JP2011226924A - Tracking apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a tracking apparatus which avoids degradation of tracking performance due to conflicts of observation values between a large target and a small target, and can generate a track with high accuracy.SOLUTION: The tracking apparatus includes: a sensor 10; a position correlation processing section 20 for generating a position correlation matrix; a cell number correlation processing section 30 for generating a cell number correlation matrix; a correlation arbitration processing section 40 for generating a correlation arbitration result using the position correlation matrix and the cell number correlation matrix; a position clustering processing section 50 for generating an observation value cluster using the correlation arbitration result; an integration cell number observation value observation processing section 60 for observing an integration cell number observation value; an integration cell number update prediction processing section 70 for generating an integration cell number estimation value and an integration cell number prediction value after updating through the use of a Kalman filter; a position speed update prediction processing section 80 for generating a position speed prediction value and a position speed estimation value of the track through the use of the Kalman filter; a delay processing section 90; and a display processing section 100.

Description

  The present invention relates to a tracking device using a radio wave sensor or an infrared sensor.

  Conventionally, in tracking devices, the resolution of a sensor is higher than the size of the target, and when the target is observed as an image or multiple points, clustering of multiple observation values observed from the same target is performed. A technique for improving the accuracy of the target track by performing tracking using the center of gravity position of the cluster as the clustering result has been proposed (see, for example, Patent Document 1).

Japanese Patent No. 4116898

  In the prior art, since the observation values of each other are exchanged in an environment where the large target and the small target are close to each other, there is a problem that the tracking performance is deteriorated and the accuracy of the target wake is deteriorated.

  This invention has been made to solve the above-described problems, and avoids deterioration in tracking performance due to mutual observations even in an environment where a large target and a small target are close to each other, The purpose is to obtain a tracking device capable of generating a highly accurate track.

  The tracking device according to the present invention generates a position correlation matrix by associating a high-resolution sensor that obtains a plurality of detection positions from the same target with a track position predicted value and a position observation value using the plurality of detection positions. Position correlation processing unit, cell number correlation processing unit for generating cell number correlation matrix by associating cell number observation value and integrated cell number prediction value using a plurality of detection positions, position correlation matrix and cell number correlation Using a matrix, a correlation mediation processing unit that generates a correlation mediation result by preferentially allocating observation values for a track with a small number of integrated cells in consideration of the wake position and the number of integrated cells, and a correlation mediation result Therefore, the position clustering processing unit that generates the observation value cluster based on the integrated cell number prediction value and the observation cell cluster integrated cell number observation value are excessive. Integrated cell number observation value monitoring processing unit that monitors the integrated cell number observation value and a motion model in which the number of integrated cells in the observation value cluster is constant. An integrated cell number update prediction processing unit that generates a cell number prediction value, and a position / velocity update prediction processing unit that generates a position / velocity prediction value and a position / velocity estimation value of a wake using a Kalman filter, using the center of gravity of the observation value cluster as an observation value; A delay processing unit that delays the position / velocity predicted value of the wake by one sampling and inputs the feedback to the position correlation processing unit, the cell number correlation processing unit, the position clustering processing unit, and the integrated cell number observation value monitoring processing unit; A display process that displays to the operator the estimated position speed and the estimated number of integrated cells of the wake finally generated from the speed update prediction processing section. And a science department.

  According to the present invention, using the results of the position correlation matrix and the cell number correlation matrix, observation values are preferentially assigned to a wake with a small integrated cell number prediction value, and the observation value cluster obtained from the observation value clustering result is obtained. Compare the observed number of integrated cells with the predicted number of integrated cells, cluster the observed values equivalent to the predicted number of integrated cells, and calculate the estimated number of integrated cells. By updating the estimated number of integrated cells based on a constant motion model, even in an environment where a large target and a small target are close to each other, the deterioration of tracking performance due to the mutual observation of each other is avoided, and high accuracy is achieved. A wake can be generated.

It is a block diagram which shows the function structure of the tracking apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows an example of the observation result by the sensor in FIG. It is explanatory drawing which shows an example of the position matrix by the position correlation process part in FIG. It is explanatory drawing which shows an example of the cell number correlation matrix by the cell number correlation process part in FIG. It is explanatory drawing which shows an example of the cell number distance matrix used by the cell number correlation process part in FIG. It is explanatory drawing which shows the calculation process of the correlation mediation matrix by the correlation mediation process part in FIG. It is explanatory drawing which shows the allocation process of the observed value by the correlation mediation process part in FIG. It is explanatory drawing which shows the calculation process of the integrated cell number observation value by the position clustering process part in FIG. It is explanatory drawing which shows operation | movement of the integrated cell number observation value monitoring process part in FIG. 1, and a position clustering process part. It is explanatory drawing which shows operation | movement of the integrated cell number observation value monitoring process part in FIG. 1, and a position clustering process part. It is a block diagram which shows the function structure of the tracking apparatus which concerns on Embodiment 2 of this invention. It is explanatory drawing which shows operation | movement of the cell number correlation gate threshold value control process part in FIG. It is a block diagram which shows the function structure of the tracking apparatus which concerns on Embodiment 3 of this invention. It is explanatory drawing which shows operation | movement of the velocity vector correction process part in FIG. It is a block diagram which shows the function structure of the tracking apparatus which concerns on Embodiment 4 of this invention. It is a block diagram which shows the function structure of the tracking apparatus which concerns on Embodiment 5 of this invention.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a functional configuration of a tracking device according to Embodiment 1 of the present invention.
1, the tracking device includes a sensor 10, a position correlation processing unit 20, a cell number correlation processing unit 30, a correlation mediation processing unit 40, a position clustering processing unit 50, and an integrated cell number observation value monitoring processing unit 60. And an integrated cell number update prediction processing unit 70, a position / speed update prediction processing unit 80, and a delay processing unit 90.
The characteristic functions of the present invention relate to the cell number correlation processing unit 30, the correlation mediation processing unit 40, the integrated cell number update prediction processing unit 70, and the position / speed update prediction processing unit 80.

The sensor 10 includes a high-resolution radio wave sensor or an infrared sensor, and acquires a plurality of detection positions from the same target.
The position correlation processing unit 20 generates a position correlation matrix by associating the track position predicted value with the position observation value using a plurality of detection positions from the sensor 10.
The cell number correlation processing unit 30 generates a cell number correlation matrix by associating the cell number observation value with the integrated cell number prediction value using a plurality of detection positions from the sensor 10.

  The correlation mediation processing unit 40 uses the position correlation matrix from the position correlation processing unit 20 and the cell number correlation matrix from the cell number correlation processing unit 30, and considers the track position and the integrated cell number predicted value, and the number of cells. Correlation mediation results are generated by preferentially allocating observation values for tracks with a small number of integrated cells corresponding to small targets with small numbers.

The position clustering processing unit 50 uses the correlation mediation result from the correlation mediation processing unit 40 to generate an observation value cluster based on the integrated cell number prediction value.
The integrated cell number observation value monitoring processing unit 60 monitors the integrated cell number observation value so that the integrated cell number observation value of the observation value cluster from the position clustering processing unit 50 does not become excessive.

  The integrated cell number update prediction processing unit 70 calculates the integrated cell number estimated value and the updated integrated cell number predicted value by the Kalman filter based on the motion model in which the integrated cell number of the observation value cluster from the position clustering processing unit 50 is constant. Generate.

  The position / velocity update prediction processing unit 80 uses the center of gravity of the observation value cluster from the position clustering processing unit 50 as an observation value, and the integrated cell number estimation value from the integrated cell number update prediction processing unit 70 and the updated integrated cell number prediction. Using the values, the predicted position speed and the estimated position speed of the wake are generated by the Kalman filter.

The delay processing unit 90 delays the position / velocity prediction value of the wake from the position / velocity update prediction processing unit 80 by one sampling, and the position correlation processing unit 20, the cell number correlation processing unit 30, the position clustering processing unit 50, and A feedback input is made to the integrated cell number observation value monitoring processor 60.
The display processing unit 100 displays the position / velocity estimated value and the integrated cell number estimated value of the wake finally generated from the position / velocity update prediction processing unit 80 to the operator.

Next, a specific operation according to the first embodiment of the present invention shown in FIG. 1 will be described with reference to FIGS.
FIG. 2 is an explanatory view showing an example of the observation result by the sensor 10, and the detection position (detection cell) corresponding to the small target (wake T1) and the large target (wake T2) and the center of gravity of the detection position calculated by the sensor 10. Points 1 to 5 (black circles) and the true shape (one-dot chain line frame) of each target (T1, T2) are shown.

First, the sensor 10 including a radio wave sensor (such as a radar) or an infrared sensor acquires a plurality of detection positions as observation results from a small target (wake T1) and a large target (wake T2) as shown in FIG. .
In FIG. 2, each detection position (detection cell) corresponding to the small target and the large target (T 1, T 2) is determined by the resolution of the sensor 10.

In addition, the sensor 10 performs clustering based on a desired clustering method for a plurality of detection positions, and obtains centroid points 1 to 5 of each detection position including one or more cells.
In the example of FIG. 2, five points obtained by clustering a plurality of detection positions are the barycentric points 1 to 5 of the detection positions calculated by the sensor 10.

  Further, the sensor 10 performs positional correlation between the time when the detection position is acquired, the position of the centroid points 1 to 5 of the detection position, and the number of detection cells required to calculate the centroid points 1 to 5 as the observation results. The data is input to the processing unit 20 and the cell number correlation processing unit 30, respectively.

Here, among the output information of the sensor 10, the time when the detection position is obtained is “time”, the positions of the gravity points 1 to 5 of the detection position are “position observation values”, and it is necessary to calculate the gravity points 1 to 5. When the number of detected cells is defined as “cell number observation value”, the sensor 10 inputs the time, the position observation value, and the cell number observation value to the position correlation processing unit 20 and the cell number correlation processing unit 30. Become.
Note that the position observation value includes the barycentric point position and the observation error covariance matrix of the barycentric point position.

Next, the position correlation processing unit 20 performs correlation between the predicted target track position value and the observed value (position observed value), and calculates a position correlation matrix.
Here, input information from the sensor 10 (time, position observation value and cell number observation value) is collectively referred to as “observation value”.

  FIG. 3 is an explanatory diagram showing an example of the calculation result of the position correlation matrix. For the track T1 corresponding to the small target and the track T2 corresponding to the large target, the observed values Z1, Z2, Z3, Z4, Z5 (correlation) Yes (1) and no correlation (0)) are indicated by flag values.

In FIG. 3, the observed values Z1 to Z5 correspond to the centroid points 1 to 5 in FIG.
Further, the position correlation flag value “1” indicates that there is a correlation between the track position predicted value and the observed value (position observed value), and the position correlation flag value “0” is the track position predicted value. Indicates that there is no correlation with the observed value (position observed value).

Hereinafter, FIG. 3 will be described with reference to FIG.
In this case, the area surrounded by the broken line of the small target wake T1 in FIG. 2 is the position correlation gate of the wake T1, and the center of the position correlation gate of the wake T1 is the track position predicted value of the wake T1.
Similarly, the position of the large target wake T2 surrounded by the broken line is the position correlation gate of the wake T2, and the center of the position correlation gate of the wake T2 is the track position predicted value of the wake T2.

  For example, when the Kalman filter is applied as a tracking filter, the size of the position correlation gate of the wakes T1 and T2 is a value obtained from the error covariance matrix of the calculated state vector or a value set in advance by the user Determined by.

  Also, the position correlation flag values “0” and “1” indicate the following conditional expression (1) indicating that the difference between the track position predicted value and the position observation value is within the position correlation gate threshold value: It is determined by whether or not it is satisfied.

  ｜ Wake position predicted value−position observed value | ≦ position correlation gate threshold value (1)

  In conditional expression (1), “||” on the left side represents a distance (norm). This distance may be any norm. For example, the distance may be set as a Euclidean distance or Mahalanobis distance (Mahalanobis Distance).

When the conditional expression (1) is satisfied, since the observed value is in the position correlation gate, the position correlation processing unit 20 sets the position correlation flag to “1”.
On the other hand, when the conditional expression (1) is not satisfied, since the observed value is outside the position correlation gate, the position correlation processing unit 20 sets the position correlation flag to “0”.

The position correlation matrix in FIG. 3 represents the correspondence between the track position predicted value and the observed value.
For the small target (wake T1), the observed value Z1 (corresponding to the center of gravity 1 in FIG. 2) is in the position correlation gate of the wake T1, so the position correlation flag is “1”.
Further, since the observed value Z5 (corresponding to the centroid point 5 in FIG. 2) is not in the position correlation gate of the wake T1, the position correlation flag becomes “0”.

On the other hand, for the large target (wake T2), since all the observed values Z1 to Z5 (corresponding to the centroid points 1 to 5 in FIG. 2) are in the position correlation gate of the wake T2, the position correlation flag is set. All the values are “1”.
That is, the observation value Z1 is also included in the position correlation gate of the wake T2, and the value of the position correlation flag is “1”, so that the wake T1 and the wake T2 are in contact.

  However, the position correlation processing unit 20 acquires the track of the time k from the delay processing unit 90 based on the time k at which the observation value was obtained from the sensor 10, and the delay value and the observation value at the time k obtained from the sensor 10. A position correlation matrix is generated from the track at time k obtained from the unit 90 and input to the correlation mediation processing unit 40.

At this time, the wake at time k obtained from the delay processing unit 90 includes the time, the predicted value of position and velocity, the error covariance matrix predicted value of position and velocity, the predicted number of integrated cells, and the integrated cell. And an error covariance matrix of number prediction values.
Note that the integrated cell number prediction value and the error covariance matrix of the integrated cell number prediction value are calculated by the integrated cell number update prediction processing unit 70 described later.

The “number of integrated cells” is the number of cells that the wake should have.
For example, in the example of FIG. 2, the number of integrated cells that the wake T1 should have is the number of detection cells (one) required to generate the center of gravity point 1, and the number of integrated cells that the wake T2 should have is This is the number of detection cells (3 + 1 + 111 + 5 = 20) required to generate 2-5.

Therefore, prior to the calculation of the position correlation matrix (FIG. 3), the cell number correlation processing unit 30 correlates the target track integrated cell number predicted value with the observed value (cell number observed value), and the number of cells. A correlation matrix is calculated.
The time, the position observation value, and the cell number observation value from the sensor 10 are collectively referred to as an “observation value”.

  FIG. 4 is an explanatory diagram showing an example of the calculation result of the cell number correlation matrix. Similar to FIG. 3, the observed values Z1, Z2, and the wake T1 corresponding to the small target and the wake T2 corresponding to the large target are shown. Z3, Z4, Z5 ("1" with correlation and "0" without correlation) are indicated by flag values.

  In FIG. 4, the cell number correlation flag value “1” indicates that there is a correlation between the integrated cell number predicted value and the observed value (cell number observed value), and the cell number correlation flag value “0” This indicates that there is no correlation between the integrated cell number predicted value and the observed value (cell number observed value).

  The cell number correlation flag values “0” and “1” indicate that the difference between the integrated cell number prediction value and the cell number observation value is within the cell number correlation gate threshold value. , Is satisfied or not.

  | Integrated cell number prediction value−Cell number observation value | ≦ Cell number correlation gate threshold value (2)

In the conditional expression (2), “||” on the left side represents the distance (norm) as in the above-described conditional expression (1), and may be set by the Euclidean distance or the Mahalanobis distance.
When the conditional expression (2) is satisfied, since the observed value is in the cell number correlation gate, the cell number correlation processing unit 30 sets the cell number correlation flag to “1”.
On the other hand, when the conditional expression (2) is not satisfied, since the observed value is outside the cell number correlation gate, the cell number correlation processing unit 30 sets the cell number correlation flag to “0”.

Hereinafter, FIG. 4 will be described with reference to FIGS. 2 and 3.
In FIG. 2, the observed cell number at the observed value Z1 is “1”, the observed cell number at the observed value Z2 is “3”, the observed cell number at the observed value Z3 is “1”, and the observed cell number at the observed value Z4. The value is “11”, and the observed cell number at the observed value Z5 is “5”.

Here, the integrated cell number prediction value for the small target wake T1 is set to “1”, the integrated cell number prediction value for the large target wake T2 is set to “13”, and the cell number correlation gate threshold is set to “3”. And set.
At this time, if the norm is defined as “Euclidean distance” in the conditional expression (2), the value on the left side of the conditional expression (2) regarding the observed value Z1 is as follows.

  | The predicted number of integrated cells in wake T1−the observed number of cells in Z1 | = | 1-1 | = 0

That is, since the cell number correlation gate threshold value is “3” or less and the conditional expression (2) is satisfied, the value of the cell number correlation flag in the observation value Z1 is “1”.
Similarly, a cell in which the values of the left side of the conditional expression (2) are combined and expressed in a table format by combining the track T1 and the observed values Z1 to Z5, and combining the track T2 and the observed values Z1 to Z5. A few distance matrix can be obtained.

FIG. 5 is an explanatory diagram illustrating an example of the calculation result of the cell number distance matrix, and the table values of the tracks T1 and T2 indicate the value (distance) on the left side of the conditional expression (2).
That is, when the value of each component of the cell number distance matrix (FIG. 5) is equal to or less than the cell number correlation gate threshold “3”, the result of summarizing the correlation flags (“1”, “0”) is This is the cell number correlation matrix (FIG. 4).

  Therefore, the cell number correlation processing unit 30 acquires the track of the time k from the delay processing unit 90 based on the time k when the observation value is obtained from the sensor 10, and the observation value at the time k obtained from the sensor 10; A cell number correlation matrix (FIG. 4) is generated from the track at time k obtained from the delay processing unit 90 and input to the correlation arbitration processing unit 40.

  The correlation mediation processing unit 40 includes a position correlation matrix at time k obtained from the position correlation processing unit 20 (FIG. 3), and a cell number correlation matrix at time k obtained from the cell number correlation processing unit 30 (FIG. 4). Is used to assign a track and an observation value based on the size of the integrated cell number prediction value and the position observation value to calculate a correlation mediation matrix.

FIG. 6 is an explanatory diagram showing the operation of the correlation mediation processing unit 40, and shows the calculation process of the correlation mediation matrix M3.
In FIG. 6, the leftmost position correlation matrix M1 corresponds to FIG. 3, and the center cell number correlation matrix M2 corresponds to FIG.
“×” represents an AND operation of the position correlation matrix M1 and the cell number correlation matrix M2 (multiplying the same row and the same row components of M1 and M2), and “=” represents the AND operation result most. This represents the right correlation mediation matrix M3 (correlation mediation result).

  As shown in FIG. 6, the correlation mediation processing unit 40 calculates a correlation mediation matrix M3 (correlation mediation result) that is an AND operation result of the position correlation matrix M1 and the cell number correlation matrix M2, and then generates a correlation mediation result. Based on the wake with a small integrated cell number prediction value, observation values are assigned sequentially.

  FIG. 7 is an explanatory diagram showing the operation of the correlation mediation processing unit 40. The track T1 (corresponding to a small target) with a small integrated cell number predicted value and the track T2 (large target) with an integrated track cell number predicted value larger than the track T1. Equivalent)), and shows the observation value assignment process.

  In FIG. 7, the correlation mediation processing unit 40 firstly calculates a position correlation for the wake T1 from the correlation mediation matrix M3 from a set (dashed line frame: flag value “1”) in which the wake T1 and the observed value Z1 are associated with each other. A matrix M1 (T1) is generated.

  Further, the correlation mediation processing unit 40 removes the components used in the position correlation matrix M1 (T1) for the wake T1 (see the one-dot chain line) from the position correlation matrix M1, and the position correlation matrix M1 for the wake T2 (T2) is generated and input to the position clustering processing unit 50 as an observed value at time k.

  Here, the case where the target number (the number of wakes) is “2” is taken as an example, but the result of the correlation mediation matrix M3 is also obtained for an arbitrary number of wakes where the target number is “3” or more as shown in FIG. The position correlation matrix for each wake can be generated by performing the observation value assignment process in the order of the wake having the smallest integrated cell number prediction value.

  Next, the position clustering processing unit 50 performs clustering of observation values based on the position, obtains the clustering result of the observation values as an “observation cluster”, calculates the sum of the cell number observations in the observation cluster, and finally The total sum value is calculated as the “integrated cell number observation value”.

FIG. 8 is an explanatory diagram showing the operation of the position clustering processing unit 50, and shows an example of a tree diagram (tree) for performing the integrated cell number observation value calculation process.
The numbers “11”, “1”, “3”, “5” at the bottom in FIG. 8 are the observed values Z4, Z3, Z2, Z5 (centroid points 4, 3, 2, 5 in FIG. 2). Corresponds to the observed cell number.

In FIG. 8, the position clustering processing unit 50 generates an observation value cluster (Z3, Z4) from the observation value Z3 and the observation value Z4, calculates the sum “12” of the cell number observation values of both, and similarly, An observation value cluster (Z2, Z5) is generated from the observation value Z2 and the observation value Z5, and the sum “8” of the cell number observation values of both is calculated.
Further, an observation value cluster (Z2, Z3, Z4, Z5) is generated from the observation value cluster (Z3, Z4) and the observation value cluster (Z2, Z5), and the sum “20” of the cell number observation values of both is generated. Ask.

  Furthermore, when the reclustering flag from the integrated cell number observation value monitoring processing unit 60 is “0”, the position clustering processing unit 50 converts the observation value cluster into the integrated cell number update prediction processing unit 70 and the position / speed update prediction processing. Each is input to the unit 80.

  In addition, when the value of the reclustering flag generated from the integrated cell number observation value monitoring processing unit 60 is “1”, it represents an instruction to perform clustering again, and when the value of the reclustering flag is “0”. Represents a command not to perform clustering again.

The observation value cluster input from the position clustering processing unit 50 to the integrated cell number update prediction processing unit 70 and the position / velocity update prediction processing unit 80 is the observation value information obtained from the sensor 10 (for example, time, position observation value). In addition to (cell number observation value), information on the center of gravity of the observation value cluster and the integrated cell number observation value is added.
Further, the position clustering processing unit 50 inputs an observation value cluster to which information on the integrated cell number observation value is added to the integrated cell number observation value monitoring processing unit 60.

  Next, the integrated cell number observation value monitoring processing unit 60 has a relationship between the integrated cell number observation value in the observation value cluster obtained from the position clustering processing unit 50 and the integrated cell number prediction value obtained from the delay processing unit 90. Determines whether or not the following conditional expression (3) is satisfied.

｜ Integrated cell count observation value −Integrated cell count prediction value | ≦ Cell count clustering threshold
... (3)

  The integrated cell number observation value monitoring processing unit 60 sets the reclustering flag to “1” when the conditional expression (3) is not satisfied, and sets the reclustering flag when the conditional expression (3) is satisfied. Set to “0”.

Hereinafter, the flow of operations of the integrated cell number observation value monitoring processor 60 and the position clustering processor 50 will be described with reference to FIGS. 9 and 10.
FIG. 9 is an explanatory diagram showing how re-clustering is performed, and shows how re-clustering is performed by comparing the predicted number of integrated cells with the observed integrated cell number based on the clustering tree diagram of FIG. Yes.

FIG. 9 shows a case where the integrated cell number observation value is “20”, the integrated cell number prediction value is “16”, and the integrated cell number observation value “20” is compared with the integrated cell number prediction value “16”. ing.
Here, when the cell number clustering threshold is set to “2” and the left side of the conditional expression (3) is calculated, | 20−16 | = 4, and the right side of the conditional expression (3) “cell number clustering is performed. Since the threshold value is larger than “2”, the conditional expression (3) is not satisfied.

Therefore, in this case, the integrated cell number observation value monitoring processing unit 60 inputs the reclustering flag “1” to the position clustering processing unit 50.
In response to this, the position clustering processing unit 50 removes the observation value having the minimum cell number observation value that satisfies the conditional expression (3) from the position clustering in order to execute the clustering again.

In the case of FIG. 9, the observation value Z2 having the cell number observation value “3” corresponds to the observation value removed to satisfy the conditional expression (3).
Therefore, as shown in FIG. 10, the position clustering processing unit 50 performs clustering again except for the observation value Z2 (cell number observation value “3”), and calculates the observation value cluster that is the calculation result as the number of integrated cells. Input to the observation value monitoring processing unit 60.
Since the integrated cell number observation value of the observation value cluster calculated again is “17” as shown in FIG. 10, when the left side of the conditional expression (3) is calculated, | 17−16 | = 1.

Therefore, in this case, since the right side of the conditional expression (3) is “2” or less, the conditional expression (3) is satisfied. Therefore, the integrated cell number observation value monitoring processing unit 60 sets the reclustering flag “0” to position clustering. Input to the processing unit 50.
In response to this, the position clustering processing unit 50 inputs the clustering result (observed value cluster at time k) to the integrated cell number update prediction processing unit 70 and the position / velocity update prediction processing unit 80.

  The integrated cell number update prediction processing unit 70 uses the integrated cell number observation value “17” in the observation value cluster at time k obtained from the position clustering processing unit 50 based on the motion model in which the target cell number is constant. An integrated cell number estimate at time k and an error covariance matrix of the integrated cell number estimate are generated by the Kalman filter.

  Further, the integrated cell number update prediction processing unit 70 calculates the integrated cell number predicted value at time k + 1 and the error covariance matrix of the integrated cell number predicted value based on the motion model in which the target cell number is constant, and the time k + 1. The position / velocity update prediction processing unit 80 calculates the calculation result related to the integrated cell number prediction value and the calculation result related to the integrated cell number estimation value at time k (the integrated cell number estimated value and the error covariance matrix of the integrated cell number estimated value). To enter.

The position / velocity update prediction processing unit 80 uses the Kalman filter to determine the centroid position of the observation value cluster at time k obtained from the position clustering processing unit 50 as a target observation value, and An error covariance matrix of the estimated position velocity is calculated.
Further, the position / velocity update prediction processing unit 80 calculates a predicted value of the wake position speed at time k + 1 and an error covariance matrix of the predicted value of the wake position speed.

  Further, the position / velocity update prediction processing unit 80 calculates the calculation result (the integrated cell number predicted value, the error covariance matrix of the integrated cell number predicted value) at the time k + 1 and the prediction of the wake position speed at the time k + 1. The calculation results relating to the values (the predicted value of the wake position speed and the error covariance matrix of the predicted value of the wake position speed) are collectively input to the delay processing unit 90 as the “wake predicted value” at time k + 1.

  Further, the position / velocity update prediction processing unit 80 calculates the calculation result (the integrated cell number estimation value, the error covariance matrix of the integrated cell number estimation value) at the time k and the estimation of the wake position speed at the time k. The calculation results regarding the values (estimated value of wake position speed, error covariance matrix of estimated value of wake position speed) are collectively input to the display processing unit 100 as the “wake estimated value” at time k.

  The delay processing unit 90 delays the “track prediction value at time k + 1” obtained from the position / velocity update prediction processing unit 80 by one sampling to generate a “track prediction value at time k”, and the position correlation processing unit 20 The cell number correlation processing unit 30, the position clustering processing unit 50, and the integrated cell number observation value monitoring processing unit 60 are fed back.

  The display processing unit 100 drives the display on the basis of the wake estimated value at time k obtained from the position / velocity update prediction processing unit 80 to display the wake position and speed vector to the operator, and the integrated cell of the wake. The number estimate is displayed as a numerical value at the same time.

  As described above, the tracking device according to Embodiment 1 (FIGS. 1 to 10) of the present invention uses the high-resolution sensor 10 that obtains a plurality of detection positions from the same target and the wake position using the plurality of detection positions. By associating the predicted value with the position observation value, the position correlation processing unit 20 that generates a position correlation matrix, and by associating the cell number observation value with the integrated cell number prediction value using a plurality of detection positions, Using the cell number correlation processing unit 30 for generating the correlation matrix, the position correlation matrix and the cell number correlation matrix, and considering the track position and the integrated cell number prediction value, the observation value is prioritized for the wake T1 having a small integrated cell number. A correlation arbitration processing unit 40 that generates a correlation mediation result by assigning to a position cluster that generates an observation value cluster based on the predicted number of integrated cells using the correlation mediation result. Stirling processing unit 50, integrated cell number observation value monitoring processing unit 60 for monitoring the integrated cell number observation value, and the integrated cell number of the observation value cluster so that the observation value cluster integrated cell number observation value does not become excessive. Based on a constant motion model, an integrated cell number update prediction processing unit 70 that generates an integrated cell number estimation value and an updated integrated cell number prediction value by a Kalman filter, and a centroid point of an observation value cluster as an observation value, a Kalman filter A position / velocity prediction value and a position / velocity update prediction processing unit 80 for generating a position / velocity prediction value, and a position / correlation processing unit 20 and a cell number correlation processing unit 30 by delaying the position / velocity prediction value of the wake by one sampling. A delay processing unit 90 that performs feedback input to the position clustering processing unit 50 and the integrated cell number observation value monitoring processing unit 60; A display processing unit 100 that displays the estimated position / velocity value of the wake and the integrated cell number estimation value finally generated from the measurement processing unit 80 to the operator.

  In this way, using the results of the position correlation matrix and the cell number correlation matrix, observation values are preferentially assigned to a track with a small integrated cell number prediction value (a small target track with a small number of cells), and the observation value clustering result Compare the observed number of integrated cells in the observed value cluster obtained from, and the predicted number of integrated cells, cluster the observed values equivalent to the predicted number of integrated cells, and calculate the observed value when calculating the estimated number of integrated cells. Based on a motion model in which the number of integrated cells in the cluster is constant, the estimated number of integrated cells is updated. Thus, a highly accurate wake can be generated.

Embodiment 2. FIG.
Although not particularly mentioned in the first embodiment (FIG. 1), as shown in FIG. 11, the cell number correlation gate threshold control processing unit 110 may be provided in the cell number correlation processing unit 30.
FIG. 11 is a block diagram showing a functional configuration of the tracking device according to Embodiment 2 of the present invention. The same components as those described above (see FIG. 1) are denoted by the same reference numerals as those described above, and detailed description thereof is omitted. .

In FIG. 11, the cell number correlation processing unit 30 is provided with a cell number correlation gate threshold value control processing unit 110 for controlling the cell number correlation gate threshold value.
The cell number correlation gate threshold value control processing unit 110 determines the cell number correlation gate threshold value of the conditional expression (2) according to the size of the integrated cell number prediction value obtained from the cell number correlation processing unit 30. Variable setting in stages.

  As a result, the cell number correlation processing unit 30 uses the cell number correlation gate threshold set by the cell number correlation gate threshold control processing unit 110 to calculate based on the conditional expression (2) as described above. And the cell number correlation matrix (FIG. 4) is calculated.

  FIG. 12 is an explanatory diagram showing the operation of the cell number correlation gate threshold value control processing unit 110. The cell number correlation gate threshold value (hereinafter simply referred to as “threshold value”) according to the integrated cell number prediction values A, B, and C. (Also referred to as “value”) ThA, ThB, and ThC are set.

In FIG. 12, the cell number correlation processing unit 30 variably sets the threshold value in accordance with the area of the integrated cell number prediction value as follows.
In the region of “0 <predicted number of integrated cells ≦ A”, the threshold ThA,
In the region of “A <predicted number of integrated cells ≦ B”, the threshold ThB,
In a region where “B <predicted number of integrated cells ≦ C”, the threshold value ThC.

Thus, the cell number correlation gate threshold is set in three stages according to the size of the integrated cell number prediction value.
In the example of FIG. 12, to simplify the description, a case where the cell number correlation gate threshold value is divided into three stages is shown. However, any N stages (N is a value) according to a user request. It goes without saying that it can be set to (natural number).

  Here, the reason why the threshold value is divided into N stages for each region of the integrated cell number prediction value is that, in the case of a large target (track T2 in FIG. 2), the left side of the conditional expression (2) (| the number of integrated cells) Since the predicted value−the observed number of cells |) is expected to be larger than the case of the small target (track T1 in FIG. 2), the cell number correlation gate threshold is also smaller in the case of the large target. This is because it is necessary to set a larger value than in the case of the target, and it is considered that the optimum value of the cell number correlation gate threshold differs depending on the large target and the small target.

  As described above, the tracking device according to Embodiment 2 (FIGS. 11 and 12) of the present invention includes the cell number correlation gate threshold value control processing unit 110 provided in the cell number correlation processing unit 30. The cell number correlation gate threshold value control processing unit 110 controls the cell number correlation gate threshold value in the cell number correlation processing unit according to the size of the integrated cell number prediction value.

In this way, by dividing the cell number correlation gate threshold value into a plurality of stages according to the size of the integrated cell number prediction value, the cell number correlation gate threshold value is set according to the large target and the small target. It becomes possible to set appropriately to some extent.
In addition, as a result of appropriately determining the cell number correlation gate threshold value, since the relationship between the observation values of the large target and the small target is further reduced, the accuracy of the wake can be further improved.

Embodiment 3 FIG.
Although not particularly mentioned in the first and second embodiments (FIGS. 1 and 11), as shown in FIG. 13, a velocity vector correction process is performed between the position / velocity update prediction processing unit 80 and the delay processing unit 90. The unit 120 may be provided.
FIG. 13 is a block diagram showing a functional configuration of the tracking device according to the third embodiment of the present invention. Components similar to those described above (see FIG. 1) are denoted by the same reference numerals as those described above, and detailed description thereof is omitted. .

In FIG. 13, a velocity vector correction processing unit 120 is inserted between the position / velocity update prediction processing unit 80 and the delay processing unit 90.
The speed vector correction processing unit 120 recalculates and corrects the direction of the speed vector of the wake so as to coincide with the center line of the target passage traveling direction.

FIG. 14 is an explanatory diagram showing the operation of the speed vector correction processing unit 120, showing the process of correcting the direction of the speed vector V1 of the wake to the speed vector V2.
In FIG. 14, a path 200 having a center line 210, a wake position vector S0 (black circle) in the correction execution region R on the path 200, a wake speed vector V1 (dashed arrow) before correction, and the center line A wake speed vector V2 (double arrow) after correction is shown in the direction of 210.

For example, in the case of a large target that covers the passage 200, it often proceeds in the direction of the passage 200, but the wake velocity vector V1 is caused by the absence of the detection cell, as shown in FIG. The situation can be different from the direction.
Under such circumstances, in order to determine whether or not it is a large target, the estimated number of integrated cells in the wake is compared with a preset threshold value (a reference value for determining a large target) It is desirable to determine that the target is a large target when the estimated number of integrated cells in the wake is equal to or greater than the threshold value, and to correct the velocity vector V1 of the passage in the direction of the center line 210 of the passage.

  The velocity vector correction processing unit 120 includes a wake position vector S0 (estimated wake position value obtained from the position / velocity update prediction processing unit 80) within the correction execution region R in the passage 200, and an integrated wake cell. When the number estimate value is equal to or greater than a preset threshold value, the magnitude of the track speed vector V1 before correction (the track speed estimate value obtained from the position / speed update prediction processing unit 80) is stored. The direction of the velocity vector V1 is corrected so as to match the direction of the center line 210 as in the corrected wake velocity vector V2 (corrected wake velocity estimated value).

That is, the speed vector correction processing unit 120 performs a wake speed estimation process by correcting the wake speed vector V1 from the uncorrected wake speed vector V1 so as to match the direction of the center line 210 of the passage 200 in the traveling direction. The corrected wake speed estimated value is calculated.
Then, the corrected wake speed estimated value is overwritten on the original wake position estimated speed value to calculate a new corrected wake position estimated speed value.

Further, the speed vector correction processing unit 120 performs position / speed prediction using a motion model in which a corrected track position / speed estimated value is assumed in advance, and calculates a corrected track position / speed predicted value.
Then, as the “track estimated value at time k” to be input to the display processing unit 100, the corrected track position / speed estimated value instead of the “track position / speed estimated value at time k” which is a component of the track estimated value at time k. Is generated.

  Further, the speed vector correction processing unit 120 replaces “the track position / speed prediction value at time k + 1”, which is a component of the track prediction value at time k + 1, as the “track prediction value at time k + 1” input to the delay processing unit 90. Then, a corrected wake position speed prediction value is generated.

  As described above, the tracking device according to Embodiment 3 (FIGS. 13 and 14) of the present invention includes the velocity vector correction processing unit 120 inserted between the position / velocity update prediction processing unit 80 and the delay processing unit 90. The speed vector correction processing unit 120 recalculates and corrects the direction of the wake speed vector V1 so as to coincide with the center line 210 in the target passage traveling direction, and the corrected wake speed vector. V2 is calculated.

That is, even if the target is a large target covering the passage 200 and the wake speed vector V1 is inconsistent with the direction of the passage 200, it is large when the estimated number of integrated cells in the wake is greater than or equal to the threshold value. The target speed is determined and the speed vector V1 of the wake is corrected in the direction of the center line 210 of the passage 200.
As a result, it is possible to improve the speed accuracy of the wake.

Embodiment 4 FIG.
Although not particularly mentioned in the first to third embodiments (FIGS. 1, 11, and 13), a clustering adjustment processing unit 130 may be provided in the position clustering processing unit 50 as shown in FIG.

FIG. 15 is a block diagram showing a functional configuration of a tracking device according to Embodiment 4 of the present invention. The same components as those described above (see FIG. 1) are denoted by the same reference numerals as those described above, and detailed description thereof is omitted. .
In FIG. 15, the position clustering processing unit 50 is provided with a clustering adjustment processing unit 130.

  When generating the observation value cluster, the clustering adjustment processing unit 130 generates a cell in the observation value cluster when the observation value cluster has an excessive number of integrated cell observations and there are a plurality of observation values with the smallest number of cells. The observation with the smallest number of cells that is farthest from the observation with the largest number is removed from the observation cluster.

  Specifically, when the position clustering processing unit 50 generates the observation value cluster again in response to the reclustering flag “1” from the integrated cell number observation value monitoring processing unit 60, the clustering adjustment processing unit 130 generates the observation value cluster again. When there are a plurality of minimum cell number observation values that satisfy the conditional expression (3), an observation value cluster is acquired from the position clustering processing unit 50.

  Subsequently, the clustering adjustment processing unit 130 removes, from the observation value cluster, the smallest cell number observation value (farthest from the largest observation value) from the observation value cluster in the observation value cluster acquired from the position clustering processing unit 50. The subsequent observation value cluster is input again to the position clustering processing unit 50.

As described above, according to the tracking device according to Embodiment 4 (FIG. 15) of the present invention, the clustering adjustment processing unit 130 is provided in the position clustering processing unit 50, and the minimum cell number observation value when performing position clustering Is removed from the position clustering process as an observation value farthest from the observation value having the largest number of cells, so that the processing of the position clustering processing unit 50 can be continued.
In addition, since the observation value farthest from the observation value having the largest number of cells is likely to be an observation value from another target, it is possible to reduce false tracking to the other target.

Embodiment 5 FIG.
In Embodiments 1 to 4 (FIGS. 1, 11, 13, and 15), no particular mention was made, but as shown in FIG. 16, between the delay processing unit 90 and the position clustering processing unit 50, A clustering start determination processing unit 140 may be provided to remove the unstable integrated cell number prediction value in the initial time zone.

FIG. 16 is a block diagram showing a functional configuration of a tracking device according to Embodiment 5 of the present invention. The same components as those described above (see FIG. 1) are denoted by the same reference numerals as those described above, and detailed description thereof is omitted. .
In FIG. 16, a clustering start determination processing unit 140 is inserted between the delay processing unit 90 and the position clustering processing unit 50.

  The clustering start determination processing unit 140 determines the position clustering start time when the track prediction value is obtained exceeding a preset sampling threshold, and starts the position clustering based on the integrated cell number prediction value. The command is input to the position clustering processing unit 50.

  That is, the clustering start determination processing unit 140 receives position clustering based on the integrated cell number prediction value by the position clustering processing unit 50 when the track prediction value for the same track number obtained from the delay processing unit 90 is input more than M samples. The process is started, and the unstable integrated cell number prediction value in the initial time zone is removed.

  As described above, according to the tracking device according to Embodiment 5 (FIG. 16) of the present invention, the clustering start determination processing unit 140 is provided between the delay processing unit 90 and the position clustering processing unit 50, and the wake prediction value When the acquisition number of cells exceeds the sampling threshold, it is determined that the position clustering start time is based on the integrated cell number prediction value, and the unstable integrated cell number prediction value in the initial time zone is removed from the target of the position clustering process. With this configuration, it is possible to stabilize the wake accuracy in the initial time zone.

Embodiment 6 FIG.
In the fifth embodiment (FIG. 16), the clustering start determination processing unit 140 determines the position clustering start time when the wake prediction value is obtained exceeding the sampling threshold value. If the absolute value of the difference between the average value of the integrated cell number prediction value and the integrated cell number prediction value at the current time is within a preset threshold, the position clustering process based on the integrated cell number prediction value It may be determined that this is the start time.

  In this case, the clustering start determination processing unit 140 in FIG. 16 uses the average value of the integrated cell number prediction value for the past M samples and the current time for the integrated cell number prediction value in the wake prediction value obtained from the delay processing unit 90. If the absolute value of the difference from the integrated cell number predicted value is within a preset threshold value, the position clustering process based on the integrated cell number predicted value is executed.

  On the other hand, when the absolute value of the difference between the average value of the integrated cell number prediction values for the past M samples and the integrated cell number prediction value at the current time exceeds a preset threshold value, The clustering start determination processing unit 140 inputs to the position clustering processing unit 50 a control signal for executing position clustering by removing the observation value farthest from the center of gravity of the observation value cluster.

  As described above, according to the tracking device according to Embodiment 6 (FIG. 16) of the present invention, the clustering start determination processing unit 140 is provided between the delay processing unit 90 and the position clustering processing unit 50, and the clustering start determination is performed. The processing unit 140 is based on the integrated cell number prediction value when the absolute value of the difference between the average value of the integrated cell number prediction value for the past M samples and the integrated cell number prediction value at the current time is within a threshold value. Since it is determined that it is the start time of the position clustering process, and the estimated number of integrated cells unstable in the initial time zone is removed from the target of the position clustering process, it is possible to stabilize the track accuracy in the initial time zone. It becomes possible.

  In the above third to sixth embodiments, the case where the present invention is applied to the configuration of the above-described first embodiment (FIG. 1) has been described. However, the present invention can be applied to other embodiments, and arbitrarily. Needless to say, they can be applied in a duplicated manner and exhibit their synergistic effects.

  1-5 barycentric points, 10 sensors, 20 position correlation processing unit, 30 cell number correlation processing unit, 40 correlation mediation processing unit, 50 position clustering processing unit, 60 integrated cell number observation value monitoring processing unit, 70 integrated cell number update prediction Processing unit, 80 position speed update prediction processing unit, 90 delay processing unit, 100 display processing unit, 110 threshold control processing unit, 120 speed vector correction processing unit, 130 clustering adjustment processing unit, 140 clustering start determination processing unit, 200 Path, 210 Centerline, M1 position correlation matrix, M2 cell number correlation matrix, M3 correlation mediation matrix, R correction execution area, S0 position vector, T1, T2 track, ThA to ThC threshold, V1, V2 velocity vector, Z1 ~ Z5 observed value.

Claims (6)

A high-resolution sensor that obtains multiple detection positions from the same target;
A position correlation processing unit that generates a position correlation matrix by associating a track position predicted value and a position observation value using the plurality of detection positions;
A cell number correlation processing unit for generating a cell number correlation matrix by associating the cell number observation value with the integrated cell number prediction value using the plurality of detection positions;
Using the position correlation matrix and the cell number correlation matrix, a correlation mediation result is generated by preferentially allocating observation values for a track having a small number of integrated cells in consideration of the track position and the integrated cell number prediction value. A correlation mediation processing unit;
Using the correlation mediation result, a position clustering processing unit that generates an observation value cluster based on the integrated cell number prediction value;
An integrated cell number observation value monitoring processor that monitors the integrated cell number observation value so that the integrated cell number observation value of the observation value cluster does not become excessive;
Based on a motion model in which the number of integrated cells of the observation value cluster is constant, an integrated cell number update prediction processing unit that generates an integrated cell number estimated value and an updated integrated cell number predicted value by a Kalman filter;
A position / velocity update prediction processing unit that generates a position / velocity prediction value and a position / velocity estimation value of a wake by a Kalman filter, using the center of gravity of the observation value cluster as an observation value;
Delay processing for delaying the position / velocity predicted value of the wake by one sampling and feeding back to the position correlation processing unit, the cell number correlation processing unit, the position clustering processing unit, and the integrated cell number observation value monitoring processing unit And
A display processing unit for displaying a position velocity estimation value and an integrated cell number estimation value of a wake finally generated from the position velocity update prediction processing unit to an operator;
A tracking device characterized by comprising: A cell number correlation gate threshold control processing unit provided in the cell number correlation processing unit;
The cell number correlation gate threshold value control processing unit controls a cell number correlation gate threshold value in the cell number correlation processing unit in accordance with a size of the integrated cell number prediction value. 2. The tracking device according to 1. A velocity vector correction processing unit inserted between the position / velocity update prediction processing unit and the delay processing unit;
The speed vector correction processing unit recalculates and corrects the direction of the speed vector of the wake so as to coincide with a center line in a target passage traveling direction. Tracking device. A clustering adjustment processing unit provided in the position clustering processing unit;
The clustering adjustment processing unit, when generating the observation value cluster, when the observation cell cluster has an excessive number of integrated cell observation values and there are a plurality of observation values having the smallest number of cells, the observation value cluster The tracking according to any one of claims 1 to 3, wherein an observation value having the smallest number of cells and having the smallest number of cells is removed from the observation value cluster. apparatus. A clustering start determination processing unit inserted between the delay processing unit and the position clustering processing unit;
The clustering start determination processing unit determines that position clustering based on the integrated cell number prediction value is started when a predicted track value is obtained exceeding a preset sampling threshold. The tracking device according to any one of claims 1 to 4. A clustering start determination processing unit inserted between the delay processing unit and the position clustering processing unit;
The clustering start determination processing unit is configured such that the absolute value of the difference between the average value of the integrated cell number prediction values for the past M samples and the integrated cell number prediction value at the current time is within a preset threshold value The tracking device according to claim 1, wherein the position clustering based on the integrated cell number prediction value is determined to be started.

Images