JP6279188B2 - Target tracking device - Google Patents

Description

  The present invention relates to a target tracking device that estimates a track including information on a target position from observation values of a sensor.

Estimate the track indicating the current position and speed of the target from the sensor observation values, and transmit the track to the other target tracking device via the network, while the track transmitted from the other target tracking device is A target tracking device that generates a highly accurate fusion track by fusing the track estimated by itself with a track transmitted from another target tracking device is provided in the following patent document. 1 is disclosed.
In this target tracking device, for the purpose of reducing the communication capacity of the network, it is determined whether or not the error of the fusion track can be greatly reduced by fusing the estimated track with the fusion track. Only when it is determined that the error can be greatly reduced, the wake estimated by itself is transmitted to another target tracking device.

JP 2010-281882 A

  Since the conventional target tracking device is configured as described above, it is possible to reduce the communication frequency of the network by reducing the transmission frequency of the wake estimated by itself. However, when deciding whether or not to transmit the wake estimated by the self, the movement state of the target is not taken into consideration, and therefore the transmission frequency of the wake may be lowered even when the target is making a turning motion. is there. If the transmission frequency of the wake is lowered while the target is making a turning motion, there is a problem that the tracking accuracy of the turning target may deteriorate due to insufficient wake data amount. It was.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to obtain a target tracking device that can reduce the communication capacity while ensuring the tracking accuracy of a turning target.

  A target tracking device according to the present invention includes a tracking processing unit that estimates a partial track including information on a target position from time-series data of observation values indicating target positions output from a sensor, and time-series data of observation values A turn determination unit that determines whether or not the target is turning, a fusion track storage unit that stores a fusion track including information on the target position, and a time of the partial track estimated by the tracking processing unit A transmission determination unit that determines whether or not to transmit the partial track estimated by the tracking processing unit from the time difference from the time of the fusion track stored in the track storage unit and the determination result of the turning determination unit; and transmission When the determination result of the determination unit indicates that the partial track is to be transmitted, the transmission / reception unit that transmits the partial track to the other target tracking device and receives the partial track transmitted from the other target tracking device; Established by the Wake Fusion Department Using the transmission or reception portions track by Shin section, in which so as to update the fusion track stored in the fusion track storage unit.

  According to the present invention, the turn determination unit that determines whether or not the target is turning from the time series data of the observation values, and the time of the partial wake estimated by the tracking processing unit and the fusion wake storage unit are stored. A transmission determination unit that determines whether or not to transmit the partial track estimated by the tracking processing unit from the time difference from the time of the fusion track and the determination result of the turning determination unit is provided, and the transmission / reception unit is a transmission determination unit When the result of the determination indicates that a partial track is to be transmitted, the partial track is transmitted to another target tracking device, so that the communication capacity is maintained while ensuring the tracking accuracy of the turning target. There is an effect that can be reduced.

It is a block diagram which shows the target tracking apparatus by Embodiment 1 of this invention. It is a hardware block diagram of the target tracking apparatus by Embodiment 1 of this invention. It is a hardware block diagram of a computer in case a target tracking apparatus is implement | achieved by software, firmware, etc. It is a block diagram which shows the turning determination part 17 of the target tracking apparatus by Embodiment 1 of this invention. 4 is a configuration diagram showing a turning start determination processing unit 71 of a turning determination unit 17. FIG. FIG. 5 is a configuration diagram showing a turning end determination processing unit 73 of a turning determination unit 17. 4 is a configuration diagram showing a turning level calculation processing unit 75 of the turning determination unit 17. FIG. It is a block diagram which shows the transmission determination part 19 of the target tracking apparatus by Embodiment 1 of this invention. 7 is a flowchart showing processing contents of a turning start determination processing unit 71. 7 is a flowchart showing processing contents of a turning end determination processing unit 73. It is a flowchart which shows the processing content of the threshold value determination process part 73h. It is explanatory drawing which shows the transmission determination concept in the transmission determination part 19. FIG. It is a conceptual diagram which shows the time change of the fusion wake error before and during turning, and the transmission frequency of partial wakes. It is a conceptual diagram which shows the time change of the fusion wake error before the turn in the case of not using a transmission threshold value, and the turn, and the transmission frequency of a partial wake. It is a block diagram which shows the target tracking apparatus by Embodiment 2 of this invention. It is a hardware block diagram of the target tracking apparatus by Embodiment 2 of this invention. It is a block diagram which shows the turning determination part 17 of the target tracking apparatus by Embodiment 2 of this invention. 4 is a configuration diagram showing a turning start determination processing unit 71 of a turning determination unit 17. FIG. FIG. 5 is a configuration diagram showing a turning end determination processing unit 73 of a turning determination unit 17. 4 is a flowchart showing processing contents of a flag update processing unit 101. 5 is a flowchart showing processing contents of an erroneous determination cancellation processing unit 103. 7 is a flowchart showing processing contents of a turning end determination processing unit 73. It is a flowchart which shows the processing content of the threshold value determination process part 73h.

  Hereinafter, in order to describe the present invention in more detail, modes for carrying out the present invention will be described with reference to the accompanying drawings.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a target tracking device according to Embodiment 1 of the present invention, and FIG. 2 is a hardware configuration diagram of the target tracking device according to Embodiment 1 of the present invention.
FIG. 1 shows an example in which M (M is an integer of 2 or more) target tracking devices 1-1 to 1-M are connected to the network 2. The internal configuration of the target tracking devices 1-1 to 1-M is the same.
1 and 2, for example, a radar, an optical camera, an infrared camera, or the like is assumed as the sensor 11. The sensor 11 observes the position of a target to be observed, and is a sensor observation that is an observation value of that position. The value is output to the tracking processing unit 12.
Although FIG. 1 shows an example in which the target tracking devices 1-1 to 1-M are mounted with the sensor 11, the sensor 11 is provided outside the target tracking devices 1-1 to 1-M. Also good.

The tracking processing unit 12 includes a sensor track storage unit 13, a sensor tracking processing unit 14, a partial track storage unit 15, and a retracking processing unit 16. From the time series data of sensor observation values output from the sensor 11, the tracking processing unit 12 As the partial track including the position information, for example, a partial track having a state vector such as the current position and speed of the target is estimated.
The sensor track storage unit 13 is realized by, for example, the storage processing circuit 41 of FIG. 2 and stores the sensor correlated tracked sensor observation value estimated by the sensor tracking processing unit 14 and the sensor correlated value.

The sensor tracking processing unit 14 is realized by, for example, the sensor tracking processing circuit 42 of FIG. 2, and the position indicated by the sensor observation value output from the sensor 11 and the sensor track stored in the sensor track storage unit 13 are determined. A process of determining whether or not the indicated position is correlated is performed.
Further, when the sensor tracking processing unit 14 determines that the position indicated by the sensor observation value is correlated with the position indicated by the sensor track, the sensor tracking value is stored in the sensor track storage unit 13 as a target correlated sensor observation value. At the same time, the target correlated sensor observation value is output to the re-tracking processing unit 16 and the turning determination unit 17.
Further, the sensor tracking processing unit 14 has a state vector such as the current position and speed of the target from the target correlated sensor observation value and the past target correlated sensor observation value stored in the sensor track storage unit 13. The sensor wake is estimated, and the sensor wake stored in the sensor wake storage unit 13 is replaced with the estimated sensor wake.

The partial wake storage unit 15 is realized by, for example, the storage processing circuit 41 in FIG. 2 and stores the partial wake estimated by the re-tracking processing unit 16 and the sensor correlation sensor observation value that is a sensor observation value. .
The retracking processing unit 16 is realized by, for example, the retracking processing circuit 43 in FIG. 2, and is stored in the partial track storage unit 15 and the position indicated by the target correlated sensor observation value output from the sensor tracking processing unit 14. A process for determining whether or not the position indicated by the partial track is correlated is performed.
When the re-tracking processing unit 16 determines that the position indicated by the target correlated sensor observation value output from the sensor tracking processing unit 14 is correlated with the position indicated by the sensor track, the target correlated sensor observation value is determined. Is stored in the partial wake storage unit 15, and the current position and speed of the target are determined from the target correlated sensor observation values and the past target correlated sensor observation values stored in the partial wake storage unit 15. A process of estimating a partial wake having a vector is performed.
Further, the re-tracking processing unit 16 replaces the partial track stored in the partial track storage unit 15 with the estimated partial track, and the target which is the determination result of the turning determination unit 17 with respect to the estimated partial track. A turning flag indicating the presence or absence of turning and a turning level indicating the target turning degree are added, and a process of outputting the turning flag and the partial track with the turning level to the transmission determination unit 19 is performed.

The turning determination unit 17 is realized by, for example, the turning determination processing circuit 44 of FIG. 2, and whether or not the target is turning from the time series data of the target correlated sensor observation values output from the sensor tracking processing unit 14. As a result of the determination, a process of outputting a turning flag indicating the presence or absence of the target turning to the re-tracking processing unit 16 is performed.
The fusion track storage unit 18 is realized by, for example, the storage processing circuit 41 in FIG. 2, and stores the fusion track estimated by the track fusion unit 21 as a fusion track including target position information.

The transmission determination unit 19 is realized by, for example, the transmission determination processing circuit 45 of FIG. 2, and is stored in the fusion track storage unit 18 and the time of the partial wake with the turn flag and the turn level output from the retracking processing unit 16. Based on the time difference from the time of the integrated wake and the turn flag attached to the partial wake, a process for determining whether or not to transmit the partial wake with the turn flag and the turn level is performed.
At this time, the transmission determination unit 19 is more in the case where the turn flag attached to the partial track indicates that the target is turning than in the case where the target is not turning. The frequency at which the determination result indicating that the partial wake with the turning flag and the turning level is transmitted is increased.

The transmission / reception unit 20 is realized by, for example, the transmission / reception processing circuit 46 of FIG. 2, and when the determination result of the transmission determination unit 19 indicates that a partial track with a turning flag and a turning level is transmitted, the network 2 The partial wake with the turning flag and the turning level is transmitted to the other target tracking device 1-M via the, and processing for outputting the partial wake to the wake fusion unit 21 is performed.
Further, when the partial wake transmitted from the other target tracking device 1 -M is received via the network 2, the transmission / reception unit 20 performs a process of outputting the partial wake to the wake fusion unit 21.

The wake fusion unit 21 includes a correlation determination unit 22 and a fusion tracking processing unit 23, and is stored in the fusion wake storage unit 18 using the partial wake with the turn flag and the turn level output from the transmission / reception unit 20. Implement a process to update the fusion track.
The correlation determination unit 22 is realized by, for example, the correlation determination processing circuit 47 of FIG. 2, and is stored in the fusion track storage unit 18 and the position indicated by the partial wake with the turn flag and the turn level output from the transmission / reception unit 20. The process which determines whether the position which the fusion wake which is shown correlates is implemented.
The fusion tracking processing unit 23 is realized by, for example, the fusion tracking processing circuit 48 of FIG. 2. When the correlation determination unit 22 determines that there is a correlation, the turning flag and the turning level output from the transmission / reception unit 20. The fusion wake having a state vector such as the current position and speed of the target is estimated from the partial wake with and the fusion wake stored in the fusion wake storage unit 18, and the fusion wake stored in the fusion wake storage unit 18. Is replaced with the estimated fusion wake.
When the fusion tracking processing unit 23 determines that there is no correlation by the correlation determination unit 22, the fusion track storage unit 18 uses the partial track with the turn flag and the turn level output from the transmission / reception unit 20 as a new fusion track. The process to record in is performed.

  The display processing unit 24 is realized by, for example, the display processing circuit 49 of FIG. 2, and includes a fusion track estimated by the fusion tracking processing unit 23, a turn flag determined to be correlated by the correlation determination unit 22, and A process for displaying a partial track with a turning level on a display or the like is performed.

In FIG. 1, a sensor 11, a sensor track storage unit 13, a sensor tracking processing unit 14, a partial track storage unit 15, a retracking processing unit 16, a turning determination unit 17, a fusion track storage unit 18, which are components of the target tracking device. Each of the transmission determination unit 19, the transmission / reception unit 20, the correlation determination unit 22, the fusion tracking processing unit 23, and the display processing unit 24 includes dedicated hardware as illustrated in FIG. 2, that is, a storage processing circuit 41 and a sensor tracking processing circuit. 42, assuming re-tracking processing circuit 43, turning determination processing circuit 44, transmission determination processing circuit 45, transmission / reception processing circuit 46, correlation determination processing circuit 47, fusion tracking processing circuit 48 and display processing circuit 49. Yes.
Here, the memory processing circuit 41 is, for example, a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), an EEPROM (Electrically Erasable Memory), or the like. Alternatively, a volatile semiconductor memory, a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, a DVD (Digital Versatile Disc), or the like is applicable.
The sensor tracking processing circuit 42, the re-tracking processing circuit 43, the turning determination processing circuit 44, the transmission determination processing circuit 45, the transmission / reception processing circuit 46, the correlation determination processing circuit 47, the fusion tracking processing circuit 48, and the display processing circuit 49 are, for example, A single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a combination thereof.

Further, the components of the target tracking device are not limited to those realized by dedicated hardware, and the target tracking device may be realized by software, firmware, or a combination of software and firmware. .
Software and firmware are stored as programs in the memory of the computer. The computer means hardware that executes a program, and includes, for example, a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, a processor, a DSP (Digital Signal Processor), and the like. .
FIG. 3 is a hardware configuration diagram of a computer when the target tracking device is realized by software, firmware, or the like.
When the target tracking device is realized by software, firmware, or the like, the sensor track storage unit 13, the partial track storage unit 15, and the fusion track storage unit 18 are configured on the memory 61 of the computer, and the sensor tracking processing unit 14, retracking A program for causing the computer to execute the processing procedures of the processing unit 16, turning determination unit 17, transmission determination unit 19, transmission / reception unit 20, correlation determination unit 22, fusion tracking processing unit 23, and display processing unit 24 is stored in the memory 61. The computer processor 62 may execute a program stored in the memory 61.

2 shows an example in which each component of the target tracking device is realized by dedicated hardware, and FIG. 3 shows an example in which the target tracking device is realized by software, firmware, or the like. Some components in the tracking device may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, or the like.
For example, the sensor 11, the transmission / reception unit 20, and the display processing unit 24 are realized by dedicated hardware, the sensor track storage unit 13, the sensor tracking processing unit 14, the partial track storage unit 15, the retracking processing unit 16, and the turning determination unit 17. The fusion track storage unit 18, the transmission determination unit 19, the correlation determination unit 22, and the fusion tracking processing unit 23 can be realized by software, firmware, or the like. However, the combination of dedicated hardware and software is arbitrary.

FIG. 4 is a block diagram showing the turning determination unit 17 of the target tracking device according to Embodiment 1 of the present invention.
In FIG. 4, the turn start determination processing unit 71 detects the target correlated sensor observation value output from the sensor tracking processing unit 14 and the target turn stored in the turn flag database (hereinafter “turn flag DB”) 72. A turning start determination process for determining the start of turning is performed using the turning flag indicating presence / absence, and the turning flag stored in the turning flag DB 72 is updated.
The turning flag DB 72 is a memory that stores a turning flag indicating the presence or absence of a target turning. It is assumed that the initial value of the turn flag stored in the turn flag DB 72 indicates “non-turn”.
The turn end determination processing unit 73 performs a turn end determination process for determining the end of the turn using the target correlated sensor observation value output from the sensor tracking processing unit 14 and the turn flag stored in the turn flag DB 72. It carries out and updates the turning flag memorize | stored in turning flag DB72.

The turning flag output processing unit 74 performs a process of outputting the turning flag stored in the turning flag DB 72 to the turning level calculation processing unit 75 and the retracking processing unit 16.
When the turning flag output from the turning flag output processing unit 74 indicates that the target is turning, the turning level calculation processing unit 75 uses the target correlated sensor observation value output from the sensor tracking processing unit 14. Then, the turning level indicating the target turning degree is calculated, and the turning level is output to the retracking processing unit 16.

FIG. 5 is a configuration diagram illustrating the turning start determination processing unit 71 of the turning determination unit 17.
In FIG. 5, an observation value database for turning start determination (hereinafter referred to as an “observation value DB for turning start determination”) 71 a stores target correlated sensor observation values for the past Nth samples output from the sensor tracking processing unit 14. Memory.
A speed database (hereinafter referred to as “speed DB”) 71b is a memory that accumulates smoothing speeds for the past Nth samples.
A smooth wake database for turning start determination (hereinafter referred to as “smooth wake DB for turning start determination”) 71c is a memory for storing smooth wakes for the past Nth samples.

When the turning flag stored in the turning flag DB 72 is “turning”, the initialization processing unit 71d stores the accumulated data of the turning start determination observation value DB 71a, the speed DB 71b, and the turning start determination smooth track DB 71c, that is, the target correlation. The completed sensor observation value, smooth speed and smooth track are deleted, and the turn start determination process is terminated. On the other hand, when the turning flag stored in the turning flag DB 72 is “non-turning”, the execution of the subsequent turning start determination process is permitted.
The linear trajectory estimation processing unit 71e assumes that the target motion state is constant-velocity linear motion, and uses the linear least square method, the Kalman filter, or the like, for example, to store the past accumulated in the observation value DB 71a for turning start determination. The target smoothed position and smoothing speed are calculated from the target correlated sensor observation values for Nth samples, and the smoothing position error covariance matrix and smoothing speed error covariance matrix are calculated.
Further, the linear trajectory estimation processing unit 71e stores the calculated smoothing speed in the speed DB 71b.

The speed re-smoothing processing unit 71f assumes that the target motion state is constant-velocity linear motion, and uses, for example, a linear least square method or a Kalman filter to smooth the past Nth samples accumulated in the speed DB 71b. From the speed, a process of calculating a re-smoothing speed that is a re-smoothed speed and a re-smoothing speed error variance matrix is performed.
Further, the speed re-smoothing processing unit 71f performs a process of replacing the speed term of the smoothing speed and smoothing error covariance matrix calculated by the linear trajectory estimation processing unit 71e with a re-smoothing speed and a re-smoothing speed error variance matrix. The smooth wake is calculated and stored in the smooth wake DB 71c for turning start determination.

The N-step prediction processing unit 71g performs a process of calculating a predicted track having a state vector such as the current position and speed of the target from the smooth tracks for the past Nth samples accumulated in the smooth start DB 71c for turning start determination. .
The Mahalanobis square distance calculation processing unit 71h calculates the predicted track from the position indicated by the predicted track calculated by the N step prediction processing unit 71g and the position indicated by the target correlated sensor observation value output from the sensor tracking processing unit 14. A process of calculating the Mahalanobis square distance between the indicated position and the position indicated by the target correlated sensor observation value is performed.

The time direction smoothing processing unit 71i performs processing for smoothing the Mahalanobis square distance calculated by the Mahalanobis square distance calculation processing unit 71h in the time direction.
The threshold value determination processing unit 71j compares the Mahalanobis square distance smoothed in the time direction by the time direction smoothing processing unit 71i with a preset threshold value, and if the Mahalanobis square distance is greater than the threshold value, the turning has started. And the turning flag stored in the turning flag DB 72 is set to “turning”. If the Mahalanobis square distance is less than or equal to the threshold value, it is determined that turning has not started, and is stored in the turning flag DB 72. The process of maintaining the “non-turning” turning flag is executed.

FIG. 6 is a configuration diagram showing the turning end determination processing unit 73 of the turning determination unit 17.
In FIG. 6, an observation value database for turning end determination (hereinafter referred to as “turning end determination observation value DB”) 73 a stores target correlated sensor observation values for the past Nth samples output from the sensor tracking processing unit 14. Memory.
A smooth wake database for turning completion determination (hereinafter referred to as “smooth wake DB for turning completion determination”) 73b is a memory for accumulating smooth wakes for the past Nth samples.
When the turning flag stored in the turning flag DB 72 is “non-turning”, the initialization processing unit 73c stores the accumulated data of the turning end determination observation value DB 73a and the turning end determination smooth track DB 73b, that is, the target correlated sensor. The observation value and the smooth track are deleted, and the turn end determination process ends. On the other hand, when the turning flag stored in the turning flag DB 72 is “turning”, the execution of the subsequent turning end determination process is permitted.

The trajectory estimation processing unit 73d assumes that the target motion state is a uniform acceleration motion, and uses, for example, a linear least square method, a Kalman filter, or the like to store the past Nth samples stored in the smooth wake DB 73b for turning completion determination. The target smoothed position, smoothing speed and smoothing acceleration are calculated as the target smooth track from the target correlated sensor observations for the minute, and the smoothing position error covariance matrix and smoothing speed error covariance matrix are calculated. carry out.
In addition, the trajectory estimation processing unit 73d stores the calculated smooth track in the smooth track DB 73b for turning completion determination.

The N-step prediction processing unit 73e performs a process of calculating a predicted track having a state vector such as the current position and speed of the target from the smooth tracks for the past Nth samples accumulated in the smooth track DB 73b for turning completion determination. .
The Mahalanobis square distance calculation processing unit 73f calculates the predicted track from the position indicated by the predicted track calculated by the N-step prediction processing unit 73e and the position indicated by the target correlated sensor observation value output from the sensor tracking processing unit 14. A process of calculating the Mahalanobis square distance between the indicated position and the position indicated by the target correlated sensor observation value is performed.

The time direction smoothing processing unit 73g performs a process of smoothing the Mahalanobis square distance calculated by the Mahalanobis square distance calculation processing unit 73f in the time direction.
The threshold value determination processing unit 73h compares the Mahalanobis square distance smoothed in the time direction by the time direction smoothing processing unit 73g with a preset threshold value, and if the Mahalanobis square distance is equal to or less than the threshold value, the turning is finished. If the turning flag stored in the turning flag DB 72 is set to “non-turning” and the Mahalanobis square distance is larger than the threshold value, it is determined that the turning has not ended, and the turning flag DB 72 is set. A process of maintaining the stored “turn” turning flag is executed.

Here, an example is shown in which the threshold determination processing unit 73h compares the Mahalanobis square distance smoothed in the time direction with the threshold value. However, the threshold determination processing unit 73h performs the smoothing Mahalanobis square distance at the current time, A difference value with the smoothed Mahalanobis square distance at the time may be calculated, and it may be determined that the turn is finished when the number of times that the difference value becomes negative exceeds a threshold value.
In this case, a difference value counter database (hereinafter referred to as “difference value counter DB”) 73i that stores the number of times that the difference value becomes negative is provided, and the number of times that the threshold determination processing unit 73h is stored in the difference value counter DB 73i. Update.

FIG. 7 is a configuration diagram showing the turning level calculation processing unit 75 of the turning determination unit 17.
In FIG. 7, if the turning flag output from the turning flag output processing unit 74 is “non-turning”, the initialization processing unit 75a turns the observation value database for turning level calculation (hereinafter referred to as the “observation value DB for turning level calculation”). If the target correlated sensor observation value stored in 75b is deleted and the turning flag is “turn”, the target correlated sensor observation value output from the sensor tracking processing unit 14 is used for turning level calculation. A process of storing in the observation value DB 75b is performed.
The observation value DB 75b for turning level calculation is a memory for accumulating the target correlated sensor observation values output from the sensor tracking processing unit 14.
The turning trajectory estimation processing unit 75c assumes that the target motion state is a uniform acceleration motion, and uses, for example, the target correlation accumulated in the turning level calculation observation value DB 75b using a linear least square method, a Kalman filter, or the like. The target acceleration is calculated from the measured sensor observation values, the target acceleration is converted into a turning level, and the turning level is output to the retracking processing unit 16.

FIG. 8 is a block diagram showing the transmission determination unit 19 of the target tracking device according to Embodiment 1 of the present invention.
In FIG. 8, the parameter change processing unit 81 changes the transmission determination system noise parameter and the transmission determination threshold according to the turning flag and the turning level attached to the partial track output from the retracking processing unit 16. Perform the process.
The fused track prediction processing unit 82 calculates a time difference between the time of the partial track output from the retracking processing unit 16 and the time of the fused track stored in the fused track storage unit 18, and the time difference and the parameter change processing unit 81. A process of calculating a prediction error covariance matrix for transmission determination is performed using the system noise parameter changed by the above.

The determination error calculation processing unit 83 outputs the prediction error covariance matrix for transmission determination calculated by the fusion track prediction processing unit 82, the fusion track stored in the fusion track storage unit 18, and the retracking processing unit 16. A process of calculating a fusion wake error, an updated fusion wake error, and a fusion wake error difference is performed from the obtained partial wakes.
Here, the fusion track error is a spectral norm of the error covariance matrix of the fusion track. The updated fusion wake error is a spectrum norm of the error covariance matrix of the fusion wake after updating with the partial wake based on the prediction error covariance matrix for transmission determination. Although the wake error is defined by the spectrum norm of the error covariance matrix, it may be a trace norm of the error covariance matrix or a determinant. The method for defining the norm is determined in advance by the user.
Further, the error covariance matrix may be an error covariance matrix with only position components or an error covariance matrix with only velocity components. The error covariance matrix can be defined by orthogonal coordinates or polar coordinates.

The determination processing unit 84 uses the fusion track error calculated by the determination error calculation processing unit 83, the updated fusion track error and the fusion track error difference, and the transmission determination threshold value changed by the parameter change processing unit 81. The process for determining whether or not to transmit the partial track output from the retracking processing unit 16 is performed.
At this time, the determination processing unit 84 is more in the case where the turn flag attached to the partial wake indicates that the target is turning than in the case where the target is not turning. The frequency of issuing a determination result that permits transmission of the partial wake with the turning flag and the turning level is increased.

Next, the operation will be described.
In the first embodiment, an example in which M target tracking devices 1-1 to 1-M are connected to the network 2 is assumed. Since the processing contents of the target tracking devices 1-1 to 1-M are the same, here, the processing content of the target tracking device 1-1 will be described as a representative.
The sensor 11 observes the position of the target to be observed, and outputs a sensor observation value that is an observation value at that position to the tracking processing unit 12.

When the sensor tracking processing unit 14 of the tracking processing unit 12 receives the sensor observation value from the sensor 11, for example, by executing a known correlation algorithm such as GNN (Global Nearest Neighbor), the position indicated by the sensor observation value Then, it is determined whether or not the position indicated by the sensor track stored in the sensor track storage unit 13 is correlated.
That is, the sensor tracking processing unit 14 determines whether or not the sensor observation value output from the sensor 11 is observing the position of the target related to the sensor track estimated in the past.

The sensor tracking processing unit 14 correlates the position indicated by the sensor observation value with the position indicated by the sensor wake stored in the sensor wake storage unit 13, and may associate the sensor observation value with the sensor wake. If it is determined that it is possible, the sensor observation value is stored in the sensor track storage unit 13 as the target correlated sensor observation value, and the target correlated sensor observation value is output to the retracking processing unit 16 and the turning determination unit 17.
Further, the sensor tracking processing unit 14 executes a known tracking algorithm such as a Kalman filter, for example, so that the target correlated sensor observation value and the past target correlated sensor observation value stored in the sensor track storage unit 13 are stored. From the above, a sensor track having a state vector such as the current position and speed of the target is estimated.
When the sensor tracking processing unit 14 estimates the sensor track, the sensor track stored in the sensor track storage unit 13 is replaced with the estimated sensor track by replacing the sensor track stored in the sensor track storage unit 13 with the estimated sensor track. Update.

Upon receiving the target correlated sensor observation value from the sensor tracking processing unit 14, the turning determination unit 17 determines whether or not the target is turning from the time series data of the target correlated sensor observation value, and the determination result As a result, a turn flag indicating whether or not the target is turning and a turn level indicating the degree of the target turn are output to the re-tracking processing unit 16.
Hereinafter, the turning determination process by the turning determination unit 17 will be specifically described.

When the turn start determination processing unit 71 of the turn determination unit 17 receives the target correlated sensor observation value from the sensor tracking processing unit 14, the target correlation completed sensor observation value and the turn flag stored in the turn flag DB 72 are obtained. The turning start determination process for determining the start of turning is performed, and the turning flag stored in the turning flag DB 72 is updated.
FIG. 9 is a flowchart showing the processing contents of the turning start determination processing unit 71.
Hereinafter, the processing content of the turning start determination processing unit 71 will be specifically described with reference to FIG.

The observation value DB 71a for turning start determination of the turning start determination processing unit 71 records the target correlated sensor observation value output from the sensor tracking processing unit 14, and the observation value DB 71a for turning start determination stores past Nth samples. The target correlated sensor observations are accumulated.
The initialization processing unit 71d performs initialization processing for turning start determination processing (step ST1 in FIG. 9).
That is, when the turning flag stored in the turning flag DB 72 is “turning”, the initialization processing unit 71d is the accumulated data of the turning start determination observation value DB 71a, the speed DB 71b, and the turning start determination smooth track DB 71c. The target correlated sensor observation value, smoothing speed, and smooth track are deleted, and the turning start determination process is terminated.
When the turning flag stored in the turning flag DB 72 is “non-turning”, the initialization processing unit 71d permits the subsequent turning start determination process to be performed. That is, the operations of the linear trajectory estimation processing unit 71e, the speed re-smoothing processing unit 71f, the N-step prediction processing unit 71g, the Mahalanobis square distance calculation processing unit 71h, the time direction smoothing processing unit 71i, and the threshold determination processing unit 71j are permitted.

The linear trajectory estimation processing unit 71e performs the linear trajectory estimation process when the number of target correlated sensor observation values accumulated in the turning start determination observation value DB 71a is Nth samples or more (step ST2: YES). Implement (step ST3).
In other words, the linear trajectory estimation processing unit 71e is accumulated in the turning start determination observation value DB 71a using, for example, a linear least square method or a Kalman filter, assuming that the target motion state is constant velocity linear motion. A linear trajectory estimation process of calculating a target smooth position and smoothing speed from the target correlated sensor observation values for the past Nth samples and calculating a smooth position error covariance matrix and a smooth speed error covariance matrix. carry out.
For example, when the target position is defined by X, Y, and Z in the Cartesian coordinate system, the smooth position and smooth speed for each axis are calculated by the linear least square method.
Hereinafter, the calculation process of the linear trajectory estimation processing unit 71e will be specifically described taking the X axis as an example.

The linear trajectory estimation processing unit 71e calculates the smooth position p hat _{x, k} with respect to the X axis and the smooth speed p dot hat _{x, k} with respect to the X axis, as shown in the following equations (1) to (5). .
Here, because of the electronic application, in the text of the specification, since the symbol “^” cannot be added on the letter, it is expressed as p hat _{x, k} . Similarly, since the symbols “·” and “^” cannot be added on the characters, they are represented as p dot hats _{x and k} .

式（１）〜（５）において、下付添え字のｋはセンサ１１の各サンプリング時刻ｔ_ｋ（ｋ＝１，２，・・・）におけるデータ、上付き添え字のｋはサンプリング時刻ｔ_ｋまでのデータの集合を示している。Ｔはベクトルもしくは行列の転置を意味する。
直線軌道推定処理部７１ｅは、Ｙ軸及びＺ軸に対しても、上記と同様の手順で平滑位置と平滑速度を算出する。直線軌道推定処理部７１ｅは、算出した平滑速度を速度ＤＢ７１ｂに格納する。

In the equations (1) to (5), the subscript k is data at each sampling time t _k (k = 1, 2,...) Of the sensor 11, and the superscript k is the sampling time t _k. The data set up to is shown. T means transposition of a vector or matrix.
The linear trajectory estimation processing unit 71e calculates a smooth position and a smooth speed in the same procedure as described above for the Y axis and the Z axis. The linear trajectory estimation processing unit 71e stores the calculated smoothing speed in the speed DB 71b.

ここで、Ａ［ｉ，ｊ］は行列Ａのｉ行ｊ列目の要素を意味する。また、ｄｉａｇ［ａ ｂ ｃ］は要素ａ，ｂ，ｃを対角要素に持つ行列を意味する。
なお、直線軌道推定処理部７１ｅは、旋回開始判定用観測値ＤＢ７１ａに蓄積されている目標相関済みセンサ観測値の数が、Ｎｔｈサンプル未満である場合（ステップＳＴ２：ＮＯの場合）、直線軌道推定処理を実施しない。この場合、ステップＳＴ１１の処理に移行する。Further, linear track estimation processing unit 71e, when defining the position of the target orthogonal coordinate system X, Y, with Z, as shown in equation (6) to (10) below, smoothed position vector p hat _k, A smooth velocity vector p dot hat _k , a smooth position error covariance matrix P _{pos, k} , and a smooth velocity error covariance matrix P _{vel, k} are defined.

Here, A [i, j] means an element in the i-th row and j-th column of the matrix A. Diag [a b c] means a matrix having elements a, b, and c as diagonal elements.
The linear trajectory estimation processing unit 71e, when the number of target correlated sensor observation values accumulated in the turning start determination observation value DB 71a is less than Nth samples (step ST2: NO), linear trajectory estimation. No processing is performed. In this case, the process proceeds to step ST11.

When the linear trajectory estimation processing unit 71e performs the linear trajectory estimation process, the speed re-smoothing processing unit 71f performs the speed re-smoothing process (step ST4).
That is, the speed re-smoothing processing unit 71f assumes that the target motion state is a constant-velocity linear motion, and uses, for example, a linear least square method, a Kalman filter, or the like to store past Nth samples accumulated in the speed DB 71b. A speed re-smoothing process of calculating a re-smoothing speed that is a re-smoothed speed and a re-smoothed speed error variance matrix is performed from the smooth speed of minutes.
In addition, the speed re-smoothing processing unit 71f performs a process of replacing the smoothing velocity and the velocity term of the smoothing error covariance matrix calculated by the linear trajectory estimation processing unit 71e with the re-smoothing velocity and the re-smoothing velocity error variance matrix. The smooth wake is calculated by the above and the smooth wake is stored in the smooth wake DB 71c for turning start determination.
Hereinafter, taking the X axis as an example, the calculation process of the speed re-smoothing processor 71f will be specifically described.

As shown in the following formulas (10) to (14), the speed re-smoothing processing unit 71f performs a re-smoothing speed V hat _{x, k} and a re-smoothing speed error variance matrix P _xV by the linear least square method with respect to the X axis. _{, K.}

The speed re-smoothing processing unit 71f calculates the re-smoothing speed and the re-smoothing speed error variance matrix in the same procedure as described above for the Y axis and the Z axis. However, the following formula (15) is used instead of the formula (12) for the Y axis, and the following formula (16) is used for the Z axis instead of the formula (12).

The speed re-smoothing processing unit 71f _converts the speed term of the smoothing speed p dot hat _{x, k} and the smoothing speed error covariance matrix P _{vel, k} calculated by the linear trajectory estimation processing unit 71e into the re-smoothing speed V hat _{x, k.} And the smoothing speed error variance matrix P _{xV, k} and the smoothed wake x hat _k at the sampling time t _k after the replacement is calculated by the following equation (17), and the smoothing error covariance matrix P _k is It calculates by Formula (18).

下付添え字がｙであればＹ軸方向、下付添え字がｚであればＺ軸方向を意味する。

If the subscript is y, it means the Y-axis direction, and if the subscript is z, it means the Z-axis direction.

Ｎステップ予測処理部７１ｇは、最も古い平滑航跡の時刻ｔ_ｂが、Ｎ_ｐサンプル前の時刻ｔ_{ｋ−Ｎｐ＋１}と同じ時刻又はＮ_ｐサンプル前の時刻ｔ_{ｋ−Ｎｐ＋１}より古い時刻であれば（ステップＳＴ５：ＹＥＳの場合）、旋回開始判定用平滑航跡ＤＢ７１ｃに蓄積されている過去Ｎｔｈサンプル分の平滑航跡から、目標の現時点における位置や速度などの状態ベクトルを有する予測航跡を算出するというＮステップ予測処理を実施する（ステップＳＴ６）。
例えば、等速直線運動モデルにしたがって、旋回開始判定用平滑航跡ＤＢ７１ｃに蓄積されているＮ_ｐサンプル前の平滑航跡から直線外挿した予測ベクトルを算出する。When the speed re-smoothing processor 71f performs the speed re-smoothing process, the N-step prediction processing unit 71g includes the smooth wake stored in the turning start determination smooth wake DB 71c as shown in the following equation (19). The time t _b of the oldest smoothed track is compared with the time t _{k−Np + 1} before N _p samples (step ST5).

If the time t _b of the oldest smoothed track is the same time as the time t _{k−Np + 1} before N _p samples or the time t _{k−Np + 1} before N _p samples, the N step prediction processing unit 71g (step ST5: In the case of YES), N-step prediction of calculating a predicted track having a state vector such as the current position and speed of the target from the smooth track for the past Nth samples accumulated in the smooth track DB 71c for turning start determination Processing is performed (step ST6).
For example, according to uniform linear motion model to calculate the prediction vector with interpolation outside straight line from the N _p samples before the smoothing track stored in the turning start determination smooth track DB71c.

Ie, N step prediction processing unit 71g, as shown in equation (20) to (23) below, the predicted vector x hat _k with interpolation straight out from the _{N p} samples before the smoothing track at the sampling time _{t k | k- Np} is calculated, and the prediction error covariance matrix P _{k | k−Np} is calculated.

If it contains the acceleration term to N _p samples smoothing vector and error covariance matrix of the prior smooth track, remove only the terms of the position and velocity, carrying out the above process.
Incidentally, N step prediction processing section 71g, among smooth track stored in the turning start determination smooth track DB71c, time _{t b} of the oldest smooth track is, _{N p} samples before time _{t k-Np + 1} newer If the time (step ST5: in the case of nO), the order N _p samples before smoothing track is not stored in the turning start determination smooth track DB71c, not performed N step prediction processing. In this case, the process proceeds to step ST11.

The Mahalanobis square distance calculation processing unit 71h is output from the sensor tracking processing unit 14 and the position indicated by the predicted track calculated by the N step prediction processing unit 71g when the N step prediction processing unit 71g performs the N step prediction processing. From the position indicated by the target correlated sensor observation value, the Mahalanobis square distance between the position indicated by the predicted track and the position indicated by the target correlated sensor observation value is calculated (step ST7).
That is, the Mahalanobis square distance calculation processing section 71h, as shown in equation (24) below, the predicted vector x hat _k predicted trajectory in the latest sampling time t _{k |} a _k-Np, the output from the sensor tracking processing unit 14 calculating the residual r _k and the position z _k indicated by the target the correlated sensor observations.

In the formula _{(25), z x, k} is the position in the X-axis indicated by the sensor observation value at the sampling time _{t k, z y, k} is the position in the Y-axis indicated by the sensor observation value at the sampling time _{t k, z z , K} is the position on the Z axis indicated by the sensor observation value at the sampling time t _k .
Mahalanobis square distance calculation processing section 71h, as shown in the following equation (27) calculates the residual covariance matrix S _k, using the residual covariance matrix S _k, the following equation (28) As shown, the Mahalanobis square distance ε _v (k) is calculated.

式（２９）において、ε_ａｖ（ｋ）はサンプリング時刻ｔ_ｋにおける時間方向平滑値、αは事前に設定された係数である。When the Mahalanobis square distance calculation processing unit 71h calculates the Mahalanobis square distance ε _v (k), the time direction smoothing processing unit 71i performs a time direction smoothing process for smoothing the Mahalanobis square distance ε _v (k) in the time direction. (Step ST8).
That is, the time direction smoothing processing unit 71i smoothes the Mahalanobis square distance ε _v (k) in the time direction as shown in the following equation (29).

In Expression (29), ε _av (k) is a time direction smooth value at the sampling time t _k , and α is a coefficient set in advance.

閾値判定処理部７１ｊは、その時間方向平滑値ε_ａｖ（ｋ）が閾値Ｔｈ_{ｔｕｒｎｂｇｎ}より大きく、式（３１）が成立する場合（ステップＳＴ９：ＹＥＳの場合）、旋回が開始していると判定して、旋回フラグＤＢ７２に記憶されている旋回フラグを“旋回”に設定する（ステップＳＴ１０）。
閾値判定処理部７１ｊは、その時間方向平滑値ε_ａｖ（ｋ）が閾値Ｔｈ_{ｔｕｒｎｂｇｎ}以下であり、式（３１）が成立しない場合（ステップＳＴ９：ＮＯの場合）、旋回が開始していないと判定して、旋回フラグＤＢ７２に記憶されている“非旋回”の旋回フラグを維持する（ステップＳＴ１１）。
閾値判定処理部７１ｊは、直線軌道推定処理部７１ｅにより直線軌道推定処理が実施されない場合や、Ｎステップ予測処理部７１ｇによりＮステップ予測処理が実施されない場合も、旋回フラグＤＢ７２に記憶されている“非旋回”の旋回フラグを維持する（ステップＳＴ１１）。When the time direction smoothing processing unit 71i smoothes the Mahalanobis square distance ε _v (k) in the time direction, the threshold determination processing unit 71j smooths the Mahalanobis square smoothed in the time direction as shown in the following equation (31). The time direction smooth value ε _av (k), which is the distance, is compared with a _preset threshold value Th _turngn (step ST9).

When the time direction smooth value ε _av (k) is larger than the threshold value Th _turnbgn and the formula (31) is satisfied (step ST9: YES), the threshold determination processing unit 71j determines that the turn has started. Thus, the turning flag stored in the turning flag DB 72 is set to “turn” (step ST10).
When the time direction smooth value ε _av (k) is _{equal to} or less than the threshold value Th _turnbgn and Equation (31) is not satisfied (step ST9: NO), the threshold determination processing unit 71j determines that the turn has not started. Then, the “non-turning” turning flag stored in the turning flag DB 72 is maintained (step ST11).
The threshold determination processing unit 71j is stored in the turning flag DB 72 even when the linear trajectory estimation processing is not performed by the linear trajectory estimation processing unit 71e or when the N step prediction processing is not performed by the N step prediction processing unit 71g. The turning flag “non-turning” is maintained (step ST11).

When the turning end determination processing unit 73 of the turning determination unit 17 receives the target correlated sensor observation value from the sensor tracking processing unit 14, the target correlation completed sensor observation value and the turning flag stored in the turning flag DB 72 are obtained. The turning end determination process for determining the end of turning is performed, and the turning flag stored in the turning flag DB 72 is updated.
FIG. 10 is a flowchart showing the processing contents of the turning end determination processing unit 73.
Hereinafter, the processing content of the turning end determination processing unit 73 will be specifically described with reference to FIG.

The observation value DB73a for turning end determination of the turning end determination processing unit 73 records the target correlated sensor observation value output from the sensor tracking processing unit 14, and the observation value DB 73a for turning end determination stores the past Nth samples. The target correlated sensor observations are accumulated.
The initialization processing unit 73c performs an initialization process of the turning end determination process (step ST21 in FIG. 10).
That is, when the turning flag stored in the turning flag DB 72 is “non-turning”, the initialization processing unit 73c is a target correlation that is accumulated data of the turning end determination observed value DB 73a and the turning end determination smooth track DB 73b. The completed sensor observation value and the smooth track are deleted, and the turning end determination process is ended.
When the turning flag stored in the turning flag DB 72 is “turning”, the initialization processing unit 73c permits the subsequent turning end determination process to be performed. That is, the operations of the trajectory estimation processing unit 73d, the N step prediction processing unit 73e, the Mahalanobis square distance calculation processing unit 73f, the time direction smoothing processing unit 73g, and the threshold determination processing unit 73h are permitted.

The trajectory estimation processing unit 73d performs trajectory estimation processing when the number of target correlated sensor observation values accumulated in the turn end determination observation value DB 73a is equal to or greater than Nth samples (step ST22: YES). (Step ST23).
That is, the trajectory estimation processing unit 73d is accumulated in the smooth wake DB for turning end determination using, for example, a linear least square method or a Kalman filter, assuming that the target motion state is a uniform acceleration motion. Calculates the smooth position, smooth velocity, and smooth acceleration of the target as the smooth track of the target from the observed values of the correlated sensors for the past Nth samples, and calculates the smooth position error covariance matrix and smooth velocity error covariance matrix. The trajectory estimation process is performed.
For example, when the target position is defined by X, Y, and Z in the Cartesian coordinate system, the target smooth position, smooth speed, and smooth acceleration for each axis are calculated by the linear least square method.
Hereinafter, taking the X axis as an example, the calculation process of the trajectory estimation processing unit 73d will be specifically described.

As shown in the following equations (32) to (34) and equations (4) and (5), the trajectory estimation processing unit 73d has a smooth position p hat _{x, k} with respect to the X axis and a smooth speed p dot hat with respect to the X axis. calculating _x, and _k, smooth acceleration p-to-dot hat _x with respect to the _X-axis, and _k.
Here, because of the electronic application, in the text of the description, the symbols “··” and “^” cannot be added on the characters, so that the p-two-dot hats _{x and k} It is written as follows.

The trajectory estimation processing unit 73d calculates a smooth position, a smooth speed, and a smooth acceleration with respect to the Y axis and the Z axis in the same procedure as described above.
The trajectory estimation processing unit 73d stores the smooth position, smooth speed, and smooth acceleration in the smooth wake DB 73b for turning completion determination as a smooth wake.

Further, when the target position is defined by X, Y, and Z of the orthogonal coordinate system, the trajectory estimation processing unit 73d, as shown in the above formulas (6) to (9), the smooth position vector p hat _k , A velocity vector p dot hat _k , a smooth position error covariance matrix P _{pos, k} , and a smooth velocity error covariance matrix P _{vel, k} are defined.
The trajectory estimation processing unit 73d performs trajectory estimation processing when the number of target correlated sensor observation values accumulated in the turn end determination observation value DB 73a is less than Nth samples (step ST22: NO). Not implemented. In this case, the process proceeds to step ST30.

Ｎステップ予測処理部７３ｅは、最も古い平滑航跡の時刻ｔ_ｂが、Ｎ_ｐサンプル前の時刻ｔ_{ｋ−Ｎｐ＋１}と同じ時刻又はＮ_ｐサンプル前の時刻ｔ_{ｋ−Ｎｐ＋１}より古い時刻であれば（ステップＳＴ２４：ＹＥＳの場合）、旋回終了判定用平滑航跡ＤＢ７３ｂに蓄積されている過去Ｎｔｈサンプル分の平滑航跡から、目標の現時点における位置や速度などの状態ベクトルを有する予測航跡を算出するというＮステップ予測処理を実施する（ステップＳＴ２５）。When the trajectory estimation processing unit 73d performs the trajectory estimation process, the N-step prediction processing unit 73e, among the smooth wakes accumulated in the smooth wake DB 73b for turning completion determination, as shown in the following equation (35), The time t _b of the oldest smoothed track is compared with the time t _{k−Np + 1} before N _p samples (step ST24).

If the time t _b of the oldest smoothed track is the same as the time t _{k−Np + 1} before N _p samples or the time t _{k−Np + 1} before N _p samples, the N step prediction processing unit 73e (step S ST24: YES), N-step prediction of calculating a predicted track having a state vector such as the current position and speed of the target from the smooth track for the past Nth samples accumulated in the smooth track DB 73b for turning completion determination Processing is performed (step ST25).

The N step prediction processing of the N step prediction processing unit 73e is the same as the N step prediction processing of the N step prediction processing unit 71g shown in FIG.
Incidentally, N step prediction processing section 73e is in the smooth track stored in the pivot end decision smooth track DB73b, time _{t b} of the oldest smooth track is, _{N p} samples before time _{t k-Np + 1} newer If the time (step ST24: in the case of nO), the order N _p samples before smoothing track is not stored in the pivot end decision smooth track DB73b, not performed N step prediction processing. In this case, the process proceeds to step ST30.

The Mahalanobis square distance calculation processing unit 73f outputs the position indicated by the predicted track calculated by the N step prediction processing unit 73e and the sensor tracking processing unit 14 when the N step prediction processing unit 73e performs the N step prediction processing. From the position indicated by the target correlated sensor observation value, the Mahalanobis square distance between the position indicated by the predicted track and the position indicated by the target correlated sensor observation value is calculated (step ST26).
The Mahalanobis square distance calculation processing in the Mahalanobis square distance calculation processing unit 73f is the same as the Mahalanobis square distance calculation processing in the Mahalanobis square distance calculation processing unit 71h shown in FIG.

When the Mahalanobis square distance calculation processing unit 73f calculates the Mahalanobis square distance ε _v (k), the time direction smoothing processing unit 73g performs time direction smoothing processing for smoothing the Mahalanobis square distance ε _v (k) in the time direction. (Step ST27).
For smoothing the Mahalanobis squared distance ε _{v (k)} in the time direction smoothing unit 73g is similar to the smoothing process Mahalanobis squared distance ε _{v (k)} in the time direction smoothing unit 71i illustrated in FIG. 5, Detailed description is omitted.

閾値判定処理部７３ｈは、その時間方向平滑値ε_ａｖ（ｋ）が閾値Ｔｈ_{ｔｕｒｎｅｎｄ}以下となり、式（３６）が成立する場合（ステップＳＴ２８：ＹＥＳの場合）、旋回が終了していると判定して、旋回フラグＤＢ７２に記憶されている旋回フラグを“非旋回”に設定する（ステップＳＴ２９）。
閾値判定処理部７３ｈは、その時間方向平滑値ε_ａｖ（ｋ）が閾値Ｔｈ_{ｔｕｒｎｂｇｎ}より大きく、式（３６）が成立しない場合（ステップＳＴ２８：ＮＯの場合）、旋回が終了していないと判定して、旋回フラグＤＢ７２に記憶されている“旋回”の旋回フラグを維持する（ステップＳＴ３０）。
閾値判定処理部７３ｈは、軌道推定処理部７３ｄにより軌道推定処理が実施されない場合や、Ｎステップ予測処理部７３ｅによりＮステップ予測処理が実施されない場合も、旋回フラグＤＢ７２に記憶されている“旋回”の旋回フラグを維持する（ステップＳＴ３０）。When the time direction smoothing processing unit 73g smoothes the Mahalanobis square distance ε _v (k) in the time direction, the threshold value determination processing unit 73h, as shown in the following formula (36), Mahalanobis square smoothed in the time direction. The time direction smooth value ε _av (k), which is the distance, is compared with a _preset threshold value Th _turnend (step ST28).

When the time direction smooth value ε _av (k) is equal to or less than the threshold value Th _turnend and the equation (36) is satisfied (in the case of YES at step ST28), the threshold determination processing unit _73h determines that the turn has ended. Thus, the turning flag stored in the turning flag DB 72 is set to “non-turning” (step ST29).
When the time direction smooth value ε _av (k) is larger than the threshold value Th _turnbgn and Equation (36) is not satisfied (step ST28: NO), the threshold determination processing unit _73h determines that the turn has not ended. Thus, the “turn” turn flag stored in the turn flag DB 72 is maintained (step ST30).
The threshold determination processing unit 73h is “turn” stored in the turning flag DB 72 even when the trajectory estimation processing unit 73d does not perform the trajectory estimation processing or when the N step prediction processing unit 73e does not perform the N step prediction processing. The turning flag is maintained (step ST30).

Here, the threshold determination processing unit 73h is an example of comparing the Mahalanobis square distance and a threshold value which is smoothed in the time direction, the threshold determination processing unit 73h is time direction smoothing value at the current time t _k ε _{av (k} ) and, before the time _{t k-1} calculates a difference value [Delta] [epsilon] _av (k) of the time direction smoothing value ε _av (k-1) in, the difference value [Delta] [epsilon] _av (k) a threshold number of times that is negative Th _If it exceeds _cnt, it may be determined that the turn has ended.
FIG. 11 is a flowchart showing the processing contents of the threshold determination processing unit 73h. Hereinafter, the processing content of the threshold determination processing unit 73h will be specifically described with reference to FIG.

閾値判定処理部７３ｈは、差分値Δε_ａｖ（ｋ）を算出すると、その差分値Δε_ａｖ（ｋ）が負であれば（ステップＳＴ３２：ＹＥＳの場合）、差分値カウンタＤＢ３７ｉに記録されているカウント値Ｃを１だけ増やすインクリメント処理を実施する（ステップＳＴ３３）。そのカウント値Ｃは、差分値Δε_ａｖ（ｋ）が負になった回数を示すものであり、差分値カウンタＤＢ３７ｉに記録されているカウント値Ｃの初期値は０である。
閾値判定処理部７３ｈは、その差分値Δε_ａｖ（ｋ）が０以上であれば（ステップＳＴ３２：ＮＯの場合）、差分値カウンタＤＢ３７ｉに記録されているカウント値Ｃを維持する（ステップＳＴ３４）。Every time the time direction smoothing processing unit 73g calculates the time direction smoothing value ε _av (k) at the current time t _k , the threshold determination processing unit 73h, as shown in the following equation (37), at the current time t _k A difference value Δε _av (k) between the time direction smooth value ε _av (k) and the time direction smooth value ε _av (k−1) at the previous time t _k−1 is calculated (step ST31 in FIG. 11).

Threshold determination processing unit 73h, when calculating a difference value [Delta] [epsilon] _av (k), if the difference value [Delta] [epsilon] _av (k) is negative (step ST32: YES), of the count stored in the difference value counter DB37i An increment process for incrementing the value C by 1 is performed (step ST33). The count value C indicates the number of times the difference value Δε _av (k) has become negative, and the initial value of the count value C recorded in the difference value counter DB 37 i is 0.
If the difference value Δε _av (k) is equal to or greater than 0 (step ST32: NO), the threshold determination processing unit 73h maintains the count value C recorded in the difference value counter DB 37i (step ST34).

Next, as shown in the following formula (38), the threshold determination processing unit 73h compares the number of times C recorded in the difference value counter DB 37i with a preset threshold Th _cnt (step ST35).
C> Th _cnt (38)
When the number of times C recorded in the difference value counter DB 37i is greater than the threshold value Th _cnt and Expression (38) is satisfied (in the case of YES at step ST35), the threshold value determination processing unit 73h determines that the turn has ended. Determination is made and the turning flag stored in the turning flag DB 72 is set to “non-turning” (step ST36).
Further, the threshold determination processing unit 73h stores in the turning flag DB 72 when the number of times C recorded in the difference value counter DB 37i is equal to or less than the threshold Th _cnt and Expression (38) is not satisfied (in the case of NO at step ST35). The turning flag for “turning” is maintained (step ST37).
When the turning flag stored in the turning flag DB 72 is set to “non-turning”, the threshold determination processing unit 73h initializes the number of times C recorded in the difference value counter DB 37i to 0.

When the operation of the turn start determination processing unit 71 or the turn end determination processing unit 73 is completed, the turn flag output processing unit 74 of the turn determination unit 17 uses the turn flag stored in the turn flag DB 72 as the turn level calculation processing unit 75 and The result is output to the retracking processing unit 16.
When the turning level calculation processing unit 75 of the turning determination unit 17 receives the turning flag from the turning flag output processing unit 74, the target correlated sensor output from the sensor tracking processing unit 14 when the turning flag is "turning". From the observed value, a turning level indicating the degree of turning of the target is calculated, and the turning level is output to the retracking processing unit 16.
Hereinafter, the processing content of the turning level calculation process part 75 is demonstrated concretely.

If the turning flag output from the turning flag output processing unit 74 is “non-turning”, the initialization processing unit 75a of the turning level calculation processing unit 75 completes the target correlation stored in the observation value DB 75b for turning level calculation. Delete sensor observations.
When the turning flag output from the turning flag output processing unit 74 is “turning”, the initialization processing unit 75a uses the target correlated sensor observation value output from the sensor tracking processing unit 14 as the turning level calculation observation value DB 75b. To store.

The turning trajectory estimation processing unit 75c of the turning level calculation processing unit 75 assumes, for example, that the target motion state is a uniform acceleration motion, and uses, for example, a linear least square method, a Kalman filter, or the like to observe the turning level calculation observation value. The target acceleration for the X axis, the Y axis, and the Z axis in the orthogonal coordinate system is calculated from the target correlated sensor observation values accumulated in the DB 75b.
When calculating the target acceleration with respect to the X axis, the Y axis, and the Z axis, the turning trajectory estimation processing unit 75c calculates the acceleration magnitude a hat as shown in the following equation (39).

When calculating the acceleration magnitude a hat, the turning trajectory estimation processing unit 75 c converts the acceleration magnitude a hat into a turning level and outputs the turning level to the retracking processing unit 16.
It is assumed that the acceleration magnitude a hat and the turning level are directly proportional to each other, and the turning level increases as the acceleration magnitude a hat increases.

The partial track storage unit 15 of the tracking processing unit 12 stores the partial track of the partial track previously estimated by the re-tracking processing unit 16 and the target correlated sensor observation value output from the sensor tracking processing unit 14. The target correlated sensor observation value in which the correlation is recognized is stored.
The partial wake is basically composed of a state vector consisting of time, position, and speed and an error covariance matrix of the state vector. In addition, a target number for identifying a target and a sensor 11 are identified. An identification number such as a sensor number is assigned.
The identification number may be, for example, a number that identifies a target, a number that identifies the type of aircraft such as an aircraft or a satellite, or a number that is information indicating whether it is an enemy or a friend.
The state vector of the partial wake may be a state vector including the position of the sensor observation value, a state vector including the position and velocity, or a state vector including even the acceleration.

When the re-tracking processing unit 16 of the tracking processing unit 12 receives the target correlated sensor observation value from the sensor tracking processing unit 14, for example, the re-tracking processing unit 16 executes the known correlation algorithm such as GNN, thereby performing the target correlated sensor observation value. It is determined whether or not the position indicated by and the position indicated by the partial track stored in the partial track storage unit 15 are correlated.
That is, the re-tracking processing unit 16 determines whether the target correlated sensor observation value output from the sensor tracking processing unit 14 is observing the position of the target related to the partial track estimated in the past. Determine.

The re-tracking processing unit 16 correlates the position indicated by the target correlated sensor observation value with the position indicated by the partial wake stored in the partial wake storage unit 15, and the target correlated sensor observation value is If it is determined that it can be associated with the partial track, the target correlated sensor observation value is stored in the partial track storage unit 15.
In addition, the re-tracking processing unit 16 executes a known tracking algorithm such as a Kalman filter, for example, so that the target correlated sensor observation value and the past target correlated sensor observation value stored in the partial track storage unit 15 are stored. From these, a partial wake having a state vector including the current position and speed of the target is estimated.
At this time, the re-tracking processing unit 16 may execute a tracking algorithm using a Kalman filter corresponding to the turning flag output from the turning determination unit 17.
For example, if the turning flag output from the turning determination unit 17 is “non-turning”, the re-tracking processing unit 16 calculates a state vector including the position and speed by a Kalman filter based on the constant velocity linear motion model, and the turning If the flag is “turn”, a state vector including position, velocity, and acceleration may be calculated by a Kalman filter based on a constant acceleration motion model.

When the re-tracking processing unit 16 estimates the partial track, the re-tracking processing unit 16 replaces the partial track stored in the partial track storage unit 15 with the estimated partial track, and the turn determination unit 17 performs the estimation on the estimated partial track. The output turning flag and turning level are added, and the partial track with the turning flag and turning level is output to the transmission determination unit 19.
When the re-tracking processing unit 16 does not correlate the position indicated by the target correlated sensor observation value output from the sensor tracking processing unit 14 and the position indicated by the partial track stored in the partial track storage unit 15, The partial track with the turning flag and the turning level is not output to the transmission determination unit 19.

When the transmission determination unit 19 receives the partial track with the turning flag and the turning level from the re-tracking processing unit 16, the time indicated by the partial track and the position indicated by the fusion track stored in the fusion track storage unit 18. Whether or not to transmit the partial wake is determined from the time difference between the time and the turn flag and the turn level attached to the partial wake.
At this time, the transmission determination unit 19 displays the partial track with the turn flag and the turn level when the turn flag attached to the partial track is “turn” rather than when the turn flag is “non-turn”. The frequency of issuing a determination result indicating that transmission is to be performed is increased.
In addition, the higher the turn level attached to the partial track, the higher the frequency of issuing a determination result indicating that the partial track with the turn flag and the turn level is transmitted.
When the transmission determination unit 19 determines to transmit the partial track, the transmission determination unit 19 outputs the partial track to the transmission / reception unit 20 and then deletes the partial track stored in the partial track storage unit 15.
Hereinafter, the transmission determination process of the partial wake by the transmission determination part 19 is demonstrated concretely.

The parameter change processing unit 81 of the transmission determination unit 19 changes the system noise parameter for transmission determination according to the turning flag and the turning level attached to the partial track output from the retracking processing unit 16, and transmits the transmission threshold value. The threshold for transmission judgment such as the difference threshold and the transmission hold threshold is changed. Assume that the initial value of the threshold value for transmission determination is set in advance by the user.
The system noise parameter for transmission judgment is a parameter for compensating for the motion error when the target motion deviates from the assumed motion model. As the system noise parameter increases, the error of the fusion wake described later The rate of increase increases.
The parameter change processing unit 81 sets the system noise parameter so that the system noise parameter is higher when the turning flag attached to the partial track is “turning” than when the turning flag is “not turning”. change.
Further, when the turning flag attached to the partial wake is “turn”, the parameter change processing unit 81 is configured so that the higher the turning level attached to the partial wake, the higher the system noise parameter. Change system noise parameters.

式（４０）において、ｔ_ｋ ^（Ｓ）は現サンプリング時刻を示し、上付き添え字の（Ｓ）は再追尾処理部１６から出力された部分航跡に関わる旨を示す変数である。
また、ｔ_ｍ−１ ^（Ｆ）は融合追尾処理部２３により融合航跡が推定された時刻を示し、上付き添え字の（Ｆ）は融合航跡に関わる旨を示す変数である。The fusion wake prediction processing unit 82 of the transmission determination unit 19 includes the time of the partial wake with the turn flag and the turn level output from the retracking processing unit 16, and the time of the fusion wake stored in the fusion wake storage unit 18. The time difference is calculated.
That is, as shown in the following equation (40), the fused track prediction processing unit 82 outputs the time t _k ^(S) of the partial track with the turning flag and the turning level output from the retracking processing unit 16, and the fused track. A time difference ΔT _k with the time t _m−1 ^(F) of the fusion wake stored in the storage unit 18 is calculated.

In Expression (40), t _k ^(S) indicates the current sampling time, and the superscript (S) is a variable indicating that the partial track output from the retracking processing unit 16 is involved.
Further, t _m-1 ^(F) indicates the time when the fusion track is estimated by the fusion tracking processing unit 23, and the superscript (F) is a variable indicating that the fusion track is involved.

ここで、Φ_ｋは状態遷移行列、Ｑ^（Ｆ）はシステム雑音共分散行列、ｑ_ｘ ^（Ｆ）はパラメータ変更処理部８１により変更されたＸ軸におけるシステム雑音パラメータ、ｑ_ｙ ^（Ｆ）はパラメータ変更処理部８１により変更されたＹ軸におけるシステム雑音パラメータ、ｑ_ｚ ^（Ｆ）はパラメータ変更処理部８１により変更されたＺ軸におけるシステム雑音パラメータである。
また、Ｐ_{ｓ，ｍ−１} ^（Ｆ）は融合航跡の誤差共分散行列である。When the fusion wake prediction processing unit 82 calculates the time difference ΔT _k between the time t _k ^(S) of the partial wake and the time t _m−1 ^(F) of the fusion wake, the following equation (41) to (44) is obtained. As shown, using the time difference ΔT _k and the system noise parameter changed by the parameter change processing unit 81, the prediction error covariance matrix P _{p, m} ^(F for transmission judgment at time t _k ^(S) of the partial track is shown. ⁾ Is calculated.

Here, Φ _k is a state transition matrix, Q ^(F) is a system noise covariance matrix, q _x ^(F) is a system noise parameter on the X axis changed by the parameter change processing unit 81, and q _y ^(F) is a parameter. The system noise parameter on the Y axis changed by the change processing unit 81, q _z ^(F), is the system noise parameter on the Z axis changed by the parameter change processing unit 81.
P _{s, m−1} ^(F) is an error covariance matrix of the fusion wake.

When the fused track prediction processing unit 82 calculates the transmission error prediction error covariance matrix P _{p, m} ^(F) , the determination error calculation processing unit 83 calculates the transmission error prediction error covariance matrix P _{p, m.} ^{From (F)} , the fusion track stored in the fusion track storage unit 18, and the partial track output from the retracking processing unit 16, the fusion track error, the updated fusion track error, and the fusion track error difference are calculated. .
Here, FIG. 12 is an explanatory diagram showing a transmission determination concept in the transmission determination unit 19.
In FIG. 12, the fusion track error indicates an error of the fusion track before being merged with the partial track output from the retracking processing unit 16, and the updated fusion track error is a portion output from the retracking processing unit 16. The error of the fusion wake after being merged with the wake is shown.
Also, the fusion track error difference indicates a difference obtained by subtracting the updated fusion track error from the fusion track error.

When the determination error calculation processing unit 83 calculates the fusion track error, the updated fusion track error, and the fusion track error difference, the determination processing unit 84 calculates the fusion track error, the updated fusion track error, the fusion track error difference, and the parameters. Using the transmission threshold value, the difference threshold value, and the transmission hold threshold value changed by the change processing unit 81, it is determined whether or not to transmit the partial track with the turning flag and the turning level output from the retracking processing unit 16.
That is, when the fusion track error difference is larger than the difference threshold, the determination processing unit 84 determines that the contribution of reducing the fusion track error by fusing the partial track output from the retracking processing unit 16 with the fusion track is large. Then, a determination result indicating that the partial track is transmitted, that is, a transmission determination flag of “transmission” and a partial track with a turning flag and a turning level are output to the transmission / reception unit 20, and are also stored in the partial track storage unit 15. The stored partial track is deleted.
When the fusion track error difference is equal to or smaller than the difference threshold, the determination processing unit 84 determines that the contribution of reducing the fusion track error by fusing the partial track with the fusion track is small, and suspends transmission of the partial track. A determination result indicating that, that is, a transmission determination flag of “transmission pending”, and a partial track with a turning flag and a turning level are output to the transmitting / receiving unit 20.

If the fused track error is larger than the transmission threshold, the determination processing unit 84 determines that the fused track needs to be merged with the partial track regardless of the size of the fused track error difference because the fused track error is large. Then, a determination result indicating that the partial track is transmitted, that is, a transmission determination flag of “transmission” and a partial track with a turning flag and a turning level are output to the transmission / reception unit 20, and are also stored in the partial track storage unit 15. The stored partial track is deleted.
If the fused wake error is less than or equal to the transmission threshold, the determination processing unit 84 determines that the fused wake need not be merged with the partial wake if the fused wake error difference is less than or equal to the difference threshold. A determination result indicating that transmission is suspended, that is, a transmission determination flag of “transmission pending”, and a partial track with a turning flag and a turning level are output to the transmission / reception unit 20. However, if the fusion track error difference is larger than the difference threshold, a “transmission” transmission determination flag may be output to the transmission / reception unit 20.

FIG. 13 is a conceptual diagram showing the temporal change of the fusion wake error before and during the turn and the transmission frequency of the partial wake.
When the target is not turning, since the system noise parameter for transmission determination is kept low by the parameter change processing unit 81 and the rate of increase of the fusion track error in the time direction is small, until the fusion track error exceeds the transmission threshold It takes a lot of time. For this reason, when the turning flag is “non-turning”, the transmission frequency of the partial wake is low.
On the other hand, when the target starts turning, the turning flag becomes “turning”, and when the parameter change processing unit 81 increases the system noise parameter for transmission determination, the rate of increase of the fusion wake error in the time direction increases. In a short time, the fusion track error exceeds the transmission threshold. For this reason, when the turning flag is “turning”, the transmission frequency of the partial wake is high.
FIG. 13 shows an example in which the transmission threshold is fixed. For this reason, when the target is turning, the frequency at which the fusion track error exceeds the transmission threshold value may increase, and the communication load on the network 2 may become too large.
In order to prevent the communication load on the network 2 from becoming excessively large, the parameter change processing unit 81 sets the turn flag to “turn” and increases the transmission threshold when the system noise parameter for transmission determination is increased. It may be. However, if the transmission threshold is set too high, the transmission frequency of the partial track cannot be increased, so it is necessary to consider the balance between the system noise parameter and the transmission threshold.

In order to avoid a situation in which the communication load in the network 2 becomes excessively large, the partial wake transmission determination may be performed without using the transmission threshold.
FIG. 14 is a conceptual diagram showing the temporal change of the fusion wake error before and during the turn when the transmission threshold is not used, and the transmission frequency of the partial wake.
In FIG. 14, the determination processing unit 84 outputs a determination result indicating that the partial track is transmitted only when the fusion track error difference is larger than the difference threshold, that is, a “transmission” transmission determination flag to the transmission / reception unit 20. An example is shown.
In the example of FIG. 14, the parameter change processing unit 81 increases the system noise parameter for transmission determination, so that the determination result indicating that the partial wake is transmitted even if the rate of increase of the fusion wake error in the time direction increases. Therefore, it is possible to avoid a situation in which the communication load on the network 2 becomes too large.

When the transmission / reception unit 20 receives the “transmission” transmission determination flag and the partial track with the turn flag and the turn level from the transmission determination unit 19, the transmission / reception unit 20 receives the turn track and the partial track with the turn level via the network 2. To the target tracking device 1-M and output the partial track to the track fusion unit 21. When the transmission / reception unit 20 receives a transmission determination flag of "transmission pending" from the transmission determination unit 19, the other target tracking device The partial wake transmission process for 1-M and the partial wake output process for the wake fusion unit 21 are not performed.
When the transmission / reception unit 20 receives a partial wake transmitted from another target tracking device 1 -M via the network 2, the transmission / reception unit 20 outputs the partial wake to the wake fusion unit 21.

When the correlation determination unit 22 of the wake fusion unit 21 receives a partial wake with a turn flag and a turn level from the transmission / reception unit 20, for example, by executing a known correlation algorithm such as GNN, the position indicated by the partial wake Then, it is determined whether or not the position indicated by the fusion track stored in the fusion track storage unit 18 is correlated.
When it is determined that the correlation determination unit 22 correlates, the fusion tracking processing unit 23 of the wake fusion unit 21 uses the partial wake output from the transmission / reception unit 20 and the fusion wake stored in the fusion wake storage unit 18. Estimating a fusion wake having a state vector including the current position and speed of the target.

That is, the fusion tracking processing unit 23 calculates the estimated vector x hat _{s, m} ^(F) of the fusion wake as indicated by the following equations (45) to (49).

また、目標の状態ベクトルを３次元直交座標系で定義すると、式（５０）におけるＤは、下記の式（５２）のように定義される。

式（５２）において、ｑ_ｘは融合追尾処理部２３により事前に設定されたＸ軸におけるシステム雑音パラメータ、ｑ_ｙは融合追尾処理部２３により事前に設定されたＹ軸におけるシステム雑音パラメータ、ｑ_ｚは融合追尾処理部２３により事前に設定されたＺ軸におけるシステム雑音パラメータである。
システム雑音パラメータｑ_ｘ，ｑ_ｙ，ｑ_ｚは目標の旋回度合いに応じて大きく設定した方が、融合航跡の追尾精度が向上するため、融合追尾処理部２３によって、部分航跡に付加されている旋回レベルに応じて設定されているものであってもよい。即ち、その旋回レベルが示す目標の旋回の度合いが大きいほど、システム雑音パラメータｑ_ｘ，ｑ_ｙ，ｑ_ｚが大きな値に設定されているものであってもよい。Here, when the fusion tracking processing unit 23 uses a Kalman filter based on a constant velocity linear motion model, the system noise covariance matrix Q ^(F) is defined by the following equations (50) and (51).

Further, when the target state vector is defined in the three-dimensional orthogonal coordinate system, D in the equation (50) is defined as the following equation (52).

In Expression (52), q _x is a system noise parameter on the X axis set in advance by the fusion tracking processing unit 23, q _y is a system noise parameter on the Y axis set in advance by the fusion tracking processing unit 23, q _z Is a system noise parameter in the Z axis set in advance by the fusion tracking processing unit 23.
When the system noise parameters q _x , q _y , and q _z are set larger in accordance with the target turning degree, the tracking accuracy of the fusion wake is improved. It may be set according to the level. That is, the system noise parameters q _x , q _y , and q _z may be set to a larger value as the degree of target turning indicated by the turning level is larger.

When the fusion track is estimated, the fusion tracking processing unit 23 performs a process of replacing the fusion track stored in the fusion track storage unit 18 with the estimated fusion track.
The fusion tracking processing unit 23 determines that the fusion wake stored in the fusion wake storage unit 18 is in a memory track state when the correlation determination unit 22 determines that there is no correlation, and the fusion wake is partially recorded. Without updating the wake, the partial wake output from the transmission / reception unit 20 is recorded in the fusion wake storage unit 18 as a new fusion wake.

  The display processing unit 24 displays on the display the fusion track estimated by the fusion tracking processing unit 23, the partial track determined to be correlated with the fusion track by the correlation determination unit 22, and the like.

As is apparent from the above, according to the first embodiment, the turn determination unit 17 that determines whether or not the target is turning from the time-series data of the sensor observation values, and the portion estimated by the tracking processing unit 12 Whether or not to transmit the partial track estimated by the tracking processing unit 12 based on the time difference between the time of the track and the time of the combined track stored in the combined track storage unit 18 and the determination result of the turning determination unit 17 is determined. The transmission determination unit 19 is provided, and when the transmission / reception unit 20 indicates that the determination result of the transmission determination unit 19 transmits the partial track, the partial track is transmitted to the other target tracking device 1-M. Thus, the communication capacity can be reduced while ensuring the tracking accuracy of the turning target.
That is, according to this Embodiment 1, when the target is turning, the transmission frequency of a partial track can be raised and the tracking accuracy of a target can be raised. On the other hand, when the target is not turning, the transmission frequency of the partial wake can be lowered to reduce the communication capacity.

Embodiment 2. FIG.
In the second embodiment, a flag integration unit that updates a turn flag attached to the fusion track stored in the fusion track storage unit 18 using the turn flag attached to the partial track output from the transmission / reception unit 20. A target tracking device provided with

FIG. 15 is a block diagram showing a target tracking device according to Embodiment 2 of the present invention, and FIG. 16 is a hardware configuration diagram of the target tracking device according to Embodiment 2 of the present invention.
15 and FIG. 16, the same reference numerals as those in FIG. 1 and FIG.
The flag integration unit 91 is realized by a flag integration processing circuit 50 configured by, for example, a semiconductor integrated circuit mounting a CPU or a one-chip microcomputer. The flag integration unit 91 uses the turning flag attached to the partial wake output from the transmission / reception unit 20 to update the turning flag attached to the fusion wake stored in the fusion wake storage unit 18.

  In FIG. 15, the sensor 11, sensor track storage unit 13, sensor tracking processing unit 14, partial track storage unit 15, retracking processing unit 16, turn determination unit 17, fusion track storage unit 18, which are components of the target tracking device, Each of the transmission determination unit 19, the transmission / reception unit 20, the correlation determination unit 22, the fusion tracking processing unit 23, the display processing unit 24, and the flag integration unit 91 includes dedicated hardware as illustrated in FIG. 16, that is, a storage processing circuit 41. , Sensor tracking processing circuit 42, re-tracking processing circuit 43, turning determination processing circuit 44, transmission determination processing circuit 45, transmission / reception processing circuit 46, correlation determination processing circuit 47, fusion tracking processing circuit 48, display processing circuit 49, and flag integration processing The one realized by the circuit 50 is assumed.

The components of the target tracking device are not limited to those realized by dedicated hardware, and the target tracking device may be realized by software, firmware, or a combination of software and firmware.
When the target tracking device is realized by software, firmware, or the like, the sensor track storage unit 13, the partial track storage unit 15, and the fusion track storage unit 18 are configured on the memory 61 shown in FIG. The processing procedures of the re-tracking processing unit 16, the turning determination unit 17, the fusion track storage unit 18, the transmission determination unit 19, the transmission / reception unit 20, the correlation determination unit 22, the fusion tracking processing unit 23, the display processing unit 24, and the flag integration unit 91 are as follows. A program to be executed by the computer may be stored in the memory 61 illustrated in FIG. 3, and the processor 62 illustrated in FIG. 3 may execute the program stored in the memory 61.

15 shows an example in which each component of the target tracking device is realized by dedicated hardware, and FIG. 16 shows an example in which the target tracking device is realized by software, firmware, etc. Some components in the tracking device may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, or the like.
For example, the sensor 11, the transmission / reception unit 20, and the display processing unit 24 are realized by dedicated hardware, the sensor track storage unit 13, the sensor tracking processing unit 14, the partial track storage unit 15, the retracking processing unit 16, and the turning determination unit 17. The fusion track storage unit 18, the transmission determination unit 19, the correlation determination unit 22, the fusion tracking processing unit 23, and the flag integration unit 91 can be realized by software, firmware, or the like. However, the combination of dedicated hardware and software is arbitrary.

FIG. 17 is a block diagram showing the turning determination unit 17 of the target tracking device according to the second embodiment of the present invention. In FIG. 17, the same reference numerals as those in FIG.
If the turn flag stored in the turn flag DB 72 is “turn” and the turn flag attached to the fusion track stored in the fusion track storage unit 18 is “turn”, the flag update processing unit 101 is “turn”. The turn flag is output to the turn start determination processing unit 71 and the erroneous determination release processing unit 103 without changing the “turn” of the turn flag, and the turn flag stored in the turn flag DB 72 is “turn”. If the turn flag attached to the combined wake is “non-turn”, the turn flag is changed to “non-turn”, and the turn flag is turned to the turn start determination processing unit 71 and erroneous determination release processing. Processing to output to the unit 103 is performed.
If the turning flag stored in the turning flag DB 72 is “non-turning”, the flag update processing unit 101 sets the turning flag to the turning start determination processing unit without changing the “non-turning” of the turning flag. 71 and the process of outputting to the erroneous determination release processing unit 103 is performed.

  The turning start time database (hereinafter referred to as “turning start time DB”) 102 is set as the turning start time when the threshold determination processing unit 71j of the turning start determination processing unit 71 first determines that the turning has started. It is a memory to store.

The erroneous determination cancellation processing unit 103 performs a process of correcting the turning flag stored in the turning flag DB 72 by using the turning flag attached to the integrated track stored in the integrated track storage unit 18.
That is, the misjudgment cancellation processing unit 103 calculates a difference time that is a time difference between the observation time of the target correlated sensor observation value output from the sensor tracking processing unit 14 and the turning start time stored in the turning start time DB 102. Perform the calculation process.
Further, the misjudgment cancellation processing unit 103 indicates that the turning flag output from the flag update processing unit 101 is “turning”, and the turning flag attached to the fusion track stored in the fusion track storage unit 18 is “turning”. "," The turning flag is output to the turning flag DB 72, the turning level calculation processing unit 75, and the retracking processing unit 16 without changing the "turning" of the turning flag.
Further, the misjudgment canceling processing unit 103 has the turning flag output from the flag update processing unit 101 “turning”, the turning flag attached to the fusion track is “non-turning”, and the difference time is If it is less than the preset release threshold, the turning flag is output to the turning flag DB 72, the turning level calculation processing unit 75, and the retracking processing unit 16 without changing the “turning” of the turning flag. To implement.
The misjudgment release processing unit 103 has the turn flag output from the flag update processing unit 101 “turn”, the turn flag attached to the fusion track is “non-turn”, and the difference time is in advance. If it is equal to or greater than the set release threshold, the turning flag is changed to “non-turning”, and the turning flag is output to the turning flag DB 72, the turning level calculation processing unit 75, and the retracking processing unit 16. To do.
Further, when the turning flag output from the flag update processing unit 101 is “non-turning”, the erroneous determination release processing unit 103 sets the turning flag to the turning flag DB 72 without changing the “non-turning” of the turning flag. Then, a process of outputting to the turning level calculation processing unit 75 and the retracking processing unit 16 is performed.

18 is a block diagram showing the turning start determination processing unit 71 of the turning determination unit 17. In FIG. 18, the same reference numerals as those in FIG.
The components of the turning start determination processing unit 71 shown in FIG. 18 are the same as those of the turning start determination processing unit 71 shown in FIG. 5, but the initialization processing unit 71d of the turning start determination processing unit 71 shown in FIG. 5 is the turning flag output from the flag update processing unit 101, whereas the turning flag is the input data of the initialization processing unit 71d of the turning start determination processing unit 71 shown in FIG. Is a turning flag stored in the turning flag DB 72.
18 includes not only the turning flag DB 72 but also the turning start time DB 102, the re-tracking processing unit 16, and the turning level calculation processing unit 75. This is different from the threshold determination processing unit 71j of the turning start determination processing unit 71 shown in FIG.

FIG. 19 is a block diagram showing the turning end determination processing unit 73 of the turning determination unit 17. In FIG. 19, the same reference numerals as those in FIG.
The turn ending determination fusion track database (hereinafter referred to as “turn end determination fusion track DB”) 73j is a memory for storing the fusion track.
When the turning flag attached to the fusion track stored in the fusion track storage unit 18 is “non-turn”, the initialization processing unit 73k deletes the smooth track stored in the turn completion determination fusion track DB 73j. Then, the turning end determination process ends.
In addition, when the turning flag attached to the fusion wake is “turn”, the initialization processing unit 73k stores the fusion wake in the turn ending determination fusion wake DB 73j, and performs the subsequent turn end determination processing. Allow.

The N-step prediction processing unit 73l is the fusion track stored in the fusion track storage unit 18 by the correlation determination unit 22 and the time of the oldest fusion track among the fusion tracks stored in the fusion track DB 73j for determining the completion of the turn. If the time difference from the time of the partial wake determined to be correlated with the time is equal to or greater than a preset threshold value, from the wake of the past Nth samples stored in the turn ending determination wake DB 73j. Then, a process for calculating a predicted wake having a state vector such as the current position and speed of the target is performed.
The Mahalanobis square distance calculation processing unit 73m is a portion determined to be correlated with the prediction track calculated by the N-step prediction processing unit 73l and the fusion track stored in the fusion track storage unit 18 by the correlation determination unit 22. From the wake, a process of calculating the Mahalanobis square distance between the prediction vector constituting the predicted wake and the partial wake is performed.

Next, the operation will be described.
Except for the flag integration unit 91 and the turning determination unit 17, the processing contents of the flag integration unit 91 and the turning determination unit 17 are mainly described in the second embodiment because they are substantially the same as those in the first embodiment.
When the flag integration unit 91 receives a turn flag and a partial wake with a turn level from the transmission / reception unit 20, the flag integration unit 91 uses the turn flag attached to the partial wake to attach to the fusion wake stored in the fusion wake storage unit 18. Update the turning flag.
In the second embodiment, the turning flag attached to the fusion wake is referred to as an “integrated turning flag”. The integrated turning flag is a turning flag common to all target tracking devices connected to the same network 2, and is a flag indicating the same turning determination result in all target tracking devices.
Hereinafter, the update process of the integrated turning flag by the flag integration unit 91 will be specifically described.

When the flag integration unit 91 receives a partial wake with a “turn” turn flag from the transmission / reception unit 20, the “non-turn” integrated turn flag attached to the fusion wake stored in the fusion wake storage unit 18. Is updated to the integrated turning flag of “turning”.
The flag integration unit 91 trusts the turning flag output from the transmission / reception unit 20, and even after receiving a partial wake with a “non-turning” turning flag from the transmission / reception unit 20, Maintain the “turn” integrated turn flag attached.
The timing at which the integrated turning flag is changed to “non-turning” is limited to the timing at which the turning end determination processing unit 73 described later determines that the turning has ended.

In addition, the flag integration unit 91 calculates an integrated turning level indicating the degree of target turning, and adds the integrated turning level to the fusion track stored in the fusion track storage unit 18.
Hereinafter, a calculation example of the integrated turning level will be specifically described.
Whenever the turn flag attached to the partial track is “turn”, the flag integration unit 91 records the turn level attached to the partial track every time the partial track is received from the transmission / reception unit 20.
The flag integration unit 91 calculates the integrated turning level using the recorded turning level at regular intervals. For example, the turning level having the highest target turning degree among the recorded turning levels may be set as the integrated turning level, or the average value of the recorded turning levels may be set as the integrated turning level.
Note that the turning level recorded by the flag integration unit 91 is deleted, for example, at the timing when the integrated turning flag is updated to “non-turning” by the turning end determination processing unit 73.

Upon receiving the target correlated sensor observation value from the sensor tracking processing unit 14, the turning determination unit 17 determines whether or not the target is turning from the time series data of the target correlated sensor observation value, and the determination result As a result, a turn flag indicating whether or not the target is turning and a turn level indicating the degree of the target turn are output to the re-tracking processing unit 16.
Most of the turning determination processing by the turning determination unit 17 is the same as that in the first embodiment, and the following description will mainly focus on the differences from the first embodiment.

When the flag update processing unit 101 of the turning determination unit 17 receives the target correlated sensor observation value from the sensor tracking processing unit 14, the turning stored in the turning flag DB 72 at the timing of receiving the target correlated sensor observation value. Implement the process to update the flag.
FIG. 20 is a flowchart showing the processing contents of the flag update processing unit 101.
Hereinafter, the processing content of the flag update processing unit 101 will be described with reference to FIG.

The flag update processing unit 101 generates a turn flag stored in the turn flag DB 72 and a fusion track stored in the fusion track storage unit 18 at a timing when the sensor correlation processed sensor observation value is received from the sensor tracking processing unit 14. Get the integrated turning flag attached.
The flag update processing unit 101 determines whether or not the acquired turn flag is “turn” (step ST41). If the turn flag is “non-turn” (step ST41: NO), Without changing the “non-turn” of the turning flag, the turning flag is output to the turning start determination processing unit 71 and the erroneous determination cancellation processing unit 103 (step ST42).
When the turning flag is “turn” (step ST41: YES), the flag update processing unit 101 determines whether or not the acquired integrated turn flag is “non-turn” (step ST43).
When the integrated turning flag is “non-turning” (in the case of YES at step ST43), the flag update processing unit 101 changes the turning flag to “non-turning” and sets the turning flag to the turning start determination processing unit. 71 and the erroneous determination release processing unit 103 (step ST44).
When the integrated turning flag is “turn” (step ST43: NO), the flag update processing unit 101 does not change the “turn” of the turn flag and sets the turn flag to the turn start determination processing unit 71. And it outputs to the misjudgment cancellation process part 103 (step ST45).

As in the first embodiment, the turning start determination processing unit 71 of the turning determination unit 17 determines whether turning has started.
However, in the second embodiment, the initialization processing unit 71d of the turning start determination processing unit 71 performs the initialization process of the turning start determination process according to the turning flag output from the flag update processing unit 101. This is different from the first embodiment.
In addition, the threshold determination processing unit 71j of the turning start determination processing unit 71 outputs the set turning flag to the turning flag DB 72 as in the first embodiment, but the time when it is determined that the turning is first started. Is recorded in the turning start time DB 102 as the turning start time.

When receiving the turning flag from the flag update processing unit 101, the erroneous determination cancellation processing unit 103 performs an erroneous determination cancellation process.
FIG. 21 is a flowchart showing the processing contents of the erroneous determination cancellation processing unit 103.
Hereinafter, the misjudgment cancellation process of the misjudgment cancellation process part 103 is demonstrated concretely, referring FIG.

First, erroneous determination release processing section 103, as shown in the following equation (53), an observation time t _a goal the correlated sensor observations output from the sensor tracking processing unit 14, is stored in the turning start time DB102 and a time difference between the turning start time _{t s} is differential time calculating the Δt _a-s (step ST51).
_{Δt a-s = t a -t} s (53)
Incidentally, observation time t _a goal the correlated sensor observations is the time at which target the correlated sensor observations is output from the sensor tracking processing unit 14.

After calculating the difference time Δt _a−s , the erroneous determination cancellation processing unit 103 determines whether or not the turning flag output from the flag update processing unit 101 is “turning” (step ST52).
When the turning flag is “non-turning” (in the case of NO at step ST52), the erroneous determination cancellation processing unit 103 sets the turning flag to the turning flag DB 72 without changing the “non-turning” of the turning flag. It outputs to the turning level calculation process part 75 and the retracking process part 16 (step ST53).
When the turn flag is “turn” (in the case of YES at step ST52), the erroneous determination release processing unit 103 sets the integrated turn flag attached to the fusion track stored in the fusion track storage unit 18 to “non-turn”. Is determined (step ST54).

If the integrated turning flag is “turning” (in the case of NO at step ST54), the erroneous determination cancellation processing unit 103 sets the turning flag to the turning flag DB 72, turning without changing the “turning” of the turning flag. It outputs to the level calculation process part 75 and the retracking process part 16 (step ST55).
When the integrated turning flag is “non-turning” (step ST54: YES), the erroneous determination release processing unit 103 has the previously calculated difference time Δt _a-s greater than or equal to a preset release threshold. It is determined whether or not there is (step ST56).

When the difference time Δt _a-s is equal to or greater than the release threshold (step ST56: YES), the erroneous determination release processing unit 103 changes the turn flag to “non-turn” and turns the turn flag. It outputs to flag DB72, turning level calculation process part 75, and re-tracking process part 16 (Step ST57).
If the difference time Δt _a-s is less than the release threshold (NO in step ST56), the erroneous determination release processing unit 103 turns the turn flag without changing the “turn” of the turn flag. It outputs to flag DB72, turning level calculation process part 75, and re-tracking process part 16 (Step ST55).

The turn end determination processing unit 73 of the turn determination unit 17 performs a turn end determination process for determining the end of the turn, and updates the integrated turn flag attached to the fusion track stored in the fusion track storage unit 18. .
FIG. 22 is a flowchart showing the processing contents of the turning end determination processing unit 73.
Hereinafter, the processing content of the turning end determination processing unit 73 will be specifically described with reference to FIG.
The initialization processing unit 73k of the turning end determination processing unit 73 acquires the fusion track stored in the fusion track storage unit 18, and determines whether or not the integrated turning flag attached to the fusion track is “turn”. Determination is made (step ST61).
When the turn flag is “non-turn” (step ST61: NO), the initialization processing unit 73k deletes the fusion track stored in the turn completion determination fusion track DB 73j, and turns completion determination processing Is finished (step ST62).
When the turn flag is “turn” (step ST61: YES), the initialization processing unit 73k stores the acquired fusion wake in the turn completion determination fusion wake DB 73j (step ST63), and thereafter. The execution of the turning end determination process is permitted. That is, the operations of the N-step prediction processing unit 73l, the Mahalanobis square distance calculation processing unit 73f, the time direction smoothing processing unit 73g, and the threshold determination processing unit 73h are permitted.

When the initialization processing unit 73k stores the fusion wake in the fusion wake DB 73j for turning completion determination, the N-step prediction processing unit 73l has the oldest fusion wake stored in the fusion wake DB 73j for turning completion determination. and time t _b, the time difference between the time t _k of the determined portions track correlation fusion track stored in the fusion track storage unit 18 is taken by the correlation determination unit 22 _(t k ^_(S) - t _b ) is calculated.
N step prediction processing unit 73l, when calculating the time difference ^{_(t k (S)} -t _b), as shown in equation (54) below, the time difference ^{_(t k (S)} -t _b) pre It is determined whether or not the threshold value TH_PRE_TIME or more is set (step ST64).
(T _k ^(S) −t _b ) ≧ TH_PRE_TIME (54)

When the time difference (t _k ^(S) −t _b ) is less than the threshold value TH_PRE_TIME and the formula (54) is not satisfied (in the case of NO at step ST64), the N step prediction processing unit 73l performs a turning end determination process. finish.
When the time difference (t _k ^(S) −t _b ) is equal to or greater than the threshold value TH_PRE_TIME and the formula (54) is satisfied (step ST64: YES), the N-step prediction processing unit 73l As shown in ^(), a past time T _tnt retroactive by a threshold TH_PRE_TIME from the time t _k ^{(S) of} the partial wake determined to be correlated by the correlation determination unit 22 is calculated.

さらに、Ｎステップ予測処理部７３ｌは、下記の式（５７）に示すように、時刻ｔ_ｊｍから部分航跡の時刻ｔ_ｋ ^（Ｓ）までの外挿時間ΔＴを算出する。

Then, N step prediction processing unit 73l from the fusion track stored in pivot end decision fusion track DB73j, to predict until time t _k partial track ^_(S), such as the position and velocity at the target at the present time A predicted track having a state vector is calculated (step ST65).
Further, as shown in the following formula (56), the N step prediction processing unit 73l selects the fusion track closest to the past time T _tnt from the fusion tracks accumulated in the turn completion determination fusion track DB 73j. Search for time t _jm .

Further, the N-step prediction processing unit 73l calculates an extrapolation time ΔT from the time t _jm to the time t _k ^(S) of the partial wake as shown in the following formula (57).

When calculating the extrapolation time ΔT, the N-step prediction processing unit 73l uses the extrapolation time ΔT to calculate a prediction vector constituting the prediction track and a prediction error covariance matrix.
The process of calculating the prediction vector and the prediction error covariance matrix is calculated according to the equations (20) to (23), similarly to the N step prediction processing unit 71g in the turning start determination processing unit 71.

When the N-step prediction processing unit 73l calculates the prediction vector and the prediction error covariance matrix constituting the prediction track, the Mahalanobis square distance calculation processing unit 73m correlates the prediction vector and the prediction error covariance matrix with the correlation determination unit 22. The Mahalanobis square distance between the prediction vector and the partial wake constituting the predicted wake is calculated from the partial wake determined to be removed (step ST66).
Hereinafter, the Mahalanobis square distance calculation processing by the Mahalanobis square distance calculation processing unit 73m will be described in detail.
The case where the partial wake is a state vector composed of position and velocity and the case where the partial wake is a state vector composed of only the position will be described separately.

[When partial track is a state vector consisting of position and velocity]
Mahalanobis square distance calculation processing unit 73m, as shown in the following equation (58), and the predicted vector for the fusion track at the current time point t _k calculated by the N step prediction processing unit 73l, the position and velocity vector of the partial track to calculate the residual _{r k.}

Next, the Mahalanobis square distance calculation processing unit 73m calculates the Mahalanobis square distance ε _v (k) using the residual r _k as shown in the following equations (59) and (60).

次に、マハラノビス平方距離算出処理部７３ｍは、下記の式（６２）（６３）に示すように、残差共分散行列Ｓ_ｋを算出し、式（６０）に示すように、残差ｒ_ｋと残差共分散行列Ｓ_ｋ用いて、マハラノビス平方距離ε_ｖ（ｋ）を算出する。

[If the state vector consists only of positions]
Mahalanobis square distance calculation processing unit 73m, as shown in the following equation (61), the predicted position vector of the fusion track at the current time point t _k calculated by the N step prediction processing unit 73l, and the position vector of the partial track to calculate the residual _{r k.}

Next, the Mahalanobis square distance calculation processing unit 73m calculates a residual covariance matrix S _k as shown in the following formulas (62) and (63), and a residual r _{k as} shown in the formula (60). And the residual covariance matrix S _k , the Mahalanobis square distance ε _v (k) is calculated.

When the Mahalanobis square distance calculation processing unit 73m calculates the Mahalanobis square distance ε _v (k), the time direction smoothing processing unit 73g performs the time direction smoothing process for smoothing the Mahalanobis square distance ε _v (k) in the time direction. (Step ST67).
For smoothing the Mahalanobis squared distance ε _{v (k)} in the time direction smoothing unit 73g is similar to the smoothing process Mahalanobis squared distance ε _{v (k)} in the time direction smoothing unit 71i illustrated in FIG. 5, Detailed description is omitted.

When the time direction smoothing processing unit 73g smoothes the Mahalanobis square distance ε _v (k) in the time direction, the threshold determination processing unit 73h, as shown in the above formula (36), the Mahalanobis square smoothed in the time direction. A threshold value determination process for comparing the time direction smooth value ε _av (k), which is the distance, with a preset threshold value Th _turnend is performed (step ST68).
When the time direction smooth value ε _av (k) is equal to or less than the threshold value Th _turnend and the expression (36) is satisfied, the threshold value determination processing unit _73h determines that the turn is finished, and stores it in the fusion wake storage unit 18. The integrated turning flag attached to the stored fusion track is set to “non-turning”.
When the time direction smooth value ε _av (k) is larger than the threshold value Th _turnbgn and Equation (36) is not satisfied, the threshold value determination processing unit _73h determines that the turn has not ended, and the “turn” integrated turn Keep the flag.

Here, the threshold determination processing unit 73h is an example of comparing the Mahalanobis square distance and a threshold value which is smoothed in the time direction, the threshold determination processing unit 73h is time direction smoothing value at the current time t _k ε _{av (k} ) and, before the time _{t k-1} calculates a difference value [Delta] [epsilon] _av (k) of the time direction smoothing value ε _av (k-1) in, the difference value [Delta] [epsilon] _av (k) a threshold number of times that is negative Th _If it exceeds _cnt, it may be determined that the turn has ended.
FIG. 23 is a flowchart showing the processing contents of the threshold determination processing unit 73h. Hereinafter, the processing content of the threshold determination processing unit 73h will be specifically described with reference to FIG.

Every time the time direction smoothing processing unit 73g calculates the time direction smoothing value ε _av (k) of the current time t _k , the threshold determination processing unit 73h, as shown in the above equation (37), at the current time t _k A difference value Δε _av (k) between the time direction smooth value ε _av (k) and the time direction smooth value ε _av (k−1) at the previous time t _k−1 is calculated (step ST71).
Threshold determination processing unit 73h, when calculating a difference value [Delta] [epsilon] _av (k), if the difference value [Delta] [epsilon] _av (k) is negative (step ST72: YES), of the count stored in the difference value counter DB37i Increment processing to increase the value C by 1 is performed (step ST73). The count value C indicates the number of times the difference value Δε _av (k) has become negative, and the initial value of the count value C recorded in the difference value counter DB 37 i is 0.
If the difference value Δε _av (k) is equal to or greater than 0 (step ST72: NO), the threshold determination processing unit 73h maintains the count value C recorded in the difference value counter DB 37i (step ST74).

Next, as shown in the above equation (38), the threshold determination processing unit 73h compares the number of times C recorded in the difference value counter DB 37i with a preset threshold Th _cnt (step ST75).
When the number of times C recorded in the difference value counter DB 37i is greater than the threshold value Th _cnt and the equation (38) is satisfied (in the case of YES at step ST75), the threshold determination processing unit 73h determines that the turn has ended. Determination is made, and the integrated turning flag attached to the fusion track stored in the fusion track storage unit 18 is set to “non-turn” (step ST76).
Further, the threshold determination processing unit 73h integrates “turn” when the number of times C recorded in the difference value counter DB 37i is equal to or less than the threshold Th _cnt and Expression (38) is not satisfied (in the case of NO at step ST75). The turning flag is maintained (step ST77).
When the integrated turning flag is set to “non-turning”, the threshold determination processing unit 73h initializes the number of times C recorded in the difference value counter DB 37i to 0.

  When the turning flag stored in the turning flag DB 72 is “turning”, the turning level calculation processing unit 75 of the turning determination unit 17 sets the target output from the sensor tracking processing unit 14 as in the first embodiment. A turn level indicating the degree of target turn is calculated from the correlated sensor observation values, and the turn level is output to the retracking processing unit 16.

When the re-tracking processing unit 16 of the tracking processing unit 12 receives the target correlated sensor observation value from the sensor tracking processing unit 14, as in the first embodiment, the position indicated by the target correlated sensor observation value and the part It is determined whether or not the position indicated by the partial track stored in the track storage unit 15 is correlated.
The re-tracking processing unit 16 correlates the position indicated by the target correlated sensor observation value with the position indicated by the partial wake stored in the partial wake storage unit 15, and the target correlated sensor observation value is When it is determined that it can be associated with the partial track, the target correlated sensor observation value is stored in the partial track storage unit 15 as in the first embodiment.
The re-tracking processing unit 16 also calculates the target current time from the target correlated sensor observation value and the past target correlated sensor observation value stored in the partial track storage unit 15 as in the first embodiment. A partial wake having a state vector including the position and velocity at is estimated.
When the re-tracking processing unit 16 estimates the partial track, the re-tracking processing unit 16 replaces the partial track stored in the partial track storage unit 15 with the estimated partial track, and the turn determination unit 17 performs the estimation on the estimated partial track. The output turning flag and turning level are added, and the partial track with the turning flag and turning level is output to the transmission determination unit 19.

When receiving the partial track with the turning flag and the turning level from the re-tracking processing unit 16, the transmission determining unit 19 stores the time at the position indicated by the partial track and the fusion track storage unit 18 as in the first embodiment. Whether or not to transmit the partial wake is determined from the time difference from the time of the position indicated by the stored fusion wake and the turn flag and turn level attached to the partial wake.
At this time, the transmission determination unit 19 determines that the partial wake with the turn flag and the turn level is greater when the turn flag attached to the partial track is “non-turn” than when the turn flag is “turn”. Increase the frequency of issuing determination results that permit transmission.
Further, the higher the turn level attached to the partial track, the higher the frequency of issuing a determination result that permits transmission of the partial track with the turn flag and the turn level.

Here, if the turning flag attached to the partial wake is “turn”, the transmission determination unit 19 increases the system noise parameter of the equation (42) as in the first embodiment to obtain the partial wake. If the integrated turning flag attached to the fusion wake stored in the fusion wake storage unit 18 is “turn”, it is assumed that the frequency of issuing the determination result indicating that the The frequency at which the determination result indicating that the partial wake is transmitted may be increased by increasing the system noise parameter of the equation (42) than when the integrated turning flag is “non-turning”.
Also, here, the transmission determination unit 19 increases the system noise parameter of the equation (42) as the turning level attached to the partial wake increases, as in the first embodiment, so that the turning flag and It is assumed that the frequency of issuing a determination result indicating that a partial wake with a turning level is to be transmitted is increased, but the integrated turning level attached to the fusion wake stored in the fusion wake storage unit 18 is The higher the system noise parameter of the equation (42), the higher the frequency, the more frequently the determination result indicating that the partial track with the turning flag and the turning level is transmitted may be increased.

As is apparent from the above, according to the second embodiment, as in the first embodiment, the communication capacity can be reduced while ensuring the tracking accuracy of the turning target.
That is, according to the second embodiment, when the target is turning, the transmission frequency of the partial wake can be increased and the tracking accuracy of the target can be increased. On the other hand, when the target is not turning, the transmission frequency of the partial wake can be lowered to reduce the communication capacity.

Further, according to the second embodiment, the flag integration unit 91 uses the turning flag attached to the partial wake output from the transmission / reception unit 20 to attach to the fusion wake stored in the fusion wake storage unit 18. The integrated turn flag is updated so that the accuracy of the turn flag attached to the partial track output from the transmitting / receiving unit 20 is not so high, but the transmission frequency according to the presence or absence of the target turn. There is an effect that a partial wake can be transmitted.
Further, even when the “turn” of the turning flag attached to the partial wake output from the transmission / reception unit 20 is incorrect, the partial wake can be transmitted at an appropriate transmission frequency without being affected by the error. There is an effect.

  In the present invention, within the scope of the invention, any combination of the embodiments, or any modification of any component in each embodiment, or omission of any component in each embodiment is possible. .

  The target tracking device according to the present invention is suitable for use in estimating the target position from the observation value of the sensor.

  1-1 to 1-M target tracking device, 2 networks, 11 sensors, 12 tracking processing unit, 13 sensor track storage unit, 14 sensor tracking processing unit, 15 partial track storage unit, 16 retracking processing unit, 17 turning determination unit , 18 Fusion track storage unit, 19 Transmission determination unit, 20 Transmission / reception unit, 21 Track fusion unit, 22 Correlation determination unit, 23 Fusion tracking processing unit, 24 Display processing unit, 41 Storage processing circuit, 42 Sensor tracking processing circuit, 43 Tracking processing circuit, 44 turning determination processing circuit, 45 transmission determination processing circuit, 46 transmission / reception processing circuit, 47 correlation determination processing circuit, 48 fusion tracking processing circuit, 49 display processing circuit, 50 flag integration processing circuit, 61 memory, 62 processor, 71 Turning start determination processing unit, 71a Observation value DB for turning start determination, 71b Speed DB, 71c Smoothing for turning start determination Wake DB, 71d initialization processing unit, 71e linear trajectory estimation processing unit, 71f speed re-smoothing processing unit, 71g N-step prediction processing unit, 71h Mahalanobis square distance calculation processing unit, 71i time direction smoothing processing unit, 71j threshold determination processing unit 72 turning flag DB, 73 turning end determination processing unit, 73a observation value DB for turning end determination, 73b smooth track DB for turning end determination, 73c initialization processing unit, 73d orbit estimation processing unit, 73e N step prediction processing unit, 73f Mahalanobis square distance calculation processing unit, 73g time direction smoothing processing unit, 73h threshold value determination processing unit, 73i difference value counter DB, 73j turning end determination fusion track DB, 73k initialization processing unit, 73l N step prediction processing unit, 73m Mahalanobis square distance calculation processing unit, 74 turning flag output processing unit, 75 Turning level calculation processing unit, 75a initialization processing unit, 75b turning level calculation observation value DB, 75c turning trajectory estimation processing unit, 81 parameter change processing unit, 82 fusion track prediction processing unit, 83 determination error calculation processing unit, 84 Determination processing unit, 91 flag integration unit, 101 flag update processing unit, 102 turning start time DB, 103 erroneous determination release processing unit.

Claims (13)

From the time series data of the observed values indicating the target position output from the sensor, a tracking processing unit that estimates a partial wake including information on the target position;
A turning determination unit for determining whether or not the target is turning from the time series data of the observation values;
A fusion wake storage unit storing fusion wakes including information on the target position;
Estimated by the tracking processing unit from the time difference between the time of the partial wake estimated by the tracking processing unit and the time of the combined wake stored in the fusion wake storage unit and the determination result of the turning determination unit A transmission determination unit for determining whether or not to transmit a partial wake;
A transmission / reception unit that transmits the partial track to another target tracking device and receives the partial track transmitted from the other target tracking device when the determination result of the transmission determination unit indicates that the partial track is transmitted When,
A target tracking device comprising: a track merging unit that updates a fused track stored in the fused track storage unit using a partial track transmitted or received by the transmission / reception unit. The tracking processing unit
A sensor wake storage unit that stores sensor wake including information on the position of the target, and an observation value indicating the position of the target;
The position indicated by the observation value output from the sensor is determined by determining whether or not the position indicated by the observation value output from the sensor is correlated with the position indicated by the sensor wake stored in the sensor wake storage unit. And the position indicated by the sensor track, the observation value output from the sensor is stored in the sensor track storage unit, and the observation value and the observation value stored in the sensor track storage unit are stored. And a sensor tracking processing unit that estimates a sensor track including information on the target position, and replaces the sensor track stored in the sensor track storage unit with the estimated sensor track, and
A partial wake storage unit that stores a partial wake including information on the target position, and an observation value indicating the target position;
Whether the position indicated by the observation value determined to be correlated with the position indicated by the sensor track by the sensor tracking processing unit and the position indicated by the partial track stored in the partial track storage unit are correlated. If the position indicated by the observed value correlates with the position indicated by the partial wake, the observed value is stored in the partial wake storage unit and stored in the partial wake storage unit. A re-tracking processing unit that estimates a partial wake including information on the target position from the observed values and replaces the partial wake stored in the partial wake storage unit with the estimated partial wake. The target tracking device according to claim 1, wherein   The turning determination unit estimates the position of the target from the time-series data of the observation values, converts the difference between the estimated position and the position indicated by the observation value output from the sensor into a Mahalanobis square distance, The target tracking device according to claim 1, wherein whether or not the target is turning is determined from the Mahalanobis square distance.   The turning determination unit estimates the position of the target from the time series data of the observation values, and converts the difference between the estimated position and the position indicated by the observation value output from the sensor into a Mahalanobis square distance. The Mahalanobis square distance is smoothed in the time direction, and the Mahalanobis square distance smoothed in the time direction is compared with a threshold value to determine whether or not the target is turning. The target tracking device described.   The transmission determination unit transmits the partial track when the turn determination unit determines that the target is turning than when the turn determination unit determines that the target is not turning. The target tracking device according to claim 1, wherein the target tracking device increases the frequency of issuing a determination result indicating the effect. When the turning determination unit determines that the target is turning, it calculates the degree of turning of the target from the time series data of the observed values,
The target tracking device according to claim 5, wherein the transmission determination unit increases the frequency of issuing a determination result indicating that the partial wake is transmitted as the degree of turning calculated by the turning determination unit increases. .   7. The target according to claim 6, wherein the wake fusion unit sets a system noise parameter used for updating the fusion wake as a larger value as the degree of turn calculated by the turn determination unit is larger. Tracking device. The wake fusion unit is
A correlation determination unit that determines whether or not a position indicated by the partial wake transmitted or received by the transmission / reception unit and a position indicated by the fusion wake stored in the fusion wake storage unit are correlated;
When the correlation determination unit determines that there is a correlation, the fusion track including the information on the target position is estimated from the position indicated by the partial track and the position indicated by the fusion track, and the fusion track storage unit When the fusion wake stored in FIG. 5 is replaced with the estimated fusion wake and the correlation determination unit determines that there is no correlation, the partial wake transmitted or received by the transmission / reception unit is used as the fusion wake. The target tracking device according to claim 1, further comprising a fusion tracking processing unit that records in the fusion track storage unit. The tracking processing unit adds a turning flag indicating the determination result of the turning determination unit to the estimated partial track, and outputs the partial track with the turning flag,
The transmission / reception unit transmits the partial track with the turning flag to another target tracking device and transmits from the other target tracking device when the determination result of the transmission determination unit indicates that the partial track is to be transmitted. Receive a partial wake with a turn flag,
The fusion wake storage unit stores the fusion wake with the turning flag,
A flag integration unit for updating a turn flag attached to the fusion track stored in the fusion track storage unit using a turn flag attached to the partial track transmitted or received by the transmission / reception unit; The target tracking device according to claim 1, wherein   The turn determination unit determines whether the target has started to turn when the turn flag attached to the fusion track stored in the fusion track storage unit indicates that the target is not turning. The target tracking device according to claim 9, wherein a turn start determination process is performed to determine whether or not.   The turning determination unit determines whether or not the target has finished turning from a difference between the fusion track stored in the fusion track storage unit and the partial track transmitted or received by the transmission / reception unit. The target tracking device according to claim 9, wherein a turn end determination process is performed.   The turn determination unit is configured to turn the target from a turn flag attached to a fusion track stored in the fusion track storage unit and a turn flag attached to a partial track transmitted or received by the transmission / reception unit. The target tracking device according to claim 9, wherein a turn end determination process is performed to determine whether or not the operation has ended.   The turning determination unit corrects the turning determination result determined from the time-series data of the observation values using a turning flag attached to the fusion track stored in the fusion track storage unit. The target tracking device according to claim 9.

Images