KR102126670B1 - Apparatus and method for tracking objects with optimizing region of interest - Google Patents

Abstract

The present invention relates to an obstacle tracking device and method for optimizing the detection area, a remote sensor for sensing the position of an object around the vehicle, a target calculating unit for identifying a target object to be tracked among the objects detected by the remote sensor , A track management unit for selecting and updating an effective gate indicating a range of a target object identified by the target calculation unit, and a storage unit for storing the location of the target and the update history of the effective gate, and according to the present invention, the surrounding state of the vehicle In order to accurately detect the obstacles around the vehicle by dynamically updating the effective gate indicating the detection area that the detection system tracks to detect as an obstacle, the distance to an obstacle that can be accurately tracked using only a lidar or radar sensor is extended. The risk of accidents can be prevented.

Description

Obstacle tracking device and method for optimizing detection area{APPARATUS AND METHOD FOR TRACKING OBJECTS WITH OPTIMIZING REGION OF INTEREST}

The present invention relates to an obstacle tracking device and method for optimizing a detection area, and more specifically, in a vehicle's surrounding state detection system, extends a distance to an obstacle that can be tracked using a rider or a radar sensor and prevents an obstacle. The present invention relates to an obstacle tracking device and method for optimizing a detection area to enable accurate tracking.

The vehicle's surrounding condition detection system detects surrounding objects through a LIDAR sensor, radar sensor, or camera sensor, and issues an alert to notify the driver of the presence of an obstacle, or controls the vehicle's driving system. Stop before hitting this obstacle, or try to drive around the obstacle.

However, when a surrounding object is detected through a camera sensor, it is highly likely to determine whether the detected object is an obstacle that needs to avoid collision, because it directly shoots the image of the object, but due to the resolution and visibility of the screen There is a limit to the distance that an image that can be discerned is limited, and there is a problem that it is difficult to detect a distance from an object using the image alone.

A lidar sensor or a radar sensor has the advantage of being able to detect objects at a relatively long distance, but since it is not susceptible to noise, rather than directly sensing the image of the object, it is an obstacle that the detected object needs to avoid collision Whether it is cognitive noise or not, it is not easy to determine whether there is a problem that the target may be omitted because the sensor cannot track the movement of the object when tracking the movement of the surrounding object.

Related prior art is Republic of Korea Patent Publication No. 10-1996-0034983 (published October 24, 1996, the name of the invention: a vehicle collision prevention device and method using a detection type optical distance measuring device).

The present invention dynamically detects obstacles around a vehicle by dynamically updating an effective gate indicating a detection area including a target object to be tracked so that a real obstacle can be accurately identified and tracked among object information around a vehicle detected by a lidar or a radar sensor. An object of the present invention is to provide an obstacle tracking device and method for optimizing a detection area, which enables accurate tracking.

The obstacle tracking device for optimizing the detection area according to an aspect of the present invention includes a remote sensor that detects the position of an object and outputs a detection signal, identifies an obstacle based on the detection signal, and estimates a position corresponding to the identified obstacle. Tracking the position of an obstacle by creating a track containing the covariance of errors, but updating the track generated by the target tracking unit and the target tracking unit that calculates and updates the detection area that is the range in which the track detects the obstacle and updates the track generated by the target tracking unit. And a track manager configured to initialize the track based on the estimated position of the included obstacle.

In the present invention, the remote sensor is characterized in that at least one of a lidar sensor or a radar sensor.

In the present invention, the target tracking unit sets the number of the tracks, the initial position and initial detection area of each track based on the position of the object indicated by the detection signal, and the detection area increases as the time for tracking the obstacle increases. Characterized in that to enlarge.

In the present invention, the target tracking unit determines whether the position of the object indicated by the detection signal is included in the detection area based on the error of the measurement value and the prediction value of the object position and the covariance of the error, and the track management unit The position estimation value included in the track is updated based on the position estimation value and the detection signal included in a track corresponding to an object included in the detection area.

In the present invention, the track management unit further includes a storage unit for storing a history in which the track is updated, and the track management unit has a distance between the object position estimation value included in the first track and the object position estimation value included in the second track is less than a preset reference value. If the first track and the second track are initialized, and all the object position estimates included in the first track are not included in the detection area corresponding to the first track, the first track is initialized, but the history stored in the storage unit is initialized. It is characterized by initializing the track on the basis of.

An obstacle tracking method for optimizing a detection area according to another aspect of the present invention includes a target estimation unit co-variance of a position estimation value and an error corresponding to an obstacle identified based on a detection signal that a remote sensor detects and outputs an object position. Generating a track, calculating and updating a detection area within a range in which the target tracking unit detects an obstacle with respect to the track, and a position of an object indicated by the detection signal among the detection signals in the target tracking unit in the detection area Selecting an included detection signal and updating a position estimation value included in the track based on a position estimation value included in a track corresponding to an object including a position of an object in the detection area and the selected detection signal and the track management unit It characterized in that it comprises a step.

In the obstacle tracking method for optimizing the detection area according to the present invention, the remote sensor is characterized in that at least one of a lidar sensor or a radar sensor.

In the obstacle tracking method for optimizing the detection area according to the present invention, in the step of selecting the detection signal, the target tracking unit determines whether the position of the object indicated by the detection signal is included in the detection area corresponding to the track. It is characterized by determining based on the error of the measured value and the predicted value of the position and the covariance of the error.

In the obstacle tracking method for optimizing the detection area according to the present invention, in the step of updating the position estimate included in the track, the track management unit estimates the object position included in the first track and the object position estimate included in the second track. When the distance therebetween is less than a preset reference value, it is characterized in that the first track and the second track are initialized based on the track history stored in the storage unit.

In the obstacle tracking method for optimizing the detection area according to the present invention, in the step of updating the position estimation value included in the track, the track management unit includes all object position estimation values included in the track in the detection area corresponding to the track If not, the track is initialized based on the history of the track stored in the storage unit.

According to the present invention, by accurately updating the effective gate indicating the detection area that the vehicle's surroundings detection system tracks to detect as an obstacle to accurately track obstacles around the vehicle, accurately tracking the position using only a lidar or radar sensor This can extend the distance to possible obstacles and prevent the risk of accidents.

1 is a block diagram of an obstacle tracking device that optimizes a detection area according to an embodiment of the present invention.
2 is a flowchart illustrating an operation of an obstacle tracking method for optimizing a detection area according to an embodiment of the present invention.

Hereinafter, an obstacle tracking apparatus and method for optimizing a detection area according to the present invention will be described in detail with reference to the accompanying drawings. In this process, the thickness of the lines or the size of components shown in the drawings may be exaggerated for clarity and convenience. In addition, terms to be described later are terms defined in consideration of functions in the present invention, which may vary according to a user's or operator's intention or practice. Therefore, the definition of these terms should be made based on the contents throughout this specification.

1 is a block diagram of an obstacle tracking device that optimizes a detection area according to an embodiment of the present invention.

As shown in Figure 1, the obstacle tracking device for optimizing the detection area according to an embodiment of the present invention includes a remote sensor 100, a target tracking unit 200, a track management unit 300 and a storage unit 500 Can be achieved.

The remote sensor 100 detects the position of an object around the vehicle and outputs a detection signal.

At this time, the remote sensor 100 may be at least one of a lidar sensor or a radar sensor. A lidar and radar sensor or a camera sensor may be used to detect obstacles at a distance in the vehicle's surroundings detection system. While the camera sensor is difficult to cope with long distance obstacles, it has the advantage that it can accurately determine whether the detected object is an obstacle. Accordingly, a lidar sensor or a radar sensor can be used to cope with distant obstacles.

The target tracking unit 200 tracks a position by identifying an obstacle based on a detection signal and generating a track including a covariance of a position estimate and an error corresponding to the identified obstacle, but the track is a range for detecting an obstacle The detection area is calculated and updated.

At this time, the target tracking unit 200 includes a detection area calculating unit 210 for calculating and updating a detection area that is a range in which a track detects an obstacle, and an effective measurement value determining unit 220 for determining whether a detection signal is included in the detection area It can be made including.

In general, Kalman filters are used to overcome sensor errors and detect the position of moving objects. The Kalman filter repeats calculating the estimated value of the position of the object by the estimated value of the position of the object of the previous time and the measured value of the position of the object, thereby offsetting the error occurring in the measurement of the position of the object to accurately position the object. It is a technique for estimating. At this time, first, the estimated value of the current time using only the measured value up to the previous time is calculated using the estimated value of the position of the object up to the previous time. After that, the estimated value of the current time using only the measured values up to the previous time is corrected using the measured value of the position of the object of the current time and the covariance of the current time calculated using only the measured values up to the previous time. Calculate the estimated value of

The target tracking unit 200 sets the number of the tracks, the initial position of each track, and the detection area based on the position of the object indicated by the detection signal. At this time, the initial position and the detection area of the track can be calculated using the Kalman filter represented by the following equation (1).

는 시간 k-1까지의 정보를 사용하여 추정한 시간 k에서의 물체의 상태값의 추정치를 나타내고,

는 시간 k-1까지의 정보를 사용하여 추정한 시간 k-1에서의 물체의 상태값의 추정치를 나타내며.

는 시간 k-1까지의 정보를 사용하여 추정한 시간 k에서의 물체의 위치의 추정치를 나타낸다. At this time,

Denotes an estimate of the state value of an object at time k estimated using information up to time k-1,

Denotes an estimate of the state value of an object at time k-1 estimated using information up to time k-1.

Denotes an estimate of the position of the object at time k estimated using information up to time k-1.

Here, each track may include a Kalman filter for tracking a specific object detected by the remote sensor 100. That is, each track includes the covariance of the error for correcting the position by the estimated value and the measured value of the position of the obstacle to be tracked, and the estimated value, measured value, and covariance calculated for each time are stored in the storage unit 400 as a history. You can.

The configuration of the Kalman filter included in the track can be expressed by Equation 2 below.

In Equation 1, x(k) represents a state value of an object at time k, and F(k-1) is a state transition model representing a change in the transition from time k-1 to time k, and z(k ) Denotes the position of the object at time k, H(k) is the observation model representing the transformation from the state of the object to the position of the object, v(k-1) is the processing noise at time k-1, w (k) means the measurement noise at time k. In addition, P(k|k) means the covariance of the error of the Kalman filter at time k calculated by considering information up to time k, and P(k|k-1) considers information up to time k-1 It means the covariance of the error of the Kalman filter at the time k calculated. Q(k-1) represents the expected covariance at time k-1.

The target tracking unit 200 updates the detection area, which is a range where the track detects an obstacle, and determines whether the position of the object indicated by the detection signal is included in the detection area corresponding to the track. Judgment is based on the covariance of errors.

At this time, the target tracking unit 200 sets the range of the detection area according to the state value of the Kalman filter included in the current track and updates the state value of the Kalman filter using the measurements included in the detection area. The target tracking unit 200 first calculates a residual based on a measurement of the position of the object and a predicted value of the position of the object, and calculates the residual covariance based on the covariance and observation model of the estimated error included in the Kalman filter. It is determined whether the object enters the detection area based on the residual covariance.

Here, the detection region is set to a region representing a specific probability value or less in a Gaussian probability distribution having a residual covariance as a variance, and this probability value is called a gate probability. Therefore, the detection region can be calculated by calculating the residual covariance and setting the gate probability value, and since the residual covariance and the detection region are optimized over time by the Kalman filter, the detection region can be optimized according to the passage of time. At this time, the tracking reliability of the track increases as the tracking time for the track object increases, and as the tracking reliability increases, the detection area may be updated to expand to a certain limit.

The calculation method of the residual and residual covariance and the inclusion conditions in the detection region of the object can be expressed by the following equation (3).

In Equation 3, v(k,i) represents the residual of object i at time k, and z(k,i) represents a measurement of the position of object i. In addition, P(k|k-1) represents the covariance of the estimated error of the Kalman filter, R(k) represents the measurement noise at time k, and S(k) represents the measurement noise at time k. r represents the range of the detection area.

The track management unit 300 updates the position estimation value included in the track based on the position estimation value and the detection signal included in the track corresponding to the object included in the detection area.

Here, the track management unit 300 calculates and updates the position estimation value of the target included in the track, the position estimation value updating unit 310, the track initialization unit that initializes the track if the track crashes or fails to follow the target ( 320) and a track updating unit 330 for updating the track according to the updated position estimate.

At this time, the track manager 300 calculates the Kalman gain based on the covariance and the residual covariance of the estimation error in order to update the position estimate, and estimates the Kalman gain, the position measurement of the object, and information up to the previous time. Based on the estimate of one location, the estimated location estimate is calculated using the information up to the current time. The update of the position estimate value can be expressed by the following equation (4).

In Equation 4, K(k) represents Kalman gain. As described above, the track management unit 300 updates the position estimation value in consideration of the measurement value over time, thereby obtaining a more accurate position estimation value.

If the distance between the object position estimate value included in the first track and the object position estimate value included in the second track is less than a preset reference value, the track management unit 300 may perform the first track and the first track based on the history stored in the storage unit 500. 2 Initialize the track.

The storage unit 500 stores a history in which tracks are updated. At this time, the history stored in the storage unit may include a covariance value of a position estimate and a measurement value and an estimation error for each time of the Kalman filter included in the track.

When updating the position estimation value as described above, in some cases, objects represented by two tracks may collide, and when the position estimation value of the object represented by the track decreases below a pre-stored reference value, the track management unit 300 includes two tracks. It is determined that the object to be represented collides, and the track can be initialized based on data included in the history of both colliding tracks.

In addition, the track management unit 300 initializes the track based on the history of the track stored in the storage unit when all the object position estimates included in the track are not included in the detection area corresponding to the track. That is, if all the objects tracked are out of the detection area or are judged to be noise or error and the tracked object disappears, the track fails to track the object, so the track manager 300 initializes the track to replace the new object. Can be followed.

As described above, the Kalman filter is used to track the obstacles that the track moves, but when the track that follows the object fails to follow or the two tracks collide with each other, the surrounding state detection system is made by resetting the track and letting the new object follow. It can improve the object identification performance.

As described above, the data included in the track that is generated and updated by the target tracking unit 200 and the track management unit 300 following the obstacle is transmitted to the vehicle control unit 400 so that the vehicle avoids the obstacle or alerts the driver. It can be used to control the vehicle.

2 is a flowchart illustrating an operation of an obstacle tracking method for optimizing a detection area according to an embodiment of the present invention.

The operation of the obstacle tracking method for optimizing the detection area according to an embodiment of the present invention will be described in detail with reference to FIG. 2.

First, the target tracking unit 200 generates a track including a covariance of a position estimation value and an error corresponding to an obstacle identified based on a detection signal that the remote sensor 100 detects and outputs the position of the object (S110).

At this time, as described above, the remote sensor may be at least one of a rider sensor or a radar sensor.

In addition, the track generated by the target tracking unit 200 includes a Kalman filter that includes a covariance of position estimates and errors. At this time, the configuration of the Kalman filter included in the track is as described in relation to Equations 1 and 2 above.

Thereafter, the target tracking unit 200 calculates and updates a detection area that is a range for detecting an obstacle with respect to the track (S120).

At this time, the size of the detection area may be set to an initial value based on the position of the object indicated by the detection signal. In addition, the detection region may be set as a region indicating a gate probability value or less in a Gaussian probability distribution having residual covariance as a variance.

Here, as the time for the track to follow the object increases, the reliability of the track increases, and as the reliability of the track increases, the detection area may be updated to enlarge the detection area.

Subsequently, the target tracking unit 200 selects a valid detection signal in which the position of the object indicated by the detection signal among the detection signals is included in the detection area (S130).

As described above, the detection signal includes a measurement of the position of the object tracked by the vehicle's surrounding search system, and the target tracking unit 200 selects an effective measurement included in the detection area from among these measurements, and updates the Kalman filter. Make it available for tracking objects.

At this time, the target tracking unit 200 determines whether or not the position of the object indicated by the detection signal is included in the detection area corresponding to the track, based on the error of the measured value of the object position and the prediction value and the covariance of the error.

At this time, the target tracking unit 200 sets the range of the detection area according to the state value of the Kalman filter included in the current track and updates the state value of the Kalman filter using the measurements included in the detection area. Here, the target tracking unit 200 first calculates a residual based on the measured value of the object's position and the predicted value of the object's position, and calculates the residual covariance based on the covariance and observation model of the estimated error included in the Kalman filter. , It is determined whether the object enters the detection area based on the residual and the residual covariance. The calculation method of the residual and residual covariance and the inclusion conditions in the detection region of the object are as expressed by Equation 3 above. Since the residual covariance and the detection region are optimized over time by the Kalman filter, the detection region can be optimized over time.

Subsequently, the track management unit 300 determines whether the distance between the object position estimation value included in the first track and the object position estimation value included in the second track is less than a preset reference value (S140). That is, it is determined whether there are tracks that are close to each other.

If it is determined in step S140 that there are tracks adjacent to each other, the process proceeds to step S160 to update the position estimate included in the track.

If it is determined in step S140 that there are tracks that are close to each other, the track manager 300 initializes the first and second tracks that are close to each other based on the history of the tracks stored in the storage 500 (S150). . That is, when the distance between the object position estimation value included in the first track and the object position estimation value included in the second track is less than a preset reference value, the track management unit 300 may control the track based on the track history stored in the storage unit 500. The first track and the second track can be initialized.

At this time, the history stored in the storage unit 500 may include covariance values of position estimates and measured values and estimated errors for each time of the Kalman filter included in the track.

When the position estimation value is updated as described above, in some cases, objects represented by two tracks may collide, and when the position estimation value of the object represented by the track decreases below a pre-stored reference value, the track management unit 300 includes two tracks. It is determined that the object to be represented collides, and the track can be initialized based on data included in the history of both colliding tracks.

Subsequently, the track manager 300 updates the selected detection signal and the estimated position value included in the track corresponding to the object including the position of the object in the detection area (S160).

At this time, the track manager 300 calculates the Kalman gain based on the covariance and the residual covariance of the estimation error in order to update the position estimate, and estimates the Kalman gain, the position measurement of the object, and information up to the previous time. Based on the estimate of one location, the estimated location estimate is calculated using the information up to the current time. The update of the position estimate is as expressed by Equation 4 above.

Thereafter, the track manager 300 determines whether the track has failed to follow the obstacle (S170). At this time, the track management unit 300 determines that the obstacle tracking has failed if all the object position estimates included in the track are not included in the detection area corresponding to the track.

If it is determined in step S170 that the obstacle following has failed, the track manager 300 may initialize the track based on the history of the track stored in the storage. At this time, the track management unit 300 first sorts data corresponding to the history of the track that failed to follow the obstacle stored in the storage unit (S180). Thereafter, the track manager 300 initializes the track that has failed to follow the obstacle (S190). Subsequently, the track manager 300 matches the data arranged in the step S180 to the track that has failed to follow the obstacle (S200 ). At this time, the track management unit 300 may make the track follow the new object by leaving only the data useful for the track to follow the new object among the sorted data.

If it is determined in step S170 that the obstacle following is successful, the track is currently successfully following the obstacle and the process is terminated while maintaining the current track. As described above, the track manager 300 may initialize the track based on the history of the track stored in the storage unit when all the object position estimates included in the track are not included in the detection area corresponding to the track.

As described above, the Kalman filter is used to track the obstacles that the track moves, but when the track that follows the object fails to follow or the two tracks collide with each other, the surrounding state detection system is made by resetting the track and letting the new object follow. It can improve the object identification performance.

The data included in the track generated and updated in the above manner may be transmitted to the vehicle control unit 400 and used to control the vehicle so that the vehicle avoids obstacles or issues an alert to the driver.

As described above, according to the vehicle brake light control apparatus and method according to the present embodiment, the vehicle's surrounding state detection system dynamically updates the effective gate indicating the region of interest to be detected as an obstacle to accurately track obstacles around the vehicle By using only a lidar or radar sensor, it is possible to extend the distance to an obstacle that can be accurately tracked and prevent the risk of an accident.

The present invention has been described with reference to the embodiment shown in the drawings, but this is only exemplary, and those skilled in the art to which the art belongs can various modifications and equivalent other embodiments from this. Will understand. Therefore, the technical protection scope of the present invention should be defined by the claims below.

100: remote sensor
200: target tracking unit
210: detection area calculator
220: effective measurement value determination unit
300: track management unit
310: position estimation value update unit
320: track initialization unit
330: track update unit
400: vehicle control unit
500: storage

Claims (10)

A remote sensor that senses the position of the object and outputs a detection signal;
Based on the detection signal, an obstacle is identified, and a track including a position estimation value corresponding to the identified obstacle and a covariance of error is generated to track the position of the obstacle, but calculating a detection area that is a range in which the track detects the obstacle. Target tracking unit to update;
A track management unit that updates the track generated by the target tracking unit and initializes the track based on an estimated position of an obstacle included in the track; And
It includes a storage unit for storing the history of the track is updated,
When the distance between the object position estimate value included in the first track and the object position estimate value included in the second track is less than a preset reference value, the track management unit may track the first track and the second track based on the track history stored in the storage unit. Obstacle tracking device comprising a.
According to claim 1,
The remote sensor is an obstacle tracking device, characterized in that at least one of a lidar sensor or a radar sensor.
According to claim 1,
The target tracking unit sets the number of the tracks, the initial position of each track, and the detection area based on the position of the object indicated by the detection signal, and updates to increase the detection area as the time for tracking the obstacle increases. Obstacle tracking device characterized in that.
◈ Claim 4 was abandoned when payment of the set registration fee was made.◈ According to claim 3,
The target tracking unit determines whether the position of the object indicated by the detection signal is included in the detection area, based on the error of the measurement value and the predicted value of the object position and the covariance of the error, and the track management unit Obstacle tracking device, characterized in that for updating the position estimate included in the track, based on the position estimate and the detection signal corresponding to the object to be included.
According to claim 1,
The track management unit initializes the first track based on the history stored in the storage unit when all the object position estimates included in the first track are not included in the detection area corresponding to the first track. .
Generating a track including a covariance of a position estimation value and an error corresponding to an obstacle identified by the target tracking unit based on a detection signal output by the remote sensor sensing and outputting the position of the object;
Calculating and updating a detection area that is a range in which the target tracking unit detects an obstacle with respect to the track;
The target tracking unit selecting a detection signal in which the position of the object indicated by the detection signal is included in the detection area among the detection signals; And
And a track management unit updating the position estimation value included in the track based on the selected detection signal and the position estimation value included in the track corresponding to the object including the position of the object in the detection area.
In the step of updating the position estimate included in the track,
When the distance between the object position estimate value included in the first track and the object position estimate value included in the second track is less than a preset reference value, the track management unit displays the first track and the second track based on the history of the track stored in the storage unit. Obstacle tracking method characterized in that the initialization.
The method of claim 6,
The remote sensor is an obstacle tracking method, characterized in that at least one of a lidar sensor or a radar sensor.
◈ Claim 8 was abandoned when payment of the set registration fee was made.◈ The method of claim 6,
In the step of selecting the detection signal,
The target tracking unit determines whether the position of the object indicated by the detection signal is included in the detection area corresponding to the track, based on the error of the measured value and the predicted value of the object position and the covariance of the error. Tracking method.
delete The method of claim 6,
In the step of updating the position estimate included in the track,
The track management unit initializes the track based on the history of the track stored in the storage unit when all the object position estimates included in the track are not included in the detection area corresponding to the track.

Images