the direction pointing to the imaging plane I1 as the x-axis, the vertical upward direction as the z-axis, and the y-axis of the coordinate system determined by the right-hand rule. Calculate the optical centers O1, O2, O3 of sensors S1, S2, S3, the imaging planes I1, I2, I3, and each target point P j At the pixel center p in the imaging plane ij (denotes the target point P j In the imaging plane I i The expression of the corresponding point on (in the world coordinate system) in the world coordinate system.

[0064] First, epipolar geometry constraints are used to eliminate some false matches.

[0065] Mark the points on the three imaging planes, and mark the points on I1 as p1,...,p5, the points on I2 as p′1,...,p′5, and the points on I3 as p″1,...,p″5. Select a point p1 on the imaging plane I1 and use the three optical centers O1, O2, and O3 to calculate the two epipolar planes R 21 :a 21 x+b 21y+c 21 z+d 21 =0, R 31 :a 31 x+b 31 y+c 31 z+d 31 =0, where R 21 Indicates the polar plane corresponding to the selected point p1 and O2, R 31 Similarly, calculate the coefficient a 21 ,b 21 ,c 21 ,d 21 The formula is as follows:

[0066]

[0067]

[0068] in, Polar plane R 21 The normal vector of is the three-dimensional coordinate of the imaging pixel in the world coordinate system. 31 The calculation method is the same and will not be repeated below.

[0069] Calculate the epipolar plane R 21 The equation of the intersection line with the imaging plane I2 is l 21 , the intersection line is the epipolar line in the epipolar geometry constraint, and its schematic diagram can be referred to Figure 4 The solution formula for the straight line l is as follows:

[0070] R 21 :a 21 x+b 21 y+c 21 z+d 21 =0

[0071] I2:α2x+β2y+κ2z+ω2=0

[0072] (a 21 :b 21 :c 21 )≠(α2:β2:κ2)

[0073] Polar line l in space 21 The equation can be expressed as:

[0074]

[0075] The equation for converting it to the imaging plane I2 of sensor S2 is:

[0076] A 21 u+B 21 v+C21 =0

[0077] Calculate the epipolar line l of the imaging points p′1,...,p′5 in sensor S2 21 The distance between each point and the extreme line is calculated, and a reasonable distance threshold dis0 is set. If the distance between each point and the extreme line is greater than the threshold dis0, it is considered that the possibility of the point being a correct matching point is very low and is discarded; if it is less than or equal to the threshold dis0, it is considered that the possibility of the point being a correct matching point is high, thereby obtaining the possible matching set M of point p1 on sensor S2. 21 ={p|p∈I2,dis(p)≤dis0}. The distance is calculated as follows:

[0078] dis(p)=|A 21 u p +B 21 v p +C 21 |

[0079] Among them, u p ,v p is the two-dimensional pixel coordinate of an imaging point p in the sensor S2 on the imaging plane I2.

[0080] Using the same calculation method as above, we can get the possible matching set M of point p1 on sensor S3: 31 ={p|p∈I3,dis(p)≤dis0}, the possible matching set M1 of point p1 ={{p1},M 21 ,M 31},by Figure 4 (1) is an example, M1 = {{p1}, {p′1, p′2}, {p″1, p″2}}; Figure 4 (2) in the example is used. The possible matching set of p2 is M2 = {{p2}, {p′3, p′4}, {p″2}}. The same algorithm steps can be used to calculate the possible matching sets M3, M4, and M5 of p3, p4, and p5. It can be seen that the new method proposed in this invention can effectively reduce the number of false matches, thereby reducing the number of positioning points and likelihood function calculations in subsequent algorithm steps and reducing the amount of calculation.

[0081] Then, the above results are used to perform further calculations to obtain the correct match and locate the target.

[0082] Based on the set M1 = {{p1}, {p′1, p′2}, {p″1, p″2}} obtained above, we still get four sets of matches. Further calculations are required to find the correct match. Take one element from each of the three subsets of M1 to obtain a set of matches. Find the predetermined location and calculate the likelihood function value.

[0083] Calculate the positioning point of target P1 by matching (p1, p′1, p″1) In this example, it is assumed that the postures of cameras S1, S2, and S3 in the real coordinate system are (0, 0), (0, θ), and (0, -θ) respectively. Figure 5 As shown, first solve the two points (p1, p′1), assuming that the azimuth angle obtained by S1 is The azimuth angle obtained by S2 is R1, R2 represent the distance between each sensor and the foot point of the direction-finding line, and m is the distance vector between the two direction-finding lines, that is, the line vector between the foot points of the two direction-finding lines. The geometric vector relationship is as follows:

[0084]

[0085] Written in matrix form:

[0086] HR-D=M

[0087] in

[0088]

[0089]

[0090] Due to the existence of measurement error or mismatch, m is not zero. Take its minimum value, that is, m is the common perpendicular of the two direction-finding lines, then there is a relationship R1⊥m, R2⊥m, that is, H T M=0, then:

[0091]

[0092] The m obtained at this time is the minimum value, so the foot point X i The coordinates of (i=1,2) can be calculated as:

[0093]

[0094] Take the average of the two perpendicular points as the coordinates of the predetermined site of the group of matching At this time, M is the common perpendicular line L 12 The same steps are used to calculate (p1, p″1) and (p′1, p″1) to obtain the predetermined site and L 13 ,L 23 ,L1,L2,L3, then the final positioning coordinates of the target for this set of matches (p1,p′1,p″1) are

[0095] Use the above-solved coordinates to re-solve the pitch angle and azimuth angle of the sensor to the target and Combined with L 12 ,L 13 ,L 23 ,L1,L2,L3 calculate the likelihood value. The calculation formula is as follows:

[0096]

[0097]

[0098] Among them Like 111 Represents the likelihood value of the set of matching points p′1, p′1, p″1.

[0099] The same steps can be used to calculate the likelihood function values ​​between all possible matches. Based on the calculated likelihood values, the Hungarian algorithm is used to select the target matching results so that the sum of the likelihood values ​​between each match is maximized. The matching principle is that one target generates only one imaging point on one sensor, and one imaging point corresponds to only one target. Its mathematical expression is:

[0100]

[0101] At this point, the matching of the identity of the observation results of the five targets by the three passive sensor observation stations has been completed.

[0102] In this embodiment, the new algorithm proposed in the present invention is used to effectively reduce the amount of calculation for target identity matching. Taking the calculation effect of p1 in the embodiment as an example, if the algorithm is not used, there will be 25 matching results. After using the algorithm, it is reduced to 4 possible matches, laying the foundation for the subsequent rapid calculation of the matching likelihood value and the selection of the correct match.

[0103] Example 2:

[0104] In this embodiment, it is assumed that there are three fixed passive observation stations S1, S2, and S3 in space, whose coordinates in three-dimensional space are (0, 0, 0), (10000, 0, 0), and (5000, 5000, 6000), respectively. There are three maneuvering targets in space, among which T1 performs a sinusoidal maneuver with an initial position of (4000, 4000, 5000), T2 performs a uniform turning maneuver with an initial position of (5000, 0, 5000), and T3 performs a uniform linear maneuver with an initial position of (6000, 0, 5000). During a period of simulation time, the situational relationship between the observation station and the target is as follows: Figure 6 .

[0105] In this embodiment, the simulation step size is set to 0.1s, with a total of 300 steps. In each step, the passive cross positioning step is calculated based on the matching results calculated in the steps of Example 1 to recalculate the positioning coordinates P1, P2, and P3 of the targets T1, T2, and T3. Figure 7 and Figure 8 shown.

[0106] In the simulation of this embodiment, 10 Monte Carlo simulations were performed, and the accuracy of target matching in all simulation steps was 99.6667%. At the same time, when the sensor angle measurement error was set to 0.5 rad, the average error of passive positioning in all directions was 23.38 meters.

[0107] The algorithm mainly requires solving the following six parts:

[0108] 1. Calculate the corresponding epipolar plane R according to the optical center and imaging point;

[0109] 2. Calculate the intersection line (epipole) l between the epipolar plane and the imaging plane;

[0110] 3. Calculate the distance dis between the point and the epipolar line on the imaging plane and take the possible matching set M;

[0111] 4. Solve the predetermined location for each match in the possible matching set and calculate the vertical length and reprojection angle error;

[0112] 5. Calculate the likelihood value Like;

[0113] 6. Use the Hungarian algorithm to match and calculate the positioning points.

Claims

1. A method for determining the identity of a target observed by multiple passive sensors and for collaborative positioning, characterized in that: The steps include: Step 1: Select and fix the world coordinate system and calculate the S of all sensors i (where i = 0, 1, ..., N-1, N is the total number of sensors) The coordinates O of the optical center of each sensor in the world coordinate system i and imaging plane I i The equation α in this world coordinate system i x+β i y+κ i z+ω i =0, select sensor S0 as the reference sensor, which is referred to as sensor S0 in the following text; Step 2: According to the coordinates p of a target imaging pixel in sensor S0 j (where j = 1, ..., M, M is the number of target points imaged in sensor S0) and the number of sensor S0 and other sensors S i Baseline O0O between optical centers i , calculated by p j ,O0,O i The unique polar plane R determined by the coordinates in the world coordinate system ij The expression a ij x+b ij y+c ij z+d ij =0, calculate coefficient a ij ,b ij ,c ij ,d ij The formula is as follows: in, Polar plane R ij The normal vector of is the three-dimensional coordinate of the imaging pixel in the world coordinate system; Step 3: Calculate the epipolar plane R ij And the corresponding sensor S i Imaging plane I i The intersection line l ij , the intersection line is the epipolar line in the epipolar geometry constraint, Polar line l ij The equation in the world coordinate system can be expressed as: Among them, a ij ,b ij ,c ij ,d ij and α i ,β i ,κ i ,ω i Represent the polar plane R ij and imaging plane I i The expression coefficients, and there is a constraint relationship (a ij :b ij :c ij )≠(α i :β i :κ i ); Convert it to sensor S i Imaging plane I i The equation is: A ij u+B ij v+C ij =0 Among them A ij ,B ij ,C ij Polar line l ij The coefficients of the expression in the imaging plane; Step 4: Calculate Sensor S i All imaging points in (i≠0) and the epipolar line l on the plane ij Due to the existence of imaging and measurement errors, the calculated epipolar lines and target imaging points are inaccurate. Therefore, the correct matching point may not be on the calculated epipolar line, or there may be multiple target points on the epipolar line. Therefore, a reasonable distance threshold dis0 is set; the distance between each point and the epipolar line is compared. If it is greater than the threshold dis0, it is considered that the possibility of the point being the correct matching point is very low and it is discarded; if it is less than or equal to the threshold dis0, it is considered that the possibility of the point being the correct match is high, thereby obtaining the possible matching set M. ij ={p|p∈I i ,dis(p)≤dis0}, used for subsequent matching; the distance calculation method is as follows: dis(p)=|A ij u p +B ij v p +C ij | Among them, u p ,v p For sensor S i (i≠0) an image point p in the imaging plane I i The two-dimensional pixel coordinates on ; Step 5: Select a new point p in sensor S0 j , and repeat steps 2-4 until all imaging points of sensor S0 obtain the corresponding possible matching set M j ={M ij ,i=1,...,n-1}; Step 6: Calculate the predetermined position; take an imaging point p in sensor S0 j , and the possible matching set M corresponding to this point j Each subset M in ij Take a point from each as a possible match, and use this group of points to solve the predetermined location; To make the description easier to understand, the following steps of the invention are all based on the three sensors S0, S1, and S2 as an example, and assume that a group of matching points is obtained as q0, q1, and q2; Combined with the camera poses q0, q1, and q2 on the corresponding imaging plane I i The two-dimensional coordinates u and v on the target are used to solve the pitch angle and azimuth of the target in the world coordinate system, and the predetermined positions are solved pairwise; Take S1 and S2 as an example to explain how to solve the predetermined location; assuming that the azimuth angle measured by S1 is The azimuth measured by S2 is R1 and R2 represent the distance between each sensor and the foot point of the direction-finding line. m is the distance vector between the two direction-finding lines, that is, the line vector between the foot points of the two direction-finding lines. The geometric vector relationship is as follows: Written in matrix form: HR-D=M in For S i The coordinate point of Due to the existence of measurement error or mismatch, m is not zero. Take its minimum value, that is, m is the common perpendicular of the two direction-finding lines, then there is a relationship R1⊥m, R2⊥m, that is, H T M=0, then: The m obtained at this time is the minimum value, so the foot point X i The coordinates of (i=1,2) can be calculated as: Take the average of the two perpendicular points as the coordinates of the predetermined site of the group of matching Step 7: Use the method in step 6 to calculate the predetermined site X of S0S1, S1S2, and S0S2 respectively 01 ,X 12 ,X 02 , then the positioning coordinates of the three sensors are: Step 8: Predetermine the target site based on the calculated possible target Solving for the pitch angle and azimuth The calculation formula is as follows: Step 9: Based on the posterior angle error of the positioning point and the length L of the common perpendicular between the direction-finding lines ij And the distance L between multiple perpendicular points on the same direction-finding line i Construct a likelihood function and calculate the likelihood values ​​of different matches: Among them Like nmp express The likelihood value of this set of points, k is the weight coefficient for balancing angle error and distance error; Step 10: Based on the calculated likelihood values, the Hungarian algorithm is used to select the target matching results to maximize the sum of the likelihood values ​​between each match. The matching principle is that one target generates only one imaging point on one sensor, and one imaging point corresponds to only one target. Its mathematical expression is: Among them, M i Indicates S i The total number of targets in , ρ nmp express The probability of this set of matches, ρ nmp ∈{0,1}; Step 11: Based on the above matching calculation results, use the vector relationship in steps 6-7 to calculate the coordinates of each target in the three-dimensional space at that moment.

Images