JP3208634B2 - Same target determination device and same target determination method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technology for judging the identity of a target from information of a target observed by a plurality of sensors or human eyes and information of a detected target already observed. In particular, it is used for a target monitoring system including an air traffic control system.

[0002]

2. Description of the Related Art Conventional similar devices of this type include, for example, IEEE Transaction on Automatic Control, Vol.
24, No.6, December 1979, pp843-854 A n Algorithm for
Tracking Multiple Target *
FIG. 36 is a block diagram showing the configuration of a conventional target tracking device shown in the above-mentioned document, and FIG. 37 is a flowchart for explaining the operation of FIG.

In FIG. 36, a target observation device 21 outputs a target observation position and an observation time, which are detection results from unnecessary signals such as a tracking target and clutter, as detection data.
22 is an observation data transfer device for transferring detection data from the target observation device, 23 is a target prediction processor for calculating prediction data of each target based on the observation time of the detection data and smooth data of each target, 24 Is a first delay circuit for delaying the prediction data by one sampling, and 25 is a motion data correlation data calculation for calculating correlation data of motion data for each target based on the detection data and the prediction data of each target. 26, a motion data correlator for selecting detection data having a possibility of correlation with each target based on the correlation data of the new target and the motion data for each target, and 27 a detection data selected by the correlator. Is a detection data hypothesis generator that generates a hypothesis composed of identification results of whether the target is a new target, a tracked target, or an unnecessary signal such as clutter. 28 calculates the hypothesis reliability based on the correlation data between the above hypothesis and the motion data. Luck A hypothesis reliability calculator based on dynamic data, 29 is a quasi-optimal hypothesis selector that reduces hypotheses based on the hypothesis reliability and the like, and 30 is a hypothesis target smoother that calculates smoothing parameters of each target in the quasi-optimized hypothesis 31 is a tracking establishment determiner that determines whether tracking has been established based on the reliability of the quasi-optimized hypothesis, and 32 is a smoothing parameter of the target for which the tracking has been established by the tracking establishment determiner. Is a wake indicator that displays.

In FIG. 37, first, at step (30)
In 1), the target observation device 21 inputs motion data such as a target position and detection data such as observation time, and then in step (302), the target prediction processor 23 uses the observation time of the detection data and smoothing data of each target. The motion data correlation parameter calculator 25 calculates the predicted data of each target based on the predicted data and the motion data correlation data calculator 25 in step (303) based on the predicted data and the detected data. The target predicted existence range is calculated as a starting point, and then, in step (304), the motion data correlator 26 generates a cluster based on the initial target predicted existence range and the target predicted existence range, whereby the possibility of correlation with the target is calculated. Then, in step (305), the detection data hypothesis generator 27 determines whether the detection data in the cluster is a new target or a tracked target. Generating hypotheses consisting identification result of whether unwanted signals of the clutter and the like, then step (30
In 6), the hypothesis reliability calculator 28 based on the motion data calculates the hypothesis reliability of the detection data based on the hypothesis and the correlation data of the motion data. Then, in step (307), the suboptimal hypothesis selector 29 calculates the hypothesis reliability. Reduce hypotheses based on reliability, etc.
Next, in step (308), the hypothesis target smoothing processor 3
0 calculates smooth specifications for the target in the hypothesis remaining in the quasi-optimization processing by the quasi-optimum hypothesis selector 29, and then, in step (309), the tracking establishment determiner 31 It is determined whether or not the tracking has been established based on the reliability of the remaining hypothesis. Then, when the tracking has been established in step (310), the smoothed data is output to the wake indicator 32.
In step (311), it is determined whether or not the processing is to be terminated, and this series of steps is repeated until the tracking is completed.

[0005]

A conventional target tracking device having the same target determination function is configured as described above.
In a tracking system having a radar fixed to one platform as a sensor, the position information of the target may be ambiguous due to an error in the target observation system, or a remote location may be located at a plurality of points where there is a delay difference in data transmission. Like observation data from sensors, the order of input target observation data is different from the order of observation times, and it is not time-series data, or like target observation data from passive sensors,
When observation data is not always obtained, it is difficult to determine the identity of the target.

[0006]

The present invention has been made to solve the above problems.
Has been made in order, the invention is eye according to claim 1
Target position information contains ambiguity due to errors in the target observation system.
The same goal determination device that can determine the identity of the goal
The purpose is to get. Further, and when the observation data of the target is not a time series, even if the observation data is not regularly, and aims at obtaining the same target determining device capable of determining the identity of the target.

According to a second aspect of the present invention, there is provided an identical target determination device capable of determining the identity of a target when movement data is not necessarily obtained as target observation data and attribute data is obtained. The purpose is.

According to a third aspect of the present invention, there is provided an identical target determination apparatus capable of determining the identity of a target from an attribute tree when only attribute data is obtained without obtaining movement data as target observation data. The purpose is to get.

According to a fourth aspect of the present invention, there is provided an identical target determination device capable of determining the identity of a target from an attribute tree when only movement data is obtained as target observation data and no attribute data is obtained. It is intended to be.

[0011]

The invention according to claim 5 provides a method for determining the same target which can determine the identity of the target from the attribute tree when the movement data is not obtained as the target observation data and only the attribute data is obtained. The purpose is to get.

The invention according to claim 6 to claim 11 is characterized in that the position information of the target includes ambiguity due to an error in the target observation system, the target observation data is not in a time series, It is intended to obtain the same target determination method that can determine the sameness of the target even when is not regular, and also when the movement data is not obtained as the target observation data and only the attribute data is obtained. .

[0014]

[0015]

In order to achieve the above object, the same target determination apparatus according to the first aspect of the present invention includes the following elements so as to determine the target identity from target information obtained by observing the target. It is. (1) a probability density function expressing means for expressing the spatial existence of the observed target by a probability density function; (2) using the transition probability, the probability density functions of both the above-mentioned observed target and the already-observed detected target are calculated. Transition probability calculating means for adjusting the time; (3) distance calculating means for calculating the distance between the reference positions of the two probability density functions adjusting the time; (4) a target value based on the distance between the reference positions; Identity determining means for determining identity.

[0017]

The same target determining apparatus according to the second aspect of the present invention includes the following elements so as to determine the identity of the target from target information obtained by observing the target. (1) motion data estimation means for estimating motion data from attribute data of an observed target; (2) probability density function expression means for expressing the spatial existence of the target by a probability density function; And a transition probability calculating means for adjusting the times of the probability density functions of both of the detected targets that have already been observed using the transition probabilities. (4) Obtaining the distance between the reference positions of the probability density functions of both of the times that have been adjusted. (5) Identity determination means for determining the identity of the target based on the distance between the reference positions.

The same target determination apparatus according to a third aspect of the present invention includes the following elements so as to determine the identity of the target from target information obtained by observing the target. (1) The overall classification of the target is expressed by a hierarchical attribute tree, and the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as the probability of belonging) is estimated from the attribute data of the observed target. (2) Similarity calculation means for comparing the affiliation probabilities of both the observed target and the already-observed detected target to determine the similarity of the target. (3) Identity determination means for determining the identity of the target based on the identification.

The same target determination apparatus according to a fourth aspect of the present invention includes the following elements so as to determine the identity of the target from target information obtained by observing the target. (1) Means for estimating the attribute data of the target from the observed movement data of the target; (2) Representing the entire classification of the target of interest in a hierarchical attribute tree, and calculating the target from the estimated attribute data. Belonging probability estimating means for estimating the probability of belonging to each node of the attribute tree (hereinafter, referred to as belonging probability). (3) Comparing the belonging probability of both the above-observed target and the already-observed detected target. (4) Identity determining means for determining the target identity based on the similarity.

[0021]

The same target determination method of the invention according to claim 5 includes the following steps for determining the sameness of the target from target information obtained by observing the target. (1) First, the overall classification of the target to be processed is represented by a hierarchical attribute tree.
A similarity that estimates the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as the belonging probability), and compares this belonging probability with the belonging probability of the already detected target to obtain the similarity of the target. (2) Next, an identity determination step of determining the identity of the target based on the similarity.

The same target determination method of the invention according to claim 6 includes the following steps for determining the sameness of the target from target information obtained by observing the target. (1) First, the spatial existence of the observed target is represented by a probability density function, and the times of the probability densities of both the observed target and the already-detected detected target are matched by using transition probabilities. (2) Next, a step of selecting a candidate for a detected target having a high degree of identity with the observed target based on the correlation coefficient of the probability density function described above, (3) Next, The overall classification of the target is expressed by a hierarchical attribute tree, and from the attribute data of the observed target, the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as belonging probability) is estimated. Calculating the similarity of the target by comparing the probability with the belonging probability of the already detected detected target; and (4) determining the identity of the target based on the similarity.

The same target determination method according to the invention according to claim 7 includes the following steps for determining the sameness of the target from target information obtained by observing the target. (1) First, the overall classification of the target to be processed is represented by a hierarchical attribute tree.
Estimating a probability that the target belongs to each node of the attribute tree (hereinafter, referred to as a member probability), and comparing the member probability with a member probability of an already-observed detected target to obtain similarity of the target; (2) Next, a step of selecting a candidate for a detected target having a high degree of identity with the observed target based on the above similarity, (3) Next, the spatial existence of the observed target is represented by a probability density function, Adjusting the times of the probability density functions of both the observed target and the already detected target using the transition probabilities to determine the degree of overlap of the probability density functions; (4) Next, the above probability density function Determining the identity of the target based on the correlation coefficient of.

The same target determination method of the invention according to claim 8 includes the following steps so as to determine the target identity from target information obtained by observing the target. (1) First, the spatial existence of the observed target is represented by a probability density function, and the times of the probability densities of both the observed target and the already-detected detected target are matched by using transition probabilities. In parallel with the step of determining the degree of overlap of the functions, (2) the overall classification of the target to be processed is represented by a hierarchical attribute tree, and from the attribute data of the observed target, the target is stored in each node of the attribute tree. A step of estimating the probability of belonging (hereinafter referred to as a member probability) and comparing the member probability with the member probability of an already-observed detected target to determine the similarity of the target; (3) Then, the above probability density Determining the identity of the target based on the correlation coefficient of the function and the similarity.

The same target judging method according to the ninth aspect of the present invention includes the following steps for judging the identity of the target from target information obtained by observing the target. (1) First, the spatial existence of the observed target is represented by a probability density function, and the times of the probability densities of both the observed target and the already-detected detected target are matched by using transition probabilities. (2) Next, a step of selecting a candidate of a detected target having a high degree of identity with the target observed based on the distance between the reference positions, (3) Then, The overall classification of the target is represented by a hierarchical attribute tree, and from the attribute data of the observed target, the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as belonging probability) is estimated. (4) Next, a step of determining the similarity of the target by comparing the detected probability and the belonging probability of the already detected target. (4) Next, determining the identity of the target based on the similarity.

The same target determination method according to the tenth aspect of the present invention includes the following steps so as to determine the target identity from target information obtained by observing the target. (1) First, the overall classification of the target to be processed is represented by a hierarchical attribute tree.
Estimating a probability that the target belongs to each node of the attribute tree (hereinafter, referred to as a member probability), and comparing the member probability with a member probability of an already-observed detected target to obtain similarity of the target; (2) Next, a step of selecting a candidate for a detected target having a high degree of identity with the observed target based on the similarity, (3) Next, the spatial existence of the observed target is represented by a probability density function. Adjusting the times of the probability density functions of both the observed target and the already detected target using transition probabilities to obtain the distance between the reference positions of the probability density functions; (4) Next, the reference position Determining the identity of the target based on the distance.

The same target determination method according to the eleventh aspect of the present invention includes the following steps so as to determine the target identity from target information obtained by observing the target. (1) First, the spatial existence of the observed target is represented by a probability density function, and the times of the probability densities of both the observed target and the already-detected detected target are matched by using transition probabilities. In parallel with the step of obtaining the distance between the reference positions of the function, (2) the overall classification of the target to be represented is represented by a hierarchical attribute tree, and the target is identified from each attribute data of the attribute tree from the observed attribute data of the target. Probability of belonging to node (affiliation probability)
Estimating the affiliation probability and comparing the affiliation probability of the already detected target with the affiliation probability of the detected target to obtain the similarity of the target. (3) Next, the distance between the reference positions of the probability density function and the similarity Determining the identity of the target based on

[0029]

[0030]

In the same target determination apparatus according to the first aspect of the present invention, the spatial existence of the observed target is determined.
Is expressed as a probability density function, and the observed target and the already observed
Based on the correlation between the probability density functions of both detected targets
To determine the identity of the target
Target location information is ambiguous due to errors in the target observation system
Can be used to determine the identity of goals.
You. Furthermore, the times of the probability density functions of both the observed target and the already detected detected target are matched using the transition probability,
By configuring to determine the identity of the target based on the correlation between the two probability density functions, even if the target observation data is not time-series or the observation data is not regular, the target identity Can be determined.

In the same target determination apparatus according to the second aspect of the present invention, the movement data and the attribute data of the observed target are also taken in, the movement data of the observation target is estimated from the attribute data, and the observed target is compared with the target already observed. The probability density functions of both the detected targets that have been observed are adjusted using the transition probability,
By configuring so as to determine the identity of the target based on the correlation between the two probability density functions, the movement data is not necessarily obtained as the observation data of the target, and when the attribute data is obtained, the identity of the target is determined. Gender can be determined.

Further, in the same target determination apparatus according to the third aspect of the present invention, the probability that the target belongs to each node of the attribute tree (belonging probability) is estimated from the attribute data of the observed target, and the observed target is determined. And, based on the similarity of the affiliation probabilities of both the detected targets that have already been observed, the identity of the target is determined, so that the movement data cannot be obtained as the target observation data, and only the attribute data is If so, the identity of the goals can be determined.

Further, in the same target determination apparatus according to the fourth aspect of the present invention, attribute data is estimated from the movement data of the observed target, and the probability that the target belongs to each node of the attribute tree (affiliation probability) is estimated. , Based on the similarity of the belonging probability of both the previously observed target and the already-detected detected target, by determining the identity of the target, only movement data is obtained as target observation data, If no attribute data is available, the identity of the target can be determined from the attribute tree.

[0034]

In the same target determination method according to the fifth aspect of the present invention, the probability that the target belongs to each node of the attribute tree (affiliation probability) is estimated from the attribute data of the observed target. By comparing the affiliation probabilities of both the target and the already-detected detected target to determine the similarity of the target, and then configuring the identity determination step of determining identity based on the similarity, When the movement data is not obtained as the target observation data and only the attribute data is obtained, the identity of the target can be determined from the attribute tree.

In the same target determination method of the invention according to the sixth to eleventh aspects, the target position information includes ambiguity due to an error in the target observation system, or the target observation data is not time-series. In the case, when the observation data is not regular, or when the movement data is not obtained as the target observation data and only the attribute data is obtained, the identity of the target can be determined.

[0037]

【Example】

Embodiment 1 FIG. Hereinafter, a first embodiment of the present invention will be described. As an example, when the average position and the variance of the target (hereinafter, referred to as an incoming target) captured by a remote sensor and the variance thereof are transmitted via a satellite line, the target (hereinafter, the target) which has been recognized by the own system. , A detected target) will be described. FIG. 1 is a configuration block diagram showing a first embodiment of the present invention. In the figure, reference numeral 1 denotes a probability density function expressing means for fetching motion data of an incoming target and expressing the motion data by a probability density function, and 2 retrieves a probability density function of a detected target from the target management database 4 and obtains a probability density of the incoming target. Correlation coefficient calculating means 3 for calculating a correlation coefficient with the function, the judgment of the identity of the incoming target and the detected target is made from the result of the correlation coefficient calculating means 2, and the judgment result is added to the target management. Identity determination means stored in the database,
Reference numeral 4 denotes a target management database that records a probability density function of an incoming target to which the identity determination result of the identity determination unit 3 has been added.

The above-mentioned probability density function expressing means 1 will be described. Position coordinates (x, y) at target observation time t
And Since the observation data contains an ambiguous aspect due to the accuracy of the observation system, the given target position (x, y) is an average position vector at that time and is actually centered on this. Is spread by a covariance matrix D shown in the following equation.

[0039]

(Equation 1)

Now, as shown in FIGS. 2 and 3, the point (x,
When the coordinate axes ξ−η are set with y) as the center, the existence probability is an average position vector (x, y), and when the probability density function p (ξ, η) of the covariance matrix D is given, the position (ξ ,
η) and (ξ + dξ, η + dη) mean that the probability that the target exists in a parallelogram region having diagonal lines is expressed by the following equation.

[0041]

(Equation 2)

That is, the existence probability in the area S is given by the following equation.

[0043]

(Equation 3)

Since the contents of the target observation data sent as a result of various observations are considered to follow the central limit theorem, assuming that the probability density function of the existence of the target is a normal distribution, it can be expressed by the following equation.

[0045]

(Equation 4)

Here,

[0047]

(Equation 5)

Next, the correlation coefficient calculating means 2 will be described. Here, an example will be described in which the dimension of the probability density function of the target represented by Expression (4) is reduced by one, and the motion of the target is assumed to be a linear motion. Assuming that the probability density functions of the incoming target and the detected target are equations (6) and (7), respectively, the correlation coefficient used as an index for determining the same target as the incoming target is calculated using equation (8).

[0049]

(Equation 6)

However,

[0051]

(Equation 7)

Assume that: Equations (6), (7), (8),
From (9)

[0053]

(Equation 8)

Is as follows.

When this value exceeds a certain threshold, the identity determination means 3 determines that the two targets are the same target, gives the result, and stores the result in the target management database 4. However, if the values of the correlation coefficients for a plurality of detected targets satisfy the condition, the detected targets having the maximum value of the correlation coefficient are determined to be the same target.

Next, the operation will be described with reference to FIG. The probability density function expressing means 1 fetches the movement data 101 of the incoming target (201) and expresses the incoming target by a probability density function. The correlation coefficient calculating means 2 of the probability density function fetches information closest to the observation time of the incoming target in the observation history of the detected target Ti recognized by the system (2
02), for the probability density function of the incoming target and the detected target Ti, the correlation coefficient of the equation (10) is calculated (203), and from step 202 to step 203, i is turned by the number of detected targets (204). . The identity determination means 3 determines whether or not the correlation coefficient satisfies the condition (205). If the correlation coefficient satisfies the condition, determines that the detected target is the same as the incoming target, and sets a target management database. Stored in (20
6) If the correlation coefficient does not satisfy the condition, the incoming target is determined to be a new target and stored in the target management database (2).
07). As described above, the target
Even if the location information of a target contains ambiguity, determine the identity of the target
A possible identical target determination device can be obtained.

Embodiment 2 FIG. Hereinafter, Example 2 of the present invention
Will be described. As an example, when the average position of the target (incoming target) captured by the remote sensor and its dispersion, the average speed and the dispersion speed are sent via the satellite line, the target that has been recognized by the local system so far. A case of determining the identity between the (detected target) and the incoming target will be described. FIG. 5 is a configuration block diagram showing a second embodiment of the present invention. In the figure, 1, 2, 3, and 4 are the same as in the first embodiment, and a description thereof will be omitted. 5 is a transition probability that matches the time of the probability density function between the incoming target and the detected target by searching for a target near the observation time of the incoming target in the history of the probability density function of the detected target and calculating the transition probability. It is a calculation means. Also in the present embodiment, similarly to the first embodiment, the existence of the target is determined by the formula (4),
It is represented by a probability density function represented by (5).

Next, the processing of the transition probability calculating means 5 will be described. First, the transition of the existence probability distribution after a certain time has elapsed from the time when the target was observed is determined. It is already known how the probability distribution changes with time. Here, a simple Markov process is considered as an example. That is, as shown in FIG. 6, the point (u,
The transition probability density at which the target in v) is at the point (x, y) at time t is represented by f (x, y, t | u, v, s). If the probability distribution density function at time t is p (x, y, t),
In the simple Markov process, it is expressed by the following equation.

[0059]

(Equation 9)

Here, Ω means the entire space (plane).
These relations are exactly the same for the transition probability density, s
The following expression holds for <τ <t.

[0061]

(Equation 10)

This is a basic equation representing the characterization of the Markov process, and is based on Kolmogorov-Cha.
It is called the pman equation. This transition density function f
If (x, y, t | u, v, s) is obtained, the expression (1)
From 1), the existence distribution probability of the target at an arbitrary time t after the time s can be obtained. Here, as an example, a transition probability density function that satisfies the following Kolmogorov-Feller equation is considered.

[0063]

[Equation 11]

In this equation,

[0065]

(Equation 12)

The quantities of the average velocities Ax and Ay and the dispersion velocities Bx ² , By ² and Bxy are introduced.

Next, Kolmogorov-Felle
Find the solution to the equation for r. Here, the solution of the Kolmogorov-Feller equation is determined for the case where the target motion is considered in the latitude, longitude, and altitude (depth) directions. That is, the number of dimensions so far is reduced by one, and the transition density function at which point u at time s transitions to point x at time t is f
If (x, t | u, s), equations (13) and (14)
Is represented by the following equation.

[0068]

(Equation 13)

Next, a spatially uniform continuous process is considered.
The formulas for defining the average speed and the dispersion speed are as follows:
(23).

[0070]

[Equation 14]

Now, in a spatially uniform continuous process, the following equation can be obtained.

[0072]

(Equation 15)

When this is substituted into equations (22) and (23), the following equation is obtained.

[0074]

(Equation 16)

Here, if x−u = z, the average speed is given by the following equation.

[0076]

[Equation 17]

Similarly, the dispersion speed is expressed by the following equation.

[0078]

(Equation 18)

That is, both the average speed and the dispersion speed are time s.
Is just a function. Therefore, Kolmogorov-F
The equation of eller is the following equations (29) and (30), which are equations given by Bachelier.

[0080]

[Equation 19]

The solution of the above equation is (s, t, u, x) →
It is known that it can be obtained by performing a variable transformation of (s ', t', u ', x'). That is,

[0082]

(Equation 20)

In other words,

[0084]

(Equation 21)

Therefore, by substituting this into Expression (29),

[0086]

(Equation 22)

Similarly,

[0088]

(Equation 23)

By substituting the above into equation (30), the following equation is obtained.

[0090]

(Equation 24)

The transition density function f (x, t | u, s) satisfying the equations (36) and (37) is as follows.

[0092]

(Equation 25)

This is the case of Bachelier, that is, Kolmogorov-Fell in the case of spatial uniformity.
This is a solution to the equation of er. That is, in this case, the transition density function has a normal distribution. From the above equations (31) to (34),

[0094]

(Equation 26)

Equation (39) is arranged as follows.

[0096]

[Equation 27]

As described above, it has been shown that, assuming a spatially uniform system, the solution of the Kolmogorov-Feller equation becomes a normal distribution of equation (42). this,
By substituting into the equation (11), the target existence probability distribution P (x, t) at the time t is obtained.

As described above, the target existence probability distribution density function given by the communication information assumes a normal distribution. That is, the following equation, where the average of the distribution is m _s and the variance is σ _s , is expressed as time s
Given the probability density at

[0099]

[Equation 28]

From the equation (11), the following equation is obtained.

[0101]

(Equation 29)

Also,

[0103]

[Equation 30]

In other words,

[0105]

(Equation 31)

Therefore, equation (44) becomes the following equation.

[0107]

(Equation 32)

As described above, when a target probability distribution density function at a certain time is given by a normal distribution, the result of the transition over time also becomes a normal distribution. That is, when Expression (42) and Expression (47) are compared,

[0109]

[Equation 33]

It can be seen that the distribution has an average and a wide variance. If the target average speed and dispersion speed are assumed to be constant from s to t, the following expression is obtained.

[0111]

(Equation 34)

Next, as shown in FIG. 7, using the transition probabilities described above, so-called time adjustment is performed in which the old probability distribution at the observation time is shifted to the new observation time. That is, when the observation time of the incoming target is different from the observation time of the detected target, the existence probability distribution density function at the oldest observation time is changed using Equation (49) to obtain a probability distribution at the same time.

Next, the correlation coefficient calculating means 2 will be described. As described above, if the probability distribution density function of the target given by the communication information is given by a normal distribution, the probability distribution density function after an elapse of an arbitrary time also becomes a normal distribution. Assuming that the probability distribution densities are equations (6) and (7), a correlation coefficient is calculated using equation (10) as an index for determining the same target as the incoming target.

When the value of equation (10) satisfies the condition, the identity determination means 3 determines that the two are the same target, and assigns the result and stores it in the target management database 4. However, if the correlation coefficients for a plurality of detected targets satisfy the condition, the detected targets having the maximum value of the correlation coefficient are set as the same target.

Next, the operation will be described with reference to FIG. The probability density function expressing means 1 fetches the movement data 101 of the incoming target (201) and expresses the incoming target by a probability density function. The transition probability calculating means 5 fetches the information closest to the observation time of the incoming target from the observation history of the detected target Ti recognized by the system (202), and performs the observation among the incoming target and the detected Ti. The probability density function with the old time is shifted to the new observation time using the transition probability of the equation (49) (209), and the correlation coefficient calculating means 2 for the probability density function is used.
Calculates the correlation coefficient of equation (10) for the probability density function of the incoming target and the detected target Ti matched to the same time (210), and performs i from step 202 to step 210 by the number of detected targets. Turn (204). The identity determination means 3 determines whether the correlation coefficient satisfies the condition (2
05) If the correlation coefficient satisfies the condition, it is determined that the detected target is the same as the incoming target, and is stored in the target management database (206). If the correlation coefficient does not satisfy the condition, the incoming target is determined. Is determined as a new target and stored in the target management database (207).

Embodiment 3 FIG. Hereinafter, Example 3 of the present invention
Will be described. As an example, when the average position of the target (incoming target) captured by the remote sensor and its dispersion, the average speed and the dispersion speed are sent via the satellite line, the target that has been recognized by the local system so far. A case of determining the identity of the (detected target) and the incoming target will be described. FIG. 9 is a configuration block diagram showing a third embodiment of the present invention. In the figure, 1, 3, 4, and 5 are the same as in the second embodiment, and a description thereof will be omitted. 6
Is a distance calculating means for calculating an average position of the probability density function combined by the transition probability calculating means 5. Also in the present embodiment, similarly to the second embodiment, the existence of the target is expressed by the probability density functions represented by Expressions (4) and (5), and the time of the probability distribution is adjusted using Expression (49). afterwards,
The average position distance calculating means 6 calculates the distance between the average position of the incoming target and the average position of the detected target by adjusting the time of the probability distribution according to the equation (48)
Calculate according to

[0117]

(Equation 35)

When the value of equation (50) satisfies the condition, the identity determination means 3 determines that the two are the same target, and assigns this result to the target management database 4.
To be stored. However, if the values of the correlation coefficients for a plurality of detected targets satisfy the condition, the detected targets having the maximum value of the correlation coefficient are determined to be the same target.

Next, the operation will be described with reference to FIG. The probability density function expressing means 1 fetches the movement data 101 of the incoming target (201) and expresses the incoming target by a probability density function. The transition probability calculating means 5 fetches the information closest to the observation time of the incoming target from the observation history of the detected target Ti recognized by the system (202), and performs the observation among the incoming target and the detected Ti. The probability density function with the old time is shifted to the new observation time using the transition probability of the equation (49) (209). The distance calculation means 6 calculates the correlation coefficient of the equation (50) for the probability density function of the incoming target and the detected target Ti adjusted to the same time (211), and performs steps 202 to 211 of the detected target. The i is turned by the number (204). The identity determination means 3 determines whether or not the correlation coefficient satisfies the condition (205). If the correlation coefficient satisfies the condition, determines that the detected target is the same as the incoming target, and sets a target management database. Stored in (20
6) If the correlation coefficient does not satisfy the condition, the incoming target is determined to be a new target and stored in the target management database (2).
07).

Embodiment 4 FIG. Hereinafter, Example 4 of the present invention
Will be described. FIG. 11 is a configuration block diagram showing a fourth embodiment of the present invention. In the figure, 1, 2, 3, 4, 5
Is the same as in the second embodiment, and a description thereof will be omitted. Reference numeral 7 denotes exercise data estimating means for retrieving a target speed corresponding to the attribute information from the target performance / specification database 8, and reference numeral 8 denotes a target performance / specification database for recording the target speed corresponding to the attribute data whose input is assumed. Is shown.

Embodiment 4 corresponds to the case where the average speed cannot be observed as described in Embodiment 2.
The average speed used for calculating the transition probability is estimated and supplemented by using attribute information on the target performance, specifications, and properties. The estimation is performed by searching the target performance / specification database 8 in which the attribute data and the speed corresponding thereto are stored in advance. Based on the result, time alignment of the probability distribution densities of the incoming target and the detected target is performed using Expression (49). Other processes are the same as in the second embodiment.

Next, the operation will be described with reference to FIG. The motion data estimating means 7 captures the motion and attribute data of the incoming target (213), estimates the average speed of the incoming target from the attribute data (214), and the probability density function expressing means 1 calculates the incoming target by a probability density function. Expressing, the transition probability calculating means 5 fetches the information closest to the observation time of the incoming target in the observation history of the detected target Ti recognized by the system (202), and calculates the information of the incoming target and the detected target Ti. Of these, the old probability density function at the observation time is shifted to the new observation time using the transition probability of the equation (49) (209), and the correlation coefficient calculation means 2 of the probability density function sets the input target and the For the probability density function of the detected target Ti, the correlation coefficient of the equation (10) is calculated (210), and step 20 is performed.
From step 2 to step 210, i is turned by the number of detected targets (204). The identity determination means 3 determines whether or not the correlation coefficient satisfies the condition (205). If the correlation coefficient satisfies the condition, determines that the detected target is the same as the incoming target, and sets a target management database. (206), and when the correlation coefficient does not satisfy the condition, the incoming target is determined to be a new target and stored in the target management database (207).

Embodiment 5 FIG. Hereinafter, Example 5 of the present invention
Will be described. FIG. 13 is a configuration block diagram showing a fifth embodiment of the present invention. In the figure, 1, 3, 4, 5, 6
Is the same as in the third embodiment, and a description thereof will be omitted. Reference numeral 7 denotes exercise data estimating means for retrieving a target speed corresponding to the attribute information from the target performance / specification database 8; Is shown.

The fifth embodiment corresponds to the case where the average speed cannot be obtained as the movement data of the incoming target as described in the third embodiment, and the attribute information relating to the performance, specifications and properties of the target is stored. It is used to estimate and compensate for the average speed used for calculating the transition probability. The estimation is performed by searching the target performance / specification database 8 in which the attribute data and the speed corresponding thereto are stored in advance. Based on the result, time alignment of the probability distribution densities of the incoming target and the detected target is performed using Expression (49). Other processes are the same as in the third embodiment.

Next, the operation will be described with reference to FIG. The motion data estimating means 7 captures the motion and attribute data of the incoming target (213), estimates the average speed of the incoming target from the attribute data (214), and the probability density function expressing means 1 calculates the incoming target by a probability density function. The transition probability calculating means 5 expresses the detected target Ti recognized by the system in
The information closest to the observation time of the incoming target in the observation history of (2) is fetched (202), and the probability density function with the oldest observation time among the incoming target and the detected target Ti is calculated using the transition probability of the equation (49). The time is changed to a new observation time (209). The distance calculation means 6 calculates the correlation coefficient of the equation (50) for the probability density function of the incoming target and the detected target Ti set at the same time (211), and proceeds from step 202 to step 2
Turn i by the number of detected targets up to 11 (204). The identity determination means 3 determines whether or not the correlation coefficient satisfies the condition (205). If the correlation coefficient satisfies the condition, determines that the detected target is the same as the incoming target, and sets a target management database. (206), and when the correlation coefficient does not satisfy the condition, the incoming target is determined to be a new target and stored in the target management database (207).

Embodiment 6 FIG. Hereinafter, Example 6 of the present invention.
Will be described. As an example, when only the attribute information of the target (incoming target) captured by the remote sensor is sent via the satellite line, the target (detected target) and the incoming target that have been recognized by the own system so far. A case in which the sameness is determined will be described. FIG. 15 is a configuration block diagram showing a sixth embodiment of the present invention. In the figure, reference numeral 4 denotes a target management database for recording the affiliation probability of the incoming target to which the sameness determination result has been given; 10 fetches the attribute data of the incoming target; A membership probability estimating means for estimating the membership probability of the target type by using the attribute tree construction means for calculating the membership probability of the entire attribute tree from the membership probabilities of the target type estimated by the membership probability estimating means 10; Similarity calculating means for calculating a similarity from the belonging probability of the attribute tree of the incoming target obtained by the constructing means 11 and the belonging probability of the attribute tree of the detected target searched from the target management database 4,
3 determines the similarity between the incoming target and the detected target based on the result of the similarity calculating means 12. If the target is not the same, the incoming target is set as a new target and the probability of belonging of the attribute tree is stored in the target management database 4. Identity determination means to be recorded,
13 is a belonging probability update unit that updates the belonging probability of the detected target when the incoming target and the detected target Ti are determined to be the same as a result of the similarity calculating unit 12, and 14 is necessary for estimating the target type. 4 shows an estimated information database for recording information.

In this embodiment, the target (target) to be handled is shown in FIG.
As shown in FIG. 6, it is expressed by a hierarchically divided hierarchy (attribute tree), and the probability that the target belongs to each node of the attribute tree (affiliation probability) is estimated from the attribute information of the target. Here, the attribute tree may have one layer. The belonging probability estimated from the attribute information is not calculated for all the nodes of the attribute tree, but a convenient group of layers is determined in advance according to the type of the attribute information, and the probability of belonging to each node is calculated. The hierarchical group is a set of nodes combined so as to cover all the goals without overlapping. That is, the events to which the target belongs to each node of the hierarchical group must be mutually exclusive and exhaust all the objects to be handled. Next, a method of estimating the target type performed by the belonging probability estimating means 10 will be described. Here, the belonging probabilities of the hierarchical group H [h1, h2,..., Hn] composed of n types of target types are obtained. The M types of attribute data are represented by the following equations.

[0128]

[Equation 36]

Here,

[0130]

(37)

It is assumed that: However, if the attribute value is a ^j1 , a ^j2 , ...
, A ^jlj must be mutually exclusive and exhaustive. For example, if the ATT ¹ and attribute data relating to the weight of the moving body,

[0132]

(38)

And so on. In addition, the attribute data ATT ^1, ATT ² at the observation time t, ···, the observed value of ATT ^M to the following equation.

[0134]

[Equation 39]

For example, attribute data ATT at time t
^When the observed value of ^j is a ^j2 , the following expression is obtained.

[0136]

(Equation 40)

When the observed value ATT _{m of the} attribute data is given, the target type is hs (s = 1, 2,..., N).
Is calculated using the following equation.

[0138]

[Equation 41]

Here, P {hs | ATT ^j _m (0)} is an initial value, and the values of all target types are usually the same.
Further, the probability P {ATT ^j _m | hs} that the attribute data becomes ATT ^j _m when the target type is assumed to be hs can be obtained in advance. Therefore, these are recorded in the estimation information database 14 in advance and searched as needed.

Next, the attribute tree construction means will be described. When the probability of belonging to a certain hierarchical group is determined using Expression (57), the probability of belonging to the parent node is determined. In the attribute tree shown in FIG.

[0141]

(Equation 42)

Assuming that P {Bj} is known, the parent probability P {Aj} is obtained from the following equation.

[0143]

[Equation 43]

Similarly, it is possible to obtain the belonging probability of a higher hierarchy. Conversely, the affiliation probabilities of the child nodes of the hierarchical group for which the affiliation probabilities have been calculated are maintained in the form of a probability when the parent's affiliation probability is assumed to be 1, ie, a conditional probability, without updating. As described above, when the belonging probabilities of a certain hierarchical group combined so as to cover all the objects to be handled without overlapping are obtained, the belonging probabilities of all the nodes of the attribute tree are obtained.

Next, the similarity calculating means 12 will be described. Here, the correlation between the incoming target and the detected target is examined.
Items below the root item

[0146]

[Equation 44]

Of the incoming target is

[0148]

[Equation 45]

If the probability of belonging of the detected target is

[0150]

[Equation 46]

Then, for example,

[0152]

[Equation 47]

Let the degree of similarity be the level A. γ _A takes a value from 0 to 1. If the similarity is lower than a predetermined threshold, it is determined that the target is not the same target. If the similarity is higher, the process goes to the next layer, and a probability distribution in that layer is output. That is, the node Ai (i =
1, 2,. . . , N) can be represented by the following equation:

[0154]

[Equation 48]

The conditional belonging probability of this node (node Ai
Is determined, the conditional probability that the target is B _ij ) is p ^c _ij
Then, the belonging probability of the incoming target of the node B _ij is given by the following equation.

[0156]

[Equation 49]

Similarly, the belonging probability of the detected target recognized by the system is as follows.

[0158]

[Equation 50]

The similarity between these distributions is given by the following equation.

[0160]

(Equation 51)

The threshold value operation performed on the A level is performed on the similarity. The similarities up to the necessary layers are obtained, and the similarities of all the layers are obtained. This is performed for all the detected targets.

The identity determination means 3 determines that the two targets are the same target when reaching the deepest hierarchy by this threshold value operation. If no correlation is obtained with any of the detected targets, the identity determination means 3 determines that the two targets are the new target. It is stored in the target management database 4. However, when a plurality of detected targets reach the same layer, the detected targets satisfying the similarity condition are determined to be the same target.

Next, the belonging probability updating means 13 will be described. When the same target is confirmed among the detected targets as a result of the identity determination means 3, the belonging probability of the detected target is updated using the following equation.

[0164]

(Equation 52)

Then, similarly to the attribute tree construction means 11, the belonging probabilities of all nodes of the attribute tree are obtained and stored in the target management database 4.

Next, the operation will be described with reference to FIG. The affiliation probability estimating means 10 fetches the attribute data of the incoming target (215), retrieves the target type estimation information from the estimation information database 14, and estimates the affiliation probability of the target type (hierarchy group) using equation (57). (216).
The attribute tree construction means 11 calculates, in the attribute tree, the belonging probabilities of a layer higher than the group of layers for which the belonging probabilities have been calculated by the formula (6).
0) (217). Next, the similarity calculation means 12 reads the detected target Ti from the target management database 4.
(218), and the similarity of the incoming target and the detected target on the attribute tree is calculated using Expression (64) or Expression (68) (219), and Step 21 is performed.
From i to step 219, i is turned by the number of detected targets (204). The identity determination means 3 determines whether the similarity satisfies the condition (220), and determines that the detected target satisfying the condition is the same as the incoming target (221). The affiliation probability updating means 13 updates the affiliation probability of the detected target determined to be the same target using Expression (69) (222),
The probability of all members belonging to the attribute tree of the detected target determined to be the same target is updated using equation (60) and stored in the target management database (223). If the similarity does not satisfy the condition in step 220, The incoming target is determined to be a new target and stored in the target management database 4 (207).

Embodiment 7 FIG. Hereinafter, Example 7 of the present invention.
Will be described. FIG. 18 is a configuration block diagram showing a seventh embodiment of the present invention. In the figure, 3, 4, 10, 1
1, 12, 13, and 14 are the same as in the sixth embodiment, and a description thereof will be omitted. Reference numeral 8 denotes a target performance / specification database for recording a target speed corresponding to attribute data assumed to be input, and reference numeral 16 denotes an attribute data estimating means for searching the target performance / specification database 8 for a target attribute corresponding to exercise data. Is shown.

The seventh embodiment corresponds to the case where attribute data cannot be observed, as described in the sixth embodiment. The attribute data is estimated by using the movement data such as the speed and acceleration of the observed target. I do. That is, the attribute data estimating means 16 estimates the attribute data by searching the target performance / specification database 8 in which the relationship between the exercise data and the attribute data restricted thereby is stored in advance. Subsequent processing is the same as in the sixth embodiment.

Next, the operation will be described with reference to FIG. The attribute data estimating means 16 takes in the exercise data 101 of the incoming target (201), and estimates the attribute data of the incoming target from the exercise data by searching the target performance / specification database 8 (224). The affiliation probability estimating means 10 retrieves the target type estimation information from the estimation information database 14 and calculates the affiliation probability of the target type (hierarchy group),
Estimation is performed using Expression (57) (216). The attribute tree construction means 11 calculates the belonging probability of a layer higher than the layer group for which the belonging probability has been calculated in the attribute tree using Expression (60) (217). The similarity calculating means 12 searches the target management database 4 for the probability of belonging of the attribute tree of the detected target Ti (218), and calculates the similarity between the incoming target and the detected target on the attribute tree by the formula (64) or the formula (64). 68) (219), and from step 218 to step 21
Turn i by the number of detected targets up to 9 (204). The identity determination means 3 determines whether the similarity satisfies the condition (220), and determines that the detected target satisfying the condition is the same as the incoming target (221). Membership probability updating means 13
Updates the belonging probability of the detected target determined to be the same target using Expression (69) (222), and calculates the total belonging probability of the attribute tree of the detected target determined to be the same target by Expression (60).
And stored in the target management database (22).
3) If the similarity does not satisfy the condition in step 220, the incoming target is determined to be a new target and stored in the target management database 4 (207).

Embodiment 8 FIG. Hereinafter, Example 8 of the present invention.
Will be described. As an example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, and the average speed and dispersion speed are sent via a satellite line, the own system has recognized it so far. A case where the identity of a target (detected target) and an incoming target is determined will be described. FIG. 20 is a schematic flowchart showing Embodiment 8 of the present invention.
FIG. 21 is an example of a detailed flowchart. In the figure, (1) First, a correlation coefficient calculation step 50 of the probability density function
Has the following steps 201 to 210. The movement data of the incoming target is fetched (201), and the incoming target is expressed by a probability density function. The information closest to the observation time of the incoming target in the observation history of the detected target is fetched (202), and the old probability density function of the observation time of the incoming target and the detected target is used to calculate the transition probability of equation (49). (209), and the probability density function of the incoming target and the detected target adjusted to the same time is calculated by the equation (1).
0) is calculated (210), and steps 202 to 210 are repeated by the number of detected targets (20).
4). (2) Next, the identity determination step 55 using the correlation coefficient of the probability density function determines whether the correlation coefficient satisfies the condition (205). If the condition is not satisfied, the incoming target is set to the new target. And updates the observation history information of the detected target (207). On the other hand, if the correlation coefficient satisfies the condition, it is determined that the detected target is the same as the incoming target, and the detected target history information is updated (206).

Embodiment 9 FIG. Hereinafter, a ninth embodiment of the present invention will be described.
Will be described. As an example, when only the attribute information of the target (incoming target) captured by the remote sensor is sent via the satellite line, the target (detected target) and the incoming target that have been recognized by the own system so far. A case in which the sameness is determined will be described. FIG. 22 is a schematic flowchart showing Embodiment 9 of the present invention, and FIG. 23 is an example of a detailed flowchart. In the figure, (1) First, similarity calculation step 52 on the attribute tree
Has the following steps 215 to 219. The attribute data of the incoming target is taken in (215), and the probability of belonging of the target type is estimated using equation (56) (21).
6) In the attribute tree, the belonging probability of the layer above the layer group for which the belonging probability has been calculated is calculated using Expression (59) (217), and the belonging probability of the attribute tree of the detected target is taken in (218). The similarity on the attribute tree between the incoming target and the detected target is calculated using Expression (63) or Expression (67) (219), and Steps 218 to 219 are repeated by the number of detected targets (204). (2) Next, identity determination step 53 using similarity
Determines whether the similarity satisfies the condition (22
0) If the condition is not satisfied, the incoming target is determined to be a new target, and the observation history information of the detected target is updated (20).
7). On the other hand, when the similarity satisfies the condition, the detected target is determined to be the same as the incoming target (221), and the belonging probability of the detected target determined to be the same target is updated using Expression (68). (222) The probability of all members belonging to the attribute tree of the detected target determined to be the same target is updated using equation (59), and the detected target history information is updated (223).

Embodiment 10 FIG. Hereinafter, a tenth embodiment of the present invention will be described. For example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, the average speed and dispersion speed, and the attribute information of the incoming target are transmitted via a satellite line, In the following, a case will be described in which the identity of a target (detected target) recognized by the own system and an incoming target is determined. FIG. 24 is a schematic flowchart showing Embodiment 10 of the present invention, and FIG. 25 is an example of a detailed flowchart. In the figure, (1) First, a correlation coefficient calculation step 50 of the probability density function
Has the following steps 201 to 210. The movement data of the incoming target is fetched (201), and the incoming target is expressed by a probability density function. The information closest to the observation time of the incoming target in the observation history of the detected target is fetched (202), and the old probability density function of the observation time of the incoming target and the detected target is used to calculate the transition probability of equation (49). Is used to shift to the new observation time (209), and the correlation coefficient of equation (10) is calculated for the probability density function of the incoming target and the detected target matched to the same time (210), and step 20 is performed.
Steps 2 to 210 are repeated by the number of detected targets (20
4). (2) Next, the same candidate selection step 51 using the correlation coefficient of the probability density function determines whether or not the correlation coefficient satisfies the condition (205). At 207, the incoming target is determined to be a new target,
Update the observation history information of the detected target (207), while
If the correlation coefficient satisfies the condition in step 205, the process proceeds to the following similarity calculation step 52 on the attribute tree. (3) Next, a similarity calculation step 52 on the attribute tree
Has the following steps 215 to 219. The attribute data of the incoming target is fetched (215), and the probability of belonging of the target type is estimated using equation (57) (21).
6) In the attribute tree, the belonging probability of a layer higher than the layer group for which the belonging probability has been calculated is calculated using Expression (60) (217), and in Step 205, the correlation coefficient of the detected target satisfying the condition is calculated. The belonging probability of the attribute tree is searched (22
6), calculate the similarity between the incoming target and the detected target on the attribute tree by using Expression (64) or Expression (68) (21).
9) Repeat steps 226 to 219 for the number of targets whose correlation coefficient satisfies the condition (225). (3) Next, identity determination step 53 using similarity
Determines whether the similarity satisfies the condition (22
0) If the condition is not satisfied, the incoming target is determined as a new target, and the observation history information of the detected target is updated (229).
On the other hand, if the similarity satisfies the condition in step 220, it is determined that the detected target is the same as the incoming target (2).
21), the belonging probability of the detected target determined to be the same target is updated using Expression (69) (222), and the attribute tree total belonging probability of the detected target determined to be the same target is calculated by Expression (60).
(227).

Embodiment 11 FIG. Hereinafter, an eleventh embodiment of the present invention will be described. For example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, the average speed and dispersion speed, and the attribute information of the incoming target are transmitted via a satellite line, In the following, a case will be described in which the identity of a target (detected target) recognized by the own system and an incoming target is determined. FIG. 26 is a schematic flowchart showing Embodiment 11 of the present invention, and FIG. 27 is an example of a detailed flowchart. In the figure, (1) First, similarity calculation step 52 on the attribute tree
Has the following steps 215 to 219. The attribute data of the incoming target is fetched (215), and the probability of belonging of the target type is estimated using equation (57) (21).
6) In the attribute tree, the belonging probability of the layer above the layer group for which the belonging probability has been calculated is calculated using Expression (60) (217), and the belonging probability of the attribute tree of the detected target is taken in (218). The similarity between the incoming target and the detected target on the attribute tree is calculated using Expression (64) or Expression (68) (219), and Steps 218 to 219 are repeated by the number of detected targets (204). (2) Next, the same candidate selection step 5 using similarity
In step 220, it is determined whether or not the similarity satisfies the condition in step 220. If the condition is not satisfied, the incoming target is determined to be a new target, and the observation history information of the detected target is updated (20).
7). On the other hand, if the similarity satisfies the condition in step 220, the process proceeds to a probability density function correlation coefficient calculation step 50. (3) Next, a correlation coefficient calculation step 5 of the probability density function
0 has the following steps 201 to 210. Intake data of incoming target is taken in (201),
Express the incoming target as a probability density function. In step 220, in the observation history of the detected target whose similarity satisfies the condition,
Fetch information closest to the observation time of the incoming target (23
0), of the incoming target and the detected target, the old probability density function of the observation time is shifted to the new observation time using the transition probability of Equation (49) (209), and the incoming target and the detected target are adjusted to the same time. For the target probability density function, the correlation coefficient of equation (10) is calculated (210), and steps 230 to 210 are calculated.
Is repeated by the number of detected targets whose similarity satisfies the condition (208). (4) Next, the identity determination step 55 using the correlation coefficient of the probability density function determines whether the correlation coefficient satisfies the condition (205). If the condition is not satisfied, the incoming target is set to the new target. The determination is made, and the observation history information of the detected target is updated (229). On the other hand, when the correlation coefficient satisfies the condition in step 205, it is determined that the detected target is the same as the incoming target (206), and the belonging probability of the detected target determined to be the same target is calculated by the equation (69). Update using (22
2) The attribute tree total belonging probability of the detected target determined to be the same target is updated using Expression (60) (227).

Embodiment 12 FIG. Hereinafter, a twelfth embodiment of the present invention will be described. For example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, the average speed and dispersion speed, and the attribute information of the incoming target are transmitted via a satellite line, In the following, a case will be described in which the identity of a target (detected target) recognized by the own system and an incoming target is determined. FIG. 28 is a schematic flowchart showing Embodiment 12 of the present invention, and FIG. 29 is an example of a detailed flowchart. As shown in the figure, the probability density function correlation coefficient calculating step 50 and the similarity calculating step 52 on the attribute tree have steps 201 to 210 and 215 to 219, respectively. (1) Probability density function correlation coefficient calculation step 50 fetches the movement data of the incoming target (201), and expresses the incoming target by a probability density function. The information closest to the observation time of the incoming target in the observation history of the detected target is captured (2
02), of the incoming target and the detected target, the old probability density function of the observation time is shifted to the new observation time using the transition probability of Expression (49) (209), and the incoming target and the detected target are adjusted to the same time. The correlation coefficient of Expression (10) is calculated for the target probability density function (210), and steps 202 to 210 are performed.
Is repeated by the number of detected targets (204). (2) The similarity calculation step 52 on the attribute tree fetches the attribute data of the incoming target (215), estimates the membership probability of the target type using equation (57) (216), and assigns the membership probability in the attribute tree. The probability of belonging to a layer above the layer group for which the probability has been calculated is calculated using equation (60) (217),
The probability of belonging to the attribute tree of the detected target is taken in (21
8) The similarity between the incoming target and the detected target on the attribute tree is calculated using Expression (64) or Expression (68) (21).
9) Repeat steps 218 to 219 by the number of detected targets (204). (3) Next, the identity determination step 56 using the correlation coefficient of the probability density function and the degree of similarity on the attribute tree is performed in the same manner as described above. It is determined whether the similarity calculated in step 219 matches a detected target satisfying the condition (2
32) If not coincident, the incoming target is determined to be a new target, and the observation history information of the detected target is updated (229).
On the other hand, if they match, it is determined that the matched detected target is the same as the incoming target (233), and the probability of belonging of the detected target determined to be the same target is updated using equation (69) (222). ), Update the attribute tree total belonging probability of the detected target determined to be the same target by using equation (60) (22)
7).

Embodiment 13 FIG. Hereinafter, a thirteenth embodiment of the present invention will be described. For example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, the average speed and dispersion speed, and the attribute information of the incoming target are transmitted via a satellite line, In the following, a case will be described in which the identity of a target (detected target) recognized by the own system and an incoming target is determined. FIG. 30 is a schematic flowchart showing Embodiment 13 of the present invention, and FIG. 31 is an example of a detailed flowchart. In the figure, (1) First, the reference distance calculation step 57 of the probability density function has the following steps 201 to 211. The movement data of the incoming target is fetched (201), and the incoming target is expressed by a probability density function. The information closest to the observation time of the incoming target in the observation history of the detected target is fetched (202), and the old probability density function of the observation time of the incoming target and the detected target is used to calculate the transition probability of equation (49). (209), and the probability density function of the incoming target and the detected target adjusted to the same time is given by the equation (5).
0) is calculated (211), and steps 202 to 211 are repeated by the number of detected targets (20).
4). (2) Next, the same candidate selecting step 58 using the distance between the reference positions of the probability density function determines whether the correlation coefficient satisfies the condition (205), and has two branches according to the result. If the correlation coefficient does not satisfy the condition, the incoming target is determined to be a new target, and the observation history information of the detected target is updated (207). On the other hand, when the correlation coefficient satisfies the condition, the process proceeds to the similarity calculation step 52 on the attribute tree. (3) The similarity calculation step 52 on the attribute tree has the following steps 215 to 219. The attribute data of the incoming target is fetched (215), the belonging probability of the target type is estimated using Expression (57) (216), and the belonging probability of the layer above the hierarchical group whose belonging probability is calculated in the attribute tree is calculated. It is calculated using equation (60) (21
7) In step 205, a search is made for the belonging probability of the attribute tree of the detected target whose correlation coefficient satisfies the condition (22).
6), calculating the similarity between the incoming target and the detected target on the attribute tree by using Expression (64) or Expression (68) (21)
9) Repeat steps 226 to 219 for the number of targets whose correlation coefficient satisfies the condition (225). (4) Next, identity determination step 53 using similarity
Determines whether the similarity satisfies the condition (22
0), when the similarity does not satisfy the condition, the incoming target is determined to be a new target, and the observation history information of the detected target is updated (2).
29). On the other hand, if the similarity satisfies the condition, the detected target is determined to be the same as the incoming target (221), and the belonging probability of the detected target determined to be the same target is updated using equation (69). (222) The attribute tree total belonging probability of the detected target determined to be the same target is updated using Expression (60) (227).

Embodiment 14 FIG. Hereinafter, a fourteenth embodiment of the present invention will be described. For example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, the average speed and dispersion speed, and the attribute information of the incoming target are transmitted via a satellite line, In the following, a case will be described in which the identity of a target (detected target) recognized by the own system and an incoming target is determined. FIG. 32 is a schematic flowchart showing Embodiment 14 of the present invention, and FIG. 33 is an example of a detailed flowchart. In the figure, (1) First, similarity calculation step 52 on the attribute tree
Has the following steps 215 to 219. The attribute data of the incoming target is fetched (215), and the probability of belonging of the target type is estimated using equation (57) (21).
6) In the attribute tree, the belonging probability of the layer above the layer group for which the belonging probability has been calculated is calculated using Expression (60) (217), and the belonging probability of the attribute tree of the detected target is taken in (218). The similarity between the incoming target and the detected target on the attribute tree is calculated using Expression (64) or Expression (68) (219), and Steps 218 to 219 are repeated by the number of detected targets (204). (2) Next, the same candidate selection step 5 using similarity
4 judges whether the similarity satisfies the condition (22)
0) If the similarity does not satisfy the condition, the incoming target is determined to be a new target, and the observation history information of the detected target is updated (207). On the other hand, when the similarity satisfies the condition, the process proceeds to the reference position distance calculation step 57 of the probability density function. (3) Step 57 for calculating distance between reference positions of probability density function
Has the following steps 201 to 211. The movement data of the incoming target is fetched (201), and the incoming target is expressed by a probability density function. In step 220, information closest to the observation time of the incoming target in the observation history of the detected target whose similarity satisfies the condition is fetched (230),
Of the incoming target and the detected target, the old probability density function of the observation time is shifted to the new observation time using the transition probability of equation (49) (209), and the probability of the incoming target and the detected target matched to the same time is calculated. For the density function, the correlation coefficient of equation (50) is calculated (211), and steps 230 to 211 are repeated for the number of detected targets whose similarity satisfies the condition (2).
08). (4) Next, in the identity determination step 59 using the distance between the reference positions of the probability density function, it is determined whether or not the correlation coefficient satisfies a condition (205). The target is determined as a new target, and the observation history information of the detected target is updated (229). On the other hand, when the correlation coefficient satisfies the condition, the detected target is determined to be the same as the incoming target (206), and the belonging probability of the detected target determined to be the same target is updated using equation (69). Then, the attribute tree total belonging probabilities of the detected targets determined to be the same target are updated using Expression (60) (227).

Embodiment 15 FIG. Hereinafter, a fifteenth embodiment of the present invention will be described. For example, when the average position and dispersion of the target (incoming target) captured by the remote sensor, the average speed and dispersion speed, and the attribute information of the incoming target are transmitted via a satellite line, In the following, a case will be described in which the identity of a target (detected target) recognized by the own system and an incoming target is determined. FIG. 34 is a schematic flowchart showing Embodiment 15 of the present invention, and FIG. 35 is an example of a detailed flowchart. As shown in the figure, the step 57 for calculating the distance between the reference positions of the probability density function and the step 52 for calculating the similarity on the attribute tree have steps 201 to 211 and 215 to 219, respectively. (1) Step 57 for calculating distance between reference positions of probability density function
Fetches the movement data of the incoming target (201), and expresses the incoming target by a probability density function. The information closest to the observation time of the incoming target in the observation history of the detected target is fetched (202), and the old probability density function of the observation time of the incoming target and the detected target is used to calculate the transition probability of equation (49). (209), and the probability density function of the incoming target and the detected target adjusted to the same time is given by the equation (5).
0) is calculated (211), and steps 202 to 211 are repeated by the number of detected targets (20).
4). (2) The similarity calculation step 52 on the attribute tree fetches the attribute data of the incoming target (215), estimates the membership probability of the target type using equation (57) (216), and assigns the membership probability in the attribute tree. The probability of belonging to a layer above the layer group for which the probability has been calculated is calculated using equation (60) (217),
The probability of belonging to the attribute tree of the detected target is taken in (21
8) Calculate the similarity between the incoming target and the detected target on the attribute tree using Expression (64) or Expression (68) (21).
9) Repeat steps 218 to 219 by the number of detected targets (204). (3) Next, the identity determination step 60 using the distance between the reference positions of the probability density function and the similarity on the attribute tree is performed in the same manner. It is determined whether the similarity calculated in step 219 matches the detected target satisfying the condition (231). If not, the incoming target is determined as the new target, and the observation history information of the detected target is determined. Update (22
9). On the other hand, if they match, it is determined that the matched detected target is the same as the incoming target (233), and the probability of belonging of the detected target determined to be the same target is updated using equation (69) (222). ), The attribute tree total belonging probabilities of the detected targets determined to be the same target are updated using equation (60) (2)
27).

[0178]

[0179]

As described above , according to the first aspect of the present invention, the spatial existence of an observed target is represented by a probability density function.
Both the observed target and the detected target already observed
Based on the probability density function correlation, the observed target and
Configured to determine the identity of already detected targets
Of the target position information due to errors in the target observation system.
The same goal that can determine the identity of the goal, even if it contains ambiguity
A determination device can be obtained. In addition, the observed target,
Times of both probability density functions of the already detected detected targets are adjusted using the transition probability, and the identity of the observed target and the already detected detected target is determined based on the correlation between the two probability density functions. Thus, the same target determination device that can determine the sameness of the target can be obtained even when the target observation data is not time-series or when the observation data is not regular.

According to the second aspect of the present invention, the movement data and the attribute data of the observed target are taken in, the movement data of the observation target are estimated from the attribute data, and the observed target and the already observed target are estimated. The time of both probability density functions of the detection target is adjusted using the transition probability, and the identity of the target is determined based on the correlation between the two probability density functions. Is not necessarily obtained, and when attribute data is obtained, it is possible to obtain an identical target determination device capable of determining the identity of a target.

According to the third aspect of the present invention, the probability that the target belongs to each node of the attribute tree (affiliation probability) is estimated from the attribute data of the observed target, Based on the similarity of the belonging probabilities of both the detected targets that have been observed, it is configured to determine the identity of the targets, and when the movement data is not obtained as the target observation data and only the attribute data is obtained, Thus, it is possible to obtain an identical target determination device capable of determining the identity of a target from an attribute tree.

According to the fourth aspect of the present invention, attribute data is estimated from the movement data of the observed target, and the probability that the target belongs to each node of the attribute tree (affiliation probability) is estimated. Based on the similarity of the affiliation probabilities of both the target and the detected target that has already been detected, it is configured to determine the identity of the target, and only the target movement data is obtained as the target observation data, and the attribute data Is not obtained, it is possible to obtain an identical target determination device that can determine the identity of the target from the attribute tree.

[0183]

According to the invention of claim 5 , from the attribute data of the observed target, the probability that the target belongs to each node of the attribute tree (affiliation probability) is estimated, A step of comparing the affiliation probabilities of both the detected targets that have been observed to determine the similarity of the target, and then configuring an identity determination step of determining identity based on the similarity described above, as target observation data When the exercise data is not obtained and only the attribute data is obtained, it is possible to obtain the same target determination method that can determine the sameness of the target from the attribute tree.

Further, according to the inventions of claims 6 to 11, when the position information of the target includes ambiguity due to an error in the target observation system, or when the observation data of the target is not a time series Also, even when the observation data is not regular, or when the movement data is not obtained as the target observation data and only the attribute data is obtained, the same target determination method that can determine the identity of the target can be obtained. .

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of the present invention.

FIG. 2 is a diagram illustrating a probability density function expressing the presence of a target.

FIG. 3 is a diagram showing an existence probability.

FIG. 4 is a flowchart illustrating the operation of the first embodiment of the present invention.

FIG. 5 is a configuration diagram showing a second embodiment of the present invention.

FIG. 6 is a diagram showing transition probabilities.

FIG. 7 is a diagram showing time adjustment according to the present invention.

FIG. 8 is a flowchart illustrating the operation of the second embodiment of the present invention.

FIG. 9 is a configuration diagram showing a third embodiment of the present invention.

FIG. 10 is a flowchart illustrating the operation of the third embodiment of the present invention.

FIG. 11 is a configuration diagram showing a fourth embodiment of the present invention.

FIG. 12 is a flowchart illustrating the operation of the fourth embodiment of the present invention.

FIG. 13 is a configuration diagram showing a fifth embodiment of the present invention.

FIG. 14 is a flowchart illustrating the operation of the fifth embodiment of the present invention.

FIG. 15 is a configuration diagram showing a sixth embodiment of the present invention.

FIG. 16 is a diagram showing a configuration of an attribute tree.

FIG. 17 is a flowchart illustrating the operation of the sixth embodiment of the present invention.

FIG. 18 is a configuration diagram showing a seventh embodiment of the present invention.

FIG. 19 is a flowchart illustrating the operation of the seventh embodiment of the present invention.

FIG. 20 is a schematic flowchart showing Embodiment 8 of the present invention.

FIG. 21 is a detailed flowchart showing Embodiment 8 of the present invention.

FIG. 22 is a schematic flowchart showing Embodiment 9 of the present invention.

FIG. 23 is a detailed flowchart showing Embodiment 9 of the present invention.

FIG. 24 is a schematic flowchart showing Embodiment 10 of the present invention.

FIG. 25 is a detailed flowchart showing Embodiment 10 of the present invention;

FIG. 26 is a schematic flowchart showing Embodiment 11 of the present invention.

FIG. 27 is a detailed flowchart showing Embodiment 11 of the present invention;

FIG. 28 is a schematic flowchart showing a twelfth embodiment of the present invention.

FIG. 29 is a detailed flowchart showing Embodiment 12 of the present invention;

FIG. 30 is a schematic flowchart showing Embodiment 13 of the present invention.

FIG. 31 is a detailed flowchart showing Embodiment 13 of the present invention;

FIG. 32 is a schematic flowchart showing Embodiment 14 of the present invention.

FIG. 33 is a detailed flowchart showing Embodiment 14 of the present invention;

FIG. 34 is a schematic flowchart showing Embodiment 15 of the present invention;

FIG. 35 is a detailed flowchart showing Embodiment 15 of the present invention;

FIG. 36 is a configuration diagram showing a conventional target tracking device.

FIG. 37 is a flowchart showing a conventional target tracking method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Probability density function expressing means 2 Correlation coefficient calculating means 3 Identity determining means 4 Target management database 5 Transition probability calculating means 6 Average position distance calculating means 7 Motion data estimating means 8 Target performance / specification database 10 Membership probability estimating means DESCRIPTION OF SYMBOLS 11 Attribute tree construction means 12 Similarity calculation means 13 Membership probability updating means 14 Estimation information database 21 Target observation device 22 Observation specification transfer device 23 Target prediction processor 24 First delay circuit 25 Motion data correlation specification calculator 26 Motion Data correlator 27 Detected data hypothesis generator 28 Hypothesis reliability calculator based on motion data 29 Suboptimal hypothesis selector 30 Target hypothesis smoothing processor 31 Tracking establishment determiner 32 Wake display 101 Motion data 102 Attribute data

──────────────────────────────────────────────────続 き Continuation of front page (56) References JP-A-6-43241 (JP, A) JP-A-5-288840 (JP, A) JP-A-1-65476 (JP, A) JP-A-6-43476 94830 (JP, A) Japanese Utility Model Hei 5-59372 (JP, U) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01S ⁷ /00-7/42 G01S 13/00-13/95 G01S 15/70 G01S 15/62

Claims (11)

(57) [Claims] 1. An identical target determining apparatus comprising the following elements for determining the sameness of a target from target information obtained as an observation result of the target: (1) determining the spatial existence of the observed target; A probability density function expressing means represented by a probability density function; (2) a transition probability calculating means for adjusting the times using the transition probabilities of both the probability density functions of the above-mentioned observed target and the already-observed detected target; Distance calculating means for calculating the distance between the reference positions of the two probability density functions at the same time; (4) identity determining means for determining the target identity based on the distance between the reference positions. 2. An identical target determination apparatus comprising the following elements for determining the identity of a target from target information obtained as a result of observation of the target. (1) Movement from attribute data of the observed target Motion data estimating means for estimating data; (2) a probability density function expressing means for expressing the spatial existence of the target by a probability density function; (3) the probabilities of both the observed target and the detected target already observed. Transition probability calculating means for adjusting the time by using the transition function with the density function; (4) distance calculating means for calculating the distance between the reference positions of the two probability density functions with the time adjusted; (5) between the reference positions Identity determination means for determining the identity of the target based on the distance of the target. 3. An identical target determination device comprising the following elements for determining the identity of a target from target information obtained as observation results of the target: (1) Classification of the entire target to be targeted Is represented by a hierarchical attribute tree, and from the attribute data of the observed target, a probability that the target belongs to each node of the attribute tree (hereinafter, referred to as a probability of belonging) is estimated. A similarity calculating means for comparing the affiliation probabilities of the observed target and the already detected detected target to obtain the similarity of the target; (3) the sameness for determining the identity of the target based on the above similarity Sex determination means. 4. An identical target determining apparatus comprising the following elements for determining the identity of a target from target information obtained as a result of observation of the target. Means for estimating the attribute data of the target; (2) expressing the overall classification of the target to be targeted by a hierarchical attribute tree, and from the estimated attribute data, the probability that the target belongs to each node of the attribute tree ( (3) Similarity calculation means for calculating the similarity of a target by comparing the membership probabilities of both the observed target and the already-detected detected target. (4) Identity determination means for determining the identity of the target based on the similarity. 5. An identical target determination method comprising the following steps of determining the identity of a target from target information obtained as observation results of the target. (1) First, the entire target Is represented by a hierarchical attribute tree. From the observed target attribute data,
A similarity that estimates the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as the belonging probability), and compares this belonging probability with the belonging probability of the already detected target to obtain the similarity of the target. (2) Next, an identity determination step of determining the identity of the target based on the similarity. 6. An identical target determination method, comprising the following steps of determining the identity of a target from target information obtained as an observation result of the target. (1) First, the spatial distribution of the observed target The existence is represented by a probability density function, the time of the probability density function of both the observed target and the already detected target is adjusted using the transition probability, and the correlation coefficient calculation for calculating the degree of overlap of the probability density functions is performed. (2) Next, the same candidate selection step of selecting a candidate for a detected target having a high degree of identity with the target observed based on the correlation coefficient of the probability density function described above, (3) Next, The overall classification is represented by a hierarchical attribute tree, and from the attribute data of the observed target, the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as the membership probability) is estimated. A similarity calculation step of comparing the rate and the belonging probability of the detected target that has already been observed to determine the similarity of the target; (4) identity determination for determining the identity of the target based on the above similarity Step. 7. An identical target determination method, comprising the following steps of determining the identity of a target from target information obtained as observation results of the target: (1) First, the entire target Is represented by a hierarchical attribute tree. From the observed target attribute data,
A similarity that estimates the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as the belonging probability), and compares this belonging probability with the belonging probability of the already detected target to obtain the similarity of the target. (2) Next, the same candidate selection step of selecting a detected target candidate having a high degree of identity with the observed target based on the above similarity, (3) Next, the spatial existence of the observed target is determined. A probability density function, the time of the probability density function of both the observed target and the detected target already observed is adjusted using the transition probability, and a correlation coefficient calculation step of obtaining the degree of overlap of the probability density functions; (4) Next, an identity determination step of determining the identity of the target based on the correlation coefficient of the probability density function. 8. An identical target determination method, comprising the following steps of determining the identity of a target from target information obtained as an observation result of the target. (1) First, the spatial distribution of the observed target The existence is represented by a probability density function, the time of the probability density function of both the observed target and the already detected target is adjusted using the transition probability, and the correlation coefficient calculation for calculating the degree of overlap of the probability density functions is performed. In parallel with the step, (2) The overall classification of the target is represented by a hierarchical attribute tree, and from the observed target attribute data, the probability that the target belongs to each node of the attribute tree (hereinafter, belonging A similarity calculation step of estimating the affiliation probability and comparing the affiliation probability with the already-observed detected target affiliation probability to obtain the similarity of the target. (3) Next, the phase of the above probability density function An identity determination step of determining the identity of the target based on the number of relations and the similarity. 9. An identical target determination method, comprising the following steps of determining the identity of a target from target information obtained as an observation result of the target. (1) First, the spatial distribution of the observed target Presence is represented by a probability density function, the time of the probability density functions of both the observed target and the already detected target is adjusted using the transition probability, and the reference position for calculating the distance between the reference positions of the probability density functions (2) Next, the same candidate selection step of selecting a candidate of a detected target having a high degree of identity with the target observed based on the above-mentioned reference position distance, (3) Next, the target target The overall classification is represented by a hierarchical attribute tree, and from the attribute data of the observed target, the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as belonging probability) is estimated. A similarity calculation step of comparing the probability with the belonging probability of the detected target that has already been observed to determine the similarity of the target; (4) identity determination for determining the identity of the target based on the similarity Step. 10. An identical target determination method comprising the following steps of determining the identity of a target from target information obtained as observation results of the target. (1) First, the entire target Is represented by a hierarchical attribute tree. From the observed target attribute data,
A similarity that estimates the probability that the target belongs to each node of the attribute tree (hereinafter, referred to as the belonging probability), and compares this belonging probability with the belonging probability of the already detected target to obtain the similarity of the target. (2) Next, the same candidate selection step using similarity for selecting a candidate of a detected target having a high degree of identity with the target observed based on the above similarity, (3) Next, The spatial existence is represented by a probability density function, and the times of the probability density functions of both the observed target and the already-detected detected target are adjusted using transition probabilities, and the distance between the reference positions of the probability density functions is calculated. A reference position distance calculation step; (4) an identity determination step of determining the target identity based on the reference position distance. 11. An identical target determination method, comprising the following steps of determining the identity of a target from target information obtained as a result of observation of the target. Presence is represented by a probability density function, the time of the probability density functions of both the observed target and the already detected target is adjusted using the transition probability, and the reference position for calculating the distance between the reference positions of the probability density functions In parallel with the distance calculation step, (2) the overall classification of the target to be represented is represented by a hierarchical attribute tree, and from the attribute data of the observed target, the probability that the target belongs to each node of the attribute tree ( Affiliation probability)
And calculating the similarity of the target by comparing this belonging probability with the belonging probability of the already detected target. (3) Then, the distance between the reference positions of the probability density function and the reference position Determining the identity of the target based on the similarity of the target.