JPH11304915A - Apparatus and method for tracking of target - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect the swiveling state of a target to be tracked quickly and precisely, in a target detecting apparatus which tracks the target to be tracked by using a radar sensor. SOLUTION: In a target tracking apparatus, a quantization part 10 which binarizes and processes a video signal so as to create a target table is installed, an intersweep correlation part 12 which detects an edge formed in the target table is installed, the machine shadow of a target to be tracked is detected on the basis of the edge detected in the target table, and the center of gravity of the target to be tracked is computed on the basis of the machine shadow. Then, a swiveling judgment part 16, in which the machine nose of the target to be tracked is extracted from the machine shadow, and which judges the swiveling state of the target to be tracked on the basis of the movement of the machine nose, is installed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a target tracking device and a target tracking method, and more particularly, to a target tracking device and a target tracking method for tracking a tracking target using a radar sensor.

[0002]

2. Description of the Related Art Conventionally, for example, "Multiple Target
Tracking with Radar Application ”(SSBlackmann, A
rtech House, Dedham, 1986), a target tracking device using a radar sensor is known. A conventional radar sensor detects a shape of a tracking target based on a video signal obtained for each scan by the radar sensor, and further detects a center of gravity of the tracking target based on the shape. And
Based on the detected center of gravity position, the motion state of the tracking target,
That is, the position and speed of the tracking target are detected.

[0003]

It is desirable that the target tracking device can accurately detect the turning state of the tracking target. However, even if the motion of the tracking target changes from a straight-line motion to a turning motion, it is difficult for the change to be reflected on the center of gravity of the tracking target. For this reason, in the above-mentioned conventional target tracking device, it was difficult to quickly detect the start of the turning motion after the tracking target started the turning motion.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and quickly changes the turning state of a tracking target.
Further, it is a first object to provide a target tracking device that can be accurately detected.

It is a second object of the present invention to provide a target tracking method capable of quickly and accurately detecting the turning state of a tracking target.

[0006]

According to a first aspect of the present invention, there is provided a target tracking device having a function of detecting a center of gravity of a tracking target from a feature of the tracking target, the target tracking device having a function different from the center of gravity. A turning state detecting means for detecting a turning state of the tracking target based on a movement of a point on the tracking target is provided.

According to a second aspect of the present invention, there is provided a target tracking apparatus comprising:
The apparatus further includes a predetermined point setting unit that detects a long axis direction of the tracking target from the image of the tracking target and sets an end of the tracking target in the long axis direction as the predetermined point.

According to a third aspect of the present invention, there is provided a target tracking device, comprising:
A second predetermined point setting means for setting the predetermined point to an arbitrary point of the hit output of the tracking target.

According to a fourth aspect of the present invention, there is provided a target tracking apparatus comprising:
The apparatus further comprises third predetermined point setting means for setting a point on the tracking target separated by a predetermined distance from the center of gravity as the predetermined point.

[0010] According to a fifth aspect of the present invention, there is provided a target tracking apparatus.
Observation, an observation value acquisition unit that obtains an observation value representing the state of the center of gravity at a predetermined time, and predicts a prediction value representing the state of the center of gravity at the predetermined time based on the movement of the tracking target before the predetermined time. Predicted value obtaining means, smoothed value obtaining means for obtaining a smoothed value by smoothing the observed value and the predicted value with a predetermined gain, and gain changing means for changing the predetermined gain according to the turning state of the tracking target And

According to a sixth aspect of the present invention, there is provided a target tracking apparatus comprising:
Gain control means for calculating the predetermined gain based on a parameter including drive noise, such that the larger the drive noise, the larger the observed value is reflected on the smoothed value; and And a driving noise control unit that sets the driving noise to a larger value when the tracking target is in a non-turning motion than when the tracking target is in a non-turning motion.

According to a seventh aspect of the present invention, there is provided a target tracking apparatus comprising:
Tracking gate setting means for predicting an area where the tracking target exists at a predetermined time based on the state of the tracking target before a predetermined time, and setting the prediction area as a tracking gate at the predetermined time; turning the tracking target Gate control means for changing the size of the tracking gate based on a state.

[0013] The target tracking device according to claim 8 of the present invention comprises:
And a sampling interval control unit for controlling a sampling interval for monitoring the tracking target based on a turning state of the tracking target.

According to a ninth aspect of the present invention, there is provided a target tracking device, comprising:
State variable vector calculating means for obtaining a state variable vector including a state variable of a predetermined order representing the motion state of the tracking target, and variable order changing means for changing the predetermined order based on the turning state of the tracking target, It is provided.

According to a tenth aspect of the present invention, in the target tracking device, the state variable vector calculating means obtains a first state variable vector including the position, velocity and acceleration of the tracking target. And second tracking filter means for obtaining a second state variable vector composed of the position and speed of the tracking target, wherein the variable degree changing means determines the second state variable vector based on the turning state of the tracking target. One of the first tracking filter means and the second tracking filter means is selectively operated.

[0016] A target tracking device according to claim 11 of the present invention is a multi-motion specification estimating means for estimating the motion specification of the tracking target corresponding to each of the plurality of motion models, and Transition probability calculating means for determining a transition probability, motion specification of the tracking target corresponding to each of the plurality of motion models, motion specification integration means for integrating based on the transition probability, and a turning state of the tracking target. And a transition probability control unit that changes the transition probability based on the transition probability.

According to a twelfth aspect of the present invention, in the target tracking apparatus, the plurality of motion models estimate a state variable vector composed of a position, a velocity and an acceleration of the tracking target, and the tracking. A second motion model for estimating a state variable vector composed of a target position and velocity.

According to a thirteenth aspect of the present invention, there is provided a target tracking apparatus, comprising: a multi-motion specification estimating means for estimating a motion specification of the tracking target corresponding to each of the plurality of motion models; Based on the turning state of the tracking target, based on the turning state of the tracking target, the state of the tracking target And reliability control means for setting the establishment probabilities for the plurality of exercise models such that the establishment probability of an appropriate exercise model for grasping is relatively increased.

According to a fourteenth aspect of the present invention, there is provided the target tracking device, wherein the plurality of motion models estimate a state variable vector composed of a position, a speed, and an acceleration of the tracking target; A second motion model for estimating a state variable vector composed of a target position and a speed, wherein the reliability control means sets the probability of establishment of the first motion model to 1 during the turning motion of the tracking target, and , The probability of establishment of the second motion model is set to 0, the probability of establishment of the first motion model is set to 0 during the non-turning motion of the tracking target, and the probability of formation of the second motion model is set to 0. It is.

A target tracking method according to a fifteenth aspect of the present invention is a target tracking method including a step of detecting a center of gravity of the tracking target from a feature of the tracking target, wherein the predetermined point on the tracking target different from the center of gravity is determined. And a turning state detecting step of detecting a turning state of the tracking target based on the movement of the tracking target.

In the target tracking method according to a sixteenth aspect of the present invention, an area where the tracking target exists at a predetermined time is predicted based on a state of the tracking target before a predetermined time, and the prediction area is determined at the predetermined time. A tracking gate setting step for setting as a tracking gate; and a gate control step for changing a size of the tracking gate based on a turning state of the tracking target.

A target tracking method according to a seventeenth aspect of the present invention includes a sampling interval control step of controlling a sampling interval for monitoring the tracking target based on a turning state of the tracking target.

A target tracking method according to claim 18 of the present invention includes a state variable vector calculating step for obtaining a state variable vector including a state variable of a predetermined order representing a motion state of the tracking target, and a turning state of the tracking target. A variable order changing step for changing the predetermined order based on the variable order.

[0024] A target tracking method according to a nineteenth aspect of the present invention is a multi-motion specification estimating step of estimating the motion specification of the tracking target corresponding to each of a plurality of motion models;
A transition probability calculating step of determining a transition probability between the plurality of movement models, and a movement specification integration step of integrating movement parameters of the tracking target corresponding to each of the plurality of movement models based on the transition probability. And a transition probability control step of changing the transition probability based on the turning state of the tracking target.

[0025] A target tracking method according to a twentieth aspect of the present invention is the multiple movement specification estimating step of estimating the movement specification of the tracking target corresponding to each of a plurality of movement models.
Based on the second motion specification integration step of integrating motion parameters of the tracking target corresponding to each of the plurality of motion models based on the probability of establishment of the plurality of motion models, based on a turning state of the tracking target. And a reliability control step of setting the establishment probabilities for the plurality of exercise models so that the establishment probability of an appropriate exercise model for grasping the state of the tracking target is relatively increased.

[0026]

Embodiments of the present invention will be described below with reference to the drawings. Elements common to the drawings are denoted by the same reference numerals, and redundant description will be omitted.

Embodiment 1 FIG. 1 shows a block diagram of the target tracking device according to the first embodiment of the present invention. As shown in FIG. 1, the target tracking device according to the present embodiment includes a quantization unit 10. A video signal detected for each scan by a radar sensor (not shown) is input to the quantization unit 10. The quantization unit 10 generates a target table by binarizing a video signal in a tracking gate to be described later.

FIG. 2 shows an example of the target table created by the quantization unit 10. In this embodiment,
The radar sensor is provided with a resolution high enough to detect the shape of the aircraft. The target table shown in FIG. 2 is an example of a table obtained when the radar sensor captures an image of an aircraft.

The video signal supplied to the quantization unit 10 is
It is composed of a plurality of pixel signals arranged on two-dimensional coordinates defined by an azimuth axis and a range axis. Quantization unit 10
Performs a binarization process by comparing individual pixel signals constituting a video signal with a predetermined threshold value. The threshold value is set to a value for determining whether or not any tracking target exists in an area corresponding to the pixel signal.

Therefore, according to the above-described binarization processing, the area on the target table is divided into an area where the tracking target is estimated to exist and an area where the tracking target is estimated not to exist. In the target table shown in FIG. 2, a blank portion is a region where it is estimated that no tracking target exists, and another portion is a region where it is estimated that a tracking target exists. Hereinafter, each pixel signal included in the blank portion in the target table is referred to as a non-detection hit output, and each pixel signal included in the non-white portion is referred to as a detection hit output.

The target table created by the quantization unit 10 is supplied to the inter-sweep correlation unit 12. The inter-sweep correlation unit performs edge detection of the detected hit output based on the continuity of the detected hit output in the target table in the azimuthal axis direction. Black triangles shown in FIG. 2 indicate edges detected by the inter-sweep correlation unit 12.

The target tracking device according to the present embodiment includes a position calculator 14. The position calculating unit 14 integrates, among the detected hit outputs in the target table, those that come into contact with each other as being caused by the same tracking target. The hatched area in FIG. 2 indicates the area integrated by the above processing. Hereinafter, this area is referred to as an integrated area, and the other areas are referred to as non-integrated areas. Position calculation unit 1
Reference numeral 4 also detects the center of gravity of the integrated area, and outputs the center of gravity as the center of gravity information Ck of the tracking target.

The center-of-gravity information Ck of the tracking target detected by the position calculation unit 14 is supplied to the turning determination unit 16 together with the edge information detected by the sweep correlation unit 12. The turning determination unit 16 assumes the turning axis of the tracking target based on the above information, and further determines whether the tracking target has rotated around the turning axis. The target tracking device of the present embodiment is characterized in that the target tracking device includes the above-described turning determination unit 16.

The turning determination unit 16 includes a nose extraction unit 18. The nose extracting unit 18 extracts a pixel corresponding to the nose portion of the tracking target from the integrated area as shown in FIG. 2 and, based on the pixel, calculates coordinates Ak of the nose portion at the sampling time tk. To detect. Nose extractor 18
Recognizes the contour of the integrated region as the image of the tracking target, estimates the long axis direction of the tracking target from the image, and detects the tip of the long axis direction as the nose of the tracking target. The coordinates Ak of the nose are composed of a distance rk from the radar sensor and an angle byk with respect to the radar sensor.

The coordinate Ak is determined by the tracking filter unit 20 for turning judgment.
Is input to In the tracking filter unit 20 for turning determination, the state variable Dk (+) including the distance rk, the angle byk, and their rates of change drk / dt and dbyk / dt, and their error covariance matrix Gk
(+) Is estimated. The turning determination tracking filter unit 20 further estimates a state variable Dk + 1 (-) and an error covariance matrix Gk + 1 (-) estimated to be realized at the next sampling. Estimated value Dk + 1 (-), Gk + at next sampling
1 (−) is supplied to the turning determination execution unit 24 after being delayed by the delay element 22 by a unit time corresponding to the sampling interval.

At the sampling time tk, the turning judgment execution unit 24 is supplied with the coordinates Ak from the nose extracting unit 18, and is calculated by the turning judgment tracking filter unit 20 at the previous sampling via the delay element 22. Dk
(−) And Gk (−) are supplied. The turning determination execution unit 24 calculates a predicted area of the nose portion estimated from Dk (-) and Gk (-) supplied as described above.

The prediction area is obtained on the assumption that the tracking target is performing a linear motion. Therefore, if the tracking target is turning, the coordinate Ak may deviate from the prediction area.
The turning determination execution unit 24 determines the nose coordinates Ak from the above viewpoint.
If the tracking target is out of the prediction area, it is determined that the tracking target is in a turning motion.If not, or if the tracking target cannot be detected or the nose portion is unknown, the tracking target Judge that you maintain your exercise.

As described above, the determination process in the turning determination execution unit 24 is performed by determining the nose coordinates Ak obtained by the latest sampling and the Dk obtained by the previous sampling.
(-) And Gk (-). Therefore, in order to perform the above-described determination processing, it is necessary that sampling has been performed at least twice. For this reason, in the present embodiment, the determination process of the turning determination execution unit 24 is executed after the coordinates of the nose portion are sampled at least twice.

The result of the judgment processing executed by the turning judgment executing section 24 is supplied to the drive noise control section 26. The drive noise control unit 26 sets the drive noise Qk to a value Q1 for turning when the turning determination execution unit 24 determines that the tracking target is performing turning motion, while the tracking target is performing turning motion. If it is determined that there is no driving noise Qk,
It is set to a value Q2 for linear motion that is smaller than Q1.

The driving noise Qk set by the driving noise control unit 26 is supplied to the tracking filter unit 28 together with the center of gravity information Ck of the tracking target detected by the position calculation unit 14. The tracking filter unit 28 calculates a prediction value vector Bk + 1 (-) of the tracking target one sample ahead and a prediction error covariance matrix Pk + 1 (-) based on the centroid information Ck of the tracking target and the driving noise Qk. Ask.

The information Bk + 1 (-) relating to the motion state quantity and the information Pk + 1 (-) relating to the error covariance calculated by the tracking filter unit 28 are converted by the delay element 30 into a unit time delay corresponding to the sampling interval. After that, it is supplied to the tracking gate calculation unit 32.

At the sampling time tk, the tracking gate calculator 32 is supplied with Bk (-) and Pk (-) calculated by the tracking filter unit 28 at the time of the previous sampling. The tracking gate calculator 32 calculates these Bk (−) and Pk
Based on k (-), a region where a video signal of a tracking target is predicted to be obtained at time tk is calculated.

Next, the contents of the arithmetic processing executed in the target tracking device of this embodiment will be described in detail using mathematical expressions. In the target tracking device, a process of estimating the position and speed of the tracking target and a process of determining whether the tracking target is performing a turning motion are executed. Among these processes, the process of detecting the position and speed of the tracking target will be described first.

The motion state of the tracking target at the sampling time tk, that is, the position and speed of the center of gravity of the tracking target at the time tk can be estimated based on the position of the center of gravity detected by the radar sensor at the time tk. Less than,
The execution unit of the above processing is referred to as an observation unit. If the motion state of the tracking target at the sampling time tk-1 immediately before the time tk is known, the motion state of the tracking target at the time tk can be estimated from the motion state at the immediately preceding time tk-1. . Hereinafter, the execution unit of the above processing is referred to as a prediction unit.

The execution result of the observation unit involves an error due to the time required for radar wave propagation, observation accuracy, and the like. on the other hand,
The execution result of the prediction unit includes an error due to a mismatch between the actual motion and the model. Therefore, the target tracking device of the present embodiment estimates the motion state of the tracking target using both the observation method and the motion model method.

The target tracking device of this embodiment uses a state variable vector calculated by the prediction unit and an observation value vector detected by the observation unit as parameters representing the motion state of the tracking target. Hereinafter, the state variable vector and the observation value vector at the sampling time tk are referred to as Bk and Ck, respectively.

The state variable vector Bk is expressed on north reference orthogonal coordinates, that is, three-dimensional coordinates in which the east direction is the positive direction of the x-axis, the north direction is the positive direction of the y-axis, and the vertical direction is the positive direction of the z-axis. Vector. State variable vector Bk
Is a six-dimensional vector representing the position and velocity of the center of gravity of the tracking target, and is represented by the following equation (1).

On the other hand, the observation value vector Ck is a vector obtained by polar coordinates having the north direction as the reference direction of the azimuth angle and expressed in a north reference orthogonal coordinate system. The observation value vector Ck is a three-dimensional vector representing the position of the center of gravity of the tracking target, and is represented by the following equations (2) and (3). In the following mathematical expressions, a suffix T added to an expression representing a vector is a symbol indicating that the vector is a vertical vector.

[0049]

(Equation 1)

The state variable vector Bk at the time tk,
Between the state variable vector Bk-1 at time tk-1,
A motion model represented by the following equation (4) can be set.

[0051]

(Equation 2)

Note that Φk and ωk shown in the above equation (4) are:
A state transition matrix and a driving noise vector, respectively. The state transition matrix is a matrix that is assumed in advance when setting the motion model. The driving noise vector ωk belongs to a normal distribution with an average “0” and a variance “Qk” as shown in the following equation (5). In the present embodiment, the variance Qk is set to a different value depending on whether the tracking target is turning or not, as described later.

[0053]

(Equation 3)

The following equation (6) can be set between the state variable vector Bk and the observed value vector Ck at the time tk. Hereinafter, this relationship is referred to as an observation model.

[0055]

(Equation 4)

Note that Hk and vk shown in the above equation (6) are:
They are an observation matrix and an observation noise vector, respectively. In the present embodiment, the observation matrix is a matrix assumed in advance. The observation noise vector belongs to a normal distribution with an average “0” and a variance “Rk” as shown in the following equation (7). Variance Rk
Is appropriately set according to the observation state.

[0057]

(Equation 5)

The target tracking device according to the present embodiment includes an observation value vector obtained by the observation unit for each sampling time;
A smoothed value vector, which will be described later, is calculated using the predicted value vector obtained by the predicting unit as input information. Hereinafter, the vector obtained at the time tk by the above method is converted to a smoothed value vector Bk (+)
Called. The smoothed value vector Bk (+) is a vector that can be recognized as having high consistency with the true state variable vector Bk.

According to the relationship of the above equation (4), the state variable vector Bk at the time tk can be predicted based on the state variable vector Bk-1 at the time tk-1. From the above relational expression, the predicted value of the state variable vector Bk at the time tk (hereinafter, referred to as the predicted value vector Bk (−)) is calculated by using the smoothed value vector Bk−1 (+) as in the following expression (8). Can be calculated.

[0060]

(Equation 6)

Similarly, the target tracking device of the present embodiment smoothes the result obtained by the observation method and the result obtained by the motion model method at each sampling time, as will be described later, to thereby obtain the error covariance. Operate on a matrix. Hereinafter, the error covariance matrix obtained at the time tk by the above method is referred to as a smoothed error covariance matrix Pk (+).

When the smoothed error covariance matrix Pk-1 (+) is obtained at the time tk-1, the error covariance matrix Pk at the time tk is predicted by substituting the matrix value into the following equation (9). be able to. Hereinafter, the error covariance matrix vector predicted by the above method is referred to as a prediction error covariance matrix Pk (-). still,
Qk-1 shown in the following equation (9) is the variance of the driving noise vector ωk-1 at time tk-1 (see the above equation (5)).

[0063]

(Equation 7)

As described above, the target tracking apparatus according to the present embodiment smoothes the result obtained by the observation method and the result obtained by the motion model method to obtain a smoothed value vector Bk (+) and a smoothed error. The variance matrix Pk (+) is calculated.
The following equation (10) shows a gain matrix Kk used for the above-described smoothing processing. Using the gain matrix Kk, the smoothed value vector Bk (+) and the smoothed error covariance matrix Pk (+) are expressed by the following equation (1).
It can be calculated by 1) and (12).

[0065]

(Equation 8)

As shown in the above equation (11), the smoothed value vector Bk (+) is calculated based on the observed value vector Ck according to the gain matrix Kk.
This is a vector biased on k or a vector biased on the predicted value vector Bk (-). The gain matrix Kk changes according to the prediction error covariance matrix Pk (-), that is, according to the variance Qk-1 of the driving noise vector ωk-1 (see the above equations (9) and (10)).

More specifically, the gain matrix Kk changes so that the larger the variance Qk-1 is, the more a smoothed value vector Bk (+) biased to the observed value vector Ck is obtained.
Therefore, according to the target tracking device of the present embodiment, by setting the variance Qk-1 of the driving noise vector ωk-1 to a large value, the smoothed value vector Bk emphasizing the observation value vector is set.
(+) Can be obtained. On the other hand, by setting the variance Qk-1 to a small value, it is possible to obtain a smoothed value vector Bk (+) emphasizing the predicted value vector.

As described above, the relationship of the above equation (6) holds between the state variable vector Bk and the observed value vector Ck. Further, the target tracking device of the present embodiment uses the predicted value vector Bk based on the motion state of the tracking target at time tk-1.
(-). Therefore, the predicted value vector Bk
By substituting (-) into the above equation (6), the following equation (1)
As shown in 3), based on the motion state of the tracking target at the time tk-1, the observation value vector Ck at the time tk is obtained.
Can be obtained.

[0069]

(Equation 9)

According to the following equation (14), based on the prediction error covariance matrix Pk (-), the prediction observation value vector Ck (-)
Can be obtained for the covariance matrix Sk (-). That is, according to the following equation (14), the covariance matrix Sk (-) for the predicted observation vector Ck (-) at the time tk can be predicted based on the state of the tracking target at the time tk-1. .

[0071]

(Equation 10)

The predicted observation value vector Ck (-) shown in the above equation (13) and the covariance matrix Sk shown in the above equation (14)
By substituting (-) into the following equation (15), it is possible to appropriately determine an area where the observation value vector Ck is predicted to exist at the time tk from the state of the tracking target at the time tk-1. The target tracking device of the present embodiment sets an area set by the above method as a tracking gate at each sampling, and creates the above-described target table for a video signal in the tracking gate.

[0073]

[Equation 11]

Next, the contents of processing executed in the target tracking device to determine whether or not the tracking target is turning will be described.

The target tracking device according to the present embodiment determines whether or not the tracking target is in a turning motion based on the motion state of the nose portion recognized on the target table. The target tracking device uses a nose state variable vector calculated by the motion model method and a nose observation value vector detected by the observation method as parameters representing the motion state of the nose portion. Hereinafter, the nose state variable vector and the nose observation value vector at the sampling time tk are referred to as Dk and Ak, respectively.

The nose state variable vector Dk and the nose observation value vector Ak can be expressed by the following equations (16) and (17), respectively. As shown in these figures, the nose observation value vector Ak represents the position of the nose.
It is a two-dimensional vector composed of two factors (rk0, byk0). On the other hand, the nose state variable vector Dk is composed of four factors (rk, byk, drk / dt, and dbyk /
dt).

[0077]

(Equation 12)

The nose state variable vector Dk at time tk
And the nose state variable vector Dk-1 at the time tk-1 can set the relationship shown in the following equation (18). Less than,
This relationship is referred to as a nose-related motion model.

[0079]

(Equation 13)

Note that Ψk and ak shown in the above equation (18)
Are the state transition matrix for the nose and the driving noise vector for the nose, respectively. State transition matrix Ψk
The drive noise model ak is assumed in advance when setting the motion model. The driving noise vector ak belongs to a normal distribution with an average “0” and a variance “Lk” as shown in the following equation (19).

[0081]

[Equation 14]

The nose state variable vector Dk at time tk
A relationship expressed by the following equation (20) can be set between the nose observation value vector Ak and the nose observation value vector Ak. Hereinafter, this relationship is referred to as an observation model regarding the nose.

[0083]

(Equation 15)

It should be noted that Nk and bk shown in the above equation (20)
Are observation matrices and observation noise vectors for the nose, respectively. In the present embodiment, the observation matrix is a matrix assumed in advance. The observation noise vector is given by the following equation (2)
As shown in 1), it belongs to a normal distribution with an average “0” and a variance “Mk”. The variance Mk is appropriately set to an appropriate value according to the observation state.

[0085]

(Equation 16)

As will be described later, the target tracking device of this embodiment smoothes the result obtained by the observation method and the result obtained by the motion model method at each sampling time, and performs the nose smoothing value vector Dk ( +) And the nose smoothing error covariance matrix Gk (+). The nose smoothing value vector Dk-1 (+) and the nose smoothing error covariance matrix Gk-1 at time tk-1
According to (+), as shown in the following equations (22) and (23), at time tk-1, the nose prediction value vector Dk (-) predicted to occur at time tk, and the nose A smooth error covariance matrix Gk (-) can be obtained.

[0087]

[Equation 17]

The target tracking apparatus according to the present embodiment internally divides the result obtained by the observation method and the result obtained by the motion model method with a predetermined gain, so that the nose smoothing value vector Dk (+) and Calculate the smoothed error covariance matrix Pk (+). The following equations (24) to (26) represent the nose gain matrix Vk and the nose gain matrix V
The nose smoothing value vector Dk (+) and the nose smoothing error covariance matrix Gk (+) calculated using k are shown.

[0089]

(Equation 18)

As described above, the relationship of the above equation (20) is established between the nose state variable vector Dk and the nose observation value vector Ak. Further, the target tracking device of the present embodiment can obtain the nose predicted value vector Dk (-) based on the nose motion state at the time tk-1. Therefore, if the nose prediction value vector Dk (-) is substituted into the above equation (20), as shown in the following equation (27), based on the motion state of the nose at the time tk-1, the time tk The predicted value Ak (-) of the nose observation value vector Ak can be obtained.

[0091]

[Equation 19]

According to the following equation (28), the covariance matrix Jk (-) for the nose prediction observation vector Ak (-) is obtained based on the nose prediction error covariance matrix Gk (-). Can be. That is, according to the following equation (28), the covariance matrix Jk (-) for the nose prediction observation vector Ak (-) at the time tk is predicted based on the nose state at the time tk-1. Can be.

[0093]

(Equation 20)

According to the nose prediction observation value vector Ak (-) shown in the above equation (27) and the covariance matrix Jk (-) shown in the above equation (28), the nose at the time tk-1 is obtained. From the state described above, an area in which the nose observation value vector Ak is expected to exist at the time tk can be appropriately determined.
The following equation (29) defines an area in which the nose observation value vector Ak is predicted to be present when the tracking target is turning. Therefore, in the present embodiment, when the nose observation value vector Ak at the time tk satisfies the condition of the following equation (29), it can be determined that the tracking target is turning, while
If the above condition is not satisfied, it can be determined that the tracking target is not turning.

[0095]

(Equation 21)

Next, the operation of the target tracking device of the present embodiment will be described with reference to FIG. FIG. 3 shows a routine that is started each time the target tracking device is requested to execute the target tracking process. In a series of processes shown in FIG.
First, the process of step 40 is performed.

In step 40, a process of creating a target table by performing a binarization process on the video signal supplied from the radar sensor is executed. The process of this step 40 sets the false alarm probability as a constant value,
This is executed while the threshold value used when binarizing each pixel signal constituting the video signal is kept constant. The quantization unit 10 shown in FIG. 1 is realized by the processing of step 40.

In step 42, a process for detecting an edge corresponding to the image of the tracking target in the target table created by the process in step 40 is executed. The inter-sweep correlation unit 12 shown in FIG. 1 is realized by the processing of step 42.

In step 44, in the target table, continuous detection hit outputs in the range axis direction or the azimuth axis direction are integrated as being caused by the same tracking target, and an observation value vector representing the center of gravity of the integrated area. C
k is calculated as shown in the above equations (2) and (3). The processing of step 44 implements the position calculation unit 14 shown in FIG.

In step 46, the coordinates of the pixel of the nose portion of the tracking target, that is, the nose observation value vector Ak is extracted from the integrated area of the detected hit output. This step 4
By the processing of No. 6, the nose extracting unit 18 shown in FIG. 1 is realized.

In step 48, the nose coordinates which are expected to be realized at the next sampling are predicted. Specifically, the nose observation value vector A detected in step 46
k, and a nose prediction value vector Dk (−) predicted based on the nose state at the time of the previous sampling,
The nose smoothing value vector D is obtained by the above equations (24) to (26).
processing for calculating k (+) and the nose smoothing error covariance matrix Gk (+), substituting the calculation results into the above equations (22) and (23), and is expected to occur at the next sampling A process of calculating the nose prediction value vector Dk + 1 (-) and the nose prediction error covariance matrix Gk + 1 (-) is executed.

The nose prediction value vector Dk + 1 (-) calculated in step 48 and the nose prediction error covariance matrix Gk +
1 (-) is used as a basis for turning determination after a delay of a unit time corresponding to the sampling interval. By the processing of step 48, the turning determination tracking filter unit 20 shown in FIG.
And a delay element 22 are realized.

In step 50, the nose predicted value vector D calculated in the previous sampling in step 48 is calculated.
By substituting k (-) and the nose prediction error covariance matrix Gk (-) into the above equations (27) and (28), the nose prediction value vector A at the current sampling time tk is obtained.
k (-) and an error covariance matrix Jk (-) are calculated. Furthermore,
In this step 50, whether the relationship of the above equation (29) holds between Ak (-) and Jk (-) calculated as described above and the nose observation value vector Ak observed by the current sampling. It is determined whether or not it is.

As a result, if it is determined that the above relationship is established, it is determined that the tracking target is in a turning motion. On the other hand, if it is determined that the above relationship does not hold,
Alternatively, if the tracking target cannot be detected or the nose portion is unknown, it is determined that the tracking target maintains the motion so far. The turning determination unit 24 shown in FIG. 1 is realized by the process of step 50.

In the target tracking device of this embodiment, whether or not the tracking target is turning is determined based on the movement of the nose of the aircraft that is the tracking target. That is, the target tracking device of the present embodiment executes the turning determination of the tracking target based on the movement of a point sufficiently distant from the center of gravity of the tracking target, that is, a point sufficiently distant from the turning center axis of the tracking target. When the tracking target is in a turning motion, a large movement is more likely to occur in the tracking target at a point farther from the center of gravity than in the vicinity of the center of gravity.
For this reason, according to the target tracking device of the present embodiment, after the tracking target starts the turning motion, the turning motion is immediately performed.
And it can detect correctly.

In step 52, a process for controlling the driving noise is executed. Specifically, in this step 52, when it is determined that the tracking target is in a turning motion, the variance Qk of the driving noise ωk shown in the above equation (5) is changed to a value Q1 corresponding to the turning motion, and the tracking is performed. When it is determined that the target is not performing the turning motion, a process of setting the variance Qk to a value Q2 (a value smaller than Q1) corresponding to the non-turning motion is executed. The driving noise control unit 26 shown in FIG. 1 is realized by the processing of step 52.

At step 54, the motion parameters of the tracking target are estimated. In this step 54, specifically, the processing of calculating the prediction value vector Bk (-) and the prediction error covariance matrix Pk (-) based on the state at the time of the previous sampling (the above equations (8) and (9) Equation), and the observed value vector C obtained by the current sampling
k, the smoothed value vector Bk (+) and the smoothed error covariance matrix Pk (+) are calculated based on the predicted value vector Bk (−) and the predicted error covariance matrix Pk (−) calculated as described above. (The above equations (10) to (12)) are executed.

As described above, according to the target tracking device of the present embodiment, the variance Qk of the drive noise ωk is set to a larger value when the tracking target is turning than when the tracking target is not turning. Is done. Further, as described above, according to the target tracking device of the present embodiment, as the variance Qk increases,
The result of the observation method is more likely to be reflected on the smoothed value vector Bk (+). Therefore, according to the target tracking device of the present embodiment, it is possible to form a state where the observation result is more important when the tracking target is performing a turning motion than when the tracking target is performing a non-rotating motion. it can.

The motion model shown in the above equation (4) is a model that accurately matches the actual movement of the tracking target when the tracking target is not turning. For this reason, when the tracking target is in a turning motion, inconsistency tends to occur between the motion model expressed by the above equation (4) and the actual motion of the tracking target. As described above, the target tracking device of the present embodiment forms a state where observation results are emphasized in such a situation. Therefore,
According to the target tracking device of the present embodiment, similarly to the case where the tracking target is not turning, the motion state of the tracking target can be detected with excellent detection accuracy even when the tracking target is turning. it can.

In step 56, the process of calculating the prediction area of the tracking gate based on the prediction value vector Bk (-) calculated in step 54 and the prediction error covariance matrix Pk (-), that is, the sampling time Based on the motion state of the tracking target at tk-1, a process of predicting the existence area of the tracking target at the sampling time tk is executed (the above-described equations (13) to (15)). In the next processing cycle, the processing of step 40 is executed for the tracking gate predicted by the processing of step 54. The tracking gate calculator 32 shown in FIG. 1 is calculated by the process of step 56.

In step 58, it is determined whether or not the target tracking processing is to be ended. As a result, if it is determined that the target tracking process should be continued, the process of step 60 is executed next. On the other hand, if it is determined that the target tracking processing should be ended, the routine shown in FIG. 3 is ended.

In step 60, a delay process for waiting for a unit time corresponding to the sampling interval to elapse is executed. When the above-described delay processing ends, the processing of step 40 is executed again. Thereafter, until the end of the target tracking processing is determined, the above steps 40 to 60 are performed at each sampling time.
Is repeatedly executed.

In the above embodiment, the tracking target is limited to the aircraft. However, the present invention is not limited to this. The tracking target may be another moving object such as a ship or a car. May be. In the above embodiment, the turning determination of the tracking target is performed based on the motion of the nose. However, the present invention is not limited to this. In this case, the processing can be executed based on the movement of an arbitrary point away from the center of gravity (for example, a point near the periphery of the image of the tracking target or a point on the tracking target separated from the center of gravity by a predetermined length).

In the above embodiment, the "turning state detecting means" according to the first aspect is determined by the turning determining unit 16.
The nose extracting unit 18 determines the "predetermined point setting means" according to claim 2, the "second predetermined point setting means" according to claim 3, and the "third predetermined point setting means" according to claim 4. "Setting means" are each realized.

Further, in the above embodiment, the position calculating section 14 provides the “observation value obtaining means” according to the fifth aspect.
The “predicted value acquiring unit”, the “smoothed value acquiring unit” according to claim 5,
"Gain changing means" are each realized.

Further, in the above embodiment, the tracking filter unit 28 calculates the gain matrix K according to the above equation (13).
7. The "gain calculating means" according to claim 6 is realized by calculating k, and the driving noise control unit 26
Thus, the "drive noise control means" according to claim 6 is realized.

Embodiment 2 Next, a second embodiment of the present invention will be described with reference to FIGS. FIG.
9 shows a block diagram of a target tracking device according to a second embodiment of the present invention. In FIG. 4, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and the description thereof will be omitted or simplified.

The target tracking device of this embodiment is characterized in that a gate control unit 62 is provided instead of the drive noise control unit 26 shown in FIG. Gate control unit 62
Is a part that outputs a threshold value dk for determining the size of the tracking gate to the tracking gate calculation unit 32 in response to the result of the rotation determination by the rotation determination unit 16.

FIG. 5 shows a flowchart of a series of processing executed in the target tracking device of the present embodiment. FIG.
Are executed repeatedly after the start of the target tracking process is requested and until the end thereof is requested, similarly to the series of processes shown in FIG. In FIG. 5,
Steps that execute the same processing as the steps shown in FIG. 3 are denoted by the same reference numerals, and description thereof will be omitted or simplified.

In the routine shown in FIG. 5, after the turning of the tracking target is determined by the processing of step 50,
The process of step 64 is performed. In step 64,
A threshold value for determining the size of the tracking gate when the tracking target is determined to be performing a turning motion by the above turning determination.
A predetermined value d1 is output as dk. On the other hand, this step 64
When the tracking target is determined to be in a non-turning motion by the turning determination, a predetermined value d2 smaller than d1 is output as the threshold value dk. The gate control unit 62 shown in FIG. 4 is realized by the processing of step 64.

The target tracking device according to the present embodiment uses the following equation (3)
The range of the observation value vector Ck satisfying the relationship 0) is calculated as the tracking gate. In the following equation (30), Ck (-) and Sk (-) are a predicted value of the observation vector Ck and an error covariance matrix for the predicted value Ck (-). These are respectively the same as in the first embodiment (1).
It is calculated by equations 3) and (14).

[0122]

(Equation 22)

According to the above equation (30), a larger tracking gate is calculated as the threshold value dk increases. Therefore, according to the target tracking device of the present embodiment, a larger tracking gate can be obtained when the tracking target is performing a turning motion than when the tracking target is performing a non-rotating motion. When the size of the tracking gate changes as described above, it is possible to appropriately fit the turning target within the tracking gate without setting an unnecessary large tracking gate during the non-turning movement of the tracking target. . For this reason, according to the target tracking device of the present embodiment, excellent tracking ability can be realized without performing an unnecessary large amount of processing.

In the above embodiment, the "tracking gate setting means" according to the seventh aspect is controlled by the tracking gate calculation section 32, and the "gate control means" according to the seventh aspect is controlled by the gate control section 62. Have been realized respectively.

Embodiment 3 Next, a third embodiment of the present invention will be described with reference to FIGS. FIG.
9 shows a block diagram of a target tracking device according to a third embodiment of the present invention. In FIG. 6, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and the description thereof will be omitted or simplified.

The target tracking device of the present embodiment is characterized in that a sampling interval control unit 66 is provided instead of the drive noise control unit 26 shown in FIG. The sampling interval control unit 66 is a unit that outputs a set value of the sampling interval Δtk to the quantization unit 10 in response to the result of the turning determination by the turning determining unit 16.

FIG. 7 shows a flowchart of a series of processes executed in the target tracking device of the present embodiment. FIG.
Are executed repeatedly after the start of the target tracking process is requested and until the end thereof is requested, similarly to the series of processes shown in FIG. In FIG. 7,
Steps that execute the same processing as the steps shown in FIG. 3 are denoted by the same reference numerals, and description thereof will be omitted or simplified.

In the routine shown in FIG. 7, after the turning of the tracking target is determined by the processing of step 50,
Step 68 is executed. In step 68,
When the tracking target is determined to be performing a turning motion by the above-described turning determination, the sampling interval Δtk is set to a predetermined value Δt1, and when the tracking target is determined to be performing a non-turning motion by the turning determination. To the sampling interval Δtk
It is set to a predetermined value Δt2 which is larger than t1. Step 6
8 implements the sampling interval control unit 66 shown in FIG.

According to the above setting, when the tracking target is in a turning motion, intensive observation is realized by narrowing the sampling interval Δtk, and is unnecessary when the tracking target is in a non-turning motion. Frequent observations can be prevented. For this reason, according to the target tracking device of the present embodiment, excellent target tracking capability can be efficiently secured.

In the above embodiment, the "sampling interval control means" according to claim 8 is realized by the sampling interval control unit 66.

Embodiment 4 Next, a fourth embodiment of the present invention will be described with reference to FIGS. FIG.
9 shows a block diagram of a target tracking device according to a fourth embodiment of the present invention. 8, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted or simplified.

The target tracking device of this embodiment is characterized in that a tracking filter control unit 70 and a second tracking filter unit 72 are provided instead of the driving noise control unit 26 and the tracking filter unit 28 shown in FIG. have. The second tracking filter unit 72 includes a tracking target position term (xk, yk, zk), a speed term (dxk / dt, dyk / dt, dzk / dt), and an acceleration term (d2xk / dt2, d2yk / dt2, d2zk / dt2) 9
The system includes a tracking filter 1 using a two-dimensional model and a tracking filter 2 using a six-dimensional model for estimating only a position term and a velocity term.

The tracking filter control unit 70 includes
The tracking filter to be used in the tracking target detection processing is determined to be one of the tracking filter 1 and the tracking filter 2 based on the determination result of Step 6. Second tracking filter section 72
Detects the motion state of the tracking target as described below by executing processing similar to that of the first embodiment using the selected tracking filter.

FIG. 9 shows a flowchart of a series of processing executed in the target tracking device of the present embodiment. The series of processing shown in FIG. 9 is repeatedly executed after the start of the target tracking processing is requested and the end thereof is requested, similarly to the series of processing shown in FIG. In FIG. 9, steps that execute the same processing as the steps shown in FIG. 3 are denoted by the same reference numerals, and description thereof is omitted or simplified.

In the routine shown in FIG. 9, after the turning of the tracking target is determined by the processing of step 50,
The process of step 74 is performed. In step 74,
The order control of the tracking filter is executed. Main step 74
Specifically, when it is determined that the tracking target is in a turning motion, the tracking filter 1 for the 9-dimensional model is determined, and when the tracking target is determined to be in a non-turning motion. The tracking filter 2 for the 6-dimensional model is
Each is selected as a tracking filter used in the tracking target detection process. The processing in step 70 shown in FIG. 8 is executed by the processing in step 74.

In this embodiment, the state variable vector B used in the 9-dimensional model targeted by the tracking filter 1
k, 1 and the state variable vector Bk, 2 of the 6-dimensional model targeted by the tracking filter 2 are given by the following equations (3
1) and (32).

[0137]

(Equation 23)

When the vector corresponding to the tracking filter 1 is represented by the suffix i = 1 and the vector corresponding to the tracking filter 2 is represented by the suffix i = 2, the motion model and the observation model corresponding to the respective tracking filter are represented. By the following equation (3)
3), (34) and the following equations (35), (36).

[0139]

(Equation 24)

Note that Φk, i and ω shown in the above equation (33)
k and i are a state transition matrix and a driving noise vector corresponding to the tracking filter i (i = 1, 2), respectively. Hk, i and vk shown in the above equation (35) are an observation matrix and an observation noise vector corresponding to the tracking filter i (i = 1, 2), respectively.

According to the subscript i for differentiating the tracking filters 1 and 2 as described above, the predicted value vectors Bk, i correspond to the 9-dimensional model and the 6-dimensional model, respectively.
(-), Prediction error covariance matrix Pk, i (-), gain matrix Kk, i,
The smoothed value vector Bk, i (+) and the prediction error covariance matrix Pk, i (+) can be expressed as in the following equations (37) to (41).

[0142]

(Equation 25)

Further, a prediction value vector Bk, i (-), a prediction error covariance matrix Pk, i (-) and an observation matrix Hk, i corresponding to each of the tracking filters i (i = 1, 2), and When the variance Rk of the observation noise vector Vk is used, a prediction value Ck (−) of the observation value vector Ck, an error covariance matrix Sk (−) for the prediction value Ck (−), and a conditional expression for determining a tracking gate are represented by: The following equations (42) to (44) can be used.

[0144]

(Equation 26)

In step 76, the above (37) to (4)
According to the expression (1), the predicted value vector Bk, i (-), the prediction error covariance matrix Pk, i (-), the gain matrix Kk, i, and the smoothed value vector Bk, corresponding to the tracking filter selected in the above step 74. i (+) and a prediction error covariance matrix Pk, i (+) are calculated. The processing in step 76 implements the second tracking filter section 72 shown in FIG.

When the tracking target is in a non-turning motion, no acceleration occurs in the tracking target in any of the x-axis direction, the y-axis direction, and the z-axis direction. Therefore, the motion state of the tracking target can be accurately grasped only by the position and the speed. On the other hand, when the tracking target is in a turning motion, an acceleration toward the turning center occurs with respect to the tracking target. Therefore, in order to accurately grasp the motion state of the tracking target, it is necessary to detect the acceleration of the tracking target.

According to the processing in step 76, when the tracking target is in a non-turning motion, the position term (x, y, z) and the speed term (dxk / dt, dy) are determined as the motion state of the tracking target.
k / dt, dzk / dt) can be obtained as a smoothed value vector Bk, 2 (+). On the other hand, when the tracking target is in a turning motion, the motion state of the tracking target is the acceleration term (d2xk / dt2, d2yk / dt2, d2zk /
dt2) can be obtained as a smooth value vector Bk, 1 (+).

Therefore, according to the above processing, when the tracking target is in a turning motion, the motion of the tracking target is not unnecessarily increased when the tracking target is in a turning motion. The state can be accurately detected. When the process of step 76 is completed, the above (42)
After the tracking gate is calculated according to the formulas (44) (step 56), the processing of step 58 is executed.

As described above, according to the target tracking device of the present embodiment, the turning state of the tracking target is accurately detected, and
By appropriately switching between the 9-dimensional model and the 6-dimensional model in accordance with the turning state of the tracking target, it is possible to efficiently realize excellent detection capability regarding the motion state of the tracking target.

In the above embodiment, the “state variable vector calculating means” according to the ninth aspect is controlled by the second tracking filter 72, and the “variable degree changing means” according to the ninth aspect is controlled by the tracking filter control unit 70. Has been realized respectively.

In the above-described embodiment, the "first tracking filter means" and the "second tracking filter means" according to the tenth aspect, by the tracking filter 1 and the tracking filter 2 constituting the second tracking filter section 72. However, the “variable order changing means” according to claim 10 is realized by the tracking filter control unit 70.

Embodiment 5 FIG. Next, FIGS. 10 and 11
Embodiment 5 of the present invention will be described with reference to FIG.
FIG. 10 shows a block diagram of a target tracking device according to the fifth embodiment of the present invention. In FIG. 10, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and description thereof will be omitted or simplified.

The target tracking device of the present embodiment differs from the drive noise control unit 26 and the tracking filter unit 28 shown in FIG. 1 in that the transition probability control unit 78 and the third tracking filter unit 8
0 is provided. In the present embodiment, the third tracking filter unit 80 estimates the motion data of the tracking target based on a multiple motion model composed of a plurality of motion models. Less than,
A method in which the third tracking filter unit 80 estimates the motion data of the tracking target will be described.

In the present embodiment, as a motion model constituting a multiple motion model, only a 9-dimensional model (hereinafter, referred to as motion model 1) for estimating the position, speed and acceleration of a tracking target, and only a position term and a speed term are used. And a six-dimensional model (hereinafter, referred to as a motion model 2) for estimating.
The motion model 1 is a model suitable for accurately grasping the state of the tracking target during the turning motion. On the other hand, exercise model 2
Is a model suitable only for accurately grasping the state of the tracking target during the non-turning motion.

As will be described later, the third tracking filter section 80
Sampling time for both motion models 1 and 2
The predicted value of the state variable vector Bk at tk is calculated. Hereinafter, the distinction between the operation value corresponding to the exercise model 1 and the operation value corresponding to the exercise model 2 is represented by using a subscript i or j.

The state variable vector Bk, 1 corresponding to the motion model 1 and the state variable vector Bk, 2 corresponding to the motion model 2 can be expressed by the following equations (45) and (46), respectively. The target tracking device of the present embodiment integrates these two state variable vectors to calculate a 9-dimensional integrated state variable vector Bk and Bk, i shown in the following equation (47).

[0157]

[Equation 27]

The order matching matrix Fi used in the above equation (47) is a vector (9) corresponding to the motion model 1.
This is a matrix for matching the order of the next order and the order of the vector (sixth order) corresponding to the motion model 2 to the ninth order which is the same as the order of the integrated state variable vector.

The motion model for the state variable vector Bk, i can be expressed by the following equations (48) and (49). On the other hand, the observation model for the state variable vector Bk, i can be expressed by the following equations (50) and (51).

[0160]

[Equation 28]

Note that Φk, i and ω shown in the above equation (48)
k and i are a state transition matrix and a driving noise vector corresponding to the motion model i (i = 1, 2), respectively. Hk, i and vk shown in the above equation (50) are an observation matrix and an observation noise vector corresponding to the motion model i (i = 1, 2), respectively.

In this embodiment, the predicted value vector Bk, i (-) for each motion model is a state transition matrix Φk, i set for each motion model, and an integrated smoothed value vector Bk (+) described later.
Is calculated using the following equation (52). The prediction error covariance matrix Pk, i (-) for each motion model is a state transition matrix Φk, i and variance Qk, i set for each motion model,
Using the integrated smoothing error covariance matrix Pk (+) described later, calculation is performed as in the following equation (53).

[0163]

(Equation 29)

The integrated smoothed value vector Bk (+) shown in the above equation (52) and the integrated smoothed error covariance matrix Pk (+) shown in the above equation (53) are the smoothed value vector calculated for each motion model. Using Bk, i (+), the smooth error covariance matrix Pk, i (+), the degree matching matrix Fi, and the like, calculation is performed as in the following equation (54) or (55).

[0165]

[Equation 30]

In this embodiment, the smoothed value vector Bk, i (+) and the smoothed error covariance matrix Pk, i (+) for each motion model
Is calculated as in the following equations (57) and (58) using the gain matrix Kk, i shown in the following equation (56).

[0167]

(Equation 31)

Further, βk, i used in the above equations (54) and (55) is the probability that the motion model i is established at the time tk. The probability βk, i is given by the following equation (59):
It is calculated by (60).

[0169]

(Equation 32)

It should be noted that the function g (a;
b; c) is a probability density function at a of the trivariate normal distribution of the mean b and the covariance matrix c.

In the calculation of the above equation (60), the transition probability Pij
Is used. The transition probability Pij is a probability that the motion state of the tracking target changes from a motion state matching the motion model j (j = 1 or 2) to a motion state matching the motion model i (i = 1 or 2). . The transition probability Pij is an important value for accurately detecting the motion state of the tracking target.

In this embodiment, the transition probability Pij is
If i = j, ie, the motion model that matches the tracking target does not change between two consecutive samplings, and if i ≠ j, ie, between two consecutive samplings, For each of the cases where the motion model conforming to the tracking target changes, it is defined as in the following equation (61).

[0173]

[Equation 33]

In the above equation (61), Δtk is a sampling interval. Λk is a target parameter set according to the state of the tracking target. According to the above equation (61), the larger the target parameter λk or the longer the sampling interval Δtk, the longer Pij (i =
j) has a small value and Pij (i ≠ j) has a large value.

As shown in the following equation (62), when it is determined that the tracking target is in a turning motion, the turning target parameter λ1 is substituted into the target parameter λk, and when the tracking target is in the non-turning motion, Is determined, the straight-ahead target parameter λ2 is substituted. The turning target parameter λ1 has a larger value than the straight traveling target parameter λ2.

[0176]

(Equation 34)

Therefore, when the tracking target is in a turning motion, the transition probability Pij (i = j) has a small value and the transition probability Pij (i ≠ j) has a large value. On the other hand, when the tracking target is in a non-turning motion, the transition probability Pij (i = j) has a large value and the transition probability Pij (i ≠ j) has a small value.

During the turning movement of the tracking target, it is considered that the probability that the motion model conforming to the tracking target changes between the motion model 1 and the motion model 2 is high. Therefore, in this case, the transition probability Pij (i = j) is a small value, and
It is appropriate that the transition probability Pij (i ≠ j) is a large value. On the other hand, during the non-turning motion of the tracking target, it is considered that the probability that the motion model conforming to the tracking target changes between the motion model 1 and the motion model 2 is low. Therefore, in this case, the transition probability Pij (i = j) is a large and small value,
It is appropriate that the transition probability Pij (i ≠ j) is a small value.

According to the setting of the above equation (62), the appropriate state described above can be formed both during the turning movement of the tracking target and during the non-turning movement. Therefore, according to the target tracking device of the present embodiment, the transition probability Pij can always be set to an appropriate value regardless of the turning state of the tracking target.

From the above equations (52) and (53),
The predicted value vector Bk, i (-) calculated for each motion model is
By using the following equations (63) and (64), respectively, the integrated prediction value vector Bk (-) and the integrated prediction error covariance matrix Pk (-) can be integrated.

[0181]

(Equation 35)

In the above equation (63), “Pij · βk-1, j”
Represents the probability that the motion model conforming to the tracking target at the time tk-1 is the motion model j and the motion model conforming to the tracking target at the time tk is the motion model i. The value of the predicted value vector Bk, i (-) for the motion model i is reflected in the integrated predicted value vector Bk (-) as the value "Pij.beta.k-1, j" increases.

"Βk-1, j" becomes a large value when j matches the number of the motion model that matches the tracking target at the time tk-1. Therefore, “Pij · βk−1, j” becomes a large value when j satisfies the above condition. If j satisfies the above condition, “Pij · βk−1, j” becomes Pij
The larger the value, the larger the value.

As described above, when the tracking target is in a turning motion, Pij is small, and Pij is small.
(I ≠ j) is set to be large. Therefore, in this case, “Pij · βk−1, j” is a large value for i where i ≠ j holds. For this reason, when the tracking target is in a turning motion, the integrated predicted value vector Bk, i (-) differs from the motion model (for example, motion model 1) that was adapted to the tracking target at the time tk-1. The predicted value vector Bk, i (-) corresponding to the motion model (for example, motion model 2) is largely reflected.

On the other hand, in this embodiment, Pij is set so that Pij (i = j) is large and Pij (i ≠ j) is small when the tracking target is in a non-turning motion. I have. Therefore, in this case, “Pij · βk-1, j” is
It becomes a large value for i where i = j holds. Therefore, when the tracking target is in a non-turning motion, the integrated predicted value vector Bk (-) includes the same motion as the motion model (for example, the motion model 1) that matches the tracking target at the time tk-1. The predicted value vector Bk (-) corresponding to the model (for example, the exercise model 1) is largely reflected.

As described above, according to the target tracking apparatus of the present embodiment, during the turning movement of the tracking target, the motion model mainly used in the calculation of the integrated predicted value vector Bk (-) tends to change. Can be formed, and during the non-turning movement of the tracking target, a state in which the movement model mainly used in the calculation of the integrated predicted value vector Bk (−) hardly changes can be formed.

In this embodiment, the observation value vector Ck
Is a predicted value Ck (−), an error covariance matrix Sk (−) for the predicted value Ck (−), and a conditional expression for determining a tracking gate are:
The integrated predicted value vector B calculated by the above equation (63)
The following equations (65) to (67) are obtained using k (−), the integrated prediction slip error covariance matrix Pk (−) calculated by the above equation (64), the observation matrix Hk, and the variance Rk of the observation noise vector Vk. ).

[0188]

[Equation 36]

FIG. 11 shows a flowchart of a series of processes executed in the target tracking device of the present embodiment. Similar to the series of processing shown in FIG. 3, the series of processing shown in FIG. 11 is repeatedly executed after the start of the target tracking processing is requested and until the end thereof is requested. In FIG. 11, steps that execute the same processing as the steps shown in FIG. 3 are denoted by the same reference numerals, and description thereof will be omitted or simplified.

In the routine shown in FIG. 11, after the turning of the tracking target is determined by the processing of step 50, the processing of step 82 is performed. In step 82, processing for controlling the set value of the transition probability Pij is executed. In this step 82, specifically, as shown in the above equation (62), when it is determined that the tracking target is in a turning motion, λ1 is set to the target parameter λk,
If it is determined that the tracking target is in a non-turning motion, λ2 is set as the target parameter λk, and (6)
A process of calculating the transition probability Pij according to the expression 1) is executed. The transition probability control unit 78 shown in FIG. 10 is realized by the processing of step 82.

According to the above setting, when the tracking target is in a turning motion, a state in which the movement model used initiatively changes easily is formed, and the followability of the target tracking device with respect to the turning target is improved. Can be. Further, according to the above setting, when the tracking target is in a non-turning motion, a state in which the movement model used initiatively does not easily change is formed, and the observation accuracy can be stabilized.

In step 84, using the transition probability Pij calculated in the process of step 82, the process of estimating the motion data of the tracking target is performed by the method based on the multiple motion model described above (the above equation (Eq. 52) to (60), (6)
3), (64)).

After the processing of step 84 is completed, the tracking gate is calculated by the above equations (65) to (67) (step 56), and then the processing of step 58 is executed.

In the above embodiment, the number of motion models constituting the multiple motion model is set to two. However, the present invention is not limited to this.
A multiple motion model may be configured by more motion models.

In the above-described embodiment, the third tracking filter unit 80 calculates the predicted value vector Bk,
By calculating i (-), the “multiple motion specification estimating means” according to claim 11 calculates the transition probability Pij, and the “transition probability calculating means” according to claim 11 calculates the prediction value vector. Bk, i (-) is integrated to obtain an integrated predicted value vector B
By calculating k (−), the “motion data integration means” according to claim 11 is realized, and the “transition probability control means” according to claim 11 is realized by the transition probability control unit 78. Has been realized.

Embodiment 6 FIG. Next, FIG. 12 and FIG.
Embodiment 6 of the present invention will be described with reference to FIG.
FIG. 12 shows a block diagram of a target tracking device according to the sixth embodiment of the present invention. In FIG. 12, the same components as those shown in FIG. 1 are denoted by the same reference numerals, and the description thereof will be omitted or simplified.

The target tracking device of this embodiment is different from the driving noise control unit 26 and the tracking filter unit 28 shown in FIG. 1 in that the reliability control unit 86 and the fourth tracking filter unit 88 are used.
It has. In the present embodiment, the fourth tracking filter unit 88 uses the above equations (52) to (58), equations (63), and (64) to track in a manner similar to that of the fifth embodiment. The target motion specifications, that is, the integrated state variable vector Bk is calculated.

In Embodiment 5 described above, the probability βk, i that the motion model i is established is calculated by the above equation (59) and (6)
It is calculated based on equation (0). The target tracking device of the present embodiment is characterized in that the reliability control unit 24 forcibly corrects the probability βk, i based on the turning determination result of the turning determining unit 16.

FIG. 13 shows a flowchart of a series of processes executed in the target tracking device of the present embodiment. The series of processing shown in FIG. 13 is repeatedly executed after the start of the target tracking processing is requested and the end thereof is requested, similarly to the series of processing shown in FIG. In FIG. 13, steps that execute the same processing as the steps shown in FIG. 3 are denoted by the same reference numerals, and description thereof will be omitted or simplified.

In the routine shown in FIG. 13, after the turning of the tracking target is determined by the processing of step 50, the processing of step 90 is executed. In step 90, the reliability control of the motion model is executed. In the present step 90, specifically, when it is determined that the tracking target is in a turning motion, the establishment probability βk, 1 of the motion model 1 is set to “1”, and the establishment probability βk, 2 to “0”, and the probability of establishment βk, 1 of the motion model 1 when it is determined that the tracking target is in the non-turning determination.
Is set to “0” and the probability of establishment βk, 2 of the motion model 2 is set to “1”.

In step 92, using the probability βk, i set by the processing in step 90, the motion data of the tracking target is estimated by a method based on the multiple motion model described above. When the processing in step 92 is completed, the processing in and after step 56 is executed.

According to the above processing, the value of the probability βk, i is forcibly set to “1” or “0” according to the motion state of the tracking target.
By doing so, the probability of establishment of the motion model 1 suitable for grasping the turning motion during the turning motion of the tracking target is increased, and the motion model suitable for grasping the straight-line motion during the non-turning motion of the tracking target. 2 can be increased. For this reason, according to the target tracking device of the present embodiment, it is possible to realize excellent tracking performance in tracking the tracking target.

In the above embodiment, the fourth tracking filter unit 88 calculates the predicted value vector Bk,
By calculating i (-), the "multiple motion specification estimating means" according to claim 13 integrates the predicted value vectors Bk, i (-) to calculate an integrated predicted value vector Bk (-). Accordingly, the "second motion specification integrating means" according to claim 13 is realized, and the "reliability control means" according to claim 13 is realized by the reliability control unit 86.

[0204]

Since the present invention is configured as described above, it has the following effects. Claim 1
According to the fifteenth aspect, the turning state of the tracking target is detected based on the movement of a point different from the center of gravity. While the tracking target is turning, a large movement is likely to occur at a point different from the center of gravity. For this reason, according to the present invention, the turning state of the tracking target can be detected quickly and accurately.

According to the second aspect of the present invention, the turning determination of the tracking target is executed based on the movement of the end of the tracking target in the long axis direction. For this reason, according to the present invention, accurate turning determination can be performed.

According to the third aspect of the present invention, the turning determination of the tracking target is executed based on the movement of an arbitrary point of the hit output of the tracking target. For this reason, according to the present invention, accurate turning determination can be easily performed.

According to the fourth aspect of the present invention, the determination of the turning of the tracking target is executed based on the movement of a point separated from the center of gravity by a predetermined distance. A point at a predetermined distance from the center of gravity indicates a change according to the turning speed of the tracking target. Therefore, according to the present invention,
The turning state of the tracking target can be accurately detected.

According to the fifth aspect of the present invention, the smoothed value representing the movement of the center of gravity is calculated by smoothing the observed value and the predicted value. The gain used for calculating the smoothed value is changed according to the turning state of the tracking target. For this reason, the smoothed value always accurately matches the state of the actual center of gravity regardless of whether the tracking target is performing the turning motion or the tracking target is performing the non-turning motion.

[0209] According to the invention of claim 6, when the tracking target is in a turning motion, the driving noise is set to a larger value than when the tracking target is in a non-turning motion. As a result, the predetermined gain becomes a value that emphasizes the observation value, and a state is formed in which the observation value is easily reflected on the smoothed value. Therefore, according to the present invention, it is possible to obtain a smooth value that accurately represents the actual state of the center of gravity regardless of the turning state of the tracking target.

According to the present invention, since the tracking gate is set according to the turning state of the tracking target, an excellent tracking ability is realized without executing unnecessary processing. be able to.

According to the eighth or seventeenth aspect of the present invention, since the sampling interval is changed according to the turning state of the tracking target, intensive observation is performed during the turning movement of the tracking target, and The calculation load can be reduced during the non-swinging motion.

According to the ninth or eighteenth aspect, the order of the state variable vector representing the motion state of the tracking target is changed in accordance with the turning state of the tracking target. Can be estimated based on an appropriate state variable according to the exercise state.

According to the tenth aspect of the present invention, when the tracking target is in a turning motion, a state variable vector including acceleration in addition to the position and speed of the tracking target can be detected, and if the tracking target is not moving. If you are in a turning motion,
A state variable vector that does not include the acceleration of the tracking target can be detected. Therefore, according to the present invention, it is possible to secure an excellent detection capability during the turning movement of the tracking target while suppressing the calculation load during the non-turning movement of the tracking target.

According to the eleventh or nineteenth aspect of the present invention, the transition probability used when integrating the motion data estimated based on the multiple motion model is changed according to the turning state of the tracking target. Therefore, the followability to the turning target can be improved.

According to the twelfth aspect of the present invention, the first motion model suitable for accurately grasping the motion state of the tracking target in the turning state and the motion state of the tracking target in the non-turning state are included in the multiplex model. Since the second motion model suitable for grasping accurately is included, the motion state of the tracking target can be grasped accurately both during the turning motion of the tracking target and during the non-turning motion.

According to the thirteenth or twentieth aspect of the present invention, the probability of establishing an appropriate motion model for grasping the state of the tracking target is increased in accordance with the turning state of the tracking target. The tracking ability for tracking can be improved.

According to the fourteenth aspect of the present invention, during the turning movement of the tracking target, the probability of establishing a motion model suitable for grasping the tracking target during the turning movement is set to 1, while on the other hand, during the non-turning movement of the tracking target. Since the probability of establishing a motion model suitable for grasping a tracking target during a non-turning motion can be set to 1, the tracking performance of tracking the tracking target can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a target tracking device according to a first embodiment of the present invention.

FIG. 2 is an example of a target table created by an inter-sweep correlation unit shown in FIG. 1;

FIG. 3 is a flowchart of a series of processes executed in the target tracking device shown in FIG.

FIG. 4 is a block diagram of a target tracking device according to a second embodiment of the present invention.

5 is a flowchart of a series of processes executed in the target tracking device shown in FIG.

FIG. 6 is a block diagram of a target tracking device according to a third embodiment of the present invention.

7 is a flowchart of a series of processes executed in the target tracking device shown in FIG.

FIG. 8 is a block diagram of a target tracking device according to a fourth embodiment of the present invention.

9 is a flowchart of a series of processes executed in the target tracking device shown in FIG.

FIG. 10 is a block diagram of a target tracking device according to a fifth embodiment of the present invention.

11 is a flowchart of a series of processes executed in the target tracking device shown in FIG.

FIG. 12 is a block diagram of a target tracking device according to a sixth embodiment of the present invention.

13 is a flowchart of a series of processes executed in the target tracking device shown in FIG.

[Explanation of symbols]

Reference Signs List 10 quantization unit, 12 inter-sweep correlation unit, 14 position calculation unit, 16 turning judgment unit, 18 nose extraction unit, 20
Tracking filter unit for turning judgment, 24 turning judgment execution unit,
26 drive noise control unit, 28 tracking filter unit, 32
Tracking gate calculation unit, 62 gate control unit, 66 sampling interval control unit, 68 tracking filter control unit, 70
2nd tracking filter section, 78 transition probability control section, 80 3rd tracking filter section, 86 reliability control section, 88 4th tracking filter section.

Claims (20)

[Claims] 1. A target tracking device having a function of detecting a center of gravity of a tracking target from an image of the tracking target, wherein the turning of the tracking target is performed based on a movement of a predetermined point on the tracking target different from the center of gravity. A target tracking device comprising turning state detection means for detecting a state. 2. The apparatus according to claim 1, further comprising a predetermined point setting unit configured to detect a long axis direction of the tracking target from a feature of the tracking target, and to set an end of the tracking target in the long axis direction as the predetermined point. 2. The target tracking device according to claim 1, wherein 3. The target tracking device according to claim 1, further comprising a second predetermined point setting unit that sets the predetermined point to an arbitrary point of the hit output of the tracking target. 4. The target tracking apparatus according to claim 1, further comprising a third predetermined point setting unit that sets a point on the tracking target separated by a predetermined distance from the center of gravity as the predetermined point. 5. Observation value obtaining means for obtaining an observation value representing the state of the center of gravity at a predetermined time by observation, and indicating the state of the center of gravity at the predetermined time based on the movement of the tracking target before a predetermined time. Predicted value obtaining means for predicting a predicted value; smoothed value obtaining means for obtaining a smoothed value by smoothing the observed value and the predicted value with a predetermined gain; and the predetermined gain according to a turning state of the tracking target. The target tracking device according to claim 1, further comprising: a gain changing unit configured to change the gain. 6. The gain calculating means for calculating the predetermined gain based on a parameter including driving noise such that the larger the driving noise is, the larger the observed value is reflected on the smoothed value. And a drive noise control unit that sets the drive noise to a larger value when the tracking target is in a turning motion than when the tracking target is in a non-turning motion. The target tracking device according to claim 1. 7. A tracking gate setting means for predicting a region where the tracking target exists at a predetermined time based on a state of the tracking target before a predetermined time, and setting the prediction region as a tracking gate at the predetermined time. The target tracking device according to claim 1, further comprising: gate control means for changing a size of the tracking gate based on a turning state of the tracking target. 8. The target tracking device according to claim 1, further comprising a sampling interval control unit that controls a sampling interval for monitoring the tracking target based on a turning state of the tracking target. 9. A state variable vector calculating means for obtaining a state variable vector including a state variable of a predetermined order representing a motion state of the tracking target, and a variable order for changing the predetermined order based on the turning state of the tracking target. The target tracking device according to claim 1, further comprising: changing means. 10. The method according to claim 1, wherein the state variable vector calculating means includes a first, a second, and a third positions, the position, the speed, and the acceleration of the tracking target.
First tracking filter means for obtaining a state variable vector of the following, and second tracking filter means for obtaining a second state variable vector composed of the position and velocity of the tracking target, wherein the variable order changing means, 10. The target tracking device according to claim 9, wherein one of the first tracking filter and the second tracking filter is selectively operated based on a turning state of the tracking target. 11. A multi-motion specification estimating means for estimating the motion data of the tracking target corresponding to each of the plurality of motion models; a transition probability calculating means for calculating a transition probability between the plurality of motion models; Motion specification integrating means for integrating the motion parameters of the tracking target corresponding to each of the plurality of motion models based on the transition probability; and a transition probability for changing the transition probability based on the turning state of the tracking target. The target tracking device according to claim 1, further comprising: control means. 12. The plurality of motion models include a first motion model for estimating a state variable vector composed of the position, velocity and acceleration of the tracking target, and a state variable composed of the position and velocity of the tracking target. The target tracking apparatus according to claim 11, further comprising: a second motion model for estimating a vector. 13. A multi-motion specification estimating means for estimating the motion data of the tracking target corresponding to each of the plurality of motion models; and calculating the motion data of the tracking target corresponding to each of the plurality of motion models. Second motion specification integrating means for integrating on the basis of the establishment probabilities of the plurality of motion models; and, based on the turning state of the tracking target, the establishment probability of an appropriate motion model for grasping the state of the tracking target. 2. The target tracking device according to claim 1, further comprising: reliability control means for setting the probability of establishment for the plurality of motion models such that the relative probability increases. 14. The plurality of motion models are: a first motion model for estimating a state variable vector composed of the position, velocity and acceleration of the tracking target; and a state variable composed of the position and velocity of the tracking target. A second motion model for estimating a vector, wherein the reliability control means sets the probability of establishment of the first motion model to 1 and the probability of formation of the second motion model to 0 during the turning motion of the tracking target. And the probability of establishment of the first motion model is set to 0 during the non-turning motion of the tracking target, and
14. The target tracking device according to claim 13, wherein the establishment probability of the second motion model is set to 0. 15. A target tracking method having a step of detecting a center of gravity of a tracking target from a feature of the tracking target, wherein the turning of the tracking target is performed based on a movement of a predetermined point on the tracking target different from the center of gravity. A target tracking method comprising a turning state detecting step of detecting a state. 16. A tracking gate setting step of predicting an area where the tracking target exists at a predetermined time based on a state of the tracking target before a predetermined time, and setting the prediction area as a tracking gate at the predetermined time. The target tracking method according to claim 15, further comprising: a gate control step of changing a size of the tracking gate based on a turning state of the tracking target. 17. Based on the turning state of the tracking target,
16. The target tracking method according to claim 15, further comprising a sampling interval control step of controlling a sampling interval for monitoring the tracking target. 18. A state variable vector calculating step for obtaining a state variable vector including a state variable of a predetermined order representing a motion state of the tracking target, and a variable order for changing the predetermined order based on a turning state of the tracking target. The target tracking method according to claim 15, further comprising: changing the target. 19. A multi-motion specification estimating step of estimating the motion data of the tracking target corresponding to each of the plurality of motion models; a transition probability calculating step of obtaining a transition probability between the plurality of motion models; A movement specification integration step of integrating the movement specifications of the tracking target corresponding to each of the plurality of movement models based on the transition probability; and a transition probability of changing the transition probability based on the turning state of the tracking target. The target tracking method according to claim 15, comprising: a control step. 20. A multi-motion specification estimating step of estimating the motion data of the tracking target corresponding to each of the plurality of motion models; and calculating the motion data of the tracking target corresponding to each of the plurality of motion models. A second motion specification integration step of integrating based on the establishment probabilities of the plurality of exercise models; and a probability of establishment of an appropriate exercise model for grasping the state of the tracking target based on the turning state of the tracking target. The target tracking method according to claim 15, further comprising: a reliability control step of setting a probability of establishment for the plurality of motion models so that is relatively high.