JP2007183112A - Target tracking device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To change the number of tracking filters used in response to a situation to allow precise tracking, in a multimotion model system for processing in parallel the plurality of tracking filters for motion models in conformity with a preset target motion. <P>SOLUTION: This tracking device is provided with: a file selection control means for judging the tracking filter corresponding to a value to be separated or assembled, based on the value expressing the situation of the motion of the target obtained from the tracking filter, and for generating a control signal based on a judged result therein; and a switch means for separating or assembling the corresponding tracking filter, based on the generated control signal. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a target tracking device that estimates target motion specifications such as a target position and speed based on observation data relating to target position information from a sensor or the like.

  When tracking a target whose motion changes according to the situation, such as aircraft turning, rocket engine injection or stopping, the target's motion can be correctly predicted by using the tracking filter of the model according to the motion, As a result, the true position and speed of the target can be estimated with high accuracy. There are a plurality of patterns in such a target movement. Therefore, a tracking filter that provides a motion model in accordance with each motion is prepared. The smoothness value, smoothing error covariance matrix, etc., which are the output of each tracking filter, with the fitness (hereinafter referred to as model reliability) that matches the motion represented by each motion model with the actual target motion. A multiple motion model method using a plurality of tracking filters in parallel while performing weighted averaging has been proposed (see, for example, Patent Document 1). In this method, when the target suddenly changes the motion, the model reliability of the tracking filter of the motion model after the change becomes high, the result of the tracking filter is mainly used, and tracking can be continued.

  As an example of the multiple motion model method, a simple method that uses a plurality of tracking filters independently, weights and averages the smoothed values obtained by tracking by each tracking filter with the model reliability, and outputs to the display device, or IMM There is a method called (Interactive Multiple Model) (see Non-Patent Document 1, for example). In the IMM, in addition to the simple method, the track represented by each tracking filter is mixed using the degree of fitness that the motion represented by each motion model matches the actual target motion, and the tracks are mixed with each other. Used for prediction processing in

JP 2002-328162 A Samuel Blackman, Robert Popoli ‘Modern Tracking Systems’ published by Artech House, 1999, P221―232

Although the conventional target tracking device enables continuation of tracking even when the target suddenly changes movement as described above, there are the following problems.
In a conventional target tracking device, target motions are assumed in advance, tracking filters for each motion model in accordance with those motions are incorporated in the target tracking device, and a fixed number of tracking filters are always used in parallel. In this case, when the error of the target position observation information from the sensor etc. is small, the Mahalanobis distance between the target predicted position and the target position observation information by the tracking filter of the model according to the target motion becomes close. The model reliability of the tracking filter of the model according to the motion is increased. However, when the error of the target position observation information is large, the Mahalanobis distance between the target predicted position and the target position observation information by the tracking filter of the model according to the target motion becomes long, and as a result, the model according to the target motion If the model reliability of the tracking filter is low, but the Mahalanobis distance between the target predicted position by the tracking filter of the motion model that does not match the actual target motion and the target position observation information is close, the model reliability of this tracking filter is temporarily Can be expensive. As a result, the movement of the target cannot be predicted correctly, and the estimation accuracy with respect to the position and speed of the target may deteriorate.
In addition, regardless of whether or not the movement coincides with the actual target motion, all of the fixed tracking filters are always used in parallel, so that a computation load is imposed on the computer.

  The present invention has been made to solve the above-mentioned problems, and is used according to the situation in a multiple motion model method for processing a plurality of tracking filters of a motion model in accordance with a predetermined target motion in parallel. An object of the present invention is to obtain a target tracking device that enables high-accuracy tracking by changing the number of tracking filters.

  The target tracking device according to the present invention uses a plurality of tracking filters having a motion model in accordance with a target motion assumed in advance based on target position observation information related to a target position observed by a sensor of the target observation device. The model likelihood for each tracking filter for the target is calculated by the model reliability calculation means based on the calculated smooth value, smoothing error covariance, and target position observation information. The model reliability is calculated by normalizing each model likelihood, and the smoothing value and smoothing error covariance calculated by the tracking filter are weighted and averaged by the model likelihood by the weighting processing means for display. In the target tracking device that creates the wake, the tracking filter corresponding to the value is cut based on the value indicating the target motion status obtained from the tracking filter. A file selection control means for determining whether or not to incorporate and generating a control signal based on the determination result, and a switch means for separating or incorporating the corresponding tracking filter based on the generated control signal It is.

According to the present invention, a fixed number of tracking filters are not always used in parallel, but the number of tracking filters to be used can be changed according to the state of target movement.
In addition, although the number of tracking filters to be used changes, a method that uses multiple tracking filters in parallel, focusing on tracking filters that are highly likely to be the target motion, has caused the target to suddenly change its motion. Even in this case, tracking can be continued. Furthermore, since the tracking filter of the motion model that does not match the actual target motion is not used at all, even if the error of the target position observation information is large, the tracking filter of the motion model that does not match the actual target motion temporarily causes The estimation accuracy of the target position and speed will not deteriorate. That is, it is possible to track with higher accuracy than the conventional target tracking device. Furthermore, since the tracking filter of the motion model that does not match the actual target motion is not used at the same time, the calculation load can be reduced as compared with the conventional target tracking device.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a functional configuration of a target tracking device common to the first to third embodiments of the present invention. In each embodiment including the first embodiment, for the sake of convenience, an example of an IMM in which the number of tracking filters is set to a maximum of three will be described. However, the present invention has no limitation on the number of tracking filters, and a plurality of tracking filters. It can be applied to all multi-motion model methods using filters.
In FIG. 1, a target observation device 200 and a display device 300 are connected to the target tracking device 100. The target observation device 200 is means for generating and outputting target position observation information related to the target position observed by the sensor. The display device 300 is means for displaying a track as a target tracking result obtained by the target tracking device 100 on the screen.
The target tracking device 100 includes tracking filters 110, 120, and 130, switch means 141, 142, and 143, a model reliability calculation unit 170, a filter selection unit 180, and a weighting unit 190.

  The tracking filters 110, 120, and 130 are filters used in the multiple motion model method, respectively, and are tracking filters for a motion model that matches a target motion assumed in advance. When the target is, for example, a civil aircraft, assuming three types of target motions, straight travel, left turn, and right turn, the tracking filter 110 represents a straight motion, the tracking filter 120 represents a left turn motion, The tracking filter 130 represents a right-turning motion. Since the tracking filters 110, 120, and 130 all have the same functional configuration, the tracking filter 110 will be mainly described below.

The tracking filter 110 includes a delay processing unit 111, a mixing processing unit 113, a wake prediction unit 114, a wake update process 115, and switch means 116 and 117.
The wake update unit 115 includes a target position observation information input from the target observation device 200, a predicted value given from the wake prediction unit 114, a smooth value indicating a motion state of the target based on a prediction error covariance matrix, and a smooth error covariance. A means for calculating a matrix. The delay processing unit 111 is a means for delaying the smoothed value and smoothing error covariance calculated by the wake update unit 141 from the time when the previous wake was updated to the time when the current wake is updated. The mixing processing unit 113 is a model calculated by the model reliability calculation processing unit 170, which will be described later, in consideration of the probability that the target motion changes to another motion in the time delayed by the delay processing of the delay processing unit 111. The prior model reliability is calculated from the reliability, and the smoothed values and smoothing error covariance matrices calculated by the track update units 115, 125, and 135, which will be described later, are weighted and averaged by the prior model reliability to obtain both the mixed value and the mixing error. A means for calculating a dispersion matrix. The wake prediction unit 114 is a unit that calculates a prediction value and a prediction error covariance matrix based on the mixing value and the mixing error covariance matrix of the tracking filter 110 calculated by the mixing processing unit 113. The switch means 116 and 117 are controlled to be opened or closed by a control signal from a filter selection control unit 180, which will be described later. The switch means 116 receives a smooth value and a smooth error covariance from the delay processing unit 111 when closed. The matrix is given to the mixing processing units 113, 123 and 133. On the other hand, the switch means 117 gives the target position observation information from the target observation device 200 to the wake update unit 115 when closed.

The model reliability calculation processing unit 170 performs model likelihood for each tracking filter based on the target position observation information, the smooth value representing the motion state of the target obtained by the tracking filters 110, 120, and 130, and the smoothing error covariance. And model reliability is calculated by normalizing each model likelihood. The filter selection control unit 180 determines whether the three tracking filters 110, 120, and 130 are separated from or incorporated into the target tracking device 100 based on the model reliability calculated by the model reliability calculation processing unit 170. And a means for generating a control signal based on the determination result. The switch unit 141 is controlled to be opened or closed by a control signal from the filter selection control unit 180. When the switch unit 141 is closed, the target position observation information, the predicted value, the prediction error covariance, and the prior model reliability are model-reliable from the tracking filter 110. It is a means given to the degree calculation processing unit 170. The weighting processing unit 190 is a unit that weights and averages the smoothed value and smoothing error covariance calculated by each tracking filter with model reliability, and creates a wake for display.
Here, in the configuration shown in FIG. 1, each part except the filter selection control part 180 and the switch means 116, 126, 136, 117, 127, 137, 141 to 143 is a means conventionally used.

Next, the operation will be described mainly focusing on the tracking filter 110.
In the target tracking device 100, when power is first turned on, the switch means 116, 126, 136, 117, 127, 137 and the outer switch means 141 in each tracking filter are controlled by a control signal from the filter selection control unit 180. ~ 143 is closed. As a result, all of the provided tracking filters 110, 120, and 130 are placed in the target tracking device 100 so as to perform the tracking operation. In the target observation device 200, when the sensor observes the target, target position observation information regarding the target position is acquired and output to the tracking filters 110, 120, and 130 of the target tracking device 100. In the tracking filters 110, 120, and 130, the input target position observation information is given to the wake update units 115, 125, and 135 via the switch means 117, 127, and 137. When the tracking filter 110 is viewed, the wake update unit 115 receives the input target position observation information, the predicted value given from the wake prediction unit 114, a smooth value representing the target motion status based on the prediction error covariance matrix, A smooth error covariance matrix is calculated and output to the model reliability calculation processing unit 170. The model reliability calculation processing unit 170 also receives information representing the target exercise status from the other tracking filters 120 and 130. In the model reliability calculation processing unit 170, based on the target position observation information, the predicted value, and the prediction error covariance described above, a model likelihood that represents the degree to which the motion represented by each motion model matches the actual target motion. The degree is calculated for each tracking filter, and the model reliability obtained by normalizing each model likelihood is calculated. The model likelihood increases as the Mahalanobis distance between the predicted value and the target position observation information is closer, and as a result, the model reliability increases.

  The model reliability for each tracking filter calculated by the model reliability calculation processing unit 170 is given to the filter selection control unit 180. The filter selection control unit 180 compares each model reliability with a threshold value to determine which of the three tracking filters 110, 120, and 130 has a model reliability value lower than the threshold value. Based on the control signal generated according to the determination result, the corresponding one of the switch means 116, 126, 136, 117, 127, 137, 141-143 is switched to open. Here, for example, if the filters other than the tracking filter 110 are disconnected, the switch means 126, 136, 127, 137, 142, 143 are opened. Now, since the tracking filter 110 is in the incorporated state, the weighting processing unit 190 uses the smoothing value and smoothing error covariance calculated by the wake updating unit 141 as the tracking filter calculated by the model reliability calculation processing unit 170. A wake for display is created by weighted averaging with the model likelihood of 110 and output to the display device 300.

  Further, in each of the tracking filters 110, 120, and 130, in the state where each switch means is closed, the following processing is performed in those tracking filters. For example, when viewed with the tracking filter 110, the smoothing value and smoothing error covariance calculated by the wake update unit 141 are given to the delay processing unit 111. The delay processing unit 111 delays the current smooth value and smooth error covariance from the time when the current track is updated to the time when the next track is updated. The delayed current smoothing value and smoothing error covariance are provided to the mixing processing unit 113 via the switch unit 116. When the switching means 126 and 136 are closed, the corresponding smoothing value and smoothing error covariance matrix are also given to the mixing processing unit 113 from the delay processing units 121 and 131 of the tracking filters 120 and 130. In the mixing processing unit 113, the model reliability calculation processing unit 170 calculated as the tracking filter 110 based on the delay time obtained by the delay processing unit 111 in consideration of the probability that the target motion changes to another motion. The prior model reliability is calculated from the model reliability. Further, the mixing processing unit 113 calculates the mixed value and the mixed error covariance matrix by weighting and averaging the smoothed value and the smoothed error covariance matrix calculated by the wake updating units 115, 125, and 135 with the prior model reliability. Next, the wake prediction unit 114 calculates a prediction value and a prediction error covariance matrix based on the mixed value and the mixing error covariance matrix calculated by the mixing processing unit 112, and the wake update unit 115 at the next wake update. The values used for calculating the smooth value and the smooth error covariance matrix are as follows.

Of the model reliability of each tracking filter calculated by the model reliability calculation unit 170, when only the value related to the tracking filter 120 is high, for example, the filter selection control unit 180 relates to the tracking filter 120. The tracking means 120, 127, 142 to be closed are incorporated into the tracking filter 120, the switches related to the remaining tracking filters 110, 130 are switched to open, and the tracking filters 110, 130 are disconnected. In this case, the weighting processing unit 190 displays the smoothed value and smoothing error covariance calculated by the wake updating unit 125 by weighted averaging with the model likelihood of the tracking filter 120 calculated by the model reliability calculation processing unit 170. Will create a wake for. That is, when a possibility arises as a target motion, the switch means is closed so that a tracking filter that has not been used so far can be incorporated into the target tracking device and used.
In the above example, the incorporation of the tracking filter in the case where the target motion observed at one time is one type has been described, but there are actually cases where a plurality of types of target motion are combined. In such a case, a corresponding number of tracking filters (eg, 110, 120) will be incorporated, and the other unrelated tracking filters (eg, 130) will operate.

  As described above, according to the first embodiment, in the multiple motion model method, the filter selection control unit 180 controls the switch means based on the model reliability calculated by the model reliability calculation unit 170, and observes The tracking filter is assumed not to be related to the motion of the target and is not used separately from the target tracking device 100, and when it is estimated that the tracking filter is related to the motion of the target, The corresponding tracking filter that has been separated up to now is controlled to be incorporated into the target tracking device. Therefore, a fixed number of tracking filters are not always used in parallel, but the number of tracking filters to be used can be changed according to the situation of the target motion. In addition, although the number of tracking filters to be used changes, a method that uses multiple tracking filters in parallel, focusing on tracking filters that are highly likely to be the target motion, has caused the target to suddenly change its motion. Even in this case, tracking can be continued. Furthermore, since the tracking filter of the motion model that does not match the actual target motion is not used at all, even if the error of the target position observation information is large, the tracking filter of the motion model that does not match the actual target motion temporarily causes The estimation accuracy of the target position and speed will not deteriorate. That is, it is possible to track with higher accuracy than the conventional target tracking device. Furthermore, since a tracking filter of a motion model that does not match the actual target motion at the same time is not used, the calculation load can be reduced as compared with the conventional target tracking device.

Embodiment 2. FIG.
The second embodiment has the same configuration as that of FIG. 1, but the filter selection control unit 180 in the second embodiment performs filter selection based on the model likelihood obtained from the model reliability calculation unit 170. It is characterized by making a judgment. As described in the first embodiment, the model reliability calculation unit 170 calculates the model likelihood for each tracking filter, normalizes each model likelihood, and calculates the model reliability. Therefore, the model likelihood representing the situation of the target motion may be used instead of the model reliability in the determination of filter selection. For example, when the model likelihood of a certain tracking filter is constantly below a preset threshold, the tracking filter is disconnected from the target tracking device. Further, when the model likelihood of all the tracking filters is low, a new tracking filter is incorporated into the target tracking device.
As described above, according to the second embodiment, since the filter selection process is performed according to the model likelihood, the necessity of the tracking filter can be determined more correctly.

Embodiment 3 FIG.
Although the third embodiment has the same configuration as that of FIG. 1, the filter selection control unit 180 in the third embodiment has a smoothing value or a smoothing value of each tracking filter representing the target motion status by the tracking filter. The filter selection is determined based on the weighted average track for display. Each tracking filter calculates and outputs a smooth value representing the state of the target motion. Therefore, the smooth value from each tracking filter may be used to determine the filter selection. The weighting processing unit 190 weights the smoothing value and smoothing error covariance calculated by the wake updating unit 125 with the model likelihood of the tracking filter 120 calculated by the model reliability calculation processing unit 170 for display. The wake is created, and the weighted average wake also represents the target motion status. Therefore, the weighted average wake may be used for the filter selection determination in the filter selection control unit 180. For example, if the direction or magnitude of the smooth value or the speed vector of the wake changes greatly, it is predicted that the target motion has changed. In addition, when the tracking filter for estimating the acceleration is provided, it is predicted that the target motion has changed even when the acceleration changes greatly. Therefore, the filter selection control unit 180 can incorporate the corresponding tracking filter by determining the filter selection from the smooth value or the result of the wake.
As described above, according to the third embodiment, the filter selection control unit 180 performs the filter selection process in accordance with the smoothed value of the tracking filter or the weighted average for display, and therefore tracking is performed more correctly. Whether or not a filter is necessary can be determined.

Embodiment 4 FIG.
FIG. 2 is a block diagram showing a functional configuration of a target tracking device according to Embodiment 4 of the present invention. In the figure, the same reference numerals are given to the same or corresponding parts as in FIG. 1, and the description thereof will be omitted in principle. In the fourth embodiment, an external information source 400 is newly provided with respect to the configuration of FIG. 1, and a filter selection control unit 181 that performs different operations is provided in place of the filter selection control unit 180.
The external information source 400 is a unit that is provided in addition to the target observation device 200 and outputs external information that can be used for estimating a target motion such as a track of the same target, model likelihood, or model reliability. The external information in this case is, for example, information calculated by another sensor or assumed route information. The filter selection control unit 181 is a means for comprehensively separating the tracking filter and determining whether to incorporate it in addition to the value representing the state of the target motion described in each of the above embodiments, including external information.

The operation will be described. In the filter selection control unit 181, as in the case of the first embodiment, the tracking filter is separated or incorporated based on each model reliability calculated by the model reliability calculation unit 170. At this time, comprehensive judgment is made including external information such as the track of the same target given from the external information source 400, model likelihood or model reliability, and a control signal corresponding to the judgment result is generated. The switch means is switched. When there is another sensor different from the target observation apparatus 200 regardless of the same kind, the track, model likelihood, model reliability, and the like by the other sensor can be used as the filter selection criterion. Further, when the assumed route is turning, a filter can be selected according to the route.
As described above, according to the fourth embodiment, in the operation of the filter selection control unit 180 of the first embodiment, the filter selection control unit 181 that performs filter selection processing comprehensively in consideration of external information is provided. As a result, the necessity of the tracking filter can be determined more correctly.

Embodiment 5 FIG.
In each of the above-described embodiments, the description has been given of the case where a tracking filter that is not used can be incorporated into the target tracking device and used when there is a possibility as a target motion. In such a case, when tracking is started based on the target position observation information in the newly incorporated tracking filter, the wakes are measured until the smoothed value and smoothing error covariance matrix calculated in the incorporated tracking filter is stabilized. Must be updated multiple times. In general, there is a problem that the tracking accuracy is not good until the smooth value and the smooth error covariance matrix are stabilized. Therefore, this fifth embodiment proposes a configuration that addresses this problem.

FIG. 3 is a block diagram showing a functional configuration of a target tracking device according to Embodiment 5 of the present invention. In the figure, the same reference numerals are given to the same or corresponding parts as in FIG. 1, and the description thereof will be omitted in principle. In the fifth embodiment, an initial value calculation unit 160 is newly provided in the configuration of FIG.
When the tracking filter is newly incorporated, the initial value calculation unit 160 is displayed for display by the smoothing value, smoothing error covariance matrix, or weighting processing unit 190 of the other tracking filter that is already incorporated and stable in operation. This is a means for calculating the smoothing value of the newly incorporated tracking filter and the initial value of the smoothing error covariance matrix based on the weighted wake parameters.

Next, the operation will be described.
In FIG. 3, for the convenience of explanation, the tracking filter 110 and the tracking filter 120 are already incorporated in the target tracking device 100, and the tracking filter 130 is newly incorporated next. In such a state, when the tracking filter 130 is newly incorporated, the initial value calculation unit 160 obtains the other tracking filters 110 and 120 that are obtained from the model reliability calculation unit 170 and are already incorporated and have stable operations. Based on the smoothed value and the smoothing error covariance matrix, the newly installed smoothing value of the tracking filter 130 and the initial value of the smoothing error covariance matrix are generated and provided to the track updating unit 135 of the tracking filter 130. The wake update unit 135 uses the initial value as a smooth value and a smooth error covariance matrix when the tracking filter 130 is incorporated. Using the smooth value, the smooth error covariance matrix, the smooth value of the other tracking filter, and the smooth error covariance matrix, the mixing processing unit 113 performs the mixing process and the wake prediction unit 114 performs the prediction process. Then, in the target observation device 200, the wake update unit 135 performs smoothing processing using the target position observation information observed by the sensor. As a result, the smoothing value of the tracking filter 130 can be stabilized at an early stage. As an initial value of the tracking filter 130 in this case, the smoothing value and smoothing error covariance matrix of the tracking filter having the highest model reliability may be used.

In this example, the initial value calculation unit 160 uses the smoothed values and smoothing error covariance matrix of the other tracking filters 110 and 120 that are already stable to generate the initial value. The initial value may be calculated using the specifications of the wake weighted for display by the unit 190. In that case, the smooth value and smooth error covariance matrix obtained by weighting and averaging the smooth value of the other tracking filter and the smooth error covariance matrix that have already been stabilized with the model reliability are the initial values.
Further, in the example of FIG. 3 described above, the case where the number of newly incorporated tracking filters is limited to one (the tracking filter 130) has been described, but even if there is one, any of the other tracking filters depends on the situation. In some cases, a plurality of tracking filters (for example, tracking filters 120 and 130) are newly incorporated at the same time. In order to cope with such a case, an initial value calculation unit may be provided for each tracking filter, or the input and output of one initial value calculation unit may be switched by the control signal of the filter selection control unit 180. .

  As described above, according to the fifth embodiment, when a tracking filter that has been in an unused state is newly incorporated in the target tracking device 100, the initial value calculation processing unit 160 has already stabilized the operation. It is newly incorporated using the smoothing value and smoothing error covariance matrix of other tracking filters, or the smoothing value and smoothing error covariance matrix calculated based on the specifications of the wake weighted for display. Since the tracking filter operates, the newly incorporated tracking filter can be stabilized at an early stage, and as a result, high-accuracy tracking is possible.

Embodiment 6 FIG.
As described in the introduction at the beginning of the fifth embodiment, when a tracking filter is newly incorporated, the tracking filter is not stable at the initial stage, and there is a problem that the tracking accuracy is deteriorated. In the sixth embodiment, it is proposed to solve this problem by a method different from that in the fifth embodiment.
FIG. 4 is a block diagram showing a functional configuration of the target tracking device common to the sixth to eighth embodiments of the present invention. In the figure, the same reference numerals are given to the same or corresponding parts as in FIG. 1, and the description thereof will be omitted in principle. In the sixth embodiment, switch means 118, 128, and 138 are provided instead of the switch means 116, 126, and 136 in the configuration of FIG. Further, a filter selection control unit 182 is provided instead of the filter selection control unit 180. Furthermore, the switch means 117, 127, 137 are abolished, and the output of the target observation device 200 is configured to be directly given to each wake update unit 115, 125, 135.
The filter selection control unit 182 determines whether the three tracking filters 110, 120, and 130 are separated from or incorporated in the target tracking device 100 based on the model reliability calculated by the model reliability calculation processing unit 170. In addition, when it is determined to be incorporated, the switch means is controlled to incorporate the corresponding tracking filter into the target tracking device 100 at a timing when the operation is stable.

Next, the operation will be described.
In the filter selection control unit 182, which tracking filter is separated from the target tracking device 100 based on each model reliability calculated by the model reliability calculation processing unit 170, or any tracking filter is incorporated in the target tracking device 100. If not, it is determined whether or not the tracking filter is incorporated. This operation is the same as that of the filter selection control unit 180 of the first embodiment, but the filter selection control unit 182 of the sixth embodiment further performs the following operation. For example, when it is determined that the tracking filter 110 incorporated so far is disconnected from the target tracking device 100, the filter selection control unit 182 issues an immediate trigger (control signal), closes the switch unit 118 to the a side, and switches the switch unit. 141 is opened. Thereby, the tracking filter 110 does not perform the mixing process, and the model reliability calculation unit converts the smoothed value, the smoothed error covariance matrix, and the predicted value and the predicted error covariance matrix by being separated from the target tracking device 100. No output after 170. Accordingly, the separated tracking filter 110 performs a wake update operation alone, generates a smooth value and a smooth error covariance according to the operation, and is in a stable state. Next, when the filter selection control unit 182 determines to incorporate the tracking filter 110 that has been separated as described above into the target tracking device 100, the filter selection control unit 182 generates a trigger (control signal) at a timing when the tracking filter 110 is stable. The switch means 118 is closed to the b side and the switch means 141 is closed. Therefore, when the tracking filter 110 is incorporated in the target tracking device 100, a smooth value and a smooth error covariance in a stable state are output from the tracking filter 110 to the model reliability calculation unit 170 and the subsequent units. The operation when the tracking filters 120 and 130 are separated and incorporated is performed in the same manner.

  As described above, according to the sixth embodiment, although the tracking filter separated from the target tracking device 100 is separated, the track update operation is performed independently, and the separated tracking filter is used as the target tracking. When incorporated in the apparatus 100, the operation is controlled so that the operation is incorporated at a stable timing. As a result, high-accuracy tracking is possible.

Embodiment 7 FIG.
Although the seventh embodiment has the same configuration as that of FIG. 4, the filter selection control unit 182 in the case of the seventh embodiment is a timing for incorporating the separated tracking filter in the timing control processing related to filter selection. As a determination criterion, the smoothing error covariance matrix of the tracking filter is used.
For example, in the tracking filter 110 that is not incorporated, if the smoothing error covariance matrix calculated by the wake updating unit 115 is small, it can be considered that tracking is stable and high-precision tracking is highly likely. . Therefore, the filter selection control unit 182 decides to incorporate the tracking filter 110 into the target tracking device 100 based on the small smoothing error covariance matrix, and switches at the timing of incorporating the same tracking filter as in the sixth embodiment. 141 is controlled.
As described above, according to the seventh embodiment, the timing for incorporating the tracking filter is determined and controlled based on the smoothing error covariance matrix of the tracking filter that is not incorporated. The necessity of the tracking filter can be determined.

Embodiment 8 FIG.
Although the eighth embodiment has the same configuration as that of FIG. 4, the filter selection control unit 182 in the second embodiment incorporates the separated tracking filter in the timing control processing related to filter selection. The model likelihood of the tracking filter is used as a timing judgment criterion.
For example, when the model likelihood related to the tracking filter 110 that is not incorporated is constantly higher than a preset threshold value or higher than the model likelihood of the tracking filter that is already incorporated, the filter selection control unit 182 performs the tracking. It is determined that the filter 110 is incorporated into the target tracking device 100, and the switch unit 141 is controlled at the timing when the same tracking filter as that of the sixth embodiment is incorporated. However, in this case, in addition to the model reliability calculation unit 170, a means for calculating the model likelihood from a prediction value and a prediction error covariance matrix independently calculated by a tracking filter that is not incorporated is provided. It is necessary to keep.
As described above, according to the eighth embodiment, in addition to the model reliability calculation unit 170, a tracking filter that is not incorporated is calculated by the means for calculating the model likelihood alone, and is not incorporated. Since the timing control for incorporating the tracking filter is performed based on the model likelihood for the filter, it is possible to more accurately determine the necessity of the tracking filter.

Embodiment 9 FIG.
FIG. 5 is a block diagram showing a functional configuration of a target tracking apparatus according to Embodiment 9 of the present invention. In the figure, the same reference numerals are given to the same or corresponding parts as in FIG. 1, and the description thereof will be omitted in principle. In the ninth embodiment, the configuration of FIG.
A model reliability recalculation unit 164 is newly provided.
The model reliability recalculation unit 164 gives a model reliability set in advance for the tracking filter when the tracking filter is incorporated as the model reliability output by the model reliability calculation unit 170, and also after the incorporation to the tracking filter. The motion model corresponding to the target motion is estimated based on the smooth value or the display wake, and the model reliability of the tracking filter corresponding to the estimated motion model is calculated to be higher, and the model reliability calculation unit 170 It is a means to give.

In FIG. 5, for example, when the tracking filter 110 is incorporated, the model reliability recalculation unit 164 gives a preset parameter value to the model reliability calculation unit 170 as the model reliability of the tracking filter 110.
In addition, the model reliability calculation unit 170 obtains a smooth value obtained from the filter selection control unit 180 or a display wake obtained by the weighting processing unit 190 for the target movement targeted by the tracking filter 110 after incorporation. To do. Next, the model reliability calculation unit 170 estimates a motion model that is likely to be based on the target motion mode represented by the obtained data, and increases the model reliability of the tracking filter corresponding to the estimated motion model. To the model reliability calculation unit 170. The model reliability calculation unit 170 outputs this higher model reliability to the subsequent stage. In this way, if the target is assumed to switch at a high altitude such as a rocket, for example, the model reliability of the tracking filter should be intentionally increased, and the tracking of the built-in tracking filter should be responded promptly. Is possible.

  As described above, according to the ninth embodiment, by providing the model reliability recalculation unit 164, the model reliability output from the model reliability calculation unit 170 is used as the model reliability when the tracking filter is incorporated. A pre-set model reliability is given, and after installation, a motion model corresponding to the target motion is estimated based on a smooth value or a display track related to the tracking filter, and tracking corresponding to the estimated motion model is performed Since the model reliability of the filter is calculated and given higher, when the tracking filter is incorporated, the model reliability of the tracking filter can be adaptively reconfigured, so tracking can be performed according to the situation, High-precision tracking is possible.

It is a block diagram which shows the function structure of the target tracking apparatus by Embodiment 1 thru | or Embodiment 3 of this invention. It is a block diagram which shows the function structure of the target tracking apparatus by Embodiment 4 of this invention. It is a block diagram which shows the function structure of the target tracking apparatus by Embodiment 5 of this invention. It is a block diagram which shows the function structure of the target tracking apparatus by Embodiment 6 thru | or Embodiment 8 of this invention. It is a block diagram which shows the function structure of the target tracking apparatus by Embodiment 9 of this invention.

Explanation of symbols

  100 target tracking device, 110, 120, 130 tracking filter, 111, 121, 131 delay processing unit, 113, 123, 133 mixing processing unit, 114, 124, 134 track prediction unit, 115, 125, 135 track update unit, 116 , 117, 118, 126, 127, 128, 136, 137, 138, 141, 142, 143 switch means, 160 initial value calculation unit, 164 model reliability recalculation unit, 170 model reliability calculation unit, 180, 181, 182 filter selection control unit, 190 weighting processing unit, 200 target observation device, 300 display device, 400 external information source.

Claims (9)

From the target position observation information on the target position observed by the sensor of the target observation device, both smooth value and smooth error representing the target motion status using multiple tracking filters with motion models that match the target motion assumed in advance. Each variance matrix is calculated, and based on the calculated smooth value, smoothing error covariance and target position observation information, the model likelihood for each tracking filter for the target is calculated by the model reliability calculation means, and each model likelihood is calculated. In a target tracking device that calculates a normalized model reliability, creates a wake for display by weighting and averaging the smoothed value and smoothing error covariance calculated by the tracking filter with a model likelihood by weighting processing means ,
File selection control means for determining whether to separate or incorporate the tracking filter corresponding to the value based on a value representing the state of the target motion obtained from the tracking filter, and to generate a control signal based on the determination result;
A target tracking device comprising switch means for separating or incorporating a corresponding tracking filter based on a generated control signal.   The file selection control means uses the model reliability or model likelihood calculated by the model reliability calculation means as a value representing the state of the target exercise, and determines whether the tracking filter is separated and incorporated. The target tracking device according to claim 1.   The file selection control means uses a smoothed value of the tracking filter or a wake weighted for display created by the weighting processing means as a value representing the state of the target motion, and determines whether the tracking filter is separated and incorporated. The target tracking device according to claim 1, wherein   The file selection control means inputs external information that can be used to estimate the target motion, such as the track of the same target, model likelihood or model reliability, from an external information source other than the target observation device, and determines the status of the target motion. The target tracking device according to any one of claims 1 to 3, wherein in addition to the value to be expressed, the tracking filter is separated and incorporated to determine whether the external information is included. When a tracking filter is newly installed, it is based on the smooth values of other tracking filters that have already been installed and stable in operation and the specifications of the track weighted for display by the smoothing error covariance matrix or weighting processing unit. A smoothing value of a newly incorporated tracking filter, an initial value calculation unit for calculating an initial value of a smoothing error covariance matrix,
5. The model reliability calculation unit calculates a model reliability corresponding to the newly incorporated tracking filter based on the calculated initial value. Item 1. The target tracking device according to item 1. When the tracking filter is disconnected, the tracking update operation is performed independently to calculate the smooth value and the smooth error covariance matrix.
The filter selection control means controls the switch means when it is judged that the separated tracking filter is to be incorporated, and incorporates the tracking filter at a timing when the operation is stable. 4. The target tracking device according to claim 1.   7. The target tracking device according to claim 6, wherein the filter selection control means determines the timing for incorporating the separated tracking filter based on the smoothing error covariance matrix of the separated tracking filter.   The filter selection control means determines the timing for incorporating the separated tracking filter based on the model likelihood calculated from the smoothed value and smoothing error covariance of the separated tracking filter. Target tracking device.   As the model reliability output by the model reliability calculation means, a pre-set model reliability for the tracking filter is given when the tracking filter is installed, and based on a smooth value or a display track related to the tracking filter after installation. And a model reliability recalculation unit that estimates a motion model corresponding to the target motion and calculates and gives a higher model reliability of the tracking filter corresponding to the estimated motion model. The target tracking device according to any one of claims 1 to 8.

Images