KR102266432B1 - Method for tracking a location of a target and method for correcting a location of target - Google Patents

Abstract

The present invention provides a method for tracking a location of a target, which includes the steps of: comparing measured trajectory data obtained by a tracking camera for tracking a flying target with pre-calculated expected trajectory data of the target using a dynamic time warping (DTW) algorithm to predict a current position of the target; and using the current position predicted by predicting the current position of the target as an input value and predicting the position of the target after a predetermined time by a Kalman filter. According to the present invention, it is possible to track the target without interruption due to a natural environment or the like, and the optical tracking equipment enables photographing by correcting the position accordingly.

Description

A method of tracking the position of a target and a method of correcting the position of a target {METHOD FOR TRACKING A LOCATION OF A TARGET AND METHOD FOR CORRECTING A LOCATION OF TARGET}

The present invention relates to a method for tracking the position of a flying target and correcting the tracking position.

During the flight test of a missile, the optical tracking equipment acquires flight images from the launch of the missile to the flight and impact section, which is a great help in directly checking and qualitatively evaluating the performance of the test object.

1, the optical tracking equipment for tracking a missile is generally PAN/TILT (horizontal/vertical controller) that can control azimuth and elevation, and an IR camera (infrared light) for tracking. camera), and a high-speed camera to obtain high-quality images.

The target location information is extracted from the image acquired from the IR camera through the target tracking algorithm, and the control algorithm sends azimuth and elevation control signals to the PAN/TILT using this location information, and the PAN/TILT moves according to the control signal. . A high-speed camera installed to face the same point as the tracking camera acquires high-quality missile tracking images as the PAN/TILT moves as it moves.

One of the problems of such a target tracking system is that the target tracking algorithm cannot detect the target when the missile is obscured by clouds, the sun, or the like. To improve this, a method of predicting the target's position when the target is obscured, aiming at the predicted point, and then tracking the target again when the target appears on the screen is used. Among them, techniques for predicting the location of a target using Kalman Filter have been studied.

However, since these methods make predictions using only the target's previous speed or location information, if the target's motion shows a different trend from the previous motion characteristics, the prediction becomes inaccurate.

Another problem is that since the tracking camera is used as an IR camera, the optical tracking equipment tracks the flame generated from the missile's propellant, so that the missile, which is the actual tracking target, cannot be tracked so as to be in the center of the screen as shown in FIG. 2 .

The matters described in the background art are intended to help the understanding of the background of the invention, and may include matters that are not already known to those of ordinary skill in the art to which this technology belongs.

Korean Patent Publication No. 10-2016-0110773

The present invention has been devised to solve the above problems, and the present invention enables a target to be tracked without interruption due to a natural environment, etc., and the position of the target is corrected accordingly to enable the optical tracking equipment to shoot; and An object of the present invention is to provide a method for correcting the position of a target.

A method for tracking a position of a target according to an aspect of the present invention uses a DTW (Dynamic Time Warping) algorithm for measurement trajectory data acquired by a tracking camera that tracks a flying target and pre-calculated expected trajectory data of the target In comparison, predicting the current position of the target and predicting the position of the target after a predetermined time by a Kalman filter using the current position predicted by the step of predicting the current position of the target as an input value including the steps of

The predicting of the current position is characterized in that the predicted angular velocity is derived together with an elevation and an azimuth of the target.

In addition, the Kalman filter is characterized in that the predicted angular velocity is used as an input value to predict the position of the target after the predetermined time.

Here, the cost function applied to the DTW algorithm is characterized in that it is defined as follows.

는 코스트 매트릭스의 성분이고,

는 계측 궤적 데이터의 방위각, 고각 및 해당 시각이며,

는 예상 궤적 데이터의 방위각, 고각 및 해당 시각임.)(here,

is a component of the cost matrix,

is the azimuth, elevation, and corresponding time of the measurement trajectory data,

is the azimuth, elevation, and time of the predicted trajectory data.)

Next, in the method of correcting the position of the target according to an aspect of the present invention, the measurement trajectory data obtained by a tracking camera that tracks a flying target and the pre-calculated expected trajectory data of the target are combined with a DTW (Dynamic Time Warping) algorithm and estimating the current position of the target by comparing using , and deriving the predicted angular velocity of the target from the current position of the target.

The method may further include deriving a position correction value for tracking the target using the predicted angular velocity derived by the step of deriving the predicted angular velocity.

In addition, the step of deriving the position correction value is characterized in that deriving a new elevation (elevation) and azimuth (azimuth) oriented toward the predicted angular velocity direction from the predicted current position of the target.

So, it is characterized in that the tracking camera is controlled to track the elevation and azimuth.

Furthermore, the method may further include predicting the position of the target after a predetermined time by using a Kalman filter using the current position predicted by the step of predicting the current position of the target as an input value.

And, the Kalman filter (Kalman filter) is characterized in that the predicted angular velocity as an input value to predict the position of the target after the predetermined time.

Meanwhile, a cost function applied to the DTW algorithm is characterized in that it is defined as follows.

는 코스트 매트릭스의 성분이고,

는 계측 궤적 데이터의 방위각, 고각 및 해당 시각이며,

는 예상 궤적 데이터의 방위각, 고각 및 해당 시각임.)(here,

is a component of the cost matrix,

is the azimuth, elevation, and corresponding time of the measurement trajectory data,

is the azimuth, elevation, and time of the predicted trajectory data.)

The present invention utilizes the predicted trajectory data known in advance when measuring the flight test of a missile and the actual measured trajectory data of the missile obtained through the optical tracking equipment, and uses only the Kalman filter in the past through the method of using the DTW algorithm and the Kalman filter together It can predict the position of the target more accurately compared to the method.

And, through the predicted position, even in a situation where the missile is obscured by clouds or sunlight, the target's position is not completely lost, and when the missile appears again, it can be tracked.

In addition, if an IR camera is used as a tracking camera, the flame is tracked, so it is impossible to track the missile, which is the actual target of tracking, in the center. However, if the tracking position is corrected by using the method proposed in the present invention, it is possible to track a position closer to the missile to be tracked so that the missile comes closer to the center.

1 shows an optical tracking device for missile tracking.
2 shows a target tracking result using an IR camera.
Fig. 3 schematically shows an algorithm for the method of the present invention.
Figure 4 shows the method of the present invention in sequence.
5 is a view showing a current point of view among predicted trajectories through the DTW algorithm.
6 shows an example of the DTW algorithm.
7 is for explaining a cost matrix.
8 and 9 illustrate a method of obtaining a warping path by DTW.
10 is a summary of the prediction/correction steps of the Kalman Filter.
11 shows a target tracking result using an IR camera according to the present invention.
12 shows a target tracking position correction relationship according to the present invention.
13 is a view showing a comparison between the tracking result according to the present invention and the tracking result by using the Kalman filter alone.

In order to fully understand the present invention, the operational advantages of the present invention, and the objects achieved by the practice of the present invention, reference should be made to the accompanying drawings illustrating preferred embodiments of the present invention and the contents described in the accompanying drawings.

In describing preferred embodiments of the present invention, well-known techniques or repetitive descriptions that may unnecessarily obscure the gist of the present invention will be reduced or omitted.

Fig. 3 schematically shows an algorithm for the method of the present invention.

Hereinafter, a method for tracking a position of a target and a method for correcting a position of a target according to an embodiment of the present invention will be described with reference to FIG. 3 .

The method for tracking the position of the target and the method for correcting the position of the target according to an embodiment of the present invention compares the measurement trajectory of the target measured by the optical tracking equipment and the predicted trajectory known in advance by utilizing the DTW algorithm and Kalman Filter, It is a method and system for figuring out where the current missile's position is, predicting the next position of the target through the azimuth, elevation, and angular velocity information of the missile at that position and Kalman Filter, and correcting the target's tracking position.

The measurement trajectory data is a trajectory (elevation, azimuth) of a target acquired through tracking by a tracking camera such as an IR camera of an optical tracking device.

The predicted trajectory data is the predicted trajectory of the target pre-stored in the system.

The system and method of the present invention compares the measured trajectory data and the expected trajectory data using a DTW (Dynamic Time Warping) algorithm, and considers the time at which the target is lost, at which point the current missile's position is in the predicted trajectory. predict cognition

The predicted trajectory data predicted by the DTW algorithm become the predicted elevation and predicted azimuth corresponding to the predicted position.

Next, the predicted angular velocity derived from this, along with the predicted elevation angle and predicted azimuth, is input to the Kalman Filter to predict where the target will be located at a future point in time after the next predetermined time.

Next, a newly calibrated tracking position is obtained by using the predicted predicted angular velocity and the tracking result. That is, a correction value (elevation angle, azimuth angle) of the target position is obtained.

And, it is possible to control the PAN/TILT of the optical tracking device and the tracking camera by the correction value of the target position.

FIG. 4 shows the method of the present invention in sequence, and an embodiment of the present invention will be sequentially described with reference to FIG. 4 .

And, Figure 5 shows the current time point among the predicted trajectories through the DTW algorithm (AZ means azimuth, EL means elevation), Figure 6 shows an example of the DTW algorithm, Figure 7 shows the cost matrix will explain

First, the target is detected by the target tracking system (S11), and the last measured trajectory data of the target and the predicted trajectory data calculated in advance by the DTW (Dynamic Time Warping) algorithm are obtained (S12).

Next, it is determined whether the target is observed in the current frame (S13), and when the target is observed, the current position on the trajectory data is determined as the position on the predicted trajectory of the last observation data (S14), otherwise, the target is In consideration of the lost time, it is possible to determine which point in time the current point corresponds to on the predicted trajectory data (S21), and through this, the current position (elevation, azimuth) and angular velocity (the slope of the expected trajectory data) can be predicted ( S15).

The DTW algorithm is an algorithm that matches two consecutive data, and first calculates the cost for all combinations that can match each data of the two consecutive data to the cost matrix (Equation 1, FIGS. 6 and 7) write Cost usually uses Euclidean distance. After that, if a path is obtained according to a small cost value in the cost matrix, a warping path as in Equation 2 can be obtained.

는 도시와 같이 코스트 매트릭스의 성분이고,

는 계측 궤적 데이터의 방위각과 고각이며,

는 예상 궤적 데이터의 방위각과 고각이다.here,

is a component of the cost matrix as shown in the city,

is the azimuth and elevation of the measurement trajectory data,

are the azimuth and elevation angles of the predicted trajectory data.

In this patent, a new type of cost function is proposed as follows. The proposed cost function uses time information as well as continuous data to solve the problem of continuously matching multiple data to one data without considering the passage of time, which is a problem with the existing cost function.

First, after moving the time axis so that the first data of two consecutive data become 0s, the time difference is calculated, and the cost is calculated by considering this value. In addition, by taking exponential to the time difference value, the penalty is designed to increase as the time difference increases.

는 코스트 매트릭스의 성분이고,

는 계측 궤적 데이터의 방위각, 고각 및 해당 시각이며,

는 예상 궤적 데이터의 방위각, 고각 및 해당 시각이다.here,

is a component of the cost matrix,

is the azimuth, elevation, and corresponding time of the measurement trajectory data,

is the azimuth, elevation, and corresponding time of the predicted trajectory data.

는 Trial and error을 통해 구해지는 사용자 지정 값이며,

,

는 다음 수학식과 같이 표현되는 t와 θ의 차원을 맞춰주기 위한 계수로서, 예를 들어 θ가 5~10인데 t가 0.001~0.005이면 차원이 차이가 나서 exp 부분의 영향이 너무 작고, θ가 0.5~1인데 t가 10~50이면 차원이 차이가 나서 exp 부분의 영향이 너무 크게 된다.Also,

is a user-specified value obtained through trial and error,

,

is a coefficient for matching the dimensions of t and θ expressed as in the following equation. For example, if θ is 5 to 10 and t is 0.001 to 0.005, the dimension is different and the influence of the exp part is too small, and θ is 0.5 Although ~1, if t is 10-50, the dimension is different and the influence of the exp part becomes too large.

A method of obtaining a warping path is shown in detail in FIGS. 8 and 9 .

As shown in FIG. 8, it can be seen that the position on the trajectory of the last measurement data is the point at which the warping path is first diagonally bent from the starting point, starting from the point where the last data of the two time series data are matched as shown in FIG. The warping path can be found by following the smallest value. Here, the 3 direction is the direction in which the index decreases, and if the current index is (i, j), the cost is the highest among (i-1, j-1), (i, j-1), and (i-1, j). A small place becomes the next warping path.

Next, FIG. 10 summarizes the prediction/correction steps of the Kalman Filter.

The Kalman filter consists of a prediction step that predicts the next state using the system model and a correction step that corrects the current state using the observed data. In the present invention, the next state is predicted by inputting the predicted angular velocity obtained through DTW as a control input in the prediction step. The current position angle value on the predicted trajectory is not used in the Kalman filter.

That is, the following Equation 6 is used in the prediction step.

는 칼만필터의 미래에 대한 추정 값,

는 DTW에 의해 구해진 값이 아닌, 칼만필터의 현재에 대한 추정 값이며,

는 DTW에 의해 구해진 예측 각속도이다.here,

is an estimate of the future of the Kalman filter,

is not a value obtained by DTW, but an estimate of the current of the Kalman filter,

is the predicted angular velocity obtained by DTW.

Using this, it is checked whether the target is observed in the current frame after S15 (S16), and if not observed, the next target state is predicted with the Kalman filter using the predicted angular velocity as above (S17).

Then, if the target is observed as a result of S16, the tracking position is corrected as described later (S22), and the current target state is corrected using the Kalman filter using the observed data (S23).

The correction (S22) of the tracking position will be described with reference to FIGS. 11 and 12 .

11 shows the target tracking result using the IR camera according to the present invention, and FIG. 12 shows the target tracking position correction relationship according to the present invention.

일 때, 미사일 진행방향의 각도를

라고 가정한다.The optical tracking system used in missile tracking uses an IR camera as a tracking camera to track the missile's flame, so the missile, the measurement target, cannot be located in the center during tracking. Therefore, the present invention proposes a method of correcting this by using the trajectory information of the missile. First, the angular velocity predicted through DTW is

When , the angle of the missile's travel direction is

Assume that

의 각도로 직선을 그어서 화염을 탐지한 바운딩 박스와 만나는 지점을 새로운 추적 위치로 정한다. 그러면 조금 더 실제 추적 대상인 미사일에 가까운 점을 추적 할 수 있게 된다.And from the center point where the flame was detected

Draw a straight line at an angle of , and set the point where it meets the bounding box that detected the flame as the new tracking position Then, it is possible to track a point that is closer to the missile that is actually being tracked.

As described above, according to the target position tracking method and system and the target position correction method and system of the present invention, the next target position can be predicted by the Kalman filter through the position and angular velocity information predicted by the DTW algorithm, and the calculated It is possible to correct the position of the target for tracking by calculating the target position correction value from the inclination.

That is, according to the location tracking method of the present invention, it is possible to track the target even when the target is detected or not, and when the target is not detected, a better tracking result is obtained compared to the case where only the Kalman filter is used.

And, when the target is detected, it is possible to correct the tracking position.

In FIG. 13 , black indicates the actual target position, green indicates a case predicted using only the Kalman filter, and blue indicates a prediction result according to the present invention.

As can be seen from this reference, in a situation in which the target is not detected after 2 seconds, both methods have similar results before 2 seconds, but after 2 seconds, it can be seen that the results of the two methods are significantly different.

Although the present invention as described above has been described with reference to the illustrated drawings, it is not limited to the described embodiments, and it is common knowledge in the art that various modifications and variations can be made without departing from the spirit and scope of the present invention. self-evident to those who have Accordingly, such modifications or variations should be said to belong to the claims of the present invention, and the scope of the present invention should be interpreted based on the appended claims.

Claims (11)

estimating a current position of the target by comparing the measured trajectory data obtained by a tracking camera that tracks a flying target with the pre-calculated predicted trajectory data of the target using a DTW (Dynamic Time Warping) algorithm; and
Using the current position predicted by the step of predicting the current position of the target as an input value and predicting the position of the target after a predetermined time by a Kalman filter,
A cost function applied to the DTW algorithm is characterized in that it is defined by the following equation,
How to track a target's location.

(here,

is a component of the cost matrix,

is the azimuth, elevation, and corresponding time of the measurement trajectory data,

is the azimuth, elevation and corresponding time of the predicted trajectory data,

is a user-specified value obtained through trial and error) The method according to claim 1,
Predicting the current position is characterized in that deriving a predicted angular velocity together with an elevation and an azimuth of the target,
How to track a target's location. 3. The method according to claim 2,
The Kalman filter is characterized in that the predicted angular velocity is used as an input value to predict the position of the target after the predetermined time,
How to track a target's location. delete estimating a current position of the target by comparing the measured trajectory data obtained by a tracking camera that tracks a flying target with the pre-calculated predicted trajectory data of the target using a DTW (Dynamic Time Warping) algorithm; and
deriving a predicted angular velocity of the target from the predicted current position of the target;
A cost function applied to the DTW algorithm is characterized in that it is defined by the following equation,
How to correct the position of the target.

(here,

is a component of the cost matrix,

is the azimuth, elevation, and corresponding time of the measurement trajectory data,

is the azimuth, elevation and corresponding time of the predicted trajectory data,

is a user-specified value obtained through trial and error) 6. The method of claim 5,
Further comprising the step of deriving a position correction value for tracking the target using the predicted angular velocity derived by the step of deriving the predicted angular velocity,
How to correct the position of the target. 7. The method of claim 6,
The step of deriving the position correction value is characterized in that deriving a new elevation and azimuth oriented in the predicted angular velocity direction from the predicted current position of the target,
How to correct the position of the target. 8. The method of claim 7,
Controlling the tracking camera to track the elevation and azimuth
How to correct the position of the target. 7. The method of claim 6,
Further comprising the step of predicting the position of the target after a predetermined time by a Kalman filter (Kalman Filter) using the current position predicted by the step of predicting the current position of the target as an input value,
How to correct the position of the target. 10. The method of claim 9,
The Kalman filter is characterized in that the predicted angular velocity is used as an input value to predict the position of the target after the predetermined time,
How to correct the position of the target. delete

Images