JP6953084B2 - Radar signal processing device and radar signal processing program - Google Patents

Description

The present invention relates to a radar signal processing device and a radar signal processing program for recognizing a vehicle traveling on a road having a plurality of lanes by a radar.

When a radar / radio wave is applied to a target and the received signal is processed, the target is extracted as a detection point, and the detection point information includes the distance, speed, and direction relative to the radar transmitter / receiver. Is done. In particular, in the case of millimeter-wave radar, the resolution of distance, velocity, and direction is high, and a plurality of detection points can be obtained from one target. Therefore, a technique for performing clustering (aggregation processing) for recognizing a plurality of detection points as one target is known (see, for example, Patent Document 1 and the like). In this clustering, the region of the detection point group determined to be the detection point from the same target is grouped as one cluster based on the distance, speed and direction of each detection point. For example, when a plurality of detection points D as shown in FIG. 9 are obtained, a cluster C is generated assuming that the specific plurality of detection points D are detection points D from the same target, and the cluster C is used as an object. Recognize the size and shape of the mark.

Clustering is also used when detecting a plurality of vehicles traveling on a road having a plurality of lanes at the same time. For example, in FIG. 10, on a two-lane road / public road W having lanes L1 and L2, the vehicle M1 is traveling in the lane L1, and the vehicle M2 is traveling in the lane L2 in front of the vehicle M1. In this case, the cluster C1 is generated based on the distance, speed and direction of the detection point D1 of the vehicle M1, and the cluster C2 is generated based on the distance, speed and direction of the detection point D2 of the vehicle M2.

Such clustering is used, for example, for traffic volume surveys. In the traffic volume survey, the generated cluster is tracked by a predetermined distance, and the current position and the movement route are acquired. Then, when the tracked cluster passes through the virtually set passing line, it is counted as if one vehicle has passed through the road W.

JP-A-2017-122646

However, in the conventional clustering, for example, when two vehicles are traveling close to each other, the detection points of these two vehicles are erroneously recognized as the detection points of one vehicle to generate a cluster. , The number of vehicles was small and was sometimes recognized.

For example, as shown in FIG. 11, when the vehicle M1 and the vehicle M2 are traveling side by side in the lane L1 and the lane L2, the distance between the detection point D1 of the vehicle M1 and the detection point D2 of the vehicle M2 is large. Get closer. In such a case, if the speeds of the vehicle M1 and the vehicle M2 are different, the detection points D1 and the detection points D2 can be clustered separately based on the speed components. However, when the vehicle M1 and the vehicle M2 are traveling at substantially the same speed, that is, when the vehicle M1 and the vehicle M2 are traveling in parallel, clustering based on the speed component cannot be performed, so that the detection point D1 And the detection point D2, one cluster C3 is generated, and the two vehicles are recognized as one vehicle. If such a misrecognition occurs, a correct survey will not be performed in a traffic volume survey or the like.

Therefore, an object of the present invention is to provide a radar signal processing device and a radar signal processing program that can be recognized by a plurality of vehicles running in parallel.

In order to achieve the above object, the invention according to claim 1 transmits radio waves from a radar transmitter / receiver to a road having a plurality of lanes, receives radio waves reflected from a vehicle traveling on the road, and receives the received radio waves. In a radar signal processing device that recognizes the vehicle based on the tracking information obtained by clustering the detection points obtained by signal processing to generate a cluster of the vehicle and tracking the cluster, the lane A lane information storage means for storing lane information including a number, the width of each lane, and the position of each lane with respect to the radar transmitter / receiver, and two adjacent lanes based on the lane information and the tracking information. When the cluster located across the lanes is specified and it is determined from the tracking information that the specified cluster is not changing lanes, it is recognized as two vehicles running in parallel in the cluster. It is characterized in that it is provided with a cluster separation means for separating into clusters for two units.

The invention according to claim 2 is the radar signal processing apparatus according to claim 1, wherein the cluster separation means calculates a boundary line for separating a set of detection points on the left and right, and the boundary line. It is characterized in that the cluster is separated by.

The invention according to claim 3 is the radar signal processing apparatus according to claim 1 or 2, wherein the cluster separating means separates the cluster when the width of the cluster is equal to or larger than a predetermined width. It is characterized by.

The invention according to claim 4 is obtained by transmitting radio waves from a radar transmitter / receiver to a road having a plurality of lanes, receiving radio waves reflected from a vehicle traveling on the road, and processing the received radio waves. In a radar signal processing program that recognizes a vehicle based on tracking information obtained by clustering detection points to generate a cluster of the vehicle and tracking the cluster, a computer is used for the number of lanes and the number of lanes. A lane information storage means for storing lane information including the width of each lane and the position of each lane with respect to the radar transmitter / receiver, and a position straddling between adjacent lanes based on the lane information and the tracking information. When the cluster in the above is specified and it is determined from the tracking information that the specified cluster is not changing lanes, it is recognized that the cluster is two vehicles running in parallel and two vehicles are used. It is characterized in that it functions as a cluster separation means for separating into clusters.

According to the first and fourth aspects of the invention, a cluster located across two adjacent lanes, that is, a vehicle not in the lane, in other words, based on the lane information and the tracking information of the cluster. For example, it is possible to identify a cluster of vehicles traveling so as to extend from one lane to the next lane. For example, if two vehicles run side by side in two adjacent lanes and the detection points of each vehicle are in close proximity, the detection points cannot be separated based on distance, speed and direction, so there are two clusters. Generated to straddle between lanes. By determining such a cluster, it becomes possible to appropriately identify a cluster that may have been erroneously recognized by the two vehicles. In addition, it is possible to determine whether or not the cluster is changing lanes based on the tracking information of the cluster. For example, when a vehicle is changing lanes, the cluster of the vehicle extends out of the running lane and is detected at a position straddling between two adjacent lanes. Therefore, by determining whether or not the lane is being changed based on the tracking information of the cluster, it is possible to appropriately identify the cluster that has been erroneously recognized by the two vehicles. Furthermore, if it is determined that the cluster is not changing lanes, the cluster is recognized as two vehicles running in parallel and separated into clusters for two vehicles, so that multiple vehicles running in parallel are used. It is possible to recognize the vehicle appropriately and accurately.

According to the invention of claim 2, since the boundary line for separating the set of detection points on the left and right is calculated, the detection points of each vehicle can be determined even when a large number of detection points are located close to each other and coexist. It becomes possible to distinguish appropriately. Further, since the clusters are separated by the calculated boundary line, it is possible to appropriately separate one cluster misrecognized from two vehicles into two clusters.

According to the third aspect of the present invention, the clusters are separated when the width of the clusters is equal to or larger than a predetermined width. That is, although it should not be in terms of traffic safety, there are rare vehicles that travel a long distance while straddling two adjacent lanes, and such a cluster of vehicles has a lane even when looking at the tracking information. Since it has not been changed, the cluster separation means according to claim 1 determines that the two vehicles have been erroneously recognized and separates the two vehicles. In the invention according to claim 2, in order to prevent such clusters of vehicles from being separated, if the width of the clusters is less than a predetermined width (for example, the width of a general compact car), one vehicle is used. Do not separate the cluster by recognizing it as a vehicle. On the contrary, when the width of the cluster is equal to or larger than the predetermined width, the cluster is separated as a cluster that may be erroneously recognized by the two vehicles. According to this, it becomes possible to appropriately recognize one vehicle traveling in a state of straddling between two adjacent lanes.

It is a schematic block diagram which shows the mobile body recognition radar apparatus which concerns on Embodiment 1 of this invention. It is a top view which shows the arrangement position of the radar transceiver of the moving body recognition radar apparatus of FIG. It is a figure which shows the 1st example of the clustering result by the moving body recognition radar apparatus of FIG. 1, (a) shows one cluster generated from two vehicles running in parallel, (b) is (b). The state in which one cluster shown in a) is separated into clusters for two vehicles is shown. It is a figure which shows the 2nd example of the clustering result by the moving body recognition radar apparatus of FIG. 1, and shows the cluster which was determined to be changing lanes from the tracking information. It is a flowchart which shows the processing flow of the moving body recognition radar apparatus of FIG. It is a flowchart which shows the process flow of the cluster separation of the mobile recognition radar apparatus of FIG. It is a figure which shows the clustering result by the mobile recognition radar apparatus which concerns on Embodiment 2 of this invention, and shows the cluster of one vehicle which keeps running so as to straddle between two adjacent lanes. It is a flowchart which shows the process flow of cluster separation of the mobile recognition radar apparatus which concerns on Embodiment 2 of this invention. It is a figure which shows the cluster by the conventional clustering. It is a figure which shows the conventional clustering result of two vehicles traveling in different lanes at different speeds. It is a figure which shows the conventional clustering result of two vehicles running in parallel in different lanes at substantially the same speed.

Hereinafter, the present invention will be described based on the illustrated embodiment.

(Embodiment 1)
1 to 6 show an embodiment of the present invention, and FIG. 1 is a schematic block diagram showing a mobile body recognition radar device 1 according to the embodiment. The mobile body recognition radar device 1 transmits and irradiates radio waves and radar, processes radio waves reflected from the moving body to obtain detection points, clusters the detection points, and determines the position and size of the moving body. It is a device that recognizes a shape or the like, and mainly includes a radar transmitter / receiver 2 and a signal processing device (radar signal processing device) 3.

Here, in this embodiment, the vehicle is a vehicle traveling on a road / public road W having a plurality of lanes, and a case where the size / type of the vehicle is recognized in a traffic volume survey or the like will be mainly described below. do. Further, it is assumed that the radio wave emitted from the radar transceiver 2 is a millimeter wave, has high resolution of distance, speed, and direction, and a plurality of detection points can be obtained from one target.

The radar transceiver 2 is a radar device that transmits and receives radio waves, digitally converts the received radio waves, and transmits the received radio waves to the signal processing device 3 as a received signal. In this embodiment, the radar transceiver 2 is arranged at a position where more detection points can be obtained depending on the size of the road W, the trajectory, and the like. For example, in this embodiment, as shown in FIG. 2, the radar transceiver 2 is arranged beside one road W behind the vehicle traveling direction, and radio waves are transmitted from the rear to the vehicle traveling on the road W. do. The number and arrangement of radar transceivers 2 are not limited to this.

The signal processing device 3 is a device that determines the size of a vehicle by processing a signal received from a radar transceiver 2 to extract detection points of each target or clustering based on the detection points. Therefore, the description of the configuration and processing equivalent to the conventional one will be omitted. The signal processing device 3 clusters to generate a cluster of vehicles based on the distance, speed, and orientation of the detection points obtained by signal processing the radio waves received by the radar transmitter / receiver 2, and tracks the generated cluster. However, when the tracked cluster passes through a passing line virtually set on the road W, it has a characteristic process of counting as if one vehicle has passed the road W. In the cluster tracking process, tracking information including the current position of the cluster and the movement route of the cluster is generated by tracking the change in the position of the cluster over time.

Further, when a cluster for one vehicle is generated from two vehicles running in parallel in two adjacent lanes, the signal processing device 3 separates the cluster into clusters for two vehicles, so that the lanes are separated. It includes an information database (lane information storage means) 31 and a cluster separation unit (cluster separation means) 32.

The lane information database (hereinafter referred to as D / B) 31 contains lane information including the number of lanes on the road W on which the radar transmitter / receiver 2 is installed, the width of each lane, and the relative position of each lane with respect to the radar transmitter / receiver 2. Remember. For example, in the case of the road W shown in FIG. 2, the number of lanes L1 and L2, the widths H1 and H2 of the lanes L1 and L2, and the relative positions of the lanes L1 and L2 with respect to the radar transceiver 2 are lane information. Is stored in the lane information D / B31. In this embodiment, the lane information D / B31 is provided inside the signal processing device 3, but it may be provided outside the signal processing device 3, or it may be installed in a data center or the like to provide a communication network such as the Internet. May be used to connect to the signal processing device 3.

The cluster separation unit 32 identifies a cluster at a position straddling between two adjacent lanes based on the lane information and the tracking information, and determines from the tracking information that the identified cluster is not changing lanes. In this case, the cluster is recognized as two vehicles running in parallel and separated into two clusters.

For example, as shown in FIG. 3, when two vehicles M1 and M2 are running in parallel in lane L1 and lane L2 at substantially the same speed, detection extracted from radio waves reflected from these vehicles M1 and M2. Since the points D1 and D2 exist at a close distance, the signal processing device 3 may mistakenly recognize the two vehicles M1 and M2 as one vehicle and generate one cluster C3. Such a misrecognized cluster C3 is located so as to straddle between two adjacent lanes L1 and L2.

First, the cluster separation unit 32 has two adjacent lanes like the cluster C3, based on the lane information read from the lane information D / B31 and the current position of the tracking information obtained by tracking the cluster C3. Identify clusters that are located across L1 and lane L2, that is, vehicles that are not in lane L1 or L2, in other words, vehicles that travel out of lane L1 or L2 into the adjacent lane. .. The position of the cluster C3 with respect to the lanes L1 and L2 may be specified based on the position of the center (center of gravity) of the cluster C3, or may be specified based on the position of the outer shape of the cluster C3.

Next, the cluster separation unit 32 determines whether or not the cluster C3 is changing lanes based on the movement route included in the tracking information of the cluster C3. The square indicated by the chain double-dashed line in the figure indicates the position where the cluster C3 existed before a predetermined time, that is, the movement path Cx3 of the cluster C3. The cluster separation unit 32 determines whether or not the cluster C3 is changing lanes by using the movement route of the cluster C3 and the current position.

In the example shown in FIG. 3, it is determined that the current position of the cluster C3 and the movement path Cx of the cluster C3 are not in the process of changing lanes because there is no significant change in the positions with respect to the lanes L1 and L2. On the other hand, as shown in FIG. 4, the cluster C4 generated from the detection point D4 of the vehicle M4 and the movement path Cx4 of this cluster C4 have a large change with respect to the lanes L1 and L2, so that the lane is being changed. It is judged.

When the cluster separation unit 32 determines that the cluster located straddling between two adjacent lanes is not in the process of changing lanes, the cluster separation unit 32 recognizes that the cluster is two vehicles running in parallel. A boundary line for separating the set of detection points D1 and D2 on the left and right is calculated, and one cluster C3 is separated into two clusters by the calculated boundary line. That is, in the example shown in FIG. 3A, the cluster C3 is located at a position straddling between two adjacent lanes L1 and L2, and the lane is not being changed. As shown in FIG. (B), it is separated into two clusters C1 and C2 corresponding to the vehicles M1 and M2. The cluster C4 shown in FIG. 4 is located at a position straddling between two adjacent lanes L1 and L2, but is not separated because the lane is being changed.

The boundary line for separating the cluster C3 into the clusters C1 and C2 for two units is a straight line extending in the distance direction passing through the midpoint of the circumscissa of the cluster C3 in the lateral direction (person width direction), or the cluster C3. It is selected by the abscissa that maximizes the interclass dispersion when the weights of the included detection points D1 and D2 are expressed as a horizontal histogram. The weights of the detection points D1 and D2 may be simply the number of the detection points D1 and D2, the product of the detection points D1 and D2 and their intensities, or a function of the product. ..

Since the above-mentioned inter-class variance is explained by, for example, "Shinichi Murakami, Image Processing Engineering, Tokyo Denki University Publishing Co., Ltd., 2nd Edition, p.31, 2007", detailed explanation is omitted here. do.

According to the mobile body recognition radar device 1 having such a configuration, as shown in FIG. 5, first, the radio waves emitted from the radar transceiver 2 and reflected from the vehicles M1 and M2 are signal processing devices as received signals. It is transmitted to 3 (step S1). Next, in the signal processing device 3, the detection point D1 of the vehicle M1 and the detection point D2 of the vehicle M2 are extracted as described above (step L2).

The signal processing device 3 performs clustering based on the distance, speed, and direction of the detection point D1 group and the detection point D2 group to generate a cluster (step S3). For example, when two vehicles M1 and M2 are running side by side in two adjacent lanes L1 and L2 at substantially the same speed, the detection points D1 and D2 are located at a close distance, and the distance is long. Since it cannot be separated by speed and orientation, the signal processing device 3 misrecognizes the two vehicles M1 and M2 as one vehicle and generates one cluster C3.

Further, the signal processing device 3 generates tracking information including the current position of the cluster C3 and the movement path of the cluster C3 by tracking the change of the position of the generated cluster C3 with time (step S4).

The cluster separation unit 32 separates one cluster C3 generated by erroneously recognizing the two vehicles M1 and M2 as one vehicle (step S5). More specifically, as shown in FIG. 6, two adjacent lanes L1 and L2 are based on the lane information read from the lane information D and B31 and the current position of the tracking information obtained by tracking the cluster C3. The cluster C3 located at a position straddling the space is specified (step S51).

When the cluster separation unit 32 identifies the cluster C3 located so as to straddle the two adjacent lanes L1 and L2 (YES in step S51), the cluster separation unit 32 is based on the movement path Cx3 included in the tracking information of the cluster C3. Then, it is determined whether or not the cluster C3 is changing lanes (step S52).

Further, when the cluster separation unit 32 determines that the cluster C3 located straddling between the two adjacent lanes L1 and L2 is not changing lanes (NO in step S52), the cluster separation unit 32 aligns the clusters C3. Recognizing that the two vehicles are running M1 and M2, the two vehicles are separated into clusters C1 and C2 (step S53). More specifically, a boundary line for separating the set of detection points D1 and D2 on the left and right is calculated, and one cluster C3 is separated into clusters C1 and C2 for two vehicles by the calculated boundary line. ..

Subsequently, as shown in FIG. 5, the signal processing device 3 determines the vehicle type based on the length of the cluster C3 and counts the vehicles (step S6).

As described above, according to the moving object recognition radar device 1, the cluster C3 is in a position straddling between two adjacent lanes L1 and L2 based on the lane information and the tracking information of the cluster C3, that is, , It is possible to identify clusters that protrude from the running lane into the adjacent lane. For example, when two vehicles M1 and M2 run in parallel in two adjacent lanes L1 and L2 and the detection points D1 and D2 of each vehicle M1 and M2 are in close proximity, they straddle between the two lanes. Cluster C3 is generated as described above. By determining such a cluster C3, it becomes possible to appropriately identify the cluster C3 that may have been erroneously recognized by the two vehicles M1 and M2.

Further, it is possible to determine whether or not the cluster C3 is changing lanes based on the tracking information of the cluster C3. For example, when the vehicle M4 is changing lanes, the cluster C4 of the vehicle M4 protrudes from the traveling lane L2 and is detected at a position straddling between two adjacent lanes L1 and L2. Therefore, by determining whether or not the lane is being changed based on the tracking information of the cluster C3, it is possible to appropriately identify the cluster C3 erroneously recognized by the two vehicles M1 and M2. Further, when it is determined that the cluster C3 is not changing lanes, the cluster C3 is recognized as two vehicles running in parallel and separated into two clusters C1 and C2, so that the cluster C3 runs in parallel. It is possible to appropriately and accurately recognize a plurality of vehicles M1 and M2.

Further, since the boundary line for separating the set of detection points of the cluster C3 on the left and right is calculated, even if a large number of detection points D1 and D2 are in close positions and coexist, the detection points D1 of each vehicle M1 and M2 are calculated. , D2 can be properly distinguished. Further, since the cluster C3 is separated by the calculated boundary line, one cluster C3 erroneously recognized by the two vehicles M1 and M2 can be appropriately separated into the clusters C1 and C2 for two vehicles. ..

(Embodiment 2)
7 and 8 show an embodiment of the present invention. In this embodiment, the method of separating the clusters of vehicles is different from that of the first embodiment, and the same reference numerals are given to the configurations equivalent to those of the first embodiment. The description is omitted by adding.

In the first embodiment, when the cluster located straddling between two adjacent lanes is not changed lanes, it is recognized that the cluster is two vehicles running in parallel, and two vehicles are recognized. Separated into clusters. However, although it should not be in terms of traffic safety, as shown in FIG. 7, there is rarely a vehicle M5 that travels a long distance while straddling two adjacent lanes L1 and L2. Since the cluster C5 of the vehicle M5 has not changed lanes even when the movement path Cx5 is viewed, in the method of the first embodiment, it is determined that the two vehicles are erroneously recognized and separated.

In this embodiment, in order to prevent the cluster C5 of the vehicle 5 from being separated, the width MW of the cluster C5 and the vehicle width (predetermined width) of a general compact car are set as shown in FIG. In comparison (step S54), if the width MW is less than the predetermined width (NO in step S54), it is recognized as one vehicle and the cluster is not separated. On the contrary, when the width MW of the cluster C5 is equal to or larger than the predetermined width (YES in step S54), the clusters are separated as clusters that may be erroneously recognized by the two vehicles. According to this, it becomes possible to appropriately recognize one vehicle traveling in a state of straddling between two adjacent lanes.

In this embodiment, the width of the cluster and the predetermined width are compared after the determination of the lane change of the cluster, but before the identification of the cluster (step S51), the cluster composed of one vehicle is performed in advance. May be excluded from the candidates for the cluster to be separated.

Although the embodiment of the present invention has been described above, the specific configuration is not limited to the above-described embodiment, and even if there is a design change or the like within a range that does not deviate from the gist of the present invention. Included in the invention. For example, in the above embodiment, one radar transmitter / receiver 2 and one signal processing device 3 corresponding thereto constitute a moving object recognition radar device 1, but the moving body recognition radar device 1 is discretely arranged on the road W. The received signals of the plurality of installed radar transmitters / receivers 2 may be collectively processed by one signal processing device 3 to recognize the vehicle at each position on the road W.

Further, by installing the following radar signal processing program on a general-purpose computer, the signal processing device (radar Shingo processing device) 3 of the moving object recognition radar device 1 may be configured.
That is, the computer,
Detection obtained by transmitting radio waves from the radar transmitter / receiver 2 to the road W having a plurality of lanes L1 and L2, receiving the radio waves reflected from the vehicles M1 and M2 traveling on the road W, and processing the received radio waves. In the signal processing device 2 that recognizes the vehicles M1 and M2 based on the tracking information obtained by clustering the points D1 and D2 to generate clusters of the vehicles M1 and M2 and tracking the clusters.
Lane information storage means (lane information D / B31) for storing lane information including the number of lanes L1 and L2, the width of each lane L1 and L2, and the relative position of each lane L1 and L2 with respect to the radar transmitter / receiver 2. ,
Based on the lane information and the tracking information, the cluster C3 located at a position straddling between the two adjacent lanes L1 and L2 was specified, and it was determined from the tracking information that the specified cluster C3 was not changing lanes. In this case, a cluster separation means (cluster separation unit 32) that recognizes the cluster C3 as two vehicles M1 and M2 running in parallel and separates the cluster C3 into two clusters C1 and C2.
The present invention may be applied as a radar signal processing program to function as a radar signal processing program.

1 Mobile recognition radar device 2 Radar transmitter / receiver 3 Signal processing device (radar signal processing device)
D, D1, D2, D3, D4, D5 Detection points C, C1, C2, C3, C4, C5 Cluster W Road L1, L2 Lanes

Claims (4)

Radio waves are transmitted from a radar transmitter / receiver to a road having a plurality of lanes, radio waves reflected from a vehicle traveling on the road are received, and the detection points obtained by signal processing the received radio waves are clustered to obtain the vehicle. In a radar signal processing device that recognizes the vehicle based on the tracking information obtained by generating the cluster and tracking the cluster.
A lane information storage means for storing lane information including the number of lanes, the width of each lane, and the position of each lane with respect to the radar transmitter / receiver.
When the cluster located at a position straddling between two adjacent lanes is specified based on the lane information and the tracking information, and it is determined from the tracking information that the specified cluster is not changing lanes. In addition, a cluster separation means that recognizes the cluster as two vehicles running in parallel and separates the cluster into two clusters.
A radar signal processing device characterized by being equipped with. The cluster separation means calculates a boundary line for separating the set of detection points on the left and right, and separates the cluster by the boundary line.
The radar signal processing device according to claim 1. The cluster separation means separates the cluster when the width of the cluster is equal to or greater than a predetermined width.
The radar signal processing apparatus according to claim 1 or 2. Radio waves are transmitted from a radar transmitter / receiver to a road having a plurality of lanes, radio waves reflected from a vehicle traveling on the road are received, and the detection points obtained by signal processing the received radio waves are clustered to obtain the vehicle. In a radar signal processing program that recognizes the vehicle based on the tracking information obtained by generating the cluster and tracking the cluster.
Computer,
A lane information storage means for storing lane information including the number of lanes, the width of each lane, and the position of each lane with respect to the radar transmitter / receiver.
When the cluster located at a position straddling between adjacent lanes is specified based on the lane information and the tracking information, and it is determined from the tracking information that the specified cluster is not changing lanes. A cluster separation means that recognizes the cluster as two vehicles running in parallel and separates the cluster into clusters for two vehicles.
Radar signal processing program to function.

Images