JP6411942B2 - Object recognition device - Google Patents

Description

  The present invention relates to an object recognition device for recognizing an object located around a host vehicle.

  As said object recognition apparatus, the technique which improves the resolution using the multi-echo (technology which acquires the distance information of several objects from one laser beam) contained in the distance information obtained by a laser radar is known (for example, , See Patent Document 1).

Special table 2014-142242 gazette

  However, since the above-mentioned object recognition device has a high-function and expensive laser radar capable of detecting multi-echo, the resolution is improved by using a single echo (a technique for obtaining only one distance information from one laser beam). There is a demand for making it happen.

  Therefore, in view of such problems, it is an object of the present invention to improve resolution without using multi-echo in an object recognition apparatus that recognizes an object positioned around the host vehicle.

In the object recognition apparatus of the present invention, the distance measurement point group acquisition unit irradiates a laser beam to each irradiation region obtained by dividing a detection region for detecting an object in a grid shape in the horizontal direction and the vertical direction in advance, and applies to each irradiation region. Then, a distance measuring point group including each distance and reflection intensity obtained by receiving the reflected light of the laser beam is acquired. The clustering means clusters the distance measuring point group. (Hereinafter, a cluster of distance measuring point groups generated by the clustering means is referred to as a cluster point group.) Next, the point group correcting means corrects the position of the cluster point group to a position set for each cluster. (Hereinafter, the cluster point group corrected by the point group correcting means is referred to as a corrected cluster point group.)
The integration unit integrates each cluster point group using the corrected cluster point group of the past time and the current time. (Hereinafter, the cluster point group generated by the integration means is referred to as an integrated cluster point group.) The corrected cluster point group of the current time obtained for each cluster is the integrated cluster point of the past time obtained for each cluster in the past. An integrated cluster point group at the current time is generated by integrating the group.

  According to such an object recognition apparatus, since an object can be recognized using not only a single distance measurement point group but also a distance measurement point group obtained in the past, the resolution can be improved without using multi-echo. Can be made.

  In addition, description of each claim can be arbitrarily combined as much as possible. At this time, a part of the configuration may be excluded.

1 is an explanatory diagram showing a schematic configuration of a driving support system 1. FIG. In embodiment, it is a schematic diagram which shows the area | region which irradiates a laser beam. It is a flowchart which shows the object recognition process which the radar control part 11 performs. It is a bird's-eye view which shows an example of a ranging point group and a clustering result. It is a three-dimensional view showing an example of a cluster point group. It is a flowchart which shows the super-resolution process of 1st Embodiment. It is a schematic diagram which shows the outline | summary of the super-resolution process of 1st Embodiment. It is a flowchart which shows the identification process of 1st Embodiment. It is a flowchart which shows the super-resolution process of 2nd Embodiment. It is a schematic diagram which shows the outline | summary of the super-resolution process of 2nd Embodiment. It is a flowchart which shows the super-resolution process of 3rd Embodiment. It is a schematic diagram which shows the example which decomposes | disassembles a cluster point group into parts point group based on a model. It is a schematic diagram which shows the outline | summary of the super-resolution process of 3rd Embodiment. It is a flowchart which shows the super-resolution process of 4th Embodiment. It is a graph which shows the theoretical relationship between distance and reflection intensity. It is a flowchart which shows the super-resolution process of 5th Embodiment. It is a schematic diagram which shows the outline | summary of the super-resolution process of 5th Embodiment. It is a flowchart which shows the super-resolution process of 6th Embodiment. It is a schematic diagram which shows the outline | summary of the super-resolution process of 6th Embodiment. It is a flowchart which shows the identification process of 7th Embodiment. It is a flowchart which shows the identification process of 8th Embodiment. It is explanatory drawing which shows an example of a correction | amendment cluster point group area | region in 8th Embodiment. It is a flowchart which shows the identification process of 9th Embodiment.

Embodiments according to the present invention will be described below with reference to the drawings.
[First Embodiment]
[Configuration of this embodiment]
FIG. 1 is an explanatory diagram showing a schematic configuration of the driving support system 1 of the present embodiment, and FIG. 2 is a schematic diagram showing a region to be irradiated with laser light. The driving support system 1 is mounted on a vehicle such as a passenger car (hereinafter also referred to as “own vehicle”), and includes a radar device 10 and a vehicle control unit 30 as shown in FIG.

The radar apparatus 10 includes a radar control unit 11, a scanning drive unit 12, and an optical unit 13.
The radar control unit 11 is configured as a known computer including a CPU 18 and a memory 19 such as a ROM or a RAM. The CPU 18 performs various processes such as an object recognition process to be described later according to a program stored in the memory 19. The radar control unit 11 may be configured by hardware such as a circuit.

  The scanning drive unit 12 is configured as an actuator such as a motor, for example, and is configured to be able to direct the optical unit 13 in an arbitrary direction in the horizontal direction and the vertical direction in response to a command from the radar control unit 11. Yes. Each time the scanning drive unit 12 receives a scanning start signal from the radar control unit 11, the scanning drive unit 12 obtains reflected light from all the areas to be irradiated with laser light so that the optical unit 13 can perform scanning for one cycle. To drive.

  The optical unit 13 emits laser light in response to a command from the radar control unit 11, and laser light from the light emitting unit 14 (indicated by a solid line arrow in FIG. 1) is reflected on the object 50. And a light receiving unit 15 that receives reflected light (indicated by broken arrows in FIG. 1).

  As a result, the scanning drive unit 12 may be configured so that the emission direction of the laser light from the light emitting unit 14 is changed to the same direction as the direction in which the light receiving unit 15 can receive the reflected light. For example, instead of the optical unit 13, the scanning drive unit 12 may be configured to drive a mirror that reflects laser light and reflected light in the optical unit 13 in an arbitrary direction.

  In this case, the laser beam is scanned in the horizontal direction by rotating a mirror having a plurality of reflecting surfaces by the scanning drive unit 12 and the angles of the reflecting surfaces are set to different angles, so that the laser beam is also vertically aligned. A configuration for scanning while shaking light may be employed. Further, a mechanism for directing a mirror having one reflecting surface in an arbitrary direction may be employed.

  Further, the scanning drive unit 12 may be configured to change the direction of only the light receiving unit 15. In this case, the light emitting unit 14 may be configured to be able to irradiate a part or the whole of the region where the light receiving unit 15 is scanned without changing the direction of the light emitting unit 14.

  As described above, the radar apparatus 10 intermittently irradiates a predetermined area in an arbitrary direction around the host vehicle (in the present embodiment, the front in the traveling direction of the host vehicle) with laser light that is an electromagnetic wave while scanning. And it is comprised as a laser radar which detects the object ahead of the own vehicle as each ranging point by receiving the reflected light, respectively.

  Here, in the radar apparatus 10 of the present embodiment, the radar control unit 11 scans the laser light emitted from the optical unit 13 within a predetermined region using the scanning drive unit 12 as described above. As shown in FIG. 2, the laser beam is irradiated intermittently at equal intervals (equal angles) while changing the range in which the laser beam is irradiated horizontally from the upper left corner to the upper right corner of this region. When reaching the upper right corner, the laser beam is irradiated again while changing the range in which the laser beam is irradiated from the region below the upper left corner by a predetermined angle to the right side in the horizontal direction.

  By repeating this operation, the radar apparatus 10 sequentially irradiates the entire region of the predetermined region with laser light. The radar apparatus 10 calculates the position of the object (ranging point) each time the laser beam is irradiated based on the timing at which the reflected light is received and the direction in which the laser beam is irradiated.

  Note that the laser beam emission direction can be specified by dividing the entire region irradiated with the laser beam into a matrix for each region irradiated with the laser beam and assigning a number to each region. For example, as shown in FIG. 2, numbers are assigned in order from the left in the horizontal direction, and these numbers are called orientation numbers. Also, numbers are assigned in order from the top in the vertical direction, and these numbers are called line numbers.

In the above [Expression 1], i represents a line number, j represents an orientation number, r represents a distance, and I represents a reflection intensity.

  Next, the vehicle control unit 30 is configured as a well-known computer including a CPU, a ROM, a RAM, and the like, and controls the behavior of the host vehicle according to a program stored in the ROM or the like, or notifies the driver. Various processes such as performing are performed. For example, when the vehicle control unit 30 receives a command from the radar apparatus 10 to perform driving support such as changing the behavior of the host vehicle (or urging the behavior to be changed), the vehicle control unit 30 sends a control signal corresponding to the command. What is necessary is just to make it output to any of a display apparatus, an audio | voice output apparatus, a braking device, a steering apparatus, etc.

[Process of this embodiment]
In the driving support system 1 configured as described above, the radar control unit 11 performs the object recognition process shown in FIG. The object recognition process is a process that is performed every time, for example, all the lines for one cycle and the distance measurement point group in all directions are obtained.

  In the object recognition processing, as shown in FIG. 3, first, a distance measuring point group is acquired (S110). The range-finding point group is a set of range-finding points for one cycle representing the coordinates of the object obtained for each irradiation region of the laser beam. Is included.

  Subsequently, clustering (S120) and tracking of each cluster specified by the clustering is performed (S130). In the clustering process, for example, a cluster corresponding to each object is obtained by clustering a plurality of distance measuring points close to each other.

  Further, in the tracking process, it is determined whether or not the position of the cluster is close to the cluster specified in the previous frame (the distance measurement point group obtained in the previous cycle). Estimated to be the obtained cluster. In this case, each cluster is linked in the time direction, each ID is given, and the speed of each cluster is calculated.

  Specifically, for example, the method disclosed in `` Multilayer LIDAR-Based Pedestrian Tracking in Urban Environments, S. Sato, M. Hashimoto, M. Takita, K. Takagi, T. Ogawa; IV2010 '' can be adopted. .

One or more clusters are obtained by such clustering. When K clusters are obtained, the parameters of each cluster C ^k are defined as follows.

In the above [Expression 2], x ^k and y ^k represent the positions of arbitrarily defined clusters. Further, w ^k and d ^{k represent} the width (length in the x-axis direction) and the depth width (length in the y-axis direction) of the cluster C ^k .

Here, FIG. 4 shows an example of distance measurement point groups and clustering results. In the present embodiment, the right direction of the host vehicle is positive on the x axis, and the front side of the host vehicle is positive on the y axis. Also, the upward direction in the vertical direction is positive on the z axis. In the example shown in FIG. 4, for example, the x coordinate of the center of the left end and the right end of the cluster is x ^k , and the y coordinate of the front end of the cluster is y ^k .

A set of ranging points constituting the cluster C ^k is a cluster point group, and is held in a memory such as a RAM as a point group composed of N ^k ranging points as shown in FIG.
In particular, in this embodiment, in addition to the point cloud information, information on the time when the point cloud is acquired is also held.

Subsequently, super-resolution processing is performed (S150). The super-resolution process is a process of correcting the position and integrating the cluster point group obtained in the past frame and the cluster point group obtained in the current frame. Note that the super-resolution processing is performed for each cluster point group obtained in each frame.

  In the super-resolution processing, as shown in FIG. 6, first, the barycentric coordinates of the cluster point group are calculated (S210). In this process, for example, the barycentric coordinates are obtained using the following equation.

Subsequently, coordinate conversion is performed (S220). In this processing, for example, translation is performed using the following equation so that the barycentric coordinates of the cluster point group become the origin. The cluster point group thus obtained is called a corrected cluster point group.

Subsequently, the corrected cluster point group obtained at the current time is integrated with the integrated cluster point group obtained at the past time to generate an integrated cluster point group at the current time (S230).

In this processing, as shown in FIG. 7, for example, at time (T-1), the integrated cluster point group held at time (T-2) and obtained at time (T-1). The correction cluster point cloud is integrated. At time T, the integrated cluster point group held at time (T-1) and the corrected cluster point group obtained at time T are further integrated. That is, if a certain object can be tracked, a higher-resolution distance measuring point can be held as an integrated cluster point group for each object.

When such processing ends, the super-resolution processing ends. Subsequently, returning to FIG. 3, the integrated cluster point cloud is stored in the memory and saved (S160).
Subsequently, identification processing is performed (S170). The identification process is a process of identifying the type of object corresponding to each integrated cluster point group.

In the identification process, as shown in FIG. 8, first, a distance measuring point effective for the identification process is selected using time information attached to the integrated cluster point group (S310). Here, for example, only the distance measurement points for the past predetermined frame (TH _CT ) from the current time are used, and the previous distance measurement points are not used.

That is, since old data may have poor reliability, it is set to be difficult to use. In addition, for example, in the integrated cluster point group, the distance measuring point whose time is closer to the present time may be identified by increasing the weight.

  Subsequently, the feature amount of each cluster is extracted using the selected distance measuring point (S320). For example, various feature amounts such as cluster width, height, aspect ratio, and relative speed obtained from the selected distance measuring point are extracted.

  Then, an identification score is calculated using the feature amount and an identification model prepared in advance (S330). For example, a known multi-class SVM (Support Vector Machine) is used. In this process, an SVM model is learned in advance for each type, and an identification score for each type is obtained using each model and the extracted feature amount. Then, the model type with the maximum score is determined as the type indicating the cluster.

When such processing ends, the identification processing ends, and the object recognition processing also ends.
[Effects of this embodiment]
In the driving support system 1 according to the first embodiment described in detail above, the radar control unit 11 irradiates a laser beam for each irradiation region obtained by dividing a detection region for detecting an object in a grid shape in the horizontal direction and the vertical direction in advance. Then, a distance measuring point group representing each distance and reflection intensity obtained by receiving the reflected light of the laser beam in each irradiation region is acquired. Further, the radar control unit 11 clusters the distance measurement point group, and generates a corrected cluster point group in which the position of the cluster point group is corrected to a position set for each cluster.

  Further, the radar control unit 11 integrates the corrected cluster point group of the current time obtained for each cluster into the integrated cluster point group of the past time obtained for each cluster in the past to obtain the integrated cluster point group of the current time. Generate. The radar control unit 11 identifies the integrated cluster point group.

  According to the driving support system 1 as described above, an object can be recognized using not only one distance measuring point group but also a distance measuring point group obtained in the past, so that the resolution can be reduced without using multi-echo. Can be improved.

  In the driving support system 1, the radar control unit 11 performs correction to match the position of the origin of the cluster point group with the position of the origin set in the integrated cluster point group obtained in the past.

According to such a driving support system 1, the position of the object can be reliably corrected.
Further, in the driving support system 1, the radar control unit 11 generates an integrated cluster point group including the acquisition time of each ranging point, uses the integrated cluster point group after performing weighting according to the acquisition time.

  According to such a driving support system 1, the object is identified after weighting according to the time at which each ranging point is acquired, and therefore, the highly accurate identification is performed using only highly reliable information. be able to.

[Second Embodiment]
Next, another type of driving support system will be described. In the present embodiment (second embodiment), only the portions different from the driving support system 1 of the first embodiment are described in detail, and the same portions as those of the driving support system 1 of the first embodiment are denoted by the same reference numerals. Therefore, the description is omitted. Further, for simplicity, the description will be made in a form in which time information is not used.

  In the driving support system of the second embodiment, the super-resolution process shown in FIG. 9 is performed instead of the super-resolution process shown in FIG. That is, in the super-resolution processing of the second embodiment, as shown in FIG. 9, first, each cluster point group is projected onto the xy plane (S360).

  Projection onto the xy plane means mapping an arbitrary point (x, y, z) in the three-dimensional space to a point (x, y) in the two-dimensional space.

Next, principal component analysis is performed on the cluster point group after projection (S370).

For example, the result of the principal component analysis is represented by a broken line arrow in FIG.

  Subsequently, rotation conversion is performed (S380). In this process, as shown in the upper and middle stages of FIG. 10, coordinate transformation is performed so that the major axis, which is the first principal component, rotates in parallel with the x-axis.

Subsequently, the correction cluster point group is integrated (S390).

In this way, as shown in FIG. 10, it is possible to integrate the cluster point groups in different directions at time (T-1) and time T after aligning them in the x-axis direction. When such processing is finished, the super-resolution processing of the second embodiment is finished.

That is, in the driving support system of the second embodiment, the radar control unit 11 estimates the direction of the cluster point group, and rotates the direction of the cluster point group in a preset direction.
According to the driving support system of the second embodiment as described above, even when the object rotates with time, it can be used together with the distance measuring point group in the past after aligning the direction of the object, and highly accurate identification can be performed. It can be carried out.

[Third Embodiment]
Next, the driving assistance system of 3rd Embodiment is demonstrated. For simplicity, the description will be made in a form in which time information is not used. In the driving support system of the third embodiment, the cluster point group is divided into a plurality of parts, the direction of each part is further estimated, and the direction of the part is rotated in a predetermined direction for each part. (Hereinafter, each divided point cloud is called a part point cloud.)
Specifically, the super-resolution processing of the third embodiment shown in FIG. 11 is performed. In the super-resolution processing of the third embodiment, as shown in FIG. 11, first, each part constituting the cluster is detected (S410), and a part point group is extracted (S420).

For example, a pedestrian is considered as an identification target. As indicated by reference numeral 41 in FIG. 12, a discriminator is prepared for each pedestrian part (for example, head, arm, torso, leg). When there is a possibility that the object is a pedestrian, each distance measuring point constituting each part indicated by reference numeral 42 is extracted for each part based on the discriminator, and a part point group indicated by reference numeral 43 is obtained. . In this process, for example, the following technique can be used.
"Pedestrian Detection Combining RGB and Dense LIDAR Data, C. Premebida, J. Carreira, J. Batista, U. Nunes; IROS2014"
Subsequently, the extracted part point group is corrected. That is, when the pedestrian is in an upright state, it is defined as a standard posture, and coordinate conversion is performed so that the position and orientation of each part overlap. Specifically, principal component analysis is first performed on each part point group in the same way as in the second embodiment (S430).

Subsequently, rotation conversion is performed on each part point group (S435). In this process, as shown in the upper and middle stages of FIG. 13, coordinate transformation is performed so that the first principal component vector rotates so as to be parallel to the y-axis.

Then, the cluster point group having the standard posture is held as a corrected cluster point group.

As a result of these processes, as shown in FIG. 13, even in the case of a pedestrian whose position of limbs varies depending on time (see the upper part of FIG. 13), the position of the limbs is unified (see the middle part of FIG. 13) Can be integrated.

  Subsequently, a correction cluster point group integration process is performed (S440).

That is, even when the orientation of each part differs at each acquisition time, the correction cluster point group is integrated (see the lower part of FIG. 13) with the positions of the parts being unified (see the middle part of FIG. 13). reference). When such processing is finished, the super-resolution processing of the third embodiment is finished.

  According to the driving support system of the third embodiment, for example, even when the direction of the parts constituting the object changes, such as when a pedestrian swings his arm, the past is performed after aligning the directions of the parts. Can be used together with the distance measuring point group in FIG.

[Fourth Embodiment]
Next, the driving assistance system of 4th Embodiment is demonstrated. For simplicity, the description will be made in a form in which time information is not used. In the driving support system of the fourth embodiment, the radar control unit 11 corrects the reflection intensity according to the distance. Then, the integrated cluster point group is identified using the reflection intensity.

  In the super-resolution processing of the fourth embodiment, as shown in FIG. 14, first, a cluster point group including a distance and a reflection intensity is acquired.

Then, the reflection intensity is normalized (S470).

That is, as shown in FIG. 15, the distance measurement point group has the same distance (r = 1) by utilizing the characteristic that the reflection intensity I is inversely proportional to the square of the distance r in principle (distance square law of the reflection intensity). The reflection intensity is corrected so that the reflection intensity is assumed to exist in

  Then, integration processing is performed using the corrected cluster point group whose reflection intensity is corrected (S480).

In this case, an identification model is learned in advance using the corrected reflection intensity, and in the identification process, the type of integrated point cloud cluster is identified using the corrected reflection intensity.

According to the driving support system of the fourth embodiment, the reflection intensity is corrected according to the distance, so that an identification model that does not depend on the distance can be constructed, and highly accurate identification can be performed.
[Fifth Embodiment]
Next, the driving assistance system of 5th Embodiment is demonstrated. For simplicity, the description will be made in a form in which time information is not used. In the driving support system of the fifth embodiment, the attitude (tilt) of the host vehicle is acquired, and the distance measuring point group is corrected according to the attitude of the host vehicle.

  Specifically, the super-resolution process shown in FIG. 16 is performed. In the super-resolution processing of the fifth embodiment, as shown in FIG. 16, first, pitching estimation is performed as an example of the posture estimation of the host vehicle (S510). The pitching estimation process is well known, and for example, the technique described in Japanese Patent Laid-Open No. 2003-043147 can be used. Information from a vehicle tilt sensor or the like may be used.

  Subsequently, rotation conversion of the distance measuring point group is performed (S520). In this process, the estimated pitching angle α is used to identify which position of the object the distance measurement point group obtained during pitching actually corresponds to and correct the distance measurement point group.

Then, it integrates using the correction | amendment cluster point group which correct | amended pitching of the own vehicle (S530).

When such processing is performed, for example, as shown in FIG. 17, a cluster point group at a different position at each time is obtained by pitching of the vehicle. For example, compared to the cluster point group when no pitching occurs at time (T-2), the cluster point at a relatively high position when the host vehicle is pitching upward at time (T-1). When the vehicle is pitched downward at time T, a cluster point group at a relatively low position is obtained.

  As shown in FIG. 17, when these correction cluster point groups are integrated, an integrated cluster point group having a higher resolution or a wider range in the vertical direction can be obtained as compared with a case where point group correction by pitching is not performed. It will be. When such processing is finished, the super-resolution processing of the fifth embodiment is finished.

  According to the driving support system of the fifth embodiment as described above, by correcting the distance measuring point group according to the posture (tilt) of the host vehicle, the integrated cluster points having a higher resolution or a wider range in the vertical direction. Since a group is obtained, highly accurate identification can be performed.

[Sixth Embodiment]
Next, the driving assistance system of 6th Embodiment is demonstrated. For simplicity, the description will be made in a form in which time information is not used. In the driving support system of the sixth embodiment, past identification results are acquired, and a corrected cluster point group is generated so as to match the shape of the object based on the type.

  Specifically, the super-resolution process shown in FIG. 18 is performed. In the super-resolution processing of the sixth embodiment, as shown in FIG. 18, first, the previous identification result is acquired (S560). Then, a point cloud model to be used in the point cloud correction process is selected according to the identification result (S570).

  Subsequently, point cloud model fitting is performed (S580). In this process, for example, as shown in FIG. 19, when a utility pole point cloud model is selected at time (T−2) and the utility pole point cloud model is selected, it most closely matches the utility pole point cloud model. First, each cluster point group at time (T-1) is corrected.

  More specifically, for example, a generally known two-point group fitting technique such as ICP (Iterative Closest Point) may be used. In this process, the cluster point group is corrected so that the positional error between the point groups is minimized, and a corrected cluster point group is generated.

Subsequently, the correction cluster point group integration process as described above is performed (S590), and the super-resolution process of the sixth embodiment is terminated.
According to the driving support system of the sixth embodiment, when the type of the object can be identified, the position is corrected according to the shape, so that the correction accuracy is improved and high-precision identification is performed. Can do.

[Seventh Embodiment]
Next, the driving assistance system of 7th Embodiment is demonstrated. In the driving support system of the seventh embodiment, the radar control unit 11 performs identification processing using the integrated cluster point group after performing weighting according to the distance of the distance measuring point at the time of acquiring the distance measuring point group.
Note that the point group correction processing will be described here using, for example, the barycentric coordinate conversion of the first embodiment.

  Specifically, first, when acquiring the cluster point group, the distance r of each ranging point is also recorded.

Next, a corrected cluster point group is obtained by coordinate conversion using the barycentric coordinates.

Then, integration is performed using a corrected cluster point group including the distance of the distance measuring point.

And in the driving assistance system of 7th Embodiment, the identification process shown in FIG. 20 is implemented. In the identification processing of the seventh embodiment, as shown in FIG. 20, first, distance measurement points effective for the identification processing are selected using distance information associated with the integrated cluster point group (S610).

In this process, for example, it is set so that only a distance measuring point within a predetermined distance (TH _r ) is used and a distance measuring point beyond that is not used. That is, since data at a certain distance may be unreliable, it is set to be difficult to use. In addition, for example, in the integrated cluster point group, a distance measuring point having a shorter distance may be identified by increasing the weight.

Thereafter, the processing of S320 and S330 shown in FIG. 8 is performed, and the identification processing of the seventh embodiment is terminated.

  According to the driving support system of the seventh embodiment, since the object is identified after weighting according to the distance of each ranging point, the highly accurate identification is performed using only highly reliable information. It can be performed.

[Eighth Embodiment]
Next, the driving assistance system of 8th Embodiment is demonstrated. In the driving support system of the eighth embodiment, the radar control unit 11 includes a laser light irradiation area (hereinafter referred to as a distance measuring point area) obtained from the distance of the distance measuring point and its area (hereinafter referred to as the distance measuring point). The area of the cluster point group irradiated with laser light (hereinafter referred to as the cluster point group area) is calculated based on the area (hereinafter referred to as the area of the cluster point group). Use to identify integrated cluster point cloud. In the driving support system of the eighth embodiment, the radar control unit 11 is a sum of the cluster point cloud region obtained this time and the cluster point cloud region in the past (hereinafter referred to as an integrated cluster point cloud region). An integrated cluster point group is identified using an area (hereinafter referred to as an integrated cluster point group area).
Note that the point group correction processing will be described here using, for example, the barycentric coordinate conversion of the first embodiment.

  Specifically, first, when acquiring a distance measuring point group, a distance measuring point area of each distance measuring point is recorded.

The distance measuring point area is set so that the area increases as the distance increases.
Next, a corrected cluster point group is obtained by coordinate conversion using the barycentric coordinates.

Subsequently, integration is performed using a corrected cluster point group including a distance measuring point area.

And in the driving assistance system of 8th Embodiment, the identification process shown in FIG. 21 is implemented. In the identification processing of the eighth embodiment, first, the integrated cluster point cloud area is calculated (S660). At this time, as shown in FIG. 22, for the corrected cluster point cloud region at time (T-1) and the corrected cluster point cloud region at time T, the total sum of the distance measuring point areas is defined as the integrated cluster point cloud area. .

Subsequently, using the integrated cluster point cloud area as a feature amount, an identification score is calculated using an identification model prepared in advance (S670). When such a process ends, the identification process of the eighth embodiment ends.

  According to the driving support system of the eighth embodiment as described above, the object is identified using the integrated cluster point cloud area. Therefore, the area of the object can be expressed with high accuracy, and high-precision identification can be performed.

[Ninth Embodiment]
Next, the driving assistance system of 9th Embodiment is demonstrated. In the driving support system of the ninth embodiment, the radar control unit 11 acquires the speed of each cluster using the tracking process (S130), and selects one model from the model previously learned for each speed according to the speed of each cluster. Select and identify using the selected model.

That is, the identification process shown in FIG. 23 is performed. In the identification processing of the ninth embodiment, as shown in FIG. 23, first, the average speed of each integrated cluster point group is calculated (S710). In this process, the speed of the host vehicle is determined from the host vehicle speed and the steering angle, and the speed v _k (t) of each cluster is determined based on the tracking process (S130). Then, the average speed of a series of past clusters is set as the average speed of the integrated cluster point group.

Subsequently, a model is selected (S720). Here, in the radar control unit 11, a plurality of models are prepared in advance according to the type of the object and the classification of the object speed. For example, in the case of a pedestrian, the change in posture is small in a stationary state, but the change in posture is large during movement. Therefore, a plurality of models are prepared according to the speed.

Subsequently, an identification score is calculated using the selected model (S730). When such a process ends, the identification process of the ninth embodiment ends.
According to the driving support system of the ninth embodiment, even if the posture changes according to the speed, it is possible to identify robustly against the posture change.

[Other Embodiments]
The present invention is not construed as being limited by the above embodiment. Further, the reference numerals used in the description of the above embodiments are also used in the claims as appropriate, but they are used for the purpose of facilitating the understanding of the invention according to each claim, and the invention according to each claim. It is not intended to limit the technical scope of The functions of one component in the above embodiment may be distributed as a plurality of components, or the functions of a plurality of components may be integrated into one component. Further, at least a part of the configuration of the above embodiment may be replaced with a known configuration having the same function. Moreover, you may abbreviate | omit a part of structure of the said embodiment as long as a subject can be solved. In addition, at least a part of the configuration of the above embodiment may be added to or replaced with the configuration of the other embodiment. In addition, all the aspects included in the technical idea specified only by the wording described in the claim are embodiment of this invention.

  In addition to the driving support system described above, the control device that is a component of the driving support system, a program for causing a computer to function as the driving support system, a medium on which the program is recorded, a driving support method, etc. The invention can also be realized.

[Correspondence between Configuration of Embodiment and Means of Present Invention]
In the above embodiment, the radar control unit 11 corresponds to the object recognition apparatus referred to in the present invention. Moreover, the process of S110 among the processes executed by the radar control unit 11 in the above embodiment corresponds to the distance measuring point group acquisition means referred to in the present invention, and the process of S120 in the above embodiment corresponds to the clustering means referred to in the present invention. Correspondingly, the process of S150 corresponds to the point cloud correction means and the integration means in the present invention, and the process of S170 corresponds to the identification means in the present invention.

  Furthermore, the processes of S210 to S220, S360 to S380, S410 to S435, S460 to S470, S510 to S520, and S560 to S580 in the above embodiment correspond to the point cloud correction means referred to in the present invention. The processes of S390, S440, S480, S530, and S590 correspond to the integration means referred to in the present invention. In addition, the processes of S170, S320 to S330, and S660 to S670 in the above embodiment correspond to the identifying means in the present invention, and the processes in S360 to S370 in the above embodiment correspond to the cluster direction estimating means in the present invention.

  Further, the process of S380 in the above embodiment corresponds to the cluster rotation correcting means in the present invention, and the process of S410 in the above embodiment corresponds to the point group dividing means in the present invention. Further, the process of S430 in the above embodiment corresponds to the part direction estimating means in the present invention, and the process of S435 in the above embodiment corresponds to the part rotation correcting means in the present invention.

  Further, the process of S470 in the above embodiment corresponds to the reflection intensity correcting means referred to in the present invention, and the process of S510 in the above embodiment corresponds to the own vehicle posture acquiring means referred to in the present invention. In addition, the process of S520 in the above embodiment corresponds to the posture correction unit in the present invention, and the process of S560 in the above embodiment corresponds to the identification result acquisition unit in the present invention.

  Furthermore, the process of S660 in the above embodiment corresponds to the area calculating means in the present invention, and the process of S710 in the above embodiment corresponds to the speed acquisition means in the present invention.

  DESCRIPTION OF SYMBOLS 1 ... Driving assistance system, 10 ... Radar apparatus, 11 ... Radar control part, 12 ... Scanning drive part, 13 ... Optical unit, 14 ... Light emission part, 15 ... Light receiving part, 18 ... CPU, 19 ... Memory, 30 ... Vehicle control Department.

Claims (13)

An object recognition device (2) that is mounted on a host vehicle and recognizes an object located around the host vehicle,
Each of the detection areas obtained by detecting the detection area for detecting an object in each of the irradiation areas obtained by previously dividing the detection area into a grid shape in the horizontal direction and the vertical direction, and receiving the reflected light of the laser light in each irradiation area Distance measurement point group acquisition means (S110) for acquiring a distance measurement point group representing a set of a plurality of distance measurement points using distance information from the object as distance measurement points;
Clustering means for clustering the distance measurement point group to obtain a cluster point group (S120);
Point group correction means (S210 to S220, S360 to S380, S410 to S435, S460 to S470, S510 to S520, S560) for generating corrected cluster point groups in which the positions of the cluster point groups are corrected to positions set for each cluster. To S580),
Integration means (S230, S390, S390, S390) for obtaining an integrated cluster point group of the current time obtained by integrating the corrected cluster point group of the current time obtained for each cluster into the integrated cluster point group of the past time obtained for each cluster in the past. S440, S480, S530, S590),
Equipped with a,
The point cloud correction means includes
Point group dividing means (S410) for dividing the cluster point group into a plurality of parts and generating a part point group representing the point group constituting each part;
Part direction estimating means (S430) for estimating the direction of the part point group for each of the plurality of parts;
Part rotation correction means (S435) for rotating the direction of the plurality of parts point groups in a preset direction;
An object recognition apparatus characterized by comprising: The object recognition apparatus according to claim 1, wherein an identification unit (S170, S320 to S330, S660 to S670) for identifying a type of an object using the integrated cluster point cloud,
An object recognition apparatus characterized by comprising: The object recognition apparatus according to claim 1 or 2,
The object recognition apparatus characterized in that the point group correction means performs correction to match the position of the origin of each cluster point group with the position of the origin set in the integrated cluster point group obtained in the past. An object recognition device (2) that is mounted on a host vehicle and recognizes an object located around the host vehicle,
Each of the detection areas obtained by detecting the detection area for detecting an object in each of the irradiation areas obtained by previously dividing the detection area into a grid shape in the horizontal direction and the vertical direction, and receiving the reflected light of the laser light in each irradiation area Distance measurement point group acquisition means (S110) for acquiring a distance measurement point group representing a set of a plurality of distance measurement points using distance information from the object as distance measurement points;
Clustering means for clustering the distance measurement point group to obtain a cluster point group (S120);
Point group correction means (S210 to S220, S360 to S380, S410 to S435, S460 to S470, S510 to S520, S560) for generating corrected cluster point groups in which the positions of the cluster point groups are corrected to positions set for each cluster. To S580),
Integration means (S230, S390, S390, S390) for obtaining an integrated cluster point group of the current time obtained by integrating the corrected cluster point group of the current time obtained for each cluster into the integrated cluster point group of the past time obtained for each cluster in the past. S440, S480, S530, S590),
Identification means (S170, S320 to S330, S660 to S670) for identifying the type of the object using the integrated cluster point cloud;
In the integrated cluster point group, area calculation means (S660) for calculating a weighted area of the irradiation region of the integrated cluster point group by weighting according to the overlapping region of each irradiation region of each ranging point;
With
The identification means identifies an object type using the weighted area.
Object recognition apparatus according to claim. The object recognition apparatus according to claim 4 ,
The object recognition apparatus characterized in that the point group correction means performs correction to match the position of the origin of each cluster point group with the position of the origin set in the integrated cluster point group obtained in the past. In the object recognition device according to claim 4 or 5 ,
The point cloud correction means includes
Cluster direction estimation means (S360 to S370) for estimating the direction of the cluster point group;
Cluster rotation correction means (S380) for rotating the direction of the cluster point group in a preset direction;
An object recognition apparatus characterized by comprising: In the object recognition device according to claim 4 or 5 ,
The point cloud correction means includes
Point group dividing means (S410) for dividing the cluster point group into a plurality of parts and generating a part point group representing the point group constituting each part;
Part direction estimating means (S430) for estimating the direction of the part point group for each of the plurality of parts;
Part rotation correction means (S435) for rotating the direction of the plurality of parts point groups in a preset direction;
An object recognition apparatus characterized by comprising: In the object recognition device according to claim 2 or any one of claims 4 to 7 ,
The distance measuring point group acquisition means is configured to acquire a distance measuring point group including a distance and a reflection intensity,
The object recognition device
Reflection intensity correction means (S470) for correcting the reflection intensity according to the distance,
The object recognizing apparatus characterized in that the identifying means identifies the type of an object using reflection intensity. The object recognition apparatus according to claim 2 or any one of claims 4 to 8 ,
The point group correction means generates a correction cluster point group including the time of acquisition of the distance measurement point group,
The identification means is used after weighting according to the acquisition time of the distance measurement points included in the integrated cluster point group. The object recognition apparatus according to claim 2 or any one of claims 4 to 9 ,
The object recognition apparatus is characterized in that the identification means is used after performing weighting according to the distance of distance measurement points included in the integrated cluster point group. The object recognition device according to claim 2 or any one of claims 4 to 10 ,
Speed acquisition means (S710) for acquiring the speed of the integrated cluster point cloud;
With
The object recognition apparatus, wherein the identification unit selects a model from a plurality of models prepared in advance according to the speed of the object, and identifies the type of the object using the selected model. The object recognition device according to any one of claims 1 to 11 ,
Own vehicle posture acquisition means (S510) for acquiring the posture of the own vehicle;
Attitude correction means (S520) for correcting the position of the distance measuring point group according to the attitude of the host vehicle;
An object recognition apparatus characterized by comprising: In the object recognition device according to any one of claims 1 to 12 ,
Identification result acquisition means (S560) for acquiring identification results in the past;
With
The point group correcting unit generates a corrected cluster point group based on the type of the object so as to correct the position according to a shape set in advance corresponding to the type. .