Target tracking device and target tracking method

The target tracking device enhances prediction accuracy by integrating observation data with a motion model that accounts for direction changes, ensuring precise tracking through a closed-loop feedback system.

JP7580637B2Active Publication Date: 2024-11-11MITSUBISHI ELECTRIC CORP
View PDF 13 Cites 0 Cited by

Patent Information

Application Number
JP2023573317
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2022-02-17
Publication Date
2024-11-11
Estimated Expiration
2042-02-17

AI Technical Summary

Technical Problem

Existing target tracking devices fail to accurately predict the state of a tracking target due to the lack of consideration for changes in the direction of movement, particularly when the target's attitude angle changes, leading to deteriorated prediction accuracy.

Method used

A target tracking device and method that incorporates a smoothing section and prediction section to estimate the target's state by combining observation position information with a predicted value, using a motion model that includes an acceleration term related to changes in the target's direction of movement, and corrects the state estimation using a closed-loop feedback mechanism.

Benefits of technology

Enables accurate prediction of the target's state by considering changes in its movement direction, thereby improving prediction accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007580637000026
    Figure 0007580637000026
  • Figure 0007580637000027
    Figure 0007580637000027
  • Figure 0007580637000028
    Figure 0007580637000028
Patent Text Reader

Abstract

This target tracking device comprises: a smoothing unit (2) that acquires observed position information indicating an observed position of a tracking target, acquires a predicted value for a tracking target state at a position observation time, being the time at which the observed position was observed, said predicted value having been predicted one sampling prior to the position observation time, and uses the observed position information and the predicted value to estimate the tracking target state at the position observation time; and a predicting unit (3) that uses, as a movement model for the tracking target, a movement model including an acceleration term related to a change in a tracking target maneuver direction, predicts the tracking target state one sampling after the position observation time on the basis of the state estimated by the smoothing unit (2), and outputs the predicted value for the tracking target state to the smoothing unit (2).
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present disclosure relates to a target tracking device and a target tracking method. [Background technology]

[0002] There is a target tracking device that predicts the state of a tracked target by using a motion model of the tracked target (for example, see Non-Patent Document 1). The state of the tracked target includes, for example, the position of the tracked target, the velocity of the tracked target, and the aerodynamic coefficient of the tracked target. The target tracking device includes a smoothing unit and a prediction unit. The smoothing unit acquires observation position information indicating the observation position of the tracked target, and acquires a predicted value from the prediction unit. The predicted value is a prediction result of the state of the tracked target at the position observation time, predicted by the prediction unit one sampling before the position observation time, which is the time when the observation position is observed. The prediction unit predicts the state of the tracked target one sampling after the position observation time from the state estimated by the smoothing unit, using a motion model of the tracked target (see equation (8)), and outputs the predicted value of the state of the tracked target to the smoothing unit. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Jingshuai Huang, Hongbo Zhang, Guojian Tang, Weimin Bao,“Radar Tracking for Hypersonic Glide Vehicle Based on Aerodynamic Model”,2017 29th Chinese Control And Decision Conference (CCDC),28-30 May 2017,p.1080-1084 Summary of the Invention [Problem to be solved by the invention]

[0004] The future state of the tracked target also changes with a change in the direction of movement of the tracked target. The motion model of the tracked target used by the prediction unit of the target tracking device disclosed in Non-Patent Document 1 has a problem in that it does not take into account changes in the direction of movement of the tracked target. For this reason, in the target tracking device, for example, when the direction of movement of the tracked target changes with a change in the attitude angle of the tracked target, the prediction accuracy of the state of the tracked target by the prediction unit may deteriorate.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a target tracking device and a target tracking method that are capable of predicting the state of a tracked target while taking into account the direction of movement of the tracked target. [Means for solving the problem]

[0006] The target tracking device according to the present disclosure comprises: The present invention includes a smoothing unit and a prediction unit that form a closed loop, and the smoothing unit Obtaining observation location information indicating an observation location of a tracking target, and The target of tracking A predicted value of the state of the tracked target at the position observation time, which is predicted one sampling before the position observation time at which the observation position is observed, is obtained, and the predicted value is combined with the observed position information. Prediction of the state of the tracked object From this, the state of the tracked target at the time of position observation is estimated. The observation location information and the location at the time of the observation The target of tracking State Prediction and from , Changes in the direction of movement of the target included in the motion model of the target acceleration term The constant acceleration vector derived from of The prediction unit corrects the From the state estimated by the smoothing section, the state of the tracked target one sampling later than the position observation time is predicted, and the predicted value of the state of the tracked target is output to the smoothing section. , said It is something. Effect of the Invention

[0007] According to the present disclosure, it is possible to predict the state of a tracking target taking into account the direction of movement of the tracking target. [Brief description of the drawings]

[0008] [Figure 1] 1 is a configuration diagram showing a target tracking device according to a first embodiment. [Diagram 2]1 is a hardware configuration diagram showing hardware of a target tracking device according to a first embodiment. [Diagram 3] 1 is a hardware configuration diagram of a computer in the case where the target tracking device is realized by software, firmware, or the like. [Figure 4] FIG. 13 is an explanatory diagram showing a constant acceleration vector. [Diagram 5] 4 is a flowchart showing a target tracking method which is a processing procedure of the target tracking device. [Figure 6] FIG. 11 is a configuration diagram showing a target tracking device according to a second embodiment. [Figure 7] FIG. 11 is a hardware configuration diagram showing hardware of a target tracking device according to a second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] In order to describe the present disclosure in more detail, embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0010] Embodiment 1 FIG. 1 is a configuration diagram showing a target tracking device according to the first embodiment. FIG. 2 is a hardware configuration diagram showing hardware of the target tracking device according to the first embodiment. The target tracking device shown in FIG. 1 includes a tracking gate unit 1, a smoothing unit 2, and a prediction unit 3.

[0011] The tracking gate unit 1 is realized by, for example, a tracking gate circuit 11 shown in FIG. The tracking gate unit 1 acquires observation position information indicating the observation position of the tracking target from, for example, a target observation device (not shown), and acquires a predicted value of the state of the tracking target from the prediction unit 3. The predicted value indicates a prediction result of the state of the tracking target at the position observation time, which is predicted one sampling before the position observation time, which is the time when the observation position is observed. The tracking gate unit 1 calculates the difference between the observed position of the tracking target indicated by the observed position information and the predicted position of the tracking target included in the predicted value. If the difference between the observed position and the predicted position is equal to or smaller than the threshold, the tracking gate unit 1 outputs the observed position information to the smoothing unit 2, and if the difference is greater than the threshold, the tracking gate unit 1 discards the observed position information. The threshold may be stored in an internal memory of the tracking gate unit 1, or may be provided from outside the target tracking device shown in FIG.

[0012] The smoothing unit 2 is realized by, for example, a smoothing circuit 12 shown in FIG. The smoothing unit 2 acquires the observation position information from the tracking gate unit 1 and acquires the predicted value from the prediction unit 3 . The smoothing unit 2 estimates the state of the tracking target at the time of position observation from the observed position information and the predicted value. The smoothing unit 2 outputs the estimated value of the state of the tracked target to the outside, and also outputs the estimated value of the state of the tracked target to the prediction unit 3.

[0013] The prediction unit 3 is realized by, for example, a prediction circuit 13 shown in FIG. The prediction unit 3 includes N prediction processing units 4-1 to 4-N and a predicted value integration unit 5. N is an integer of 1 or more. The prediction unit 3 obtains an estimate of the state of the tracked target from the smoothing unit 2. The prediction unit 3 uses a motion model of the tracked target that includes an acceleration term related to changes in the direction of movement of the tracked target, and predicts the state of the tracked target one sampling later than the position observation time from the state estimated by the smoothing unit 2. The prediction unit 3 outputs a predicted value of the state of the tracking target to both the tracking gate unit 1 and the smoothing unit 2.

[0014] The prediction processing units 4-n (n=1,...,N) are provided with a motion model of the tracked target, which includes terms relating to gravity and aerodynamic force, and an acceleration term relating to a change in the maneuvering direction of the tracked target. The motion models provided in the prediction processing units 4-1 to 4-N are different from each other. 1, the motion model of the tracked target includes terms relating to gravity and aerodynamic forces, and an acceleration term relating to a change in the direction of movement of the tracked target. However, the motion model of the tracked target only needs to include an acceleration term relating to a change in the direction of movement of the tracked target, and does not necessarily need to include terms relating to gravity and aerodynamic forces. The prediction processing unit 4-n uses a motion model to predict the state of the tracked target one sampling later than the position observation time point from the state estimated by the smoothing unit 2, and outputs the predicted value of the state of the tracked target to the prediction value integration unit 5. In the target tracking device shown in Fig. 1, the prediction processing unit 4-n is provided with a motion model. However, this is merely an example, and the motion model may be provided from outside the target tracking device.

[0015] The predicted value integration unit 5 obtains predicted values ​​from each of the prediction processing units 4-1 to 4-N. The predicted value integrating unit 5 integrates the N predicted values, and outputs the integrated predicted value to the tracking gate unit 1 and the smoothing unit 2, respectively.

[0016] In Fig. 1, it is assumed that each of the components of the target tracking device, that is, a tracking gate unit 1, a smoothing unit 2, and a prediction unit 3, is realized by dedicated hardware as shown in Fig. 2. That is, it is assumed that the target tracking device is realized by a tracking gate circuit 11, a smoothing circuit 12, and a prediction circuit 13. Each of the tracking gate circuit 11, the smoothing circuit 12 and the prediction circuit 13 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination of these.

[0017] The components of the target tracking device are not limited to those realized by dedicated hardware, and the target tracking device may be realized by software, firmware, or a combination of software and firmware. The software or firmware is stored as a program in the memory of a computer. The computer means hardware that executes the program, and includes, for example, a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, a processor, or a DSP (Digital Signal Processor).

[0018] FIG. 3 is a hardware configuration diagram of a computer in the case where the target tracking device is realized by software, firmware, or the like. When the target tracking device is realized by software, firmware, or the like, a program for causing a computer to execute the respective processing procedures in the tracking gate unit 1, the smoothing unit 2, and the prediction unit 3 is stored in the memory 31. Then, a processor 32 of the computer executes the program stored in the memory 31.

[0019] 2 shows an example in which each of the components of the target tracking device is realized by dedicated hardware, and Fig. 3 shows an example in which the target tracking device is realized by software, firmware, etc. However, this is merely one example, and some of the components in the target tracking device may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0020] First, the principle of estimating the state of a tracked target by the target tracking device shown in FIG. 1 will be described. State vector x indicating the state of the tracked target k For example, as shown in the following equation (1), the position vector x of the tracking target is k ,y k ,z k , the velocity vector of the tracked object x dot k ,y dot k ,z dot k and aerodynamic coefficients α, β, and γ, where k is an integer equal to or greater than 1. State vector xk and the position vector x k is the same as x k However, the position vector x k is a scalar. The position vector y k ,z k is also a scalar. k ,y k ,z k Each of these represents, for example, the three-dimensional position of the tracked target in a north-based rectangular coordinate system.

[0021] In the text of the specification, the letters (x k ,y k ,z k ) cannot have a “·” symbol on it, so the velocity vector x dot k ,y dot k ,z dot k It is written as follows. α is an aerodynamic coefficient indicating lift, β is an aerodynamic coefficient indicating drag, and γ is an aerodynamic coefficient indicating aerodynamic force.

[0022] TIFF0007580637000001.tif12166

[0023] The observation vector z indicates the observation position information given to the tracking gate unit 1 from a target observation device (not shown). k is a vector indicating the three-dimensional observation position of the tracked target, as shown in the following equation (2), and the observation vector z k x, an element of k ,y k ,z k Each of is a scalar.

[0024] TIFF0007580637000002.tif11166

[0025] The motion model used in the prediction processing unit 4-n (n=1, . . . , N) is expressed as the state vector x k This can be expressed as a first-order differential equation for

[0026] TIFF0007580637000003.tif10166

[0027] In equation (3), w k is a driving noise vector, and has the properties shown in the following equations (4) and (5).

[0028] TIFF0007580637000004.tif39166

[0029] In equation (3), f(x k ) is expressed using aerodynamic coefficients α, β, and γ as shown in the following equation (6). In equation (5), Q k is the drive noise error covariance matrix. In equation (5a), 0 m×n is an m-by-n zero matrix, Q k pv are the driving noise matrix parameters for position and velocity, Q α ,Q β ,Q γ Each of is a driving noise matrix parameter (scalar value) related to the aerodynamic coefficient. diag[A,B,C] is a diagonal matrix with elements A, B, and C. In this example, it has 3 rows and 3 columns.

[0030] TIFF0007580637000005.tif103166 In equation (6), μ is the gravitational constant, ρ is the air density, r is the distance, and V is the velocity.

[0031] The motion model shown in equation (3) can be expressed in a discrete system as shown in equation (7) below.

[0032] TIFF0007580637000006.tif33166

[0033] The motion model used in the prediction processor 4-n (n=1, . . . , N) is a driving noise vector w k and the transformation matrix Γ' of the constant acceleration vector k-1 Each of these is extended by

[0034] TIFF0007580637000007.tif15166In equation (9), u k is the sampling time t k is a constant acceleration vector included in the N motion models, and is expressed as the following equation (10). Gamma' k-1 is the sampling time t k-1 is a transformation matrix of a constant acceleration vector in and is expressed as the following equation (11).

[0035] TIFF0007580637000008.tif35166 In formula (11), F k is the constant acceleration vector u k From the coordinate system defined above, the state vector x k is the coordinate transformation matrix to the coordinate system defined by I m×n is an m-by-n identity matrix.

[0036] FIG. 4 is an explanatory diagram showing a constant acceleration vector. In Figure 4, O is the origin of the coordinate system O-xyz, which has the target observation device as its origin, X is the x-axis of the coordinate system O-xyz, where the positive x-axis is in the eastward direction, Y is the y-axis of the coordinate system O-xyz, where the positive y-axis is in the northward direction, and Z is the z-axis of the coordinate system O-xyz, where the positive z-axis is in the upward direction. α1 is a constant acceleration vector in the positive y-axis direction, and α2 is a constant acceleration vector in the negative y-axis direction. α3 is a constant acceleration vector in the positive x-axis direction, and α4 is a constant acceleration vector in the negative x-axis direction. α5 is a constant acceleration vector in the positive direction of the z axis, and α6 is a constant acceleration vector in the negative direction of the z axis. If a constant acceleration vector α7=0 is defined in addition to the constant acceleration vectors α1 to α6, the number of motion models will be seven. One axis of the constant acceleration can be the north direction of the north-based orthogonal coordinate system, an estimated velocity vector, or LOS (Line Of Sight), etc. LOS is a vector in the line of sight direction when looking at the tracked target from the target observation device.

[0037] Next, at sampling time t k In the case where the following equation (12) is true, the motion model is defined as the following equation (13).

[0038] TIFF0007580637000009.tif23166

[0039] Here, the motion model Ψ k,a We assume that the transition of is Markovian. That is, the motion model Ψ k,a is the sampling time t k-1 The motion model of Ψ k-1,a is determined by the sampling time t k-2 Motion model Ψ up to k-2,a The above does not depend on the above. For example, if the minimum sampling interval is t s (=t k -t k-1 ) the transition probability p k,ab is expressed as shown in the following equation (14). In an N×N matrix, the element in row a and column b is p ab The above is expressed as the transition probability matrix Π between the motion models, and the sampling interval is t k -t k-1 =mt s When m is expressed as (m=1,2,...), the sampling time t k The transition probability matrix Π between the motion models in k is expressed as the following equation (15). k The elements of row a and column b of are the transition probability p k,ab It is.

[0040] TIFF0007580637000010.tif21166

[0041] Sampling time t k Observation information up to Z k Based on the sampling time t k Motion model in Ψ k,a Reliability μ k,a (+) is the conditional probability density function P[Ψ k,a |Z k ], it can be expressed as the following equation (17).

[0042] TIFF0007580637000011.tif42166In equation (17), v k,a is the normal distribution approximation P[Z|Ψ k,a ,Z k-1 ] is approximated by a multivariate normal distribution.

[0043] Sampling time t k-1 Observation information up to Z k-1 Based on the sampling time t k Motion model in Ψ k,a Prior confidence level μ k,a (-) is the conditional probability density function P[Ψ k,a |Z k-1 ], it can be expressed as the following equation (20).

[0044] TIFF0007580637000012.tif33166

[0045] Sampling time t k Observation information up to Z k The constant acceleration vector estimate u based on k-1 If (+) is defined as in the following equation (21), it can be expressed as in the following equation (22). Hereinafter, the estimated value of the constant acceleration vector u k-1 (+) is called the estimated acceleration vector.

[0046] TIFF0007580637000013.tif33166

[0047] Sampling time t k-1 Observation information up to Z k-1 The constant acceleration vector estimate u based on k-1 If (-) is defined as in the following equation (23), it can be expressed as in the following equation (24). Hereinafter, the estimated value of the constant acceleration vector u k-1 (-) is called the predicted acceleration vector.

[0048] TIFF0007580637000014.tif34166

[0049] The observation vector z shown in equation (2) k Observation model Z k is expressed as the following equation (25).

[0050] In equation (25), H is the sampling time t k is the observation matrix in v k is the sampling time t k Observation vector z k The observation noise vector is a three-dimensional normally distributed white noise with a mean of 0, as shown in the following equation (27).

[0051] TIFF0007580637000016.tif43166In equation (28), R k is the sampling time t k is an observation noise covariance matrix in , which is a value independent of the motion model. The driving noise vector and the observation noise vector are independent of each other.

[0052] From sampling time t1 to sampling time t kThe accumulation of observation vectors up to is expressed as follows:

[0053] TIFF0007580637000017.tif10166

[0054] According to the theory of the Kalman filter, the observation model shown in equation (25) is used to estimate the sampling time t k The state vector x when the observations are obtained k The estimated value of X k (+) is calculated using the following equations (30) to (38).

[0055] TIFF0007580637000018.tif148166

[0056] TIFF0007580637000019.tif101166

[0057] X-Hat k,a (-) is the sampling time t k-1 Observation information up to Z k-1 and the motion model Ψ k,a Based on and X k The average value is the conditional average value of the sampling time t k-1 Observation information up to Z k-1 and the motion model Ψ k,a Based on the sampling time t k This corresponds to a predicted vector that estimates the true value of X-Hat k,a (+) is the sampling time t k Observation information up to Z k and the motion model Ψ k,a Based on and X k The average value is the conditional average value of the sampling time t k Observation information up to Z k and the motion model Ψ k,a Based on the sampling time t k This corresponds to a smoothed vector that estimates the true value of X-Hat k(-) is the sampling time t k-1 Observation information up to Z k-1 Based on X k The average value is the conditional average value of the sampling time t k-1 Observation information up to Z k-1 Based on the sampling time t k This corresponds to a predicted vector that estimates the true value of X-Hat k (+) is the sampling time t k Observation information up to Z k Based on X k The average value is the conditional average value of the sampling time t k Observation information up to Z k Based on the sampling time t k This corresponds to a smoothed vector that estimates the true value of In the specification, the letters (X k,a (-), X k,a (+),X k (-), X k Since you cannot put a “^” sign on top of (+), k,a (-), X-hat k,a (+), X-hat k (-), X-hat k It is written as (+).

[0058] P k,a (-) is X-hat k,a The prediction error covariance matrix for each motion model shows the error covariance matrix of (-). P k,a (+) is X-hat k,a (+) is the smoothed error covariance matrix for each motion model showing the error covariance matrix of P k (-) is X-hat k The error covariance matrix of (-) is the prediction error covariance matrix by N motion models. P k (+) is X-hat k (+) is the smoothed error covariance matrix of N motion models. K shown in equation (34)k is the sampling time t k In addition, similarly to the case where a Kalman filter is applied, it is assumed that the initial values ​​X0(+) and P0(+) are separately determined. P k,a (-) is the motion model Ψ as shown in equation (31). k,a Since this value does not depend on K k and P shown in equation (36) k,a (+) also has a motion model Ψ k,a This value does not depend on

[0059] Next, the operation of the target tracking device shown in FIG. 1 will be described. FIG. 5 is a flowchart showing a target tracking method which is a processing procedure of the target tracking device. A prediction processor 4-n (n=1, . . . , N) calculates N constant acceleration vectors α1 to α N Among them, any constant acceleration vector α n The motion model Ψ as shown in Eq. (13) has the following structure: k,a (a = 1, ,N).

[0060] The prediction processing unit 4-n obtains from the smoothing unit 2 an estimated value of the state of the tracking target at the time of position observation. The prediction processing unit 4-n uses the motion model Ψ k,a The state of the tracked target one sampling later than the position observation time point is predicted from the estimated value of the state of the tracked target using (step ST1 in FIG. 5). The prediction processing unit 4-n outputs the predicted value of the state of the tracking target to the predicted value integration unit 5.

[0061] The state prediction process of the tracking target performed by the prediction processor 4-n will be specifically described below. The prediction processing unit 4-n receives the smoothed vector XHAT from the smoothing unit 2 as an estimate of the state of the tracking target. k,a (+) or smooth vector X-hat k (+) and the smoothing error covariance matrix P k Get (+) and. The prediction processor 4-n calculates x k-1 (+) as smooth vector X hat k,a (+) or smooth vector X-hat k By substituting (+) into equation (30), the predicted value of the state of the tracking target is obtained as the predicted vector X k,a Calculate (-). Moreover, the prediction processing unit 4-n calculates P k-1 (+) as the smoothing error covariance matrix P k By substituting (+) into equation (31), the prediction error covariance matrix P k,a Calculate (-). The prediction processor 4-n calculates the prediction vector X k,a (-) and the prediction error covariance matrix P k,a (-) is output to the predicted value integration unit 5.

[0062] The prediction value integration unit 5 receives N prediction vectors XHAT from the prediction processing units 4-1 to 4-N as prediction values ​​of the state of the tracking target. k,a (-) and obtain N prediction vectors XHat k,a Integrate (-) (step ST2 in FIG. 5). The prediction vector X, which is the predicted value after integration k (-) is x shown in equation (32) k-1 (+) as smooth vector X hat k,a (+) or smooth vector X-hat k This corresponds to the value calculated by substituting (+) into equation (32). The prediction value integration unit 5 calculates the prediction vector X k,a (-) is output to smoothing unit 2, and the predicted vector X hat k (-) is output to the tracking gate unit 1 and the smoothing unit 2 respectively.

[0063] The prediction value integration unit 5 also receives N prediction error covariance matrices P k,a (-) and obtain the N prediction error covariance matrix P k,a Integrate (-). That is, the prediction value integration unit 5 calculates the prediction error covariance matrix P k,aBy substituting (-) into equation (33), the integrated prediction error covariance matrix P k Calculate (-). The prediction value integration unit 5 calculates the prediction error covariance matrix P k (-) is output to smoothing unit 2.

[0064] The tracking gate unit 1 receives, for example, an observation vector z k Get the. The tracking gate unit 1 also receives a predicted vector Xhatt from the predicted value integration unit 5 as a predicted value that is predicted one sampling before the position observation time. k-1 Get (-). The tracking gate unit 1 detects the observation vector z k The observed position of the tracked target (x k ,y k ,z k ) and the prediction vector X k-1 The difference between the predicted position of the tracking target, which is the (-) element, is calculated. If the difference between the observed position and the predicted position is equal to or smaller than the threshold (step ST3 in FIG. 5: YES), the tracking gate unit 1 calculates the observed vector z k to the smoothing unit 2 (step ST4 in FIG. 5). If the difference between the observed position and the predicted position is greater than the threshold (step ST3 in FIG. 5: NO), the tracking gate unit 1 calculates the observed vector z k (step ST5 in FIG. 5).

[0065] The smoothing unit 2 receives the observation vector z k The predicted value integration unit 5 obtains a predicted vector XHat as a predicted value that is predicted one sampling before the position observation time. k-1 Get (-). The smoothing part 2 is the observation vector z k and the prediction vector X k-1 (-), the state of the tracking target is estimated (step ST6 in FIG. 5). The smoothing unit 2 estimates the state of the tracked object by the smoothing vector XHat k,a(+) or smooth vector X-hat k (+) and the smoothing error covariance matrix P k (+) is output to the outside. In addition, the smoothing unit 2 smooths the vector X hat k,a (+) or smooth vector X-hat k (+) and the smoothing error covariance matrix P k (+) is output to the prediction unit 3.

[0066] The process of estimating the state of the tracked target by the smoothing unit 2 will now be described in detail. First, the smoothing unit 2 performs the sampling at time t k Observation information up to Z k By substituting into equation (16), the sampling time t k Motion model in Ψ k,a Reliability μ k,a Calculate (+). The smoothing unit 2 is a reliability μ k,a By substituting (+) into equation (22), the estimated acceleration vector u k-1 Calculate (+). Next, the smoothing unit 2 receives the prediction error covariance matrix P k,a Get (-). The smoothing part 2 uses the prediction error covariance matrix P k,a (-) and the observation noise covariance matrix R k By substituting each of these into equation (34), the gain matrix K k Calculate.

[0067] Next, the smoothing unit 2 receives the predicted vector XHat from the predicted value integration unit 5. k,a (-) and the predicted vector X hat k Get each of (-). The smoothing section 2 is the observation information Z k and the prediction vector X k,a By substituting each of (-) into equation (35), we obtain the smooth vector X k,a Calculate (+). In addition, the smoothing unit 2 uses x k-1 (-) as the predicted vector x hat k(-) into equation (37), and the estimated acceleration vector u k-1 By substituting (+) into equation (37), we obtain the smooth vector Xhat k Calculate (+).

[0068] Next, the smoothing unit 2 calculates the prediction error covariance matrix P k,a (-) and the gain matrix K k By substituting each of these into equation (36), we obtain the error covariance matrix P k,a Calculate (+). In addition, the smoothing unit 2 calculates the smoothing error covariance matrix P k,a (+) and gain matrix K k By substituting each of these into equation (38), we obtain the error covariance matrix P k Calculate (+).

[0069] After that, the target tracking device shown in FIG. 1 repeatedly performs the processes of steps ST1 to ST6 until tracking of the tracking target is completed.

[0070] In the above-mentioned embodiment 1, the target tracking device is configured to include a smoothing unit 2 that acquires observation position information indicating the observation position of the tracking target, and acquires a predicted value of the state of the tracking target at the position observation time, which is predicted one sampling before the position observation time when the observation position is observed, and estimates the state of the tracking target at the position observation time from the observation position information and the predicted value, and a prediction unit 3 that predicts the state of the tracking target one sampling after the position observation time from the state estimated by the smoothing unit 2 using a motion model including an acceleration term related to a change in the maneuvering direction of the tracking target as a motion model of the tracking target, and outputs the predicted value of the state of the tracking target to the smoothing unit 2. Therefore, the target tracking device can predict the state of the tracking target taking into account the maneuvering direction of the tracking target.

[0071] Embodiment 2 In the second embodiment, a target tracking device in which a smoothing unit 6 corrects an acceleration term included in a motion model based on observed position information and a predicted value that is predicted one sampling before the position observation time point will be described.

[0072] Fig. 6 is a configuration diagram showing a target tracking device according to embodiment 2. In Fig. 6, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 7 is a hardware configuration diagram showing the hardware of a target tracking device according to embodiment 2. In Fig. 7, the same reference numerals as in Fig. 2 denote the same or corresponding parts, and therefore description thereof will be omitted. The target tracking device shown in FIG. 6 includes a tracking gate unit 1, a smoothing unit 6, and a prediction unit .

[0073] The smoothing unit 6 is realized by, for example, a smoothing circuit 14 shown in FIG. 1, the smoothing unit 6 acquires observed position information from the tracking gate unit 1, and acquires a predicted value of the state of the tracked target from the prediction unit 7. The predicted value indicates a predicted result of the state of the tracked target at the position observation time, which is predicted one sampling before the position observation time, which is the time when the observation position is observed. Similar to the smoothing unit 2 shown in FIG. 1, the smoothing unit 6 estimates the state of the tracking target at the time of position observation from the observed position information and the predicted value. The smoothing unit 6 corrects the acceleration term included in the motion model based on the observed position information and the predicted value. The smoothing unit 6 outputs the estimated value of the state of the tracking target to the outside. The smoothing unit 6 outputs the estimated value of the state of the tracking target and the corrected acceleration term to the prediction unit 7.

[0074] The prediction unit 7 is realized by, for example, a prediction circuit 15 shown in FIG. The prediction unit 7 includes N prediction processing units 4-1 to 4-N and a predicted value integration unit 8. The prediction unit 7 obtains from the smoothing unit 6 the estimated value of the state of the tracking target and the corrected acceleration term. A prediction unit 7 predicts the state of the tracked target one sampling after the position observation time point from the state estimated by the smoothing unit 6 using a motion model of the tracked target including the corrected acceleration term. The prediction unit 3 outputs a predicted value of the state of the tracking target to both the tracking gate unit 1 and the smoothing unit 6.

[0075] The predicted value integration unit 8 obtains predicted values ​​from each of the prediction processing units 4-1 to 4-N. The predicted value integrating unit 8 integrates the N predicted values, and outputs the integrated predicted value to both the tracking gate unit 1 and the smoothing unit 6.

[0076] In Fig. 6, it is assumed that each of the components of the target tracking device, that is, the tracking gate unit 1, the smoothing unit 6, and the prediction unit 7, is realized by dedicated hardware as shown in Fig. 7. That is, it is assumed that the target tracking device is realized by a tracking gate circuit 11, a smoothing circuit 14, and a prediction circuit 15. Each of the tracking gate circuit 11, the smoothing circuit 14 and the prediction circuit 15 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination of these.

[0077] The components of the target tracking device are not limited to those realized by dedicated hardware, and the target tracking device may be realized by software, firmware, or a combination of software and firmware. When the target tracking device is realized by software, firmware, or the like, a program for causing a computer to execute the respective processing procedures in the tracking gate unit 1, the smoothing unit 6, and the prediction unit 7 is stored in a memory 31 shown in Fig. 3. Then, a processor 32 shown in Fig. 3 executes the program stored in the memory 31.

[0078] 7 shows an example in which each of the components of the target tracking device is realized by dedicated hardware, and FIG 3 shows an example in which the target tracking device is realized by software or firmware, etc. However, this is merely one example, and some of the components in the target tracking device may be realized by dedicated hardware, and the remaining components may be realized by software or firmware, etc.

[0079] Next, a description will be given of the operation of the target tracking device shown in Fig. 6. Since the components other than the smoothing unit 6 and the predicting unit 7 are the same as those of the target tracking device shown in Fig. 1, only the operations of the smoothing unit 6 and the predicting unit 7 will be described here.

[0080] The smoothing unit 6 receives the observation vector z k The predicted value integration unit 8 obtains the predicted vector X k Get (-). The smoothing unit 6 calculates x k (-) as the predicted vector x hat k (-) into equation (47), and the observation vector z k By substituting into equation (47), the innovation vector Δ k Calculate.

[0081] TIFF0007580637000020.tif10166

[0082] The smoothing unit 6 receives the prediction error covariance matrix P k-1 Get (-). The smoothing unit 6 is a P tilde filter shown in the following formula (48). k-1 (-) as the prediction error covariance matrix P k-1 (-) into equation (48) and the innovation vector Δ k By substituting into equation (48), the prediction residual covariance matrix V k Calculate. In the text of the specification, the letter P is used for electronic filing purposes. k-1 Since you cannot put a "~" symbol above (-), use the P tilde k-1 It is written as (-).

[0083] TIFF0007580637000021.tif11166

[0084] The smoothing unit 6 performs the P tilde k-1 (-) as the prediction error covariance matrix P k-1By substituting (-) into equation (49), the predicted position error covariance matrix M k Calculate. The smoothing unit 6 calculates the prediction residual covariance matrix V k From the predicted position error covariance matrix M k By subtracting k Calculate.

[0085] TIFF0007580637000022.tif21166

[0086] The smoothing unit 6 calculates the predicted position error covariance matrix M k The diagonal elements of M(x,y,z) and the prediction residual covariance matrix N k From the diagonal components N(x,y,z) of the equation, the constant acceleration vector u corresponding to the delay amount on each of the three axes constituting the Cartesian coordinate system is obtained. k-1 (-) correction coefficient λ k Calculate (x,y,z).

[0087] TIFF0007580637000023.tif16166

[0088] The smoothing unit 6 calculates a constant acceleration vector u k-1 (-) correction coefficient λ k Using (x,y,z), the constant acceleration vector u k-1 Correct (-). u k '(-) is the constant acceleration vector after correction.

[0089] TIFF0007580637000024.tif31166

[0090] The prediction unit 7 predicts the corrected constant acceleration vector u k Using a motion model of the tracked target including '(-), the state of the tracked target is predicted from the state estimated by the smoothing unit 6. The prediction value integration unit 8 of the prediction unit 7 receives N prediction vectors XHAT from each of the prediction processing units 4-1 to 4-N as predicted values ​​of the state of the tracking target. k,a (-) and obtain N prediction vectors XHat k,a Integrate (-).

[0091] That is, the predicted value integration unit 8 calculates the corrected constant acceleration vector u k '(-) into the following equation (54) and obtain x shown in equation (54). k-1 (+) as smooth vector X hat k,a (+) or smooth vector X-hat k By substituting (+) into equation (54), the prediction vector X hat, which is the prediction value after integration, is obtained. k Calculate (-). In addition, the predicted value integration unit 8 calculates the corrected constant acceleration vector u k '(-) into the following equation (55) to obtain the prediction error covariance matrix P k,a By substituting (-) into equation (55), the integrated prediction error covariance matrix P k Calculate (-).

[0092] TIFF0007580637000025.tif43166

[0093] In the above-mentioned second embodiment, the target tracking device shown in Fig. 6 is configured so that the smoothing unit 6 corrects the acceleration term included in the motion model based on the observed position information and the predicted value of the state of the tracking target at the time of position observation. Therefore, the target tracking device shown in Fig. 6 can predict the state of the tracking target taking into account the maneuvering direction of the tracking target, similar to the target tracking device shown in Fig. 1, and can improve the estimation accuracy of the smoothing unit 2 more than the target tracking device shown in Fig. 1.

[0094] In addition, the present disclosure allows free combination of the respective embodiments, modification of any of the components of each embodiment, or omission of any of the components of each embodiment. [Industrial Applicability]

[0095] The present disclosure is suitable for a target tracking device and a target tracking method. [Explanation of symbols]

[0096] 1 tracking gate unit, 2 smoothing unit, 3 prediction unit, 4-1 to 4-N prediction processing units, 5 prediction value integration unit, 6 smoothing unit, 7 prediction unit, 8 prediction value integration unit, 11 tracking gate circuit, 12 smoothing circuit, 13 prediction circuit, 14 smoothing circuit, 15 prediction circuit, 31 memory, 32 processor.

Claims

1. The present invention includes a smoothing unit and a prediction unit that form a closed loop, the smoothing unit acquires observation position information indicating an observation position of a tracking target, and acquires a predicted value of a state of the tracking target at a position observation time point that is predicted one sampling prior to a position observation time point at which the observation position of the tracking target is observed, and estimates the state of the tracking target at the position observation time point from the observation position information and the predicted value of the state of the tracking target; correcting a constant acceleration vector derived from an acceleration term related to a change in the movement direction of the tracked target, which is included in a motion model of the tracked target, based on the observed position information and a predicted value of the state of the tracked target at the time of position observation; the prediction unit predicts a state of the tracking target one sampling later than the position observation time point from the state estimated by the smoothing unit, and outputs a predicted value of the state of the tracking target to the smoothing unit. Target tracking device.

2. The prediction unit is The target tracking device according to claim 1, characterized in that a state of the tracked target one sampling later than the position observation time point is predicted from the state estimated by the smoothing unit using a motion model based on the constant acceleration vector corrected by the smoothing unit, and the predicted value of the state of the tracked target is output to the smoothing unit.

3. The plurality of motion models used by the prediction unit each include In addition to the acceleration term related to the change in the direction of movement of the tracked target, it includes terms related to gravity and aerodynamic force.

2. A target tracking device according to claim 1.

4. the prediction unit includes a plurality of prediction processing units and a prediction value integration unit, which form the closed loop together with the smoothing unit; each of the plurality of prediction processing units predicts a state of the tracking target from the state estimated by the smoothing unit using a corresponding one of a plurality of mutually different motion models, and outputs a predicted value for each of the states of the tracking target to the prediction value integration unit; The prediction value integration unit integrates the predicted values ​​output from the plurality of prediction processing units, and outputs the integrated value to the smoothing unit.

2. A target tracking device according to claim 1.

5. 2. The target tracking device according to claim 1, further comprising a tracking gate unit that acquires the observed position information and a predicted value of the state of the tracked target at the position observation time, and outputs the observed position information to the smoothing unit if a difference between the observed position of the tracked target indicated by the observed position information and the predicted position of the tracked target included in the predicted value of the state of the tracked target is equal to or smaller than a threshold.

6. A target tracking method for a target tracking device including a smoothing unit and a prediction unit that form a closed loop, comprising: The smooth portion is Obtaining observation location information indicating an observation location of a tracking target, and obtaining a predicted value of a state of the tracking target at a location observation time point, which is a time point at which the observation location of the tracking target is observed, predicted one sampling prior to the location observation time point, and estimating the state of the tracking target at the location observation time point from the observation location information and the predicted value of the state of the tracking target; correcting a constant acceleration vector derived from an acceleration term related to a change in the movement direction of the tracked target, which is included in a motion model of the tracked target, based on the observed position information and a predicted value of the state of the tracked target at the time of position observation; The prediction unit, predicting a state of the tracking target one sampling later than the position observation time point from the state estimated by the smoothing unit, and outputting a predicted value of the state of the tracking target to the smoothing unit; Target tracking method.

Citation Information

Patent Citations

  • Tracking filter

    JP1986169786A

  • Tracking apparatus for radar

    JP1989073275A

  • Target tracking device

    JP1994347542A

  • Target tracking method and radar implementing the method

    JP1997189761A

  • Tracking device

    JP1997297176A