KR20210079992A - Multi-sensor fusion method and apparatus - Google Patents

Abstract

Disclosed are a method and a device for fusing a multi-sensor. The method for fusing a multi-sensor comprises: a step (a) of obtaining sensing information from a plurality of sensors mounted on an own vehicle; a step (b) of determining an adjustment point for a target object for association with one target object by using each detection information differently detected by the plurality of sensors; and a step (c) of fusing each sensing information of the plurality of sensors by using the adjustment point.

Description

Multi-sensor fusion method and apparatus

The present invention relates to a multi-sensor fusion method and apparatus.

Object tracking refers to a technique for tracking the movement path of an object based on data measured from a sensor. Such object tracking is applied to a vehicle and is widely used in an active safety system such as an automatic emergency braking system (AEBS).

However, conventional techniques do not suggest a solution to a problem that occurs when different points are recognized (sensed) according to sensors mounted at different locations.

(01) Republic of Korea Patent Publication No. 10-2018-0079880 (2018.07.11.)

The present invention is to provide a multi-sensor fusion method and apparatus.

Another object of the present invention is to provide a multi-sensor fusion method and apparatus capable of measuring a desired point of an object using information measured by multiple sensors mounted at different locations.

In addition, the present invention is to provide a multi-sensor fusion method and apparatus capable of sensor fusion in consideration of the accuracy of the sensor.

Another object of the present invention is to provide a multi-sensor fusion method and apparatus capable of measuring a desired point of an object through multiple sensors in a partial area affecting the own vehicle.

According to one aspect of the present invention, a multi-sensor fusion method may be provided.

According to an embodiment of the present invention, the method comprising: (a) acquiring detection information from a plurality of sensors mounted on the own vehicle; (b) determining an adjustment point for the target object for association with one target object by using the respective detection information detected differently by the plurality of sensors; and (c) fusing respective sensing information of a plurality of sensors using the adjustment point. A multi-sensor fusion method may be provided.

Before step (b), deriving an error characteristic value using a sensor accuracy model; and calculating a combined state and covariance by applying a Kalman filter using the error characteristic value, wherein the adjustment point may be calculated using the combined state and covariance.

The deriving of the error characteristic value may include: configuring a plurality of occupied areas in consideration of an area detectable by the plurality of sensors (FOV); determining an interpolated sensor and GPS position on the surface of the target object using a mounting position of the sensor of the own vehicle, a reference point of the target object, and a GPS mounting position of the target object; calculating an ordinate error and an abscissa error using the interpolated sensor and GPS position; and deriving an error characteristic value such that the average value of the vertical error and the horizontal error in each of the occupied areas becomes zero (0).

The calculating of the combined state and covariance may include calculating the combined state and covariance by selecting at least four occupancy areas affecting the own vehicle among the plurality of occupancy areas.

In the step (b), the Mahalanobis distance may be calculated using the combined state and covariance, and the adjustment point may be calculated using the distance at which the calculated Mahalanobis distance is the minimum.

The step (c) may include calculating a distance between at least two interpolated sensor positions and an adjustment point on the target object surface; and

It may include fusing the calculated distance with the sensing information.

According to another aspect of the present invention, a multi-sensor fusion device may be provided.

According to an embodiment of the present invention, an acquisition unit for acquiring detection information from a plurality of sensors mounted on the own vehicle, respectively; an adjustment unit configured to determine an adjustment point for the target object in consideration of association to one target object by using each detection information by the plurality of sensors; And a multi-sensor fusion device including a fusion unit that fuses respective sensing information of a plurality of sensors using the adjustment point may be provided.

The adjustment unit derives an error characteristic value using the sensor accuracy model, and calculates a combined state and covariance by applying a Kalman filter using the error characteristic value, wherein the adjustment point uses the combined state and covariance can be calculated by

The adjusting unit configures a plurality of occupancy zones in consideration of an area detectable by the plurality of sensors (FOV), and determines a mounting position of a sensor of the own vehicle, a reference point of the target object, and a GPS mounting position of the target object. determine the interpolated sensor and GPS position on the surface of the target object using the An error characteristic value can be derived so that the average value of is zero (0).

The adjustment unit may calculate the combined state and covariance by selecting at least four occupancy zones affecting the own vehicle from among the plurality of occupancy zones.

The adjustment unit may calculate a Mahalanobis distance using the combined state and covariance, and calculate the adjustment point using a distance at which the calculated Mahalanobis distance is a minimum.

The fusion unit may calculate a distance between at least two interpolated sensor positions and an adjustment point on the surface of the target object, and fuse the calculated distance with the sensing information.

By providing a multi-sensor fusion method and apparatus according to an embodiment of the present invention, a desired point of an object can be measured using information measured by multiple sensors mounted at different locations.

In addition, in the present invention, sensor fusion is possible in consideration of the accuracy of the sensor.

In addition, the present invention can measure a desired point of an object through multiple sensors in a partial area that affects the own vehicle.

1 is a flowchart illustrating a multi-sensor fusion method according to an embodiment of the present invention.
2 is a diagram illustrating a sensor mounting position according to an embodiment of the present invention.
3 is a view for explaining an occupied area according to an embodiment of the present invention.
4 is a diagram illustrating a sensor accuracy model according to an embodiment of the present invention.
5 is a view for explaining a selection area according to an embodiment of the present invention.
6 is a diagram illustrating a filter process according to an embodiment of the present invention.
7 is a diagram illustrating an adjustment point according to an embodiment of the present invention.
8 is a block diagram schematically illustrating an internal configuration of a sensor fusion device according to an embodiment of the present invention.

As used herein, the singular expression includes the plural expression unless the context clearly dictates otherwise. In this specification, terms such as “consisting of” or “comprising” should not be construed as necessarily including all of the various components or various steps described in the specification, some of which components or some steps are It should be construed that it may not include, or may further include additional components or steps. In addition, terms such as "...unit" and "module" described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software, or a combination of hardware and software. .

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a multi-sensor fusion method according to an embodiment of the present invention, FIG. 2 is a diagram illustrating a sensor mounting position according to an embodiment of the present invention, and FIG. 3 is an embodiment of the present invention FIG. 4 is a diagram illustrating a sensor accuracy model according to an embodiment of the present invention, and FIG. 5 is a diagram illustrating a selection zone according to an embodiment of the present invention. 6 is a diagram illustrating a filter process according to an embodiment of the present invention, and FIG. 7 is a diagram illustrating an adjustment point according to an embodiment of the present invention.

In step 110 , the sensor fusion device 100 obtains detection information for detecting an object from a plurality of sensors respectively attached to the own vehicle.

According to the position and angle of each sensor attached to the own vehicle, even if each sensor detects the same object, it may be recognized as a different object. Hereinafter, for convenience of understanding and explanation, it is assumed that each sensor attached to the own vehicle is a radar and will be mainly described. 2 is a diagram illustrating a position of a radar sensor attached to a vehicle. For example, as shown in FIG. 2 , it is assumed that four radar sensors are installed, such as the front left side of the vehicle, the front right side of the vehicle, the rear left side of the vehicle, and the rear right side of the vehicle. The positions of the radar sensors shown in FIG. 2 are only shown for convenience of understanding and description, and it is natural that the positions and the number of radars may be different.

In step 115, the sensor fusion device 100 derives an error characteristic value using the sensor accuracy model.

According to an embodiment of the present invention, the sensor fusion device 100 may configure an occupied area in consideration of a detectable area (FOV) of a radar as shown in FIG. 3 .

In FIG. 3 , {X,Y} denotes a global coordinate system, and {x,y} denotes a coordinate system of the own vehicle. Based on the coordinate system of the vehicle, the occupancy area may be partitioned on the x-axis by a multiple of the length (total length) of the own vehicle, and the y-axis may be partitioned on the basis of the vehicle's coordinate system. An example of the occupied area in FIG. 3 is only an example, and it is of course that the partitioning method of the occupied area on the x-axis and the y-axis may vary.

As shown in FIG. 4 , the divided occupied area sensed by the radar sensor may analyze the radar accuracy by comparing the radar sensor data with the GT.

This will be explained first.

According to an embodiment of the present invention, the sensor fusion device 100 uses a first straight line connecting the reference point of the radar and the virtual polygon box (target vehicle detection area) and a second straight line connecting the radar and the DGPS mounting point. Find interpolated radar and DGPS positions (points) on the target vehicle surface.

Next, the sensor fusion apparatus 100 may calculate a vertical-axis radar error and a horizontal-axis radar error using the interpolated radar and DGPS positions (points), respectively.

In more detail, as follows.

) 및 출력(

) 변수를 포함하는 레이더 측정 모델은 수학식 1과 같이 정의될 수 있다. At the moment of measurement (k), the state (

) and output (

) The radar measurement model including the variable may be defined as in Equation 1.

C는 항등 행렬(identity matrix)이며,

는 종방향 상대 거리를 나타내고,

는 횡방향 상대 거리를 나타내며,

는 종방향 상대 속도를 나타내고,

는 횡방향 상대 속도를 나타내며, m은 횡방향 상대 위치 구역 인덱스를 나타내고, n은 종방향 상대 위치 구역 인덱스를 나타내고, M은 Y축의 구역 개수를 나타내고, N은 X축의 구역 개수를 나타낸다. here,

C is the identity matrix,

represents the longitudinal relative distance,

represents the lateral relative distance,

represents the longitudinal relative velocity,

denotes the lateral relative velocity, m denotes the transverse relative position zone index, n denotes the longitudinal relative position zone index, M denotes the number of zones on the Y-axis, and N denotes the number of zones on the X-axis.

)가 각 구역에서 제로 평균 백색 가우시안 분포 특성을 가지는 것을 가정하기로 한다. In an embodiment of the present invention, radar measurement accuracy (

) is assumed to have a zero-mean white Gaussian distribution characteristic in each region.

)은 에러 특성에 기초하여 설정될 수 있다. 따라서, 위치 오차의 평균값이 "0"이 되도록 각 레이더는 각 구역에서 캘리브레이션될 수 있다. 그러므로, 제로 평균 백색 가우시안 분포 특성을 가지는 레이더 에러는 수학식 2와 같이 나타낼 수 있다. The radar measurement accuracy covariance (

) can be set based on the error characteristics. Accordingly, each radar may be calibrated in each zone so that the average value of the position error becomes “0”. Therefore, a radar error having a zero-mean white Gaussian distribution characteristic can be expressed as Equation (2).

Also, the covariance can be expressed as Equation (3).

Here, GT denotes ground data and RADAR denotes calibrated radar data.

In step 120, the sensor fusion device 100 calculates the combined state and covariance by applying the Kalman filter using the error characteristic value. Through this, the sensor fusion device 100 according to an embodiment of the present invention can solve the problem of ambiguity with respect to each divided occupied area.

As already described above, in an embodiment of the present invention, an area detectable by a radar sensor is configured as a plurality of occupied regions, and each occupied region may be defined by a multi-model. The multi-model described below should be understood as each divided occupancy area.

In an embodiment of the present invention, the problem of ambiguity between each occupied area can be solved through a filter using an error characteristic value.

First, the sensor fusion device 100 according to an embodiment of the present invention may select four zones that affect the own vehicle among a plurality of divided occupied zones as shown in FIG. 5 .

Basically, 4 zones that affect your vehicle can be selected. However, as shown in FIG. 5 , when an unknown area is included, four areas closest to the car vehicle may be selected among the occupied areas except for the unknown area.

Assuming that the vehicle moves at a constant relative speed in the longitudinal and lateral directions, respectively, the model for following the target vehicle can be expressed as a discrete-time state-space model. If this is expressed as an equation, it can be expressed as in equation (4).

는 시스템 잡음을 나타낸다. 전술한 바와 같이, 센서 정확도 모델은 공분산(

)이 사용되는 KF 필터를 통해 시스템 잡음 공분산(

)을 조정함으로써 각 구역에서 최소값에 근접한 추정 에러를 획득할 수 있다. here,

is the system noise. As mentioned above, the sensor accuracy model is based on the covariance (

) with the system noise covariance (

), it is possible to obtain an estimation error close to the minimum value in each zone.

)을 각각 설정할 수 있다. Therefore, the sensor fusion device 100 calculates the system noise covariance (

) can be set individually.

The multi-model filter is operated as shown in FIG. 6 .

In an embodiment of the present invention, the sensor fusion apparatus 100 may perform mixing, filtering, and merging processes as shown in FIG. 6 corresponding to four selected areas for a target vehicle.

For convenience of understanding and explanation, this will be described in more detail.

In step 610, the sensor fusion apparatus 100 mixes the predicted state and the covariance at the previous sample time using a mixing probability.

The mixing probability can be expressed as Equation (6).

는 k시간에서 모델 i의 모델 확률을 나타내고,

이며, 모델 인덱스는

이다. here,

represents the model probability of model i at k time,

and the model index is

to be.

는 정규화 인자를 나타내고,

는 모델 개수를 나타내며,

는 타겟 차량이 점유한 점유 구역의 인덱스를 나타내고,

는 모델 i에서 모델 j로 전이할 Markov 확률을 나타내고, 상태 전이 확률 행렬(

)는 수학식 7과 같이 나타낼 수 있다. Also,

represents the normalization factor,

represents the number of models,

represents the index of the occupied area occupied by the target vehicle,

represents the Markov probability of transition from model i to model j, and the state transition probability matrix (

) can be expressed as in Equation 7.

)와 믹스된 초기 공분산(

)은 수학식 8과 같이 나타낼 수 있다. The mixed initial state of model j (

) and mixed initial covariance (

) can be expressed as in Equation 8.

는 이전 샘플 시간(k-1)에서 모델 j의 예측된 상태를 나타내고,

는 이전 샘플 시간(k-1)에서 모델 j의 예측된 공분산을 나타낸다. here,

represents the predicted state of model j at the previous sample time (k-1),

denotes the predicted covariance of model j at the previous sample time (k-1).

In step 615, the sensor fusion device 100 estimates the current state and covariance based on the Kalman filter on the predicted state and covariance of model j at the previous sample time.

The prediction and correction process can be expressed as Equations (9) and (10).

는 j 조건부 필터의 공분산을 나타내고, I는 항등 행렬을 나타낸다. 최적의 IMM-KF(칼만 필터) 게인은 모델링된 레이더 정확도 특성(에러 특성값)을 사용하여 각 점유 구역에 맞게 설계될 수 있다. here,

denotes the covariance of the j conditional filter, and I denotes the identity matrix. An optimal IMM-KF (Kalman filter) gain can be designed for each occupied zone using the modeled radar accuracy characteristic (error characteristic value).

In multiphase 620, the sensor fusion device 100 outputs the combined state and covariance by merging the estimated current state and covariance.

)는 가우시안 가정 하에 수학식 11과 같이 나타낼 수 있다. The likelihood function of each model (j) at time k (

) can be expressed as in Equation 11 under the Gaussian assumption.

The model probability of each model can be calculated as in Equation 12.

이다. here,

to be.

)와 공분산(

)은 모델 j로부터의 상태 추정치의 차이를 설명하기 위해 공분산 행렬에 대한 증분과 함께 표준 가우시안 혼합 평균 및 공분산 공식을 사용하여 수학식 13과 같이 계산될 수 있다. combined state (

) and covariance (

) can be calculated as in Equation 13 using the standard Gaussian mixed mean and covariance formula with increments to the covariance matrix to account for the difference in the state estimate from model j.

Referring back to FIG. 1 , in step 125 , the sensor fusion device 100 calculates an adjustment point using the calculated combined state and covariance.

This will be described in more detail. 7 is a diagram illustrating a method of fusing different surface points of the same target object by a plurality of radar sensors according to an embodiment of the present invention. As shown in FIG. 7 , when different surface positions of the same target object are calculated by a plurality of sensors, the connection of the target objects must be carefully performed so that the same target object is not recognized as different targets.

In a multi-radar network environment in which each radar sensor performs its own measurement and maintains a detection area, association is a method of determining whether another sensor represents the same target object in an overlapping area.

For correlation, in an embodiment of the present invention, a sensor index corresponding to a minimum distance and a target object index may be found through comparison of the sensed target object.

In order to solve the association problem for the overlapping area of each radar, Equation 13 is used in an embodiment of the present invention.

)와 공분산(

)은 k 시간에 다른 센서(

)를 사용하여 센서 예측치(

)와 공분산 행렬(

)로 나타낼 수 있다. For simplicity, the combined state calculated in Equation 13 (

) and covariance (

) is the other sensor (

) using the sensor prediction (

) and the covariance matrix (

) can be expressed as

를 포함하는 Mahalanobis 거리는 수학식 14와 같이 계산될 수 있다. To find an adjacent target object location, an embodiment of the present invention uses the Mahalanobis distance method.

The Mahalanobis distance including ? can be calculated as in Equation (14).

는 일반적인 유클리드 거리와 달리 레이더 특성을 고려하기 위해 사용될 수 있다. here,

can be used to consider radar characteristics, unlike the general Euclidean distance.

This is an associative structure in which multiple sensors and each radar sensor detect multiple target objects.

In an embodiment of the present invention, a sensed target object index associated with an associated sensor index having a minimum distance through MOMS minimization may be found. This can be expressed as Equation (15).

는 센싱된 레이더 셋을 나타낸다. 따라서, 연관된 타겟 객체의 포즈는

,

이다. here,

denotes the sensed radar set. Thus, the pose of the associated target object is

,

to be.

In an autonomous driving environment, two radar sensors constitute an overlapping area. Therefore, we solved the association and fusion problems for the overlapping regions composed of two different sensors.

,

)를 이용하여 조정점을 획득한다. The sensor fusion device 100 performs a pose (

,

) to obtain the adjustment point.

,

)는 동일한 객체에 대한 값이나 동일한 위치(포인트)는 아니다. 이는 복수의 레이더가 동일한 타겟 객체에 대해 다른 점을 계산하는 비-일치 지점의 문제이다. The pose of the associated target object (

,

) are values for the same object, but not the same location (point). This is a problem of non-matching points where multiple radars calculate different points for the same target object.

Accordingly, in an embodiment of the present invention, as shown in FIG. 7 , it is possible to obtain an adjustment point affecting the reference point of the own vehicle.

In an ideal situation, several points on the surface of the target object can be adjusted as surface points that influence the reference point of the own vehicle.

)는 수학식 16과 같이 나타낼 수 있다. The adjusted pose of the target vehicle surface (

) can be expressed as in Equation 16.

와

는 정규화된 조정 가중치 행렬을 나타낸다.here,

Wow

denotes a normalized adjustment weight matrix.

및

과 같이 주어진 경우,

와

는 도 7에 도시된 바와 같이 쉽게 획득될 수 있다. For example, geometric information

and

If given as

Wow

can be easily obtained as shown in FIG.

)와 자 차량의 DGPS 장착 지점 위치(

)를 알고 있으므로, 도 7과 같이 기하학적 증명(즉, 두 위치에 의해 정의된 선)에 의해 타겟 객체 표면의 조정된 조정점(

)를 계산할 수 있다. That is, the sensor fusion device 100 has two integrated radar positions (

) and the location of the DGPS mounting point on your vehicle (

), as shown in Fig. 7, the adjusted control point of the target object surface by geometric proof (i.e., a line defined by two positions) (

) can be calculated.

)와 조정점(

) 사이의 거리(

)가 계산될 수 있다. 계산된 거리(

)는 조정점에 대한 보간된 레이더 위치의 가중치 행렬로 표현될 수 있다. 이를 수학식으로 나타내면 수학식 17과 같이 나타낼 수 있다. That is, as shown in FIG. 7, the interpolated rate position (

) and the control point (

) the distance between (

) can be calculated. Calculated distance (

) can be expressed as a weighting matrix of interpolated radar positions with respect to the control point. If this is expressed as an equation, it can be expressed as Equation 17.

,

가 즉각적으로 계산될 수 있다.

,

can be calculated immediately.

The normalized adjusted weight matrix can be expressed as Equation (18).

와

은 보간된 레이더 위치(

,

)와 조정점(

)의 거리에 관한 비율로 계산될 수 있다. Normalized Adjusted Weight Matrix

Wow

is the interpolated radar position (

,

) and the control point (

) can be calculated as a ratio with respect to the distance.

) 및 종축/횡축 상대 속도(

)와 관련될 수 있다. The computed normalized adjusted weight matrix is the ordinate and abscissa relative distances (

) and longitudinal/horizontal relative speed (

) may be related to

In step 130, the sensor fusion device 110 fuses the sensing information for the sensing area overlapped by the plurality of sensors using the adjustment point.

)를 융합하기 위해 이전 획득된 정규화된 조정된 가중치 행렬(

)과 공분산 행렬(

)을 이용할 수 있다. The sensor fusion device 100 is associated with the target object pose (

), the previously obtained normalized adjusted weight matrix (

) and the covariance matrix (

) can be used.

Sensor fusion errors may increase if the sensor calculates different surface points of the same target object (eg, if the target object is larger than the sensing area or very close to the sensor). This is because the sensor calculates different positions for the same target object. Since overlapping regions (regions) detected by different sensors are independent, a state (coordinates) estimated by the fusion algorithm can be obtained as shown in Equation (19).

The predicted adjusted covariance may be calculated as in Equation (20).

8 is a block diagram schematically illustrating an internal configuration of a sensor fusion device according to an embodiment of the present invention.

Referring to FIG. 8 , the sensor fusion device 100 according to an embodiment of the present invention includes a plurality of sensors 810 , an acquisition unit 815 , an adjustment unit 820 , a fusion unit 825 , a memory 830 and and a processor 835 .

The plurality of sensors 810 may be attached to respective positions of the vehicle, and may detect respective sensing areas. Regions sensed by the plurality of sensors 810 may overlap.

The acquisition unit 815 is a means for acquiring detection information from a plurality of sensors mounted on the own vehicle, respectively.

The adjustment unit 820 is a means for determining an adjustment point for a target object in consideration of association with one target object using respective detection information by a plurality of sensors.

The fusion unit 825 is a means for fusing respective sensing information of a plurality of sensors using an adjustment point.

Detailed functions of the adjustment unit 820 and the fusion unit 825 are the same as those described with reference to FIGS. 1 to 7 , and thus overlapping descriptions will be omitted.

The memory 830 is a means for storing program codes (instructions) necessary for performing the sensor fusion method according to an embodiment of the present invention.

The processor 835 is a means for controlling internal components (eg, the sensor 810 , the memory 830 , etc.) of the sensor fusion device 100 according to an embodiment of the present invention.

In addition, the instructions executed by the processor 835 may perform the sensor fusion method as described with reference to FIGS. 1 to 7 . Since this is the same as that described above, the overlapping description will be omitted.

The apparatus and method according to an embodiment of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the computer readable medium may be specially designed and configured for the present invention or may be known and available to those skilled in the computer software field. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floppy disks. - Includes magneto-optical media and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

So far, the present invention has been looked at focusing on the embodiments thereof. Those of ordinary skill in the art to which the present invention pertains will understand that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is indicated in the claims rather than the foregoing description, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

100: multi-sensor fusion device
810: sensor
815: Acquisition Department
820: adjustment unit
825: fusion unit
830: memory
835: processor

Claims (14)

(a) acquiring detection information from a plurality of sensors mounted on the own vehicle, respectively;
(b) determining an adjustment point for the target object for association with one target object by using the respective detection information detected differently by the plurality of sensors; and
(c) fusing each sensing information of a plurality of sensors using the adjustment point.
According to claim 1,
Before step (b),
deriving an error characteristic value using the sensor accuracy model; and
Further comprising the step of calculating a combined state and covariance by applying a Kalman filter using the error characteristic value,
and the adjustment point is calculated using the combined state and covariance.
3. The method of claim 2,
The step of deriving the error characteristic value is,
configuring a plurality of occupied zones in consideration of an area detectable by the plurality of sensors (FOV);
determining an interpolated sensor and GPS position on the surface of the target object using a mounting position of the sensor of the own vehicle, a reference point of the target object, and a GPS mounting position of the target object;
calculating an ordinate error and an abscissa error using the interpolated sensor and GPS position; and
and deriving an error characteristic value such that the average value of the vertical error and the horizontal error in each of the occupied zones becomes zero (0).
4. The method of claim 3,
Calculating the combined state and covariance comprises:
and calculating the combined state and covariance by selecting at least four occupancy zones affecting the own vehicle among the plurality of occupancy zones.
According to claim 1,
Step (b) is,
and calculating a Mahalanobis distance using the combined state and covariance, and calculating the adjustment point using a distance at which the calculated Mahalanobis distance is a minimum.
According to claim 1,
The step (c) is,
calculating a distance between at least two interpolated sensor positions and an adjustment point on the target object surface;
A multi-sensor fusion method comprising the step of fusing the calculated distance with the sensing information.
A computer-readable recording medium product on which a program code for performing the method according to any one of claims 1 to 6 is recorded.
an acquisition unit configured to acquire sensing information from a plurality of sensors mounted on the own vehicle;
an adjustment unit configured to determine an adjustment point for the target object for association with one target object by using each detection information differently detected by the plurality of sensors; and
A multi-sensor fusion device including a fusion unit that fuses respective sensing information of a plurality of sensors using the adjustment point.
9. The method of claim 8,
The adjustment unit,
The error characteristic value is derived using the sensor accuracy model, and the combined state and covariance are calculated by applying the Kalman filter using the error characteristic value,
and the adjustment point is calculated using the combined state and covariance.
10. The method of claim 9,
The adjustment unit,
A plurality of occupancy zones are configured in consideration of an area detectable by the plurality of sensors (FOV), and the target using the mounting position of the sensor of the own vehicle, the reference point of the target object, and the GPS mounting position of the target object determine the interpolated sensor and GPS position on the surface of the object, calculate the ordinate error and the abscissa error using the interpolated sensor and GPS position, and the average value of the ordinate error and the abscissa error in each of the occupied areas is zero A multi-sensor fusion device, characterized in that the error characteristic value is derived to be (0).
11. The method of claim 10,
The adjustment unit,
and calculating the combined state and covariance by selecting at least four occupancy zones affecting the own vehicle among the plurality of occupancy zones.
9. The method of claim 8,
The adjustment unit,
and calculating a Mahalanobis distance using the combined state and covariance, and calculating the adjustment point using a distance at which the calculated Mahalanobis distance is a minimum.
9. The method of claim 8,
The fusion part,
A multi-sensor fusion device, comprising: calculating a distance between at least two interpolated sensor positions and an adjustment point on a target object surface, and fusing the calculated distance with the sensing information.
a plurality of sensors mounted on the own vehicle; and
After determining an adjustment point for the target object in consideration of association to one target object using each detection information obtained by the plurality of sensors, the detection information of each of the plurality of sensors is obtained using the adjustment point A vehicle comprising a fusion multi-sensor fusion unit.

Images