JP2001337162A - Target-tracking apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To distinguish a track established by tracing and to distinguish to which target the track belongs to. SOLUTION: A cluster, newly providing integration part 7 integrates existing clusters and prepares, a new cluster, and a target number corresponding type track preparation part 9 prepares a new track at current time in the cluster and determines a target number. A hypothesis preparation part 10 prepares a new hypothesis at the present time, and a hypothesis reducing part 16 evaluates the hypothesis and reduces the number of hypotheses. A cluster separation part 17 separates clusters and reconstructs the hypothesis, a tracking establishment deciding part 18 decides as to whether tracking of a track has been established, and a target number providing part 19 gives a target number to the established track. A target number corresponding type track determining part 21 selects the best hypothesis and outputs it together with the target number.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a target tracking device for estimating the wakes of a plurality of targets based on an observation vector of a target position intermittently obtained by a sensor such as a radar, targeting an aircraft or a ship. It is.

[0002]

2. Description of the Related Art In order to obtain trajectories of a plurality of targets from an observation vector of a target position obtained by a sensor, the position of the current time is predicted by tracking filter processing for each existing trajectory obtained up to the previous time. Further, an expected area (hereinafter, this area is referred to as a gate) of the observation vector at the current time for each track is calculated, and the correlation processing between the observation vector actually acquired in the gate and each track is performed. By doing so, the target track at the current time is estimated.

Here, in practice, there are cases where an erroneous signal other than the target is obtained from the sensor, or conversely, the sensor fails to observe the target, and the target observation vector cannot be obtained. In addition, when a plurality of targets are densely packed in a narrow space, observation vectors of a plurality of tracked targets can be obtained in a gate of one track, or observation of a new target which has not been detected so far can be obtained. Sometimes a vector is obtained. Even in such a situation, in order to continuously track a plurality of targets correctly, it is necessary to identify which observation vector corresponds to each track, that is, to judge the correlation with high accuracy.

Conventionally, as a device meeting such a demand, for example, the following target tracking device has been proposed. FIG. 6 is a block diagram showing the overall configuration of a conventional target tracking device disclosed in Japanese Patent Application Laid-Open No. 8-271617.
In the figure, 1 is a target tracking device, 2 is a target observation device as a sensor for observing the position of the target in space and obtaining an observation vector, and 3 is a state of the target which displays wake information on a display and is tracking. Is a target display device.

[0005] In the target tracking device 1, reference numeral 4 denotes a gate calculation unit for calculating the expected area of the observation vector at the current time with respect to all the tracks included in the existing cluster. The observation vector selection unit 6 selects an observation vector included in the gate of each track calculated by the gate calculation unit 4 from the entire observation vector at the current time input to 1, and 6 selects the entire cluster in the target tracking device 1. The shown in-system cluster table 7 integrates existing clusters and creates new clusters based on the relationship between the output of the observation vector selection unit 5 and the existing clusters shown in the in-system cluster table 6. Department.

In the target tracking device 1 shown in FIG.
Reference numeral 8 denotes an intra-cluster observation vector table showing observation vectors belonging to each cluster among the input observation vectors at the current time, and reference numeral 51 denotes a data group indicating the intra-cluster observation vector table 8 and the status of an existing hypothesis in the cluster. And a wake and hypothesis creation unit that creates a new wake and hypothesis in the cluster corresponding to the observation vector at the current time.

Further, in the target tracking device 1 shown in FIG. 6, reference numeral 11 denotes an intra-cluster hypothesis status data group, reference numeral 12 denotes an intra-cluster hypothesis table defining all the hypotheses in the existing cluster, and reference numeral 13 denotes an existing cluster. In the hypothesis internal track table, which defines all the tracks in the hypothesis for each hypothesis, 14 is the intra-cluster track-observation vector table which defines the observation vectors that constitute the tracks for all the tracks in the cluster. It is.

Further, in the target tracking device 1 of FIG. 6, 16 evaluates the likelihood of each hypothesis in the cluster,
A hypothesis reducing unit 17 that reduces the number of hypotheses based on the evaluation result determines whether the cluster after the hypothesis reduction is separable, and reconstructs the hypothesis of each cluster after the separation if the cluster is separable. In a cluster separation unit, when a plurality of hypotheses exist in a cluster, 52 selects one of the best hypotheses from among them to determine the number of targets and their wakes, and displays information of each wake on a target display device. 3 is a track determination unit to be sent to the third track.

Next, the operation will be described. FIG. 7 is a diagram illustrating a simple example of a situation in which it is difficult to determine the correlation between each track and an observation vector in target tracking. In the figure,
_Tk-1,1 and _Tk-1,2 are the existing tracks obtained up to the previous time, and _{Pk, 1} and _{Pk, 2} are the tracks Tk _-1,1 and _{Tk-1,2, respectively. Gk, 1} and _{Gk, 2} are the existing trajectories T
_This is a gate that indicates the expected area of the observation vector at the current time for _k-1,1 and _Tk-1,2 . Further, Z _{k, 1} and Z _{k, 2}
Indicates the position of the observation vector actually acquired by the sensor at the current time.

In the situation shown in FIG. 7, it is necessary to consider a plurality of possibilities in determining the correlation as to which observation vector corresponds to each track. For example, the observation vector Z _{k, 1} corresponds to the wake T _k-1,1 and the observation vector Z _{k, 1}
If it can be considered that _{k, 2} corresponds to the wake T _k-1,2 , the observation vector Z _{k, 1} is a false signal or a new target,
It can also be considered that observation vector Z _{k, 2} corresponds to wake T _k-1,1 . There are several other possibilities.

If an erroneous correlation is selected from these,
There is a risk that tracking may be missed as a result, for example, because the predicted position or the gate range after the next time goes out of order. Therefore, FIG.
In a situation where the correlation between the wake and the observation vector is not uniquely determined, as in the case of, the tracking process is continued while holding the various possibilities described above as hypotheses, and the one with a high probability after the next time is selected. The postponement-type method is effective. The conventional target tracking device aims at realizing such a method.

In the above-described method, as in the case of FIG. 7, it is possible to form a hypothesis by using a group of wakes sharing an observation vector in a gate as one unit. For this reason, in the conventional target tracking device, the entire wake in the space is divided into each of the wake groups as described above, and the process is performed. This processing unit is called a cluster.

Hereinafter, the operation will be described in detail with reference to FIG. First, the target observation apparatus 2 to the current time t _k, by entering the observation vector to the target tracking device 1, the gate calculator 4
Is a gate that is an expected area of the observation vector at the current time t _{k with} respect to all the tracks defined in the intra-cluster track-observation vector table 14 of the existing cluster group input up to the previous time t _k−1. Is calculated. Next, the observation vector selection unit 5 determines which observation vector can be correlated with each track by checking which observation vector exists in the gate of each track.

Next, the cluster new integration unit 7 first creates a new cluster composed of the observation vectors that cannot be correlated with the wakes of any of the clusters, and defines them in the in-system cluster table 6. I do. If a certain observation vector exists in the gates of a plurality of tracks belonging to different clusters, these clusters are integrated as one cluster and redefined in the intra-system cluster table 6. Further, the in-cluster hypothesis status data group 11 of the cluster after integration is reconstructed from the in-cluster hypothesis status data group 11 of each cluster before integration. further,
The cluster newly-integrated unit 7 includes a cluster table 6 in the system.
For each cluster defined in (1), the entire observation vector that can be correlated with one or more tracks in the cluster is written into the intra-cluster observation vector table 8.

Next, a track and hypothesis creation unit 51 is created for each cluster defined in the intra-system cluster table 6 by the previous time t _k−1 and defined in the intra-cluster track-observation vector table 14. and the existing track in the cluster was, from the observation vector group in the cluster at the present time t _k, which is described in the cluster within the observation vector table 8, to create a track that corresponds to the current time, cluster in the wake - in the observation vector table 14 with redefining cluster hypothesis table 12, the relationship between the track of the current time t _k that situation and the creation of an existing hypothesis in the cluster defined in the hypothesis in track table 13, the current time t _k in the cluster
Is created and redefined into the intra-cluster hypothesis table 12 and the intra-hypothesis track table 13.

First, the creation of a wake will be described. Here, the track of the current time t _k, can be defined as from the time up to the present time t _k, a time-series data constituting choose observation vector of at most 1 (0 or 1) .
However, selection of zero observation vectors means that no wake was detected at that time. Also, 2
Two tracks are regarded as the same track only when all of the observation vectors constituting both tracks are the same.

The track at the current time t _k is created as follows based on the state of the track in the cluster at the previous time t _k-1 . First, when a track at the previous time tk _-1 (hereinafter referred to as an existing track) exists in the cluster, that is, when this cluster is not a newly created cluster this time, each existing track in the cluster is to, to create a wake of the current time t _k according to the following procedure. (1) For each observation vector existing in the gate of the existing track, the current time t _k at which the observation vector is added to the existing track
Create a wake for (2) as an existing track has not been detected, to create a wake at the present time t _k which is directly holding the existing track.

By the above operation, a plurality of tracks are generally created from one existing track. Next, all of the cluster, for each observation vector of the current time t _k in the cluster,
To create a wake of the current time t _k according to the following operation. (3) A wake newly started from the observation vector in the cluster (hereinafter, referred to as a new wake) is created. With this operation, the current time t _k of the observation vector and the same number of new track in the cluster is created.

Next, creation of a hypothesis will be described. Here, the hypothesis of the current time t _k, in each cluster, the wake of the current time t _k in the cluster can be defined as a set of wake configured to select zero or more. Two hypotheses are considered to be the same hypothesis only if all of the tracks that comprise them are the same. Each hypothesis is a representation of the interpretation of any of or choose a combination of the track is right in the cluster at the current time t _k. However, an observation vector in a cluster that does not belong to any wake included in the hypothesis corresponds to treating as an erroneous signal in this hypothesis. In particular, the hypothesis selecting zero or wake is the case of interpreting the erroneous signal all observation vectors up to the present time t _k in the cluster.

The criterion for creating a hypothesis is “select a track in a cluster so that a plurality of tracks belonging to the same hypothesis do not share the same observation vector”. When creating a hypothesis, it is necessary to identify all wake combinations that meet the above criteria and create a different hypothesis.

[0021] However, when you actually create a hypothesis of the current time t _k can be in accordance with the following procedure. First, when the cluster for which the hypothesis is to be created is a cluster newly created this time, a hypothesis that a wake composed of one observation vector registered in this cluster is selected and a wake is assumed to be 1
What is necessary is just to create two hypotheses of not selecting any. When the hypothesis creation target cluster is an existing cluster, as shown in the following example, each hypothesis (existing hypothesis) at the previous time t _k−1 in the cluster satisfies the above-described hypothesis criteria. It should be extended to the hypothesis of the multiple of the current time t _k so.

For example, the hypothesis X at the previous time t _k-1
It is assumed that ^{k-1, r} is composed of two existing tracks T _k-1,1 and T _{k- 1,2} , and these tracks satisfy the above-described hypothesis criteria. Furthermore, the existing track T _{k-1, 1} observation vector Z _k at the current time t _k is within a gate _{of, 1,} Z _{k, 2} can be obtained and, in the gate of an existing track T _{k-1, 2} It is assumed that the observation vector Z _{k, 2} has been obtained.

That is, the situation was the same as that of FIG.
And At this time, the existing hypothesis X ^{k-1, r} Of the above hypothesis
By expanding to meet the criteria, the following 11
Current time t_kCan create a hypothesis. Where ｛T
_{k-1, i} , Z_{k, j} ｝ Indicates existing track T_{k- 1, i} Observation vector
Z_{k, j} Track with を T_{k-1, i} ｝ Indicates existing track T
_{k-1, i}To keep as it was not detected
Trace, Z_{k, j} ｝ Is the observation vector Z_{k, j} New start
Represents a new wake. X^{k, 1} = ｛｛T_k-1,1 ｝, ｛T_k-1,2 ｝｝ X^{k, 2} = ｛｛T_k-1,1 ｝, ｛T_k-1,2 , Z_{k, 2} ｝｝ X^{k, 3} = ｛｛T_k-1,1 , Z_{k, 1} ｝, ｛T_k-1,2 ｝｝ X^{k, 4} = ｛｛T_k-1,1 , Z_{k, 1} ｝, ｛T_k-1,2 , Z_{k, 2} ｝｝ X^{k, 5} = ｛｛T_k-1,1 , Z_{k, 2} ｝, ｛T_k-1,2 ｝｝ X^{k, 6} = ｛｛T_k-1,1 ｝, ｛T_k-1,2 ｝, ｛Z_{k, 1} ｝｝ X^{k, 7} = ｛｛T_k-1,1 ｝, ｛T_k-1,2 ｝, ｛Z_{k, 2} ｝｝ X^{k, 8} = ｛｛T_k-1,1 ｝, ｛T_k-1,2 ｝, ｛Z_{k, 1} ｝, ｛Z_{k, 2} ｝｝ X^{k, 9} = ｛｛T_k-1,1 ｝, ｛T_k-1,2 , Z_{k, 2} ｝, ｛Z_{k, 1} ｝｝ X^{k, 10}= ｛｛T_k-1,1 , Z_{k, 1} ｝, ｛T_k-1,2 ｝, ｛Z_{k, 2} ｝｝ X^{k, 11}= ｛｛T_k-1,1 , Z_{k, 2} ｝, ｛T_k-1,2 ｝, ｛Z_{k, 1} ｝｝

[0024] Thus, in the example, generally, a plurality of hypotheses of current time t _k is derived from one of the hypotheses of the previous time t _k-1.
Further, each of the hypotheses at the current time t _k
Be a parent who produces multiple hypotheses.

Next, returning to FIG. 6, the hypothesis reducing unit 16 determines, for each cluster, the likelihood of each hypothesis in the cluster defined in the intra-cluster hypothesis status data group 11 (the probability that each hypothesis is true). Is evaluated, and a process for reducing the number of hypotheses is performed based on the result. The purpose of this processing is not only to keep the amount of data to be processed in the apparatus constant, but also to extract only important hypotheses from the wake and hypotheses created by the hypothesis creation unit 51.

As a specific method for reducing the number of hypotheses, the following method can be applied to the conventional target tracking device disclosed in Japanese Patent Application Laid-Open No. Hei 8-271617, for example. (1) Only the hypothesis with the highest evaluation value is left. (2) Discard hypotheses below the evaluation criteria. (3) A plurality of hypotheses having the same content for the past N times are integrated into one. (4) The number of tracks is the same, and the content of each track (position, speed, etc.)
Integrate almost identical tracks.

Here, the hypothesis reduction processing of the above (3) corresponds to performing the following operation based on the definition of the wake and the hypothesis described above. First, observation vectors for the past N times from the latest are left from each track in the cluster, and data older than that is deleted. As a result, tracks having the same observation vector configuration can be regarded as the same track. Next, by the above operation, the hypotheses having the same wake configuration are regarded as the same hypothesis and are integrated into one. At this time, the evaluation value of the hypothesis after integration is calculated by the sum of the evaluation values of the hypothesis group before integration. The above operation reduces the number of tracks and hypotheses in the cluster.

Next, the cluster separating unit 17 determines whether or not each cluster can be separated by performing the above-described hypothesis reduction processing. If the cluster can be separated, the cluster is divided into a plurality of clusters. To separate. here,
A cluster is strictly defined as follows.

That is, the wake T_iAnd T_jIs at least one
When two observation vectors are shared, the wake T_iAnd T_jLike
Called similar wake, T_i~ T_jWrite Furthermore, the wake T_iWhen
T_jAt T_i= T_i1~ T_i2~ ... ~ T_in= T_j  T_i1, T_i2, ..., T_inOnly if
, T_i≡T_j  Is defined. This relationship satisfies the reflex, symmetry, and transitivity laws.
It is an equivalence relation to be added. All wakes are given by this equivalence relation
Can be divided into several disjoint classes (subsets)
You. This class is called a cluster.

According to the above definition, the cluster separating unit 17
It is determined whether each cluster can be separated into a plurality of clusters, and if so, the cluster is separated into a plurality of clusters.
For each cluster after separation, the hypothesis in the cluster is reconstructed based on the hypothesis of the cluster before separation,
The intracluster hypothesis situation data group 11 is reconstructed. Cluster separation reduces the size of each hypothesis,
The effect that the amount of processing required for updating the hypothesis is reduced in the entire apparatus is obtained.

Further, when a cluster composed of only one track can be separated from a certain cluster, and all the observation vectors constituting the track are treated as erroneous signals in the cluster before separation. If no hypothesis exists, this wake means that all hypotheses have been adopted in the cluster before separation. Therefore, this track is a track in which the uncertainty relating to the correlation has been eliminated. In particular, when the track is first separated into clusters only by one track since the track was started with the observation vector of the first detection, it is a case where the existence of the target to be tracked by the track has been determined. Typically, such a condition is defined as tracking the target has been "established."

However, it should be noted that the track once established once in this manner is again branched into a plurality of tracks according to the observation vector after the next time and a hypothesis is created. The process at this time is to update the already established track,
The purpose is to maintain it. Further, the cluster of the track once established may be integrated with other clusters by the processing of the cluster new integration unit 7.

Next, the wake determination unit 52 examines the intra-cluster hypothesis status data group 11 of each cluster, and finds the current time t _K in the cluster defined in the intra-cluster hypothesis table 12.
The number of targets and the wake are determined by selecting one of the best hypotheses, for example, the hypothesis with the highest probability evaluation value calculated by the hypothesis reduction unit 16, and outputs it to the target display device 3. . Incidentally, the above-mentioned JP-A-8-27161
Although not explicitly described in Japanese Patent Application Publication No. 7 (1994), usually, the target tracker uses the intra-cluster track-observation vector table 14 for the time-series data of the track determined as described above.
Is read, and a tracking filter such as a Kalman filter is applied to calculate a smoothed value such as the position and speed of the target at the latest time, and the calculation results are sent to the target display device 3 and displayed on the display. It is common.

[0034]

The conventional target tracking device is configured as described above. Although the tracking is already started from a track which has been established and a certain observation vector, the tracking has not been established yet. Since the process is not clearly distinguished from the track, the operator cannot identify whether the displayed track is an established track or an unestablished track. Was. In addition, since the conventional target tracking device does not perform processing to clearly distinguish each created wake according to which target is to be tracked, the operator sets the previous track at the previous time t _k−1 There is a problem that it is difficult to determine which of the displayed tracks corresponds to which track at the current time t _k (whether it is a result of tracking the same target).

Further, when reducing the number of hypotheses, the past N
Tracks created by trying to track different targets regardless of which target each track in the cluster corresponds to, since the number of hypotheses is reduced by integrating hypotheses with the same time content There was a problem that they were identified with each other.

Further, when a plurality of targets are close to each other and it is difficult to determine the correlation, the number of tracks output and the position and speed of each track change frequently at each time, so that the operator can grasp the actual situation. There was a problem that it was difficult to do.

The present invention has been made in order to solve the above-described problems, and can identify whether a track has already been tracked or has not been tracked. An object of the present invention is to obtain a target tracking device that can identify which target is a wake created for tracking.

It is still another object of the present invention to provide a target tracking apparatus which does not cause inconsistency in equating wakes created when tracking different targets when reducing the number of hypotheses.

It is still another object of the present invention to obtain a target tracking device that outputs stable track information with little change in position, speed, etc. when determining the smoothed value of the position and speed of each track to be output.

[0040]

A target tracking apparatus according to the present invention includes an intra-cluster hypothesis table defining all the hypotheses existing in an existing cluster, and an intra-cluster hypothesis table for each hypothesis in the existing cluster. A hypothetical intra-track table that defines the trajectory of the trajectory, an intra-cluster trajectory-observation vector table that defines the observation vectors that make up the trajectory for all the trajectories in the existing cluster, and all the trajectories in the existing cluster. For the wake,
For each track defined in the intra-cluster track-target number table defining the target number associated with the track and the above-described intra-cluster track-observation vector table, each track which is the expected area of the observation vector at the current time. The gate calculation unit that calculates the gate of the current time and the input observation vector at the current time are input to check which observation vector is present in the gate of each track calculated by the gate calculation unit, and correlate with each track. An observation vector selection unit that selects an observation vector that can be
From the relationship between the output of the observation vector selection unit and the existing clusters, a new cluster is created using observation vectors that cannot be correlated with the wakes of any clusters. , A cluster new integration unit that integrates these clusters and reconstructs the intra-cluster hypothesis table, the hypothesis intra-track table, the intra-cluster trajectory-observation vector table and the intra-cluster trajectory-target number table From the observation vector at the current time in the cluster and the track in the existing cluster defined in the above-mentioned track in the cluster-observation vector table, a wake corresponding to the current time is created, and In addition to redefining the vector table, the target time defined in the track-target number table in the cluster corresponds to the current time created. A target number corresponding type track generation unit that determines the target number of the track that has been set, and redefines the track in the cluster-target number table, the hypothesis table in the cluster, and the existing in the cluster defined in the hypothesis track table. A new hypothesis in the cluster corresponding to the observation vector at the current time is created based on the hypothesis status and the wake at the current time created by the trajectory creation unit corresponding to the target number, and the hypothesis table in the cluster and the A hypothesis creation unit for redefining the track table, the hypothesis table in the cluster, the probability of each hypothesis in the cluster redefined in the hypothesis track table are evaluated, and based on this evaluation result, A hypothesis reduction unit that reduces the number of hypotheses, and determines whether the cluster after the hypothesis reduction is separable,
If separation is possible, the hypothesis of each cluster after separation is reconstructed, and the cluster in which the intra-cluster hypothesis table, the hypothesis intra-track table, the intra-cluster trajectory-observation vector table and the intra-cluster trajectory-target number table are reconstructed When a separation unit and a cluster consisting of only one track are separated, it is determined whether or not tracking of this track has been established, and a new target number assigned to the track for which tracking has been established is determined. A tracking establishment determining unit defined in the track in the cluster-target number table, a target number assigning unit for assigning a new target number to the track in which the tracking is established, and a hypothesis defined in the intracluster hypothesis table And a target number corresponding type track determining unit for determining the number of tracks and the track at the current time, and outputting the associated track with the target number defined in the above-mentioned track-target number table. Thing A.

The target tracking device according to the present invention includes a hypothesis table in which all hypotheses in an existing cluster are defined, and a hypothesis table in which all wakes in the hypothesis are defined for each hypothesis in the existing cluster. The in-cluster track table, the intra-cluster track-observation vector table that defines the observation vectors that constitute the tracks for all the tracks in the existing cluster, and the wakes for all the tracks in the existing cluster A track number within the cluster defining a track number flag and a track state flag table indicating a track state flag indicating whether tracking is established or not, and a track number within the cluster-tracking observation table A gate calculation unit for calculating the gate of each track, which is an expected area of the observation vector at the current time, for all the tracks, and the gate An observation vector selection unit that checks which observation vector exists in the gate of each wake calculated by the output unit and selects an observation vector that can be correlated with each wake, and an output of the observation vector selection unit and an existing From the relationship between clusters, a new cluster is created with observation vectors that cannot be correlated with the wakes of any clusters. If a certain observation vector exists in the gates of multiple wakes belonging to different clusters, these clusters are A new cluster integration unit that integrates and reconstructs the intra-cluster hypothesis table, the hypothesis intra-track table, the intra-cluster trajectory-observation vector table, and the intra-cluster trajectory-target number and wake state flag table; From the observation vector at the time and the tracks in the existing cluster defined in the above Create a track corresponding to the time, redefine the track within the cluster-observation vector table, and use the target number defined in the track within the cluster-target number table and the track state flag table to correspond to the created current time. The target number and the track state flag of the track that has been set are determined and redefined in the above-mentioned track in the cluster—target number and track state flag table, and the target number assigned to the created new track and tracking are established. A track status flag indicating that the track has not been changed is redefined in the track-in-cluster target number table and the track status flag in the above-mentioned cluster, and a target number and track status flag corresponding track generating unit; A target number assigning unit for assigning a target number, the intra-cluster hypothesis table, the status of an existing hypothesis in a cluster defined in the hypothesis intra-track table, and the target number and wake state flag pair A hypothesis creator that creates a new hypothesis in the cluster corresponding to the observation vector at the current time from the wake at the current time created by the adaptive wake creator, and redefines the hypothesis table in the cluster and the hypothesis trajectory table. And a hypothesis reduction unit that evaluates the likelihood of each hypothesis in the cluster that is redefined in the hypothesis intra-cluster hypothesis table and the hypothesis intratraffic table, and reduces the number of hypotheses in the cluster based on the evaluation result. It is determined whether or not the cluster after the hypothesis reduction is separable, and if the cluster is separable, the hypothesis of each cluster after the separation is reconstructed, and the intra-cluster hypothesis table, the hypothesis intra-track track, and the intra-cluster track- A cluster separation unit for reconstructing the observation vector table and the intra-cluster wake-target number and wake state flag table;
When a cluster composed of only one track is separated, it is determined whether or not tracking of this track has been established, and a track state flag changed with respect to the track for which tracking has been established is used as the track within the cluster. A tracking establishment determination unit defined in the target number and wake state flag table,
A track state flag changing unit for changing to a track state flag indicating that tracking has been established; and selecting one hypothesis from the hypotheses defined in the above-mentioned intra-cluster hypothesis table, and determining the number of targets and the track at the current time. A track number corresponding to a target number and a track state flag, which is output in association with the target number and the track state flag defined in the track-target number in the cluster and the track state flag table, is provided.

In the target tracking apparatus according to the present invention, the hypothesis reducing unit calculates an evaluation value of the likelihood of each hypothesis in the cluster, and deletes an observation vector older than the latest N time from each track in the cluster. The trajectories in the cluster are classified for each trajectory having the same observation vector configuration and target number, and trajectories classified into the same group are regarded as the same trajectory. Into one hypothesis,
The evaluation value of the post-integration hypothesis is calculated from the evaluation value of the pre-integration hypothesis group, and only one track is selected from each group of tracks that are classified for each track having the same observation vector configuration and target number. Delete the rest.

In the target tracking apparatus according to the present invention, the hypothesis reducing unit calculates an evaluation value of the likelihood of each hypothesis in the cluster. Obtain the number of the previous hypothesis, classify the hypotheses in the cluster for each hypothesis with the same number of hypotheses N times before,
Calculate the total value of the evaluation values of the hypotheses in the group for each group of the classified hypotheses, select the group of the hypothesis with the maximum total value of the evaluation values, and select the group other than the selected hypothesis group. The hypothesis that belongs to the cluster is deleted from the cluster, and the wake that is not included in any of the hypotheses remaining in the cluster is deleted from the cluster.

[0044] In the target tracking device according to the present invention, the target number-corresponding trajectory determining section selects a plurality of hypotheses with high evaluation values of the hypotheses from the hypotheses in each cluster and includes the hypotheses in the selected hypotheses. In the track to be tracked, all tracks whose tracking has already been established are extracted, and for the extracted track whose tracking has been established, the position of the target current time by tracking filter processing,
Calculate smooth values such as speed, evaluate the likelihood of the extracted track that has been tracked, classify the extracted track for each track with the same target number, and for each group of the classified tracks, The smoothed values such as the position and speed for each track in the group are weighted and integrated with the evaluation value of each track, and the integrated smoothed value corresponding to each target number is output together with the target number.

In the target tracking device according to the present invention, the track number determining unit corresponding to the target number and the track state flag selects a plurality of hypotheses having high evaluation values from the hypotheses in each cluster, and selects the selected plurality of hypotheses. Of the trajectories included in the hypothesis, all tracings whose tracking has already been established are extracted. Calculate, evaluate the likelihood of the extracted track that has been tracked, classify the extracted track for each track with the same target number, and for each group of the classified track, The smoothed values such as the position and the speed are weighted and integrated with the evaluation value of each track, and the integrated smoothed values corresponding to the respective target numbers are output together with the target numbers and the track state flags.

[0046]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below. Embodiment 1 FIG. FIG. 1 is a block diagram showing an overall configuration of a target tracking device according to Embodiment 1 of the present invention. In FIG. 1, reference numeral 1 denotes a target tracking device, 2 denotes a target observation device as a sensor for observing the position of the target in space to obtain an observation vector, and 3 denotes a state of the target being tracked by displaying a wake on a display. It is a target display device shown.

In the target tracking device 1 of FIG. 1, reference numeral 4 denotes the current time t _k for all the tracks included in the existing cluster.
Gate determining unit for calculating a presence expected region of the observation vector, 5 from the entire observation vector at the present time t _k which is inputted from the target observation device 2 to the target tracking device 1, the gate calculator 4
An observation vector selection unit for selecting an observation vector included in the gate of each wake calculated by the following;
Cluster table showing the entire cluster in the system, 7
Is a new cluster integration unit that integrates existing clusters based on the relationship between the output of the observation vector selection unit 5 and the existing clusters shown in the in-system cluster table 6, and creates a new cluster.

In the target tracking device 1 shown in FIG.
8 out of the observation vector at the present time t _k which is the input, a cluster within the observation vector table showing the observation vector belonging to each cluster for each cluster, the existing track and in the observation vector table 8 and cluster Cluster 9 enter the target number, to create a new track in the cluster corresponding to the observation vector of the current time t _k, the target number corresponding type track generating unit for determining a target number of its wake, 10 and the existing hypothesis in the cluster from the wake of the current time t _k that you created above, it is a hypothesis generation unit to create a new hypothesis in the cluster corresponding to the observation vector of the current time t _k.

Further, in the target tracking device 1 of FIG. 1, reference numeral 11 denotes an intra-cluster hypothesis situation data group, 12 denotes an intra-cluster hypothesis table defining all the hypotheses existing in the existing cluster, and 13 denotes an existing cluster hypothesis table. In the hypothesis-in-track table, which defines all the tracks in the hypothesis for each hypothesis, 14 is the intra-cluster trajectory-observation that defines the observation vectors that constitute the wakes for all the trajectories in the existing cluster Vector table, 15 for all tracks in the existing cluster,
It is an in-cluster track-target number table that defines a target number associated with a track.

Further, in the target tracking device 1 of FIG. 1, 16 evaluates the likelihood of each hypothesis in the cluster, and based on this evaluation result, a hypothesis reduction section for reducing the number of hypotheses in the cluster, and 17 A cluster separation unit that determines whether a cluster after hypothesis reduction is separable and reconstructs a hypothesis of each cluster after separation if separation is possible. A cluster 18 composed of only one track is separated. In this case, a tracking establishment determining unit that determines whether tracking of the track has been established, 19 is a target number assigning unit that assigns a new target number to the track, and 20 is used in the target tracking device 1 4 is a target number table in the system in which all target numbers are defined.

Further, in the target tracking device 1 of FIG. 1, reference numeral 21 denotes a case where a plurality of hypotheses exist in a cluster.
One of the best hypotheses is selected from among them, the number of targets and the wakes thereof are determined, and information on each wake including information on the target number assigned to each wake is transmitted to the target display device 3. This is the type track determination unit. In FIG. 1, the same components as those of the conventional target tracking device shown in FIG. 6 are denoted by the same reference numerals.

Next, the operation will be described. First, time t
_kFrom the target observation device 2 to the target tracking device 1
When the gate calculation unit 4 inputs the_k-1Until
Track within a cluster of existing clusters input to
Current time for all tracks defined in vector table 14
t _kWhich is the expected region of the observation vector
Is calculated. Next, the observation vector selection unit 5
Find out which observation vector exists in the gate
The observation vector that can be correlated with each track.
You.

Next, for the observation vector that cannot be correlated with the track of any cluster, the cluster new integration unit 7 first creates a new cluster composed of the observation vector, and defines it in the in-system cluster table 6. I do. If a certain observation vector exists in the gates of a plurality of tracks belonging to different clusters, these clusters are integrated as one cluster and redefined in the intra-system cluster table 6. Further, the post-integration hypothesis situation data group 11 after the integration is reconstructed from the intra-cluster hypothesis situation data group 11 of each cluster before the integration. Further, for each cluster defined in the intra-system cluster table 6, the cluster newly-integrated unit 7 writes the entire observation vector that can be correlated with one or more tracks in the cluster to the intra-cluster observation vector table 8. .

Next, the target number-corresponding trajectory creating unit 9 performs a process for each of the clusters defined in the intra-system cluster table 6.
The existing track in the cluster created at the previous time t _k−1 and defined in the intra-cluster track-observation vector table 14 and the observation vector group in the cluster at the current time t _k described in the intra-cluster observation vector table 8 from, to create a wake that corresponds to the current time t _k, cluster in the wake - to redefine the observation vector table 14. The method of creating the wake in this process is the same as the process of the wake and hypothesis creation unit 51 in the conventional target tracking device of FIG. 6 described above.

[0055] That is, first, if the existing track is in a cluster, for each existing track, to create a wake of the current time t _k according to the following procedure. (1) For each observation vector existing in the gate of the existing track, the current time t _k at which the observation vector is added to the existing track
Create a wake for (2) as an existing track has not been detected, to create a wake at the present time t _k which is directly holding the existing track. In addition, for each observation vector of the current time t _k in the cluster, to create a wake at the present time by the following operation. (3) Create a new track starting from the observation vector in the cluster.

[0056] Incidentally, this cluster by cluster established integration unit 7, when a cluster is established to the current time t _k, by the operation of the above (3), within a cluster, one defined in the cluster Starting from the observation vector of 1
Tracks are created and defined in the intra-cluster track-observation vector table 14.

Next, the target number corresponding type track generation unit 9
Track within raster-within the cluster defined in target number table 15
Based on the target number of the existing track
Time t _k Determine the target number for each track of
Intra-raster track-target number Table 15 is redefined. here,
The target number is the number of each track kept in the device.
Attempt to track any of the targets determined to be in
It is defined as an identification number indicating whether the track is
For the target correspondence, for example, a positive integer is given. Target number compatible type
Current time t by the track creator 9_k Number for the track of the ship
Is simple. That is, the current time t_k Wake of
Is the current time t_k (1)
In the track created from the existing track by the procedure of (2) or
In some cases, the target number for this track is the target of the existing track.
Inherit the number as it is. Also, the current time t_k The wake of
If the track is a new track created by the procedure of (3) above,
Means that the track is a track for which tracking has not yet been established
For example, “0” is assigned as a temporary target number to be executed.

Next, the hypothesis creating unit 10 determines, for each cluster defined in the intra-system cluster table 6, the state of the existing hypotheses in the cluster defined in the intra-cluster hypothesis table 12 and the from the relationship between the wake of the current time t _k that was created, to create a new hypothesis of the current time in the cluster,
The hypothesis table 12 in the cluster and the trajectory table 13 in the hypothesis are redefined. The method of creating a hypothesis in this process is described in FIG.
Track and hypothesis creation unit 51 in the conventional target tracking device of FIG.
This is the same as the case of the processing in. That is, if the cluster for which the hypothesis is to be created is a cluster newly created this time, the hypothesis that a wake composed of one observation vector registered in this cluster is selected, and no wake is selected Are created, and these hypotheses are set as initial hypotheses in this cluster.
2. Set in the hypothesis in-tracing track table 13. If the cluster for which the hypothesis is to be created is an existing cluster, each existing hypothesis in the cluster is referred to as “tracks in the cluster so that multiple tracks belonging to the same hypothesis do not share the same observation vector. The hypothesis at the current time is created by extending to a plurality of hypotheses so as to satisfy the criterion of “selection”. These hypotheses are defined in the intra-cluster hypothesis table 12 and the hypothesis intra-track track table 13.

Next, the hypothesis reduction section 16 evaluates the likelihood of each hypothesis in the cluster indicated in the intra-cluster hypothesis status data group 11 for each cluster, and uses the result to reduce the number of hypotheses. Perform processing. Here, as a specific hypothesis reduction method, for example, the following method described in the conventional target tracking device can be applied. (1) Only the hypothesis with the highest evaluation value is left. (2) Discard hypotheses below the evaluation criteria. In addition, for example, the following method can be applied. (3) Only a predetermined number of hypotheses are left in descending order of the evaluation value. (4) Select hypotheses in descending order of evaluation value until the cumulative evaluation value reaches a predetermined criterion, and discard others.

Next, the cluster separating unit 17 determines whether or not each cluster can be separated into clusters by performing the above-described hypothesis reduction processing. To separate. Further, for each cluster after separation, the hypothesis in the cluster is reconstructed based on the hypothesis of the cluster before separation, and is redefined as the intra-cluster hypothesis status data group 11.

Next, the tracking establishment determination unit 18 determines that tracking has been established for a track included in the cluster only when a cluster composed of only one track is separated by the processing of the cluster separation unit 17. It is determined whether or not it is necessary to determine whether or not it has been performed, and if necessary, the determination is made. Here, when the track separated into clusters by only one track has a numerical value other than “0” indicating the temporary target number as the target number, the tracking establishment determination for this track is unnecessary. Conversely, if the track has "0" indicating a temporary target number,
It is necessary to determine the tracking establishment of this wake. The criterion for determining whether tracking has been established for this track is whether or not this track has been adopted by all hypotheses in the cluster before separation. In other words, if the wake is included in all the hypotheses in the cluster before separation,
It is determined that tracking has been established. When it is determined that the tracking has been established, the tracking establishment determining unit 18 sends the track to the target number assigning unit 19.

Next, the target number assigning section 19 refers to the in-system target number table 20 when the tracking establishment determining section 18 sends a newly established track, and the target tracking device 1 Any positive integer target number that has not been used yet, such as the smallest number, is selected and registered in the in-system target number table 20, and the target number is added to the transmitted track. Assignment and tracking establishment determination unit 18
Return to The tracking establishment determination unit 18 redefines the target number given to the intra-cluster track-target number table 15 for this track.

Next, the target number-corresponding trajectory determination unit 21 examines the in-cluster hypothesis status data group 11 of each cluster and finds out the hypothesis of the current time t _k in the cluster defined in the intra-cluster hypothesis table 12. From among them, the best one hypothesis, for example, the hypothesis with the highest evaluation value of the certainty calculated by the hypothesis reduction unit 16 is selected. Then, for the selected hypothesis, the information in the intra-cluster trajectory-observation vector table 14 is read for each trajectory in the hypothesis defined in the hypothesis trajectory table 13, and a tracking filter is applied to apply the tracking filter. Calculate the smoothed value of the position and speed of. These calculation results are attached to the target number defined in the track-in-cluster track-target number table 15 for each track and sent to the target display device 3.

As described above, according to the first embodiment, a positive integer unique target number is assigned to each track for which tracking has been established, and an existing track to which a target number has already been assigned is continued. in the wake of the current time t _k to create Te is configured to inherit the target number of the existing track, on the other hand, for a wake that have not yet established tracking is, is the target number of provisional "0" Is given to clarify the distinction from the track that has already been established, and furthermore, the track number transmitted to the target display device 3 is output with the target number of each track added. On the display device 3 side, it is possible to distinguish and display the track whose tracking has already been established and the track whose tracking has not been established, or to select and display only the track whose tracking has been established. , Display the target number of each track Doo is an effect that it can be obtained. Or a result, the operator already and knowing which one established the track is, for example, any of the previous time t _k-1 to the displayed track and track displayed in the current time t _k corresponding The effect of being able to know the temporal connection such as is obtained.

Embodiment 2 FIG. 2 is a block diagram showing an overall configuration of a target tracking device according to Embodiment 2 of the present invention. In FIG. 2, 1 is a target tracking device, 2 is a target observing device as a sensor for observing the position of the target in space and obtaining an observation vector, and 3 is displaying a wake on a display and displaying the state of the target being tracked. It is a target display device shown.

In the target tracking apparatus 1 shown in FIG. 2, reference numeral 4 denotes a gate calculation unit for calculating the expected area of the observation vector at the current time for all the tracks included in the existing cluster.
Current time from the entire observation vector of t _k, the observation vector selector for selecting observation vector contained in the gate of each track which is calculated by the gate determining section 4 is input from the target observation device 2 to the target tracking device 1, 6 Is an in-system cluster table showing the entire cluster in the target tracking device 1, and 7 is to integrate existing clusters from the relationship between the output of the observation vector selection unit 5 and the existing clusters shown in the in-system cluster table 6. , And a new cluster integration unit that creates a new cluster.

In the target tracking device 1 shown in FIG.
8 out of the observation vector at the present time t _k which is inputted in the observation vector table clusters define the observation vector belonging to each cluster for each cluster, 31 and cluster the observation vector table 8, the existing track and the goal of the cluster A number and a wake state flag are input, a new wake in the cluster corresponding to the observation vector at the current time t _k is created, and a target number and a wake state flag corresponding to the target number and the wake state flag of the wake are determined. The type track generation unit 10 calculates the current time t _k from the existing hypothesis in the cluster and the track generated at the current time t _k described above.
Is a hypothesis creator that creates a new hypothesis in the cluster corresponding to the observation vector.

Further, in the target tracking device 1 of FIG. 2, reference numeral 11 denotes an intra-cluster hypothesis status data group, 12 denotes an intra-cluster hypothesis table defining all the hypotheses existing in the existing cluster, and 13 denotes an existing cluster hypothesis table. In the hypothesis-in-track table, which defines all the tracks in the hypothesis for each hypothesis, 14 is the intra-cluster trajectory-observation that defines the observation vectors that constitute the wakes for all the trajectories in the existing cluster A vector table, 32, for all tracks in the existing cluster,
It is a track in a cluster-target number and a track state flag table in which a target number and a track state flag associated with a track are defined.

Further, in the target tracking device 1 of FIG. 2, 16 evaluates the likelihood of each hypothesis in the cluster, and based on the evaluation result, a hypothesis reduction unit for reducing the number of hypotheses in the cluster, and 17 A cluster separation unit that determines whether a cluster after hypothesis reduction is separable and reconstructs a hypothesis of each cluster after separation if separation is possible. A cluster 18 composed of only one track is separated. In this case, a tracking establishment determining unit that determines whether or not tracking of the track has been established, and a track state flag changing unit 33 that changes the track state flag for the track for which tracking has been established.

Further, in the target tracking device 1 of FIG. 2, reference numeral 19 denotes a target number assigning unit for assigning a new target number to the track, and reference numeral 20 denotes the entire target number used in the target tracking device 1. In the system target number table 34 shown, when a plurality of hypotheses exist in a cluster, one of the best hypotheses is selected from among the hypotheses to determine the number of targets and their trajectories, and is assigned to each wake. The target number and wake state flag-corresponding wake determination unit sends information of each wake including the information of the target number and the wake state flag to the target display device 3. In FIG. 2, the conventional target tracking device shown in FIG.
The same components as those of the target tracking device of the first embodiment shown in FIG. 1 are denoted by the same reference numerals.

Next, the operation will be described. First, time t
_kFrom the target observation device 2 to the target tracking device 1
When the gate calculation unit 4 inputs the_k-1Until
Track within a cluster of existing clusters input to
Current time for all tracks defined in vector table 14
t _kWhich is the expected region of the observation vector
Is calculated. Next, the observation vector selection unit 5
Find out which observation vector exists in the gate
The observation vector that can be correlated with each track.
You.

Next, for the observation vector that cannot be correlated with the track of any cluster, the cluster new integration unit 7 first creates a new cluster composed of the observation vector, and defines it in the in-system cluster table 6. I do. If a certain observation vector exists in the gates of a plurality of tracks belonging to different clusters, these clusters are integrated as one cluster and redefined in the intra-system cluster table 6. Further, the post-integration hypothesis situation data group 11 after integration is reconstructed from the intra-cluster hypothesis situation data group 11 of each cluster before integration. Further, the cluster newly integrating unit 7 writes, for each cluster defined in the intra-system cluster table 6, the entire observation vector that can be correlated with one or more wakes in the cluster into the intra-cluster observation vector table 8. .

Next, the target number and wake state flag correspondence type wake creating unit 31 is created at the previous time t _k−1 for each cluster defined in the in-system cluster table 6, and the in-cluster wake-observation vector is created. A track corresponding to the current time t _k is created from the existing track in the cluster defined in Table 14 and the observation vector group in the cluster at the current time t _k described in the intra-cluster observation vector table 8, and It is redefined in the coastal track-observation vector table 14. The method of creating the wake in this process is the same as the process of the wake and hypothesis creation unit 51 in the conventional target tracking device of FIG. 6 described above. That is, first, when there are existing tracks in the cluster, the current time t _{k is} calculated for each existing track by the following operation.
Create a wake for (1) For each observation vector existing in the gate of the existing track, a track at the current time is created by adding the observation vector to the existing track. (2) Assuming that an existing track has not been detected, a track at the current time is created while maintaining the existing track. further,
For each observation vector of the current time t _k in the cluster, to create a wake of the current time t _k according to the following procedure. (3) Create a new track starting from the observation vector in the cluster.

[0074] Incidentally, this cluster by cluster established integration unit 7, when a cluster is established to the current time t _k, by the operation of the above (3), within a cluster, one defined in the cluster Starting from the observation vector of 1
Tracks are created and defined in the intra-cluster track-observation vector table 14.

Next, the target number and wake state flag corresponding wake creating unit 31 calculates the target number and wake state flag of the existing wake in the cluster defined in the intra-cluster wake-target number and wake state flag table 32. based, each track of the target number of the current time t _k described above created, to determine the track state flag, cluster track - the target number and the track status flag table 3
Redefine to 2. Here, the target number is defined as an identification number indicating which track each track held in the apparatus is a track to be tracked to which of the targets determined to actually exist. For the target correspondence, for example, a positive integer is given. The wake state flag is defined as an identification flag indicating whether or not each wake is a track for which tracking has already been established. For example, when the wake is a track for which tracking has already been established, the flag “1” is set. If tracking is not established, a flag “0” is given.

[0076] method of determining the target number and track status flag for the current time t _k of the track by the target number and track state flag corresponding type track creation unit 31 is as follows. First,
Track of the current time t _k is when a track that is created from an existing track by the procedure described in creating a wake of the current time t _k (1) or (2), the target number and track state of this track Inherits the target number and the track state flag of the existing track as it is. Then, track the current time t _k is the case, a new track created in the above steps (3), a new target number to be acquired by sending this track to the target number giving unit 19, the track Target number. Further, the wake state flag of the new wake gives a flag “0” meaning that this wake is a wake for which tracking has not yet been established.

The target number assigning unit 19 refers to the in-system target number table 20 when the new wake is sent from the wake creating unit 31 corresponding to the target number and wake state flag, and in the target tracking device 1. An arbitrary positive integer target number that has not been used yet, for example, the smallest number is selected, registered in the in-system target number table 20, and the target number is assigned to the new track sent. Then, it returns to the target number and wake state flag correspondence type wake creating unit 31. The target-number-and-track-state-correspondence-type track generation unit 31 stores the target number assigned to the new track and the track state flag indicating that tracking has not been established in the track-in-cluster target-target number and track state flag table 32. Defined in

Next, the hypothesis creator 10 determines, for each cluster defined in the intra-system cluster table 6, the state of the existing hypotheses in the cluster defined in the intra-cluster hypothesis table 12 and the from the relationship between the wake of the current time t _k that you created, create a new hypothesis of the current time t _k in the cluster, the cluster within the hypothesis table 12, to redefine the hypothesis in the wake table 13. The method of creating a hypothesis in this processing is based on the wake and hypothesis creation unit 5 in the conventional target tracking device of FIG.
This is the same as the case of the processing in step 1. That is, if the cluster for which the hypothesis is to be created is a cluster newly created this time, the hypothesis that a wake composed of one observation vector registered in this cluster is selected, and no wake is selected Are created, and these hypotheses are set as initial hypotheses of this cluster, and the intra-cluster hypothesis table 1
2. Defined in Hypothesis Intra-Road Table 13. If the cluster for which the hypothesis is to be created is an existing cluster, each existing hypothesis in the cluster is referred to as `` tracks in the cluster so that the same observation vector is not shared among multiple tracks belonging to the same hypothesis. A hypothesis at the current time is created by expanding to a plurality of hypotheses so as to satisfy the criterion of “selection”, and these are defined in the intra-cluster hypothesis table 12 and the hypothesis intra-track table 13.

Next, the hypothesis reduction unit 16 evaluates the likelihood of each hypothesis in the cluster defined in the intra-cluster hypothesis status data group 11 for each cluster, and uses the result to reduce the number of hypotheses. Perform processing. Here, as a specific hypothesis reduction method, for example, the following method described in the conventional target tracking device can be applied. (1) Only the hypothesis with the highest evaluation value is left. (2) Discard hypotheses below the evaluation criteria. In addition, for example, the following method can be applied. (3) Only a predetermined number of hypotheses are left in descending order of the evaluation value. (4) Select hypotheses in descending order of evaluation value until the cumulative evaluation value reaches a predetermined criterion, and discard others.

Next, the cluster separating unit 17 determines whether or not each cluster can be separated into clusters by performing the above-described hypothesis reduction processing. To separate. Further, for each cluster after separation, the hypothesis in the cluster is reconstructed based on the hypothesis of the cluster before separation, and is redefined as the intra-cluster hypothesis status data group 11.

Next, the tracking establishment judging section 18 determines that the tracking of the track included in the cluster is performed only when the cluster composed of only one track is separated by the processing of the cluster separating section 17. It is checked whether it is necessary to determine whether or not it has been established. If so, the determination is made. Here, when the track state flag of a track separated into clusters by only one track is “1”, the tracking establishment determination for this track is unnecessary. Conversely, when the wake state flag of the wake is “0”, it is necessary to determine the tracking establishment of this wake. The criterion for determining whether tracking has been established for this track is whether or not this track has been adopted by all hypotheses in the cluster before separation. That is, it is determined that the tracking has been established only when the wake is included in all the hypotheses in the cluster before separation. When it is determined that the tracking has been established, the tracking establishment determining unit 18 sends the track to the track state flag changing unit 33.

The wake state flag changing unit 33 changes the wake state flag of this wake from "0" to "1" when a wake whose tracking has been newly established is sent from the tracking establishment determining unit 18. It returns to the establishment determining means 18. The tracking establishment determination means 18 determines that the track in the cluster
The changed wake state flag is redefined in the target number and wake state flag table 32.

Next, the target number and wake state flag-corresponding wake determination unit 34 examines the in-cluster hypothesis situation data group 11 of each cluster and finds the current time t in the cluster defined in the intra-cluster hypothesis table 12. _{From the k} hypotheses, the best one hypothesis, for example, the hypothesis with the highest evaluation value of the certainty calculated by the hypothesis reduction unit 16 is selected. The hypotheses described above selected for each track in the hypothesis defined hypothesis the track table 13, a cluster in the track - reads the information of the observation vector table 14, the current time t _k by applying a tracking filter Calculate the smoothed value of the target position and speed.
The calculated result is sent to the target display device 3 with the target number of each track and the track state flag defined in the track-target number and track state flag table 32 in the cluster.

As described above, according to the second embodiment, a positive integer unique target number is assigned to each track starting from an observation vector at each time, and an existing target number already assigned a target number is assigned. The current time t _k at which the wake is continuously created
Is configured to inherit the target number of the existing track, and according to the track state flag given for each track, the track is a track for which tracking has already been established or has not been established yet. Since it is configured to clearly distinguish whether the track is a track, and further configured to output the track number to be sent to the target display device 3 with the target number of each track and the track state flag. , On the target display device 3 side,
It is possible to display a track whose tracking has already been established and a track whose tracking has not been established separately, and to select and display only a track whose tracking has been established. Further, an effect is obtained that the target number of each track can be displayed regardless of whether the track is an established track or an unestablished track. As a result, the operator knows which track has already been established, and for example, which of the track displayed at the previous time t _k-1 and the track displayed at the current time t _k corresponds to The effect of being able to know the temporal connection such as is obtained.

Embodiment 3 The overall configuration of the target tracking device according to this embodiment may be either of FIG. 1 of the first embodiment and FIG. 2 of the second embodiment. FIG. 3 is a flowchart showing a processing method of the hypothesis reducing unit 16 according to Embodiment 3 of the present invention.

In FIG. 3, ST1 is an evaluation value calculating step for calculating an evaluation value of the likelihood of each hypothesis in the cluster, and ST2 is an observation vector for deleting an observation vector older than the latest N time from each track in the cluster. A deletion step, ST3, a track classification step of classifying the tracks in the cluster for each track having the same observation vector configuration and target number;
ST4 regards the tracks classified into the same group as the same track, and as a result, sets a plurality of hypotheses having the same track configuration as one.
ST5 is an evaluation value integration step of calculating the evaluation value of the hypothesis after integration from the evaluation value of the hypothesis group before integration, and ST6 is for each track having the same target number as the observation vector configuration. This is a track deletion step of selecting only one track from each group of tracks classified as described above and deleting the rest.

Next, the operation will be described. Step ST
In step 1, the likelihood of each hypothesis in the cluster is evaluated, and the evaluation value is calculated. In step ST2,
For each track in the cluster, it is checked whether the track has an observation vector at a time older than the latest N time, and if so, the observation vector at the old time is deleted from the track. In step ST3, all tracks in the cluster are regarded as tracks belonging to the same group, if all observation vectors constituting the tracks are the same and target numbers assigned to the tracks are the same. Classify into.

At step ST4, the tracks that belong to the same group are regarded as the same track by the classification of the tracks, so that a plurality of hypotheses in the cluster having the same track configuration are determined. Merge into one. Here, when integrating hypotheses into one, only one of a plurality of hypotheses having the same wake configuration may be left in the cluster, and the other hypotheses may be deleted. Step ST5
In, the sum of the evaluation values calculated in step ST1 for the plurality of hypotheses to be integrated by the operation in step ST4 is calculated, and this calculation result is set as the evaluation value of the post-integration hypothesis. In step ST6, for each track group obtained by the operation in step ST3, only one track in the group is left in the cluster, and other tracks are deleted.

As described above, according to the hypothesis reducing section 16 of the third embodiment, not only the condition that the configuration of the observation vectors for the past N times is the same, but also the target number is the same. Depending on the conditions, multiple tracks are identified and the number of hypotheses is reduced, so the number of hypotheses can be calculated without inconsistency such as identifying tracks created when trying to track different targets. The effect of reduction can be obtained.

Embodiment 4 The overall configuration of the target tracking device according to this embodiment may be either of FIG. 1 of the first embodiment and FIG. 2 of the second embodiment. FIG. 4 is a flowchart showing a processing method of the hypothesis reducing unit 16 according to Embodiment 4 of the present invention.

In FIG. 4, ST1 is an evaluation value calculating step for calculating an evaluation value of the likelihood of each hypothesis in the cluster, and ST11 is N times before each hypothesis in the cluster that became the parent that generated this hypothesis. A parent hypothesis number acquiring step for acquiring a hypothesis number, ST12 is a hypothesis classification step of classifying the hypotheses in the cluster for each hypothesis having the same hypothesis number N times before, and ST13 is a hypothesis classification step for each group of the classified hypotheses. ST14 is a hypothesis group selecting step of selecting a group of hypotheses having the largest total of the evaluation values, and ST15 is a step other than the selected hypothesis group. A hypothesis deleting step of deleting a hypothesis belonging to the group from the cluster, that ST16 is included in any of the hypotheses remaining in the cluster It is a wake deletion step of deleting a no wake from within the cluster.

Next, the operation will be described. Step ST
In step 1, the likelihood of each hypothesis in the cluster is evaluated, and the evaluation value is calculated. In step ST11, with respect to each hypothesis in the cluster, the number of the hypothesis that is the parent of the hypothesis that generated this hypothesis N times before is acquired.

Here, the processing in step ST11 is as follows.
It has the following meaning. That is, the conventional technique
As mentioned in the description of the technique, the target tracking device of the present invention
Is the previous time t_k-1Current time t from one hypothesis_kMultiple
Hypotheses are created and these current times t_kFrom each hypothesis,
In addition, multiple hypotheses for the next time are created,
Hypotheses are created like a tree. So the example
For example, the current time t_kHypotheses and the sources that create them
Time t before_k-1Can be regarded as a parent-child relationship
And for any hypothesis at the current time, the parent of that hypothesis
Time t before _k-1Hypothesis is always one.

[0094] In addition, this relationship if traced past N time amount, for any hypothesis of the current time t _k, it is possible to identify the N time before the parent of the hypothesis. Processing in step ST11, for each hypothesis of the current time t _k, As indicated above N
This means specifying the parent before the time. In order to realize the processing of step ST11, in the processing of the hypothesis creation unit 10 at each time, a newly created hypothesis is assigned a serial number starting from 1, for example, and the hypothesis creation source is assigned to each hypothesis. What is necessary is to store the information of the history of the above-mentioned hypothesis number for the past N times and store it in the cluster.

In step ST12, all the hypotheses in the cluster are classified into a plurality of hypothesis groups, assuming that the hypotheses having the same number of the hypotheses obtained N times before in step ST11 belong to the same group. .
In step ST13, the evaluation values of the hypotheses in the group are totaled for each group of the hypotheses obtained in step ST12. In step ST14, a group having the largest total value of the evaluation values obtained in step ST13 is selected from all the hypothesis groups. Step S
At T15, hypotheses belonging to groups other than the hypothesis group selected at step ST14 are deleted from within the cluster. In step ST16, tracks in the cluster that are not included in any of the hypotheses that have remained in the cluster by the processing in step ST15 are extracted, and these tracks are deleted from the cluster.

As described above, according to the hypothesis reducing unit 16 of the fourth embodiment, it is possible to create a hypothesis regarding the possibility of correlation for the past N times and to identify a plurality of wakes as the same. Since no operation is performed, it is possible to reduce the number of hypotheses without inconsistency such as confusing tracks having different target numbers.

Embodiment 5 FIG. The overall configuration of the target tracking device according to this embodiment may be either of FIG. 1 of the first embodiment and FIG. 2 of the second embodiment. FIG. 5 is a flowchart showing a processing method of the target number-corresponding track determination unit 21 or the target number / track state flag-corresponding track determination unit 34 according to Embodiment 5 of the present invention.

In FIG. 5, ST21 is a step of selecting a plurality of hypotheses for selecting a plurality of hypotheses having a high evaluation value among the hypotheses in each cluster, and ST22 is a track included in the plurality of selected hypotheses, which has already been tracked. calculated but confirmed track extracting all those which are established track, ST23 whereas track tracking already established that the extracted, the position of the current time t _k of the target by tracking filter, the smoothed value of the speed, etc. The tracking filter processing step, ST24, is a track evaluation value calculation step for evaluating the likelihood of the extracted track whose tracking has been established.

In FIG. 5, ST25 is a track classification step for classifying the extracted tracks for each track having the same target number, ST26 is a position for each track in the group for each group of the classified tracks, A smoothing data integration step of weighting and integrating smoothed values such as speed with the evaluation value of each track, and ST27 displays the integrated smoothed value corresponding to each target number together with the target number and the wake state flag on the target display device 3. This is a wake information output step to be sent.

Next, the operation will be described. Step ST
At 21, a plurality of hypotheses with high evaluation values are selected from the hypotheses in each cluster. At this time, for example, a method of selecting all hypotheses whose hypothesis evaluation value is equal to or higher than a predetermined criterion, or selecting a hypothesis having a high hypothesis evaluation value in order and selecting a plurality of hypotheses until the cumulative evaluation value reaches the predetermined criterion And the like. In step ST22, the above step S
In the wakes included in at least one of the hypotheses selected in T21, all the wakes for which tracking has already been established are extracted. In step ST23, in step ST22
Is smoothed using a Kalman filter or the like for each of the wakes extracted in step (1), and smoothed values such as the position and speed at the current time are calculated.

In step ST24, for each track extracted in step ST22, an evaluation value for evaluating the likelihood of the track is calculated. Here, the evaluation value of the wake can be defined by, for example, the sum of the evaluation values of the likelihood of the hypothesis including the wake. In step ST25, the entirety of the tracks extracted in step ST22 are classified for each track having the same target number. In step ST26, for each track group classified for each track having the same target number, the smoothed value calculated in step ST23 for each track in the group is calculated in step ST23.
By performing weight integration with the evaluation value of each track calculated in 24, a smooth value unique to each group is calculated. The smooth value unique to each group calculated here means a smooth value corresponding to each target number. In step ST27, corresponding to each target number, the target number and the smoothed value, or the target number, the wake state flag and the smoothed value are output to the target display device 3 as wake information.

As described above, according to the target number-corresponding track determining unit 21 or the target number and track state flag-corresponding track determining unit 34 of the fifth embodiment, it is not one of the best hypotheses in the cluster. In order to output to the target display device 3 a smoothed value obtained by weighting and integrating the influences of a plurality of higher-order hypotheses with the reliability of the hypothesis corresponding to each target number, the smooth position and speed of the wake displayed at each time are displayed. Is less fluctuated, and an effect that a stable display can be obtained is obtained.

[0103]

As described above, according to the present invention, the intra-cluster hypothesis table, the hypothesis intra-track table, the intra-cluster path-observation vector table, the intra-cluster path-target number table, and the observation vector selector From the relationship between the output of the existing cluster and the existing cluster, create a new cluster with observation vectors that can not be correlated with the wake of any cluster, and if a certain observation vector exists in the gate of multiple wakes belonging to different clusters A new cluster integration unit that integrates these clusters and reconstructs a hypothesis table within the cluster, a track table within the hypothesis, a track within the cluster-observation vector table and a track within the cluster-target number table, and observation of the current time within the cluster Vector and
A track corresponding to the current time is created from the tracks in the existing cluster defined in the intra-cluster track-observation vector table and redefined in the intra-cluster track-observation vector table, and the intra-cluster track-target number. Based on the target numbers defined in the table, the target number of the track corresponding to the created current time is determined, and the target number corresponding type track generating unit for redefining the track in the cluster-target number table, a hypothesis table in the cluster, From the status of the existing hypothesis in the cluster defined in the hypothesis internal track table and the track at the current time created by the target number corresponding type track generation unit,
Creates a new hypothesis in the cluster corresponding to the observation vector at the current time, and redefines the hypothesis table in the cluster and the hypothesis tracing table described above, and the hypothesis creation table in the cluster and the hypothesis tracing table The probability of each hypothesis in the cluster being evaluated is evaluated, and a hypothesis reduction unit that reduces the number of hypotheses in the cluster and whether the cluster after the hypothesis reduction is separable is determined based on the evaluation result. A cluster separation unit that reconstructs the hypothesis of each cluster after separation when separation is possible, and reconstructs an intra-cluster hypothesis table, a hypothesis intra-track table, a intra-cluster path-observation vector table and an intra-cluster path-target number table. When a cluster consisting of only one track is separated, it is determined whether or not tracking of this track has been established, and a new target number assigned to the track for which tracking has been established is assigned to the intra-cluster navigation. A tracking establishment determination unit defined in the target number table, and one hypothesis selected from the hypotheses defined in the intra-cluster hypothesis table, and the number and trajectory of the target at the current time are determined; By providing a target number corresponding type track determination unit that outputs the target numbers defined in the table, it is possible to know which track has already been established, and to check the track displayed at the previous time. There is an effect that it is possible to know a temporal connection such as which one of the wakes displayed at the current time corresponds.

According to the present invention, an intra-cluster hypothesis table,
Hypothesis intra-track table, intra-cluster track-observation vector table,
Tracks within the cluster-target number table and track state flag table;
From the relationship between the output of the observation vector selection unit and the existing cluster, a new cluster is created with observation vectors that cannot be correlated with the wake of any cluster, and a certain observation vector is placed in the gate of multiple wakes belonging to different clusters. A cluster new integration unit that integrates these clusters when present, and reconstructs a hypothesis table within the cluster, a track table within the hypothesis, a track-intra-cluster-observation vector table and a track within the cluster-target number and wake state flag table. From the observation vector at the current time in the cluster and the trajectory in the existing cluster defined in the intra-cluster trajectory-observation vector table,
Create a track corresponding to the current time, redefine the track in the cluster-observation vector table, and use the target number defined in the track in the cluster-target number table and the track state flag table to correspond to the created current time. The target number and track state flag of the track are determined and redefined in the track-in-cluster track-target number and track state flag table, and the target number assigned to the created new track and tracking are established. A wake status flag indicating that there is no wake
The target number and wake state flag responsive wake generator that is redefined in the target number table and wake state flag, the intra-cluster hypothesis table, the status of existing hypotheses in the cluster defined in the hypothesis trajectory table, the target number and wake Creates a new hypothesis in the cluster corresponding to the observation vector at the current time based on the trajectory at the current time created by the trajectory creation unit corresponding to the status flag, and creates a hypothesis table that redefines the hypothesis table in the cluster and the hypothesis trajectory table And a hypothesis reduction unit that evaluates the likelihood of each hypothesis in the cluster that is redefined in the cluster hypothesis table and the hypothesis intratraffic table, and reduces the number of hypotheses in the cluster based on the evaluation result. It is determined whether the hypothesized cluster is separable or not. If the cluster is separable, the hypothesis of each separated cluster is reconstructed, and the intra-cluster hypothesis table, hypothesis intra-track track, intra-cluster trajectory-observation vector And a cluster separation unit that reconstructs the intra-cluster track-target number and track state flag table, and determines whether tracking of this track has been established when a cluster consisting of only one track is separated. Is determined,
A tracking establishment determining unit that defines a track state flag changed for a track for which tracking has been established in a track-in-cluster target number and track state flag table, and one hypothesis from the hypotheses defined in the intracluster hypothesis table. Select the number of the target and the wake at the current time, and correspond to the target number and the wake state flag which are output together with the target number and the wake state flag defined in the track within the cluster-target number and wake state flag table. Equipped with a type track determination unit, it is possible to know which track has already been established, and to determine whether the track displayed at the previous time corresponds to the track displayed at the current time, etc. There is an effect that it is possible to know a natural connection.

According to the present invention, the hypothesis reducing unit calculates an evaluation value of the likelihood of each hypothesis in the cluster, deletes an observation vector older than the latest N time from each track in the cluster, The wakes are classified for each trajectory having the same observation vector configuration and target number, and trajectories classified into the same group are regarded as the same trajectory. Merge, calculate the hypothesis evaluation value after integration from the hypothesis group evaluation value before integration, and calculate only one wake from each wake group classified for each trajectory with the same observation vector configuration and target number. By selecting and deleting the rest, there is an effect that the number of hypotheses can be reduced without inconsistency such as equating wakes created in pursuit of different targets.

According to the present invention, the hypothesis reducing unit calculates an evaluation value of the likelihood of each hypothesis in the cluster, and, for each hypothesis in the cluster, the hypothesis N times before the parent that generated this hypothesis. Are obtained, the hypotheses in the cluster are classified for each hypothesis having the same hypothesis number N times before, and for each group of the classified hypotheses, the total value of the evaluation values of the hypotheses in the group is calculated, Select the group of the hypothesis with the highest evaluation value, delete the hypothesis belonging to the group other than the selected hypothesis group from the cluster, and keep track of the hypothesis that is not included in any of the remaining hypotheses in the cluster Is removed from the cluster, the number of hypotheses can be reduced without inconsistency such as confusing tracks with different target numbers.

According to the present invention, the target number-corresponding wake determination unit selects a plurality of hypotheses having a high evaluation value from among the hypotheses in each cluster, and calculates the wakes included in the selected plurality of hypotheses. All the tracks that have already been tracked are extracted, and for the extracted track that has already been tracked, the smoothing value of the target current time position, speed, etc. is calculated by tracking filter processing, and the extracted track has been tracked. Evaluate the likelihood of the track, classify the extracted tracks for each track with the same target number,
For each group of categorized tracks, smoothed values such as position and speed for each track in the group are weighted and integrated with the evaluation value of each track, and the integrated smoothed value corresponding to each target number is assigned to the target number. By outputting the information, the smooth position and speed of the wake displayed at each time are less likely to fluctuate, so that a stable display can be obtained.

According to the present invention, the target number and wake state flag-corresponding wake determination unit selects a plurality of hypotheses having a high evaluation value from among the hypotheses in each cluster, and includes them in the selected plurality of hypotheses. In the track to be tracked, all tracks whose tracking was already established were extracted, and for the extracted track whose tracking had been established, the smoothing value of the target current time position, speed, etc. was calculated by tracking filter processing and extracted. Evaluate the likelihood of a track that has been established for tracking, classify the extracted tracks into tracks with the same target number, and for each group of classified tracks, calculate the smoothed value of the position, speed, etc. for each track in the group. By integrating weights with the evaluation values of each track and outputting the integrated smoothed value corresponding to each target number along with the target number and the track state flag, the smooth position and speed of the track displayed at each time fluctuate. Do DOO less, there is an effect that it is possible to obtain a stable display.

[Brief description of the drawings]

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

FIG. 2 is a block diagram showing an overall configuration of a target tracking device according to a second embodiment of the present invention.

FIG. 3 is a flowchart illustrating a processing method of a hypothesis reducing unit according to Embodiment 3 of the present invention;

FIG. 4 is a flowchart illustrating a processing method of a hypothesis reducing unit according to a fourth embodiment of the present invention.

FIG. 5 is a flowchart showing a processing method of a target number-corresponding wake determination unit or a target number and wake state flag-corresponding wake determination unit according to Embodiment 5 of the present invention;

FIG. 6 is a block diagram showing an overall configuration of a conventional target tracking device.

FIG. 7 is a diagram illustrating an example of a situation in which it is difficult to determine a correlation between each track and an observation vector.

[Explanation of symbols]

1 target tracking device, 2 target observation device, 3 target display device, 4 gate calculation unit, 5 observation vector selection unit, 6
Cluster table in the system, 7 New cluster integration section, 8
Intra-cluster observation vector table, 9 target number corresponding wake generator, 10 hypothesis generator, 11 intra-cluster hypothesis status data group, 12 intra-cluster hypothesis table, 13 hypothesis inner trajectory table,
14 Tracks in Cluster-Observation Vector Table, 15 Tracks in Cluster-Target Number Table, 16 Hypothesis Reduction Unit, 17 Cluster Separation Unit, 18 Tracking Establishment Judgment Unit, 19 Target Number Assignment Unit, 20 Target Number Table in System, 21 Target Number Corresponding track determination unit, 31 Target number and track state flag Corresponding track generator, 32 Track within cluster-target number and track state flag table, 33 Track state flag changer, 34
A track determination unit corresponding to a target number and a track state flag.

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Yoshio Kosuge 2-3-2 Marunouchi, Chiyoda-ku, Tokyo F-term within Mitsubishi Electric Corporation (reference) 5J070 AC01 AE02 AE04 AH04 AH19 AH31 AH50 AK22 BB06 BB15 BB16 BG30

Claims (6)

[Claims] 1. An intra-cluster hypothesis table defining all the hypotheses in an existing cluster, a hypothesis in-track table defining all the tracks in the hypothesis for each hypothesis in the existing cluster, Intra-cluster trajectory-observation vector table that defines the observation vectors that make up the trajectory for all trajectories in the cluster, and target numbers associated with the trajectories for all the trajectories in the existing cluster A gate calculation unit that calculates the gate of each track, which is the expected area of the observation vector at the current time, for all the tracks defined in the track within the cluster-target number table and the track within the cluster-observation vector table described above. From the input observation vector at the current time, it is checked which observation vector is present in the gate of each wake calculated by the gate calculation unit, and a view that can be correlated with each wake is obtained. From the relation between the output of the observation vector selection unit and the existing cluster, an observation vector that cannot be correlated with the wake of any cluster is used to create a new cluster. In the case where a plurality of wakes belonging to different clusters exist in the gates of the wakes, these clusters are integrated and the above-mentioned intra-cluster hypothesis table, the above-mentioned hypothesis inside trajectory table, the above-mentioned intra-cluster wake-observation vector table and the above-mentioned intra-cluster wake-target A new cluster integration unit that reconstructs the number table, an observation vector at the current time in the cluster, and a track corresponding to the current time from the track in the existing cluster defined in the above-mentioned track in the cluster-observation vector table. Create and redefine the above intra-cluster track-observation vector table,
Based on the target number defined in the intra-cluster track-target number table, the target number of the track corresponding to the created current time is determined, and the target-number-corresponding track generation is re-defined in the intra-cluster track-target number table. From the status of existing hypotheses in the cluster defined in the above-mentioned intra-cluster hypothesis table, the above-mentioned intra-hypothesis trajectory table, and the trajectory at the current time created by the trajectory creation unit corresponding to the target number. Create a new hypothesis in the corresponding cluster,
The hypothesis table in the cluster, a hypothesis creating unit for redefining the hypothesis intra-track table, and the likelihood of each hypothesis in the cluster redefined in the hypothesis intra-cluster hypothesis table,
Based on this evaluation result, a hypothesis reduction unit that reduces the number of hypotheses in the cluster, determines whether the cluster after hypothesis reduction is separable, and if it can be separated, reconstructs the hypothesis of each cluster after separation. A cluster separation unit that reconstructs the intra-cluster hypothesis table, the hypothesis intra-track table, the intra-cluster trajectory-observation vector table, and the intra-cluster trajectory-target number table; In this case, it is determined whether or not tracking of this track has been established, and a tracking establishment determining unit that defines a new target number assigned to the track for which tracking has been established in the above-mentioned track / target number table in the cluster. A target number assigning unit for assigning a new target number to the track for which tracking has been established; and selecting one hypothesis from the hypotheses defined in the above-mentioned intra-cluster hypothesis table, and determining the number of targets at the current time. Sailing A target tracking device, comprising: a target number corresponding type track determining unit that determines a track and outputs the track together with a target number defined in the intra-cluster track-target number table. 2. An intra-cluster hypothesis table defining all the hypotheses in an existing cluster, a hypothesis in-track table defining all the wakes in the hypothesis for each hypothesis in the existing cluster, For all the tracks in the cluster, the intra-cluster track-observation vector table that defines the observation vectors that make up the track, and for all the tracks in the existing cluster, the target number associated with the track, The intra-cluster track-target number table and the track state flag table that define the track state flag indicating whether the tracking is an established track, and all the tracks defined in the above-mentioned intra-cluster track-observation vector table A gate calculation unit that calculates a gate of each wake that is an expected area of the observation vector at the current time; An observation vector selection unit that checks which observation vector exists in the gate of the observation vector and selects an observation vector that can be correlated with each wake; and, based on the relationship between the output of the observation vector selection unit and the existing cluster, A new cluster is created from observation vectors that cannot be correlated with the wakes of a cluster. If a certain observation vector exists in the gates of multiple wakes belonging to different clusters, these clusters are integrated, and the above-mentioned intra-cluster hypothesis table is created. A new cluster integration unit for reconstructing the hypothetical intra-track table, the intra-cluster track-observation vector table and the intra-cluster track-target number and wake state flag table, the observation vector at the current time in the cluster, and the cluster Domestic Track-Creates a track corresponding to the current time from tracks in existing clusters defined in the observation vector table. Then, the target number and the wake state of the wake corresponding to the current time created are redefined according to the trajectory in the cluster-observation vector table, and the target number defined in the trajectory in the cluster-target number table and the wake state flag table. The flag is determined, the track in the cluster-target number and the track state flag table are redefined, and the target number assigned to the created new track and the track whose tracking is not established are determined. A target number and wake state flag responsive tracing section for redefining the wake state flag to be shown in the track-target number table in the cluster and the wake state flag, and a target number assignment for assigning a new target number to a new track Section, the hypothesis table in the cluster, the status of the existing hypothesis in the cluster defined in the trajectory in the hypothesis table, and the trajectory creation section corresponding to the target number and the wake state flag. A hypothesis generator that creates a new hypothesis in the cluster corresponding to the observation vector at the current time from the track at the current time and redefines the hypothesis table in the cluster, the hypothesis track table, and the hypothesis table in the cluster , Assessing the likelihood of each hypothesis in the cluster that has been redefined in the hypothesis internal track table,
Based on this evaluation result, a hypothesis reduction unit that reduces the number of hypotheses in the cluster, determines whether the cluster after hypothesis reduction is separable, and if it can be separated, reconstructs the hypothesis of each cluster after separation. A cluster separation unit that reconstructs the intra-cluster hypothesis table, the hypothesis intra-track table, the intra-cluster trajectory-observation vector table, and the intra-cluster trajectory-target number and wake state flag table, and only one trajectory. When a cluster to be tracked is separated, it is determined whether or not tracking of this track has been established, and the track state flag changed for the track for which tracking has been established is described in the above-mentioned track within cluster-target number and track state flag table. A tracking establishment determination unit defined in (1), a track status flag changing unit that changes a track whose tracking has been established into a track status flag indicating that tracking has been established, One hypothesis is selected from the hypotheses defined in the hypothesis table, the number of targets and wakes at the current time are determined, and the intra-cluster trajectory-target number and the target number and wake status defined in the wake status flag table are set. A target tracking device comprising a target number and a track state flag-corresponding track determination unit that outputs a flag with a flag attached thereto. 3. A hypothesis reducing unit calculates an evaluation value of the likelihood of each hypothesis in the cluster, deletes an observation vector older than the latest N time from each wake in the cluster, and observes the wake in the cluster. As a result of classifying each track with the same vector composition and target number and considering the tracks classified into the same group as the same track, multiple hypotheses with the same track configuration are merged into a single hypothesis. The evaluation value of the subsequent hypothesis is calculated from the evaluation value of the hypothesis group before integration, and only one wake is selected from each wake group classified for each wake having the same observation vector configuration and target number. 3. The target tracking device according to claim 1, wherein the rest is deleted. 4. A hypothesis reducing unit calculates an evaluation value of the likelihood of each hypothesis in the cluster, and obtains, for each hypothesis in the cluster, the number of the hypothesis N times before the parent that generated this hypothesis. Then, the hypotheses in the cluster are classified for each hypothesis having the same number of the hypothesis before the N time, and the total value of the evaluation values of the hypotheses in the group is calculated for each group of the classified hypotheses. Is selected, the hypothesis that belongs to a group other than the selected hypothesis group is deleted from the cluster, and the wake that is not included in any of the remaining hypotheses in the cluster is selected. The target tracking device according to claim 1, wherein the target tracking device is deleted from within the cluster. 5. A trajectory determination unit corresponding to a target number selects a plurality of hypotheses having a high evaluation value of the hypotheses from the hypotheses in each cluster, and the trajectory included in the selected plurality of hypotheses has already been tracked. All of the established tracks are extracted, and smoothed values such as the current position and speed of the target are calculated for the extracted track that has been tracked by tracking filter processing. Of the extracted tracks are classified into tracks with the same target number, and for each group of the classified tracks, the smoothed value of the position, speed, etc. for each track in the group is calculated for each track. 2. The target tracking device according to claim 1, wherein the integrated values are weighted and integrated, and the integrated smoothed value corresponding to each target number is output together with the target number. 6. A trajectory determination unit corresponding to a target number and a wake state flag selects a plurality of hypotheses having a high evaluation value from among the hypotheses in each cluster, and selects a trajectory included in the plurality of selected hypotheses. Then, all the tracks whose tracking has already been established are extracted, and smoothing values such as the position and speed of the target at the current time are calculated by the tracking filter processing on the extracted track which has already been established, and the extracted tracking Evaluate the likelihood of the established track, classify the extracted tracks into tracks with the same target number, and for each group of the classified tracks, smooth values such as position and speed for each track in the group 3. The target tracking device according to claim 2, wherein the integrated values are weighted and integrated with the evaluation value of each track, and the integrated smoothed value corresponding to each target number is output together with the target number and the track state flag.