JP2017096840A - Vehicular target detection apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a vehicular target detection apparatus capable of determining a ghost generated by the reflection of another vehicular side surface, and determining the ghost without setting a virtual wall.SOLUTION: A vehicular target detection apparatus includes: a radar sensor; a target detection part for clustering plural detected points into clusters and correlating a cluster with a target; a detected point categorization part for categorizing, as a ghost candidate, a target corresponding to a detected point of an echo number indicating a second-in-order arrival sequence, or thereafter, and categorizing, as a detected point of a reflection surface, a detected point of an echo number indicating that has arrived earlier than a detected point of a cluster corresponding to the target of the ghost candidate; a reflection surface estimation part for estimating a position of the reflection surface and an orientation thereof; and a ghost determination part for determining that a target of a ghost candidate is a ghost, in a case where there is a target with the same speed as a speed of a target of a ghost candidate exists at a position symmetrical with the target of the ghost candidate with the reflection surface defined as reference.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a vehicle target detection apparatus.

  Patent Document 1 describes a vehicle target detection device. This device is a radar device that outputs a radio wave such as a millimeter wave, receives a reflected wave, and detects a target, and is actually detected by a reflected wave from a target and a tunnel wall. Distinguish from untargeted targets (ghosts). Specifically, this device acquires the reflection characteristic of the side wall of the vehicle, and sets the virtual wall to the side of the detected target when the acquired reflection characteristic of the wall is equal to or greater than a predetermined threshold. To do. When the target is set at a position symmetrical to the virtual wall with respect to the target set to the side of the virtual wall, the speed of the target at the position symmetrical to the virtual wall is predetermined. It is determined whether or not the target is in the range, and when the speed of the target at a position symmetrical to the virtual wall is within the predetermined range, the target at the position symmetrical to the virtual wall is determined to be a ghost.

JP 2009-133761 A

  Since the vehicle target detection device described in Patent Document 1 uses the reflection characteristics of the side wall of the vehicle, there is another vehicle in front of the side of the vehicle, and radio waves are reflected by the other side of the vehicle. It is difficult to distinguish the ghost that occurs in the target from the actual target. Furthermore, in the vehicle target detection device described in Patent Document 1, in order to improve the accuracy of ghost determination, it is necessary to suppress the deviation between the virtual wall set on the side of the target and the actual wall.

  An object of the present invention is to provide a vehicle target detection apparatus that can determine a ghost generated by reflection on the side of another vehicle and determine a ghost without setting a virtual wall.

  A vehicle target detection device according to the present invention is a vehicle target detection device that detects a target, and transmits the electromagnetic wave around the vehicle and receives the reflected wave, thereby detecting the position where the electromagnetic wave is reflected. A plurality of radar sensors acquired as detection points, and an object for clustering a plurality of detection points into one or a plurality of clusters based on the positions of the plurality of detection points acquired by the radar sensor and associating the clusters with targets A target detection unit, a detection point acquired by a radar sensor, and an echo number indicating the arrival order of the reflected wave are associated with each transmission, and a cluster including a detection point associated with an echo number indicating the second and subsequent arrival orders is associated. The corresponding target is classified as a ghost candidate, and an echo number indicating that it has arrived earlier than the detection point belonging to the cluster corresponding to the ghost candidate target is associated. A detection point classification unit that classifies the detected detection points as detection points corresponding to the reflection surface that generates the target of the ghost candidate, and a detection point corresponding to the reflection surface classified by the detection point classification unit The reflection surface estimation unit for estimating the position and orientation of the reflection surface, and the reflection surface estimated by the reflection surface estimation unit with the same speed as the speed of the target ghost candidate classified by the detection point classification unit. And a ghost determination unit that determines that the target of the ghost candidate is a ghost when the target exists at a position symmetrical to the target of the ghost candidate as a reference.

  According to the present invention, it is possible to determine a ghost generated by the reflection of the other vehicle side surface, and it is possible to determine a ghost without setting a virtual wall.

It is a block diagram explaining the structure of a vehicle provided with the vehicle target detection apparatus which concerns on embodiment. It is a figure explaining operation | movement of a target detection part. It is a table | surface explaining the correspondence of the detection point and echo number regarding the transmission which detected the same ghost candidate. It is a figure explaining operation | movement of a reflective surface estimation part. It is a figure explaining operation | movement of a ghost determination part. It is a flowchart of the target detection process of the target detection apparatus for vehicles shown in FIG.

  Embodiments of the present invention will be described below with reference to the drawings. In the following description, the same or equivalent elements will be denoted by the same reference numerals, and redundant description will be omitted.

  FIG. 1 is a block diagram illustrating a configuration of a vehicle 2 including a vehicle target detection device 1 according to an embodiment. As shown in FIG. 1, the vehicle target detection apparatus 1 is mounted on a vehicle 2 such as a passenger car. The vehicle target detection device 1 detects a target using electromagnetic waves as described later.

  The vehicle 2 includes a radar sensor 10 and an ECU (Electronic Control Unit) 20. The radar sensor 10 and the ECU 20 are respectively connected to a network that communicates using a CAN (Controller Area Network) communication circuit, and can perform mutual communication.

  The radar sensor 10 is a detector that detects a target around the vehicle 2, and more specifically, a detector that detects a target in the traveling direction of the vehicle 2. A target is a tangible object and is a stationary object on a road such as a wall or a moving object such as a preceding vehicle. The radar sensor 10 includes an electromagnetic wave transmission circuit, a reception circuit, a calculation circuit that performs various calculations, and a storage unit. The electromagnetic wave is, for example, a laser beam or a millimeter wave. The radar sensor 10 transmits an electromagnetic wave over a range of set detection angles around the vehicle 2 and receives a reflected wave reflected by the target. The radar sensor 10 detects the position of the target based on the detection angle (reception angle) of the reflected wave and the detection distance obtained from the propagation time of the electromagnetic wave required from transmission to reception. The detection distance may be obtained from the intensity of the reflected wave. For each transmission, the radar sensor 10 acquires a plurality of positions where the electromagnetic waves are reflected as detection points. That is, the detection point is an observation point at a position where the electromagnetic wave is reflected. The radar sensor 10 outputs the detection result to the ECU 20.

  The ECU 20 includes a target detection unit 21, a detection point classification unit 22, a reflection surface estimation unit 23, a ghost determination unit 24, and a target tracking unit 25.

  The target detection unit 21 classifies the plurality of detection points acquired by the radar sensor 10 for each target. The target detection unit 21 clusters the plurality of detection points into one or a plurality of clusters based on the positions of the plurality of detection points acquired by the radar sensor 10. A cluster is a collective term for a plurality of detection points, and is a group. Clustering is to group a plurality of detection points. The target detection unit 21 estimates a detection point group based on reflected waves reflected from the same target, and sets the estimated detection point group as one cluster. Specifically, the target detection unit 21 calculates distances between all detection points in the detection point group based on the positions of the plurality of detection points, and performs clustering based on the obtained distances. The target detection unit 21 clusters detection points having the shortest distance among all detection points. After that, the target detection unit 21 clusters the one having the smallest distance among all combinations of the remaining detection points and the center of gravity of the cluster. The target detection unit 21 repeatedly performs the above-described processing until immediately before the distance between the clusters exceeds a predetermined threshold. The target detection unit 21 sets each cluster after the processing as a detection point group corresponding to one target.

  Next, the target detection unit 21 associates the cluster with the target tracked up to the previous time. As an example, the target detection unit 21 performs the Munkres algorithm using the spatial distance or the Mahalanobis distance as a reference. FIG. 2 is a diagram for explaining the operation of the target detection unit 21. FIG. 2 illustrates an example in which a series of electromagnetic waves are transmitted forward from the vehicle 2 in a situation where the other vehicles A and B that have been tracked up to the previous time are present in front of the vehicle 2. A line extending radially from the vehicle 2 indicates a transmission direction. The target detection unit 21 detects and clusters a plurality of detection points P by receiving the reflected electromagnetic waves. Here, the first cluster GA, the second cluster GB, and the third cluster GC exist by clustering. The target detection unit 21 associates the first cluster GA with the other vehicle A based on the previous position of the other vehicle A and the positions of the first cluster GA, the second cluster GB, and the third cluster GC. Similarly, the target detection unit 21 associates the second cluster GB with the other vehicle B. The target detection unit 21 regards a cluster that cannot be associated with the previous target as a new target to be tracked, and associates the cluster with the new target to be tracked. In the example illustrated in FIG. 2, the third cluster GC is associated with a new target to be tracked. The association means that when a cluster is specified, a target (other vehicle A, another vehicle B, new, etc.) can be specified. For example, the identifiers of both are associated in a table format. Say.

  The detection point classification unit 22 determines whether or not each target detected by the target detection unit 21 is a ghost candidate. If there is a ghost candidate in the target, a ghost is generated for each target. The detection points corresponding to the reflecting surface as a factor are classified from a plurality of detection points. In an environment in which other vehicles such as walls and preceding vehicles exist around the vehicle 2, electromagnetic waves transmitted from the radar sensor 10 may be reflected several times and returned to the radar sensor 10 (occurrence of multipath). That is, a plurality of reflected waves having different reception timings may be obtained by one transmission. A ghost refers to a target detected in a location that does not actually exist due to the occurrence of multipath. A ghost candidate is a target that may be a ghost. A reflective surface is a surface that reflects electromagnetic waves. As described above, since a plurality of reflected waves having different reception timings may be obtained by one transmission, an echo number for identifying the arrival order of the reflected waves is assigned to each reflected wave. The echo number is a number indicating the arrival order of reflected waves. When a plurality of reflected waves having different reception timings are obtained for one transmission, the order of arrival is “0”, “1”, “2”, “3”. An echo number is assigned. The echo number may be assigned by the radar sensor 10 or the ECU 20.

  First, the detection point classification unit 22 reads the echo numbers for all corresponding detection points for each target detected by the target detection unit 21 and detects the second and subsequent echo numbers indicating the arrival order. A target corresponding to a cluster including a point is classified as a ghost candidate. The echo number indicating the order of arrival after the second indicates that the reflected wave is other than the earliest arrival, and the echo numbers are assigned in the order of “0”, “1”, “2”. If it exists, it means that the echo number is 1 or more. As a result, when all the echo numbers of the detection points belonging to the cluster indicate that they arrived earliest, the target corresponding to the cluster is classified as not a ghost candidate. On the other hand, if the detection points belonging to the cluster include detection points with echo numbers indicating the second and subsequent arrival orders, the target corresponding to the cluster is classified as a ghost candidate.

  Next, the detection point classification unit 22 identifies a transmission that detects a detection point belonging to the cluster for a cluster corresponding to a target classified as a ghost candidate, and includes all detection points detected by the identified transmission. Create a table showing the relationship with the echo number. FIG. 3 is a table for explaining the correspondence between detection points and echo numbers for transmissions where the same ghost candidate is detected. The beam in the figure means transmission. For example, the detection point classification unit 22 associates the echo number “0” with the position (xa0, ya0) of the detection point regarding the reflected wave that has arrived first with respect to the beam a. The detection point classifying unit 22 associates the echo number “1” with the position (xa1, ya1) of the detection point regarding the reflected wave that has arrived second with respect to the beam a. The detection point classification unit 22 associates the echo number “2” with the position (xa2, ya2) of the detection point regarding the reflected wave that has arrived third for the beam a. Similarly, the detection point classifying unit 22 associates the detection points and the echo numbers with each of the beams b to e. By this operation, the table shown in FIG. 3 is generated.

  Next, in the table shown in FIG. 3, the detection point classification unit 22 is provided with an echo number indicating that it has arrived earlier than the detection point belonging to the cluster corresponding to the target classified as a ghost candidate. Identify points. From detection points closer to the vehicle 2 than detection points belonging to a cluster corresponding to a target classified as a ghost candidate, that is, detection points belonging to a cluster corresponding to a target classified as a ghost candidate in the table of FIG. However, there is a possibility that a detection point with an early echo number is a detection point on the reflecting surface that causes a ghost.

  For example, in the table shown in FIG. 3, when the detection point belonging to the cluster corresponding to the target classified as the ghost candidate in the beam a is the detection point of the echo number “1”, the detection point classification unit 22 identifies the detection point of the echo number “0”. Similarly, when the detection point belonging to the cluster corresponding to the target classified as the ghost candidate in the beam b is the detection point of the echo number “1”, the detection point classification unit 22 sets the echo number “0”. Identify the detection points. Similarly, when the detection point belonging to the cluster corresponding to the target classified as the ghost candidate in the beam c is the detection point of the echo number “2”, the detection point classification unit 22 sets the echo number “0”. And the detection point of "1" is specified. Similarly, if the detection point belonging to the cluster corresponding to the target classified as a ghost candidate in the beam d is the detection point of the echo number “1”, the detection point classification unit 22 uses the echo number “0”. Identify the detection points. Similarly, when the detection point belonging to the cluster corresponding to the target classified as the ghost candidate in the beam e is the detection point of the echo number “2”, the detection point classification unit 22 sets the echo number “0”. And the detection point of "1" is specified.

  The detection point classifying unit 22 identifies a detection point associated with an echo number indicating that it has arrived earlier than the detection point belonging to the cluster corresponding to the target of the ghost candidate identified as described above. The target is classified as a detection point corresponding to the reflecting surface that generates the mark. In the example of FIG. 3, the detection point of the echo number “0” in the beam a, the detection point of the echo number “0” in the beam b, the detection points of the echo numbers “0” and “1” in the beam c, and the echo in the beam d The detection point with the number “0” and the detection points with the echo numbers “0” and “1” in the beam e are classified as detection points corresponding to the reflection surface that generates the target of the ghost candidate.

  The reflection surface estimation unit 23 estimates the position and orientation of the reflection surface based on the position of the detection point corresponding to the reflection surface classified by the detection point classification unit 22. FIG. 4 is a diagram illustrating the operation of the reflection surface estimation unit 23. FIG. 4 shows the positional relationship between the cluster GD corresponding to the target D classified as a ghost candidate and the cluster GE, which is a set of detection points earlier in arrival order. If the target ghost candidate D is a ghost and the reflecting surface R that causes ghost is locally flat, as shown in FIG. Is reflected by the reflecting surface R, which is a flat surface, and arrives at the actual target F, is reflected by the target F, is reflected by the reflecting surface R, and returns to the vehicle 2. That is, the cluster GF in which the detection points of the target F are collected is detected by the reflection surface R as if it exists at the position of the cluster GD. The reflection surface estimation unit 23 estimates the position and orientation of the reflection surface R by linearly approximating the positions of the detection points belonging to the cluster GE. The reflection surface estimation unit 23 uses a least square method or the like as a linear approximation. Note that the reflection surface estimation unit 23 may estimate the reflection surface R from the point group by the RANSAC method or the like based on the position of the detection point belonging to the cluster GE.

  The ghost determination unit 24 determines that the target of the ghost candidate is a ghost when there is a target that is paired with the target of the ghost candidate using the reflection surface R estimated by the reflection surface estimation unit 23 as a reference. To do. More specifically, the ghost determination unit 24 uses the reflection surface R estimated by the reflection surface estimation unit 23 as a reference for the target having the same speed as that of the ghost candidate target classified by the detection point classification unit 22. If it exists at a position symmetrical to the target of the ghost candidate, it is determined that the target of the ghost candidate is a ghost. The ghost determination unit 24 may include the degree of coincidence of the target shapes in the determination. FIG. 5 is a diagram for explaining the operation of the ghost determination unit 24. The ghost determination unit 24 moves the cluster GD corresponding to the target ghost candidate D to a symmetrical position with respect to the reflection surface R, and searches for the target corresponding to the moved cluster GD. Here, as shown in FIG. 5, it is assumed that a target F (cluster GF ′) that is not a ghost candidate is detected at a corresponding position by a different beam. In this case, the ghost determination unit 24 compares the speed of the target D of the ghost candidate with the speed of the target F. The ghost determination unit 24 determines that the target D of the ghost candidate is a ghost of the corresponding target F when the speeds of both are symmetric with respect to the reflection surface R and have the same magnitude. . The ghost determination unit 24 removes the target D determined to be a ghost from the targets detected by the target detection unit 21.

  The target tracking unit 25 estimates the state of the target from the target detected by the ghost determination unit 24 from the target detected by the target detection unit 21. The state of the target is the position, speed, shape, and the like. The target tracking unit 25 sequentially updates the target state estimated based on the past (previous) cluster information based on the latest (current) cluster information. Specifically, the target tracking unit 25 detects the target from which the target determined by the ghost determination unit 24 has been removed from the state of the target tracked by the target tracking unit 25 until the previous time (immediately before). Predict the state of the (latest) target. Then, the target tracking unit 25 corrects the predicted state of the latest target using the cluster information observed this time. A Kalman filter can be used to predict the state of the target. As parameters of the Kalman filter, the position information of the cluster may be an observation value, and the position, speed, acceleration, posture angle, posture angular velocity, and the like of the target may be used as state variables.

  The vehicle target detection apparatus 1 includes the radar sensor 10 and the ECU 20 described above.

  Next, the target detection process of the vehicle target detection apparatus 1 will be described. FIG. 6 is a flowchart of the target detection process of the vehicle target detection apparatus 1. The target detection process shown in FIG. 6 is started when an operation start signal of the vehicle target detection device 1 is acquired.

  As shown in FIG. 6, the radar sensor 10 of the vehicle target detection device 1 transmits electromagnetic waves over a range of set detection angles around the vehicle 2 and reflects the target as data acquisition processing (S10). Receive the reflected wave.

  Next, the target detection unit 21 of the vehicle target detection device 1 determines whether or not a target has been detected as a target detection determination process (S12). The target detection unit 21 determines whether a target is detected based on the detection result of the data acquisition process (S10). When a detection point is not acquired, it determines with the target not being detected, and complete | finishes the target detection process shown in FIG. On the other hand, when the detection points are acquired, the target detection unit 21 classifies the detection points for each target. The target detection unit 21 clusters the plurality of detection points into one or a plurality of clusters based on the positions of the plurality of detection points acquired by the radar sensor 10. Then, the target detection unit 21 associates the cluster with the target tracked up to the previous time.

  Next, the detection point classification | category part 22 of the target detection apparatus 1 for vehicles selects one cluster from the cluster clustered by the target detection determination process (S12) as a cluster selection process (S14).

  Next, as the echo number determination process (S16), the detection point classification unit 22 assigns the detection points (second and subsequent arrival orders) of the echo number “1” or more to the cluster selected in the cluster selection process (S14). It is determined whether or not the detection point of the indicated echo number belongs. If it is determined that detection points with an echo number “1” or higher belong, a ghost candidate setting process (S18) is executed.

  As the ghost candidate setting process (S18), the detection point classification unit 22 sets the target of the cluster selected in the cluster selection process (S14) as a ghost candidate. When the processing of the ghost candidate setting process (S18) is completed, and when it is determined in the echo number determination process (S16) that the detection point of the echo number “1” or higher does not belong, the determination completion process ( S20) is executed.

  The detection point classification unit 22 determines whether or not the echo number determination process (S16) has been executed for all clusters clustered in the target detection determination process (S12) as the determination completion process (S20). If it is determined that the echo number determination process (S16) has not been executed for all clusters, the process returns to the cluster selection process (S14), and the echo number determination process (S16) and the ghost candidate setting process (S18) are executed. Is done. Thus, the cluster selection process (S14) to the determination completion process (S20) are repeatedly executed until it is determined that the echo number determination process (S16) has been executed for all clusters. When it is determined that the echo number determination process (S16) has been executed for all the clusters, the ghost candidate determination process (S22) is executed.

  As the ghost candidate determination process (S22), the detection point classification unit 22 determines whether there is a target for which a ghost candidate is set. When it is determined that there is a target for which a ghost candidate is set, a reflection surface estimation process (S24) is executed.

  First, the detection point classifying unit 22 classifies the detection points corresponding to the reflection surface R, which is a factor causing ghost, for each target from the plurality of detection points as the reflection surface estimation process (S24). The detection point classifying unit 22 selects a detection point associated with an echo number indicating that it has arrived earlier than a detection point belonging to the cluster corresponding to the target of the ghost candidate as a reflection surface R that generates the target of the ghost candidate. Are classified as detection points corresponding to. This process is performed for each ghost candidate target. That is, the detection point of the reflective surface R corresponding to each ghost candidate target is classified. And the reflective surface estimation part 23 of the vehicle target detection apparatus 1 detects as a reflective surface estimation process (S24) corresponding to the reflective surface R classified by the detection point classification | category part 22 for every target of a ghost candidate. Based on the position of the point, the position and orientation of the reflecting surface R are estimated.

  Next, the ghost determination unit 24 of the vehicle target detection apparatus 1 performs the ghost determination process (S26) with an object having the same speed as the speed of the ghost candidate target classified in the ghost candidate determination process (S22). If the target is present at a position symmetrical to the target of the ghost candidate with reference to the reflective surface R estimated in the reflective surface estimation process (S24), it is determined that the target of the ghost candidate is a ghost. To do. The ghost determination unit 24 removes the target D determined to be a ghost from the targets detected by the target detection unit 21.

  Next, the target tracking unit 25 of the vehicle target detection device 1 performs a ghost in the ghost determination process (S26) from the target detected in the target detection determination process (S12) as the tracking process (S28). The target state is estimated for the target from which the target determined to be removed. When the tracking process (S28) ends, the target detection process shown in FIG. 6 ends. When the target detection process is completed, if the signal for ending the operation of the vehicle target detection apparatus 1 is not acquired, the process starts again from the data acquisition process (S10). As described above, the target detection process shown in FIG. 6 is repeatedly executed until an operation end signal of the vehicle target detection device 1 is acquired.

  As described above, in the vehicle target detection device 1 according to the present embodiment, the target detection unit 21 has one or a plurality of clusters based on the positions of the plurality of detection points acquired by the radar sensor 10. The cluster and the target are associated with each other, the detection point classification unit 22 associates the detection point acquired by the radar sensor 10 with the echo number for each transmission, and indicates the second and subsequent arrival orders. The target corresponding to the cluster including the detection point associated with the number is classified as a ghost candidate, and an echo number indicating that it arrived earlier than the detection point belonging to the cluster corresponding to the target of the ghost candidate is associated. The detection point is classified as a detection point corresponding to the reflection surface R that generates the target of the ghost candidate, and the reflection surface estimation unit 23 corresponds to the reflection surface R. The position and orientation of the reflecting surface R are estimated based on the position of the detection point, and the ghost determination unit 24 determines that the target having the same speed as that of the ghost candidate target If the target is symmetric with the target, it is determined that the target of the ghost candidate is a ghost. As described above, according to the vehicle target detection device 1, the reflection surface R can be estimated without determining whether the factor causing the ghost is the wall or the other vehicle. Ghosts generated by reflection can be determined. Furthermore, according to the vehicle target detection apparatus 1, since the ghost is determined using the reflection surface R, the ghost can be determined without setting a virtual wall.

  The present invention can be implemented in various forms based on the above-described embodiments and various modifications and improvements based on the knowledge of those skilled in the art. Moreover, it is also possible to comprise the modification of the following Example using the technical matter described in embodiment mentioned above. You may use combining the structure of each embodiment suitably.

  DESCRIPTION OF SYMBOLS 1 ... Vehicle target detection apparatus, 2 ... Vehicle, 10 ... Radar sensor, 21 ... Target detection part, 22 ... Detection point classification | category part, 23 ... Reflection surface estimation part, 24 ... Ghost determination part, 25 ... Target tracking Department.

Claims (1)

A vehicle target detection device for detecting a target,
A radar sensor that obtains a plurality of positions where the electromagnetic waves are reflected as detection points by transmitting electromagnetic waves around the vehicle and receiving reflected waves; and
A target detection unit that clusters a plurality of detection points into one or a plurality of clusters based on positions of the plurality of detection points acquired by the radar sensor, and associates the clusters with the targets;
The detection point acquired by the radar sensor and the echo number indicating the arrival order of the reflected wave are associated for each transmission, and the target corresponding to the cluster including the detection point associated with the echo number indicating the second and subsequent arrival orders. Are detected as ghost candidates, and a detection point associated with an echo number indicating that it has arrived earlier than a detection point belonging to a cluster corresponding to the target of the ghost candidate is a reflection that generates the target of the ghost candidate. A detection point classification unit for classifying the detection points corresponding to the surface;
A reflection surface estimation unit that estimates the position and orientation of the reflection surface based on the position of the detection point corresponding to the reflection surface classified by the detection point classification unit;
A target having the same speed as the target speed of the ghost candidate classified by the detection point classifying unit is symmetric with the target of the ghost candidate with respect to the reflective surface estimated by the reflective surface estimation unit. If present, a ghost determination unit for determining that the target of the ghost candidate is a ghost;
A vehicle target detection apparatus comprising:

Images