JP2023032080A - Object recognition device - Google Patents

Abstract

To provide an object recognition device capable of properly recognizing the shape of an object while reducing the amount of memory used.SOLUTION: An object recognition device 10 applies to a vehicle that includes an object detection device 22 for detecting objects. While a vehicle is running, the object recognition device extracts a predetermined number of detection points per object out of the multiple detection points detected in time series by the object detection device and recognizes the shape of the object based on the predetermined number of detection points. The object recognition device 10 is configured to determine that the detected surface of the object is curved based on multiple detection points detected in time series by the object detection device while the vehicle is running, and based on the determination result, changes the execution mode of the object recognition performed by the predetermined number of detection points.SELECTED DRAWING: Figure 1

Description

The present invention relates to an object recognition device that recognizes objects around a vehicle.

2. Description of the Related Art Conventionally, there has been known a device that detects an object around a vehicle using, for example, electromagnetic waves or sound waves and recognizes the object based on the detection result. For example, the device of Patent Document 1 detects a plurality of detection points on an object around the vehicle in time series while the vehicle is running, and generates a plurality of line segments (vectors) connecting the detection points in different order. Recognize the shape of a planar object based on

Japanese Patent No. 4809690

While the vehicle is running, a plurality of detection points are detected in time series, and if all the detection points are stored in the memory, the amount of memory used will increase. Therefore, it is conceivable to extract a predetermined number of detection points for each object by thinning out a plurality of detection points, and to recognize the shape of the object from the predetermined number of detection points. This makes it possible to reduce the amount of memory used for object recognition.

However, when the surface to be detected of the object is a curved surface that is convex toward the own vehicle, there is a limit on the number of detection points. There is a difference between As a result, there is concern that the accuracy of recognizing the shape of the object may decrease. In addition, if the convex curved surface of the object cannot be correctly recognized, there is a concern that when the own vehicle passes near the object, it cannot be recognized even if the own vehicle is too close to the object.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an object recognition apparatus capable of appropriately recognizing the shape of an object while reducing the amount of memory used.

Means for solving the above-mentioned problems is applied to a vehicle equipped with an object detection device for detecting an object. An object recognition device for extracting detection points and recognizing the shape of an object based on a predetermined number of the detection points, wherein the object recognition device detects the shape of an object based on the plurality of detection points detected in time series by the object detection device while the vehicle is running. a determination unit that determines whether the surface to be detected of the object is curved; and a mode change unit that changes the mode of object recognition performed by the predetermined number of detection points based on the determination result of the determination unit. And prepare.

Various objects exist around the vehicle while the vehicle is running, and the objects are detected by the object detection device. In this case, in order to reduce the amount of memory used, a predetermined number of detection points are extracted for each object from a plurality of detection points detected in time series by the object detection device, and the object is detected based on the predetermined number of detection points. It is conceivable to recognize shapes. However, in this case, if the surface to be detected of the object is a curved surface that protrudes toward the vehicle, there is a concern that the accuracy of object recognition may decrease.

In the above configuration, when the vehicle is running, it is determined that the surface to be detected of the object is curved based on the plurality of detection points detected in time series by the object detection device. The embodiment of the object recognition performed by the detection point of the score was changed. In this case, object recognition can be performed in different manners depending on whether the surface to be detected of the object is curved or not curved, and the shape of the object can be properly recognized. As a result, it is possible to properly recognize the shape of the object while reducing the amount of memory used.

FIG. 1 is an overall configuration diagram of a travel control system; FIG. 4 is a diagram showing a detection point cloud acquired from a planar object; The figure which shows the detection point group acquired from a curved-surface-shaped object. 4 is a flowchart showing a processing procedure of driving support processing; 4 is a flowchart showing a processing procedure of curvature radius calculation processing; The figure which shows the angle information in each detection point. The figure which shows the distance between detection points, a separation distance, and a curvature radius. The figure which shows an example of the division|segmentation in a driving assistance process. The figure which shows an example which applied driving assistance control to parking assistance control. The figure which shows an example which applied driving assistance control to parking assistance control.

An embodiment in which an object recognition device according to the present invention is applied to a vehicle-mounted travel control system 100 will be described below with reference to the drawings. The cruise control system 100 recognizes objects existing around the own vehicle, and based on the recognition result, performs driving support control (driving assist) while preventing the own vehicle 50 from getting too close to the object.

As shown in FIG. 1 , a cruise control system 100 according to this embodiment includes an ECU 10 as an object recognition device, sensors 20 and a controlled device 30 . The sensors 20 include, for example, a camera sensor 21, a radar sensor 22, a yaw rate sensor 23, a vehicle speed sensor 24, and the like.

The camera sensor 21 is, for example, a monocular camera, and images the surroundings of the vehicle. In this embodiment, the host vehicle 50 has a front camera, a rear camera, a right side camera, and a left side camera as the camera sensors 21 . The front camera is installed near the upper end of the windshield of the vehicle 50 or the like, and captures an image of a predetermined area in front of the vehicle. The rear camera is attached to the rear part of the own vehicle 50 (for example, near the trunk), and images a predetermined area behind the own vehicle. The right side camera is attached to the right side of the vehicle 50 (for example, near the door mirror) and captures an image of a predetermined area on the right side of the vehicle. The left side camera is attached to the left side of the vehicle 50 (for example, near the door mirror) and captures an image of a predetermined area on the left side of the vehicle. An image captured by each camera is output to the ECU 10 .

The radar sensor 22 is a ranging sensor that acquires distance information of objects existing around the vehicle by transmitting exploration waves to the vicinity of the vehicle and receiving reflected waves of the exploration waves. In this embodiment, the host vehicle 50 has a front radar, a rear radar, a right side radar, and a left side radar as the radar sensors 22 . The front radar is mounted on the front surface of the vehicle 50 so that its optical axis faces the front of the vehicle. The rear radar is attached to the rear surface of the own vehicle 50 so that its optical axis faces the rear of the own vehicle. The right side radar is mounted on the right side of the vehicle 50 so that its optical axis faces the right side of the vehicle. The left side radar is mounted on the left side of the vehicle 50 so that its optical axis faces the left side of the vehicle.

Each radar scans a directional transmission wave in the millimeter wave band, which is a search wave, at regular intervals. The azimuth of the object, the relative speed of the object with respect to the own vehicle 50, and the like are acquired as distance information. Each radar calculates and acquires the distance to the object from the time of transmission of the search wave and the time of reception of the reflected wave. Further, each radar calculates and acquires the azimuth of an object based on the phase difference between reflected waves received by a plurality of antennas. Furthermore, each radar calculates and obtains the relative velocity of the object from the frequency of the reflected wave reflected by the surface of the object, which is changed by the Doppler effect. Distance information captured by each radar is output to the ECU 10 . In this embodiment, the radar sensor 22 corresponds to the "object detection device".

The yaw rate sensor 23 is configured as a known yaw rate sensor that detects the turning angular velocity of the own vehicle 50 . The vehicle speed sensor 24 detects the rotational speed of the wheels, that is, the running speed of the vehicle 50 . Detection results by these sensors 23 and 24 are output to the ECU 10 .

The ECU 10 is a control device having a well-known microcomputer composed of CPU, ROM, RAM, flash memory, and the like. The ECU 10 detects objects around the vehicle. The ECU 10 detects objects based on images acquired from the camera sensor 21 and also detects objects based on distance information acquired from the radar sensor 22 .

Specifically, the ECU 10 detects an object existing in the image by performing well-known image processing such as template matching (pattern recognition) on the image acquired from the camera sensor 21 . In this embodiment, as a template for specifying an object, a whole-body dictionary is stored in which the features of each object are patterned. The ECU 10 calculates the distance to the object based on the ratio between the actual size of the object stored in the dictionary and the image size, which is the size of the object on the image. The ECU 10 also calculates the orientation of the object on the image with respect to the origin, with the center of the lower end of the image as the origin. The ECU 10 calculates the relative position and presence area of the object from the distance to the object and the azimuth of the object calculated from the image, and acquires this information as detection point information.

In addition, the ECU 10 calculates the relative position of the object, the existence area, etc. of the object based on the distance to the object and the orientation of the object included in the distance information obtained from the radar sensor 22, and obtains this information as detection point information.

The ECU 10 recognizes the shape of an object existing around the vehicle based on the detected plurality of detection points, and adjusts the driving force and braking force of the vehicle 50 using the controlled device 30 . The controlled device 30 has an accelerator device 31 as an acceleration device, a brake device 32 as a deceleration device, and a steering device 33 . The accelerator device 31 applies driving force to the host vehicle 50 according to a driver's accelerator operation or a control command from the ECU 10 . The braking device 32 applies a braking force to the own vehicle 50 according to a driver's braking operation or a control command from the ECU 10 . The steering device 33 controls the steering of the host vehicle 50 according to the driver's steering wheel operation or control commands from the ECU 10 .

By the way, as shown in FIG. 2, when there is a planar object B1 having a planar surface to be detected, which is the outer surface of the vehicle 50, as an object existing around the vehicle, the ECU 10 detects the planar object B1. Driving support control of the own vehicle 50 is performed based on a plurality of detection points for the shaped object B1. When the own vehicle 50 travels on the left side of the planar object B1, the ECU 10 detects the left side of the planar object B1 with the radar sensor 22 (right side radar) and detects a plurality of detection points (black circles, white circles). get. The ECU 10 thins out a predetermined number of detection points (black circles) for each object from a plurality of detection points detected in time series, and stores information on the predetermined number of detection points in the memory 11 . In this embodiment, three detection points are extracted for each object, and in the example shown in FIG. D11 and a detection point D6 detected between the detection points D1 and D2 are extracted. A detection point D6 is a detection point detected at a middle timing between the detection timing of the detection point D1 and the detection timing of the detection point D2.

Based on the detection points stored in the memory 11, the ECU 10 recognizes the shape of the planar object B1. In the example shown in FIG. 2, of the detection points D1, D6, and D11 stored in the memory 11, the detection point D1 and the detection point D6, which are obtained in different order, are connected by a line segment SE1. The detection point D11 is connected by a line segment SE2. The shape of the planar object B1 is recognized from the line segments SE1 and SE2. Then, the ECU 10 performs driving support control based on the recognized shape of the planar object B1, thereby causing the own vehicle 50 to travel while preventing the own vehicle 50 from getting too close to the planar object B1. For example, the ECU 10 uses the steering device 33 to assist steering while performing acceleration/deceleration control using the accelerator device 31 and the brake device 32 .

By limiting the number of detection points stored in the memory 11 to a predetermined number, the amount of memory used for object recognition is reduced. As shown in FIG. 2, when the object existing around the vehicle is a planar object B1, even if the number of detection points stored in the memory 11 is a predetermined number, the planar object B1 recognized by the detection points There is no difference between the shape of the object B1 and the actual shape of the planar object B1.

However, as shown in FIG. 3, when the object existing around the own vehicle is a curved object B2 having a curved surface to be detected that is convex toward the own vehicle 50, it is recognized by the detection point. A difference occurs between the shape of the curved object B2 and the actual shape of the curved object B2, and the recognition accuracy of the shape of the curved object B2 is lowered. Specifically, with the vehicle 50 as a reference, the line segments SE1 and SE2 are located on the far side of the surface to be detected of the curved object B2, so that a part of the curved object B2 on the vehicle 50 side is detected. , a portion (hereinafter referred to as a missing portion) KB that is not recognized by the ECU 10 is generated. For example, since the front and rear corners of a vehicle are convex curved surfaces, when the own vehicle 50 passes through the front and rear corners of another vehicle, the front and rear corners of the other vehicle are projected onto the surface to be detected. Then, a missing portion KB is generated near this corner. In this case, even if driving support control is performed based on the recognized shape of the curved object B2, there is a concern that the host vehicle 50 may come too close to the curved object B2.

Therefore, in the present embodiment, it is determined that the surface to be detected of the curved object B2 is curved based on a plurality of detection points detected in time series while the host vehicle 50 is running, and based on the determination result, Therefore, the embodiment of object recognition performed by a predetermined number of detection points is changed. Specifically, when the surface to be detected of the curved object B2 is determined to be curved, the predetermined number of detection points, that is, the number of detection points used for object recognition is increased. As a result, the shape of the curved object B2 can be properly recognized while reducing the amount of memory used.

FIG. 4 shows a flowchart of the driving assistance process of this embodiment. The ECU 10 repeatedly executes the driving support process at predetermined intervals.

When the driving support process is started, the detection point of the object is detected by the radar sensor 22 in step S11. By repeatedly performing the driving support process, the process of step S11 is repeatedly performed. As a result, when the object to be detected is the curved object B2 while the own vehicle 50 is running, a plurality of detection points thereof are detected in time series.

In step S13, a curvature radius calculation process is performed. In the curvature radius calculation process, the curvature radius R of the surface to be detected of the curved object B2 is calculated based on the plurality of detection points detected in step S11. In addition, in this embodiment, the process of step S13 corresponds to a "calculation part."

FIG. 5 shows a flowchart of the curvature radius calculation process. When the curvature radius calculation process is started, in step S21, a detection point group DA, which is a plurality of detection points used for calculating the curvature radius R, is set.

As shown in FIG. 6, the curved object B2 has curved surfaces at the corners (corners) of two planes perpendicular to each other, and as indicated by the arrow YA in FIG. When the curved object B2 is detected while traveling nearby, the plurality of detected detection points includes the detection points (black circles, white circles) at which the curved surface is detected and the detection points (x) at which the plane is detected. In step S21, among the plurality of detected detection points, the detection points where the curved surface is detected are set as the detection point group DA. The detection point information stored in the memory 11 includes angle information JB indicating the angle of the vector connecting each detection point and the position of the radar sensor 22 when the detection point was detected. As shown in FIG. 6, the angle information JB of the detection points where the plane is detected is substantially constant, whereas the angle information JB of the detection points where the curved surface is detected is different. In step S21, a detection point group DA is set based on the degree of variation of the angle information JB. In the example shown in FIG. 6, the detection points D1 to D11 are set as the detection point group DA.

In step S22, a start point and an end point are set from the detection point group DA set in step S21. Each of the start point and the end point is one of a plurality of detection points included in the detection point group DA, and is set so as to satisfy the following two conditions. (1) A start point is a detection point detected earlier than an end point. (2) At least one other detection point is included between the start point and the end point. In addition, in this embodiment, the start point and the end point correspond to "predetermined two points".

In step S23, the line segment distance W between the start point and the end point is calculated. FIG. 7 shows the line segment distance W when the start point is set at the detection point D2 and the end point is set at the detection point D10.

In step S24, the separation distance H is calculated. As shown in FIG. 7, a straight line LA passing through the start point and the end point is drawn, and the separation distance H is the distance from the detection point between the start point and the end point to the straight line LA. If a plurality of detection points are included between the start point and the end point, the longest distance from those detection points to the straight line LA is selected as the separation distance H. In the example shown in FIG. 7, the distance H from the detection point D6 to the straight line LA is selected.

In step S25, it is determined whether or not the separation distance H calculated in step S24 is zero. If the detection point used to calculate the separation distance H is located on the straight line LA and the separation distance H is zero, the process proceeds to step S29. On the other hand, if the detection point used to calculate the separation distance H is not positioned on the straight line LA, the process proceeds to step S26.

In step S26, the radius of curvature R is calculated. As shown in FIG. 7, when the separation distance H is not zero, a circle CA passing through the three detection points of the starting point, the end point, and the detection point used to calculate the separation distance H is determined, and the radius of this circle CA is the curvature. It has a radius R. In the circle CA, the line segment distance W indicates the length of the chord for the arc between the start and end points, and the separation distance H indicates the distance from the chord to the corresponding arc. In this case, the radius of curvature R can be expressed using the line segment distance W and the separation distance H as shown in Equation (1) below.

In step S27, it is determined whether or not the radius of curvature R calculated in step S26 is the minimum. At this time, if the curvature radius R calculated this time is smaller than the minimum value of the curvature radii R up to the last time, it is assumed that the current curvature radius R is the minimum. If the curvature radius R is the minimum, in step S28, the curvature radius R calculated this time is stored in the memory 11, the curvature radius R stored in the memory 11 is updated, and the process proceeds to step S29. On the other hand, if the radius of curvature R is not the minimum, the process proceeds to step S29 without storing the radius of curvature R calculated this time.

In step S29, it is determined whether or not the radius of curvature R has been calculated for all combinations of starting points and ending points. If the radius of curvature R has not been calculated for all combinations, the process returns to step S22. On the other hand, when the radius of curvature R is calculated for all combinations, the radius of curvature calculation process ends.

Returning to FIG. 4, when the radius of curvature calculation process ends, the process proceeds to step S14. In step S14, based on the plurality of detection points detected in step S11, it is determined that the surface to be detected of the curved object B2 is curved. Specifically, in step S28, it is determined whether or not the last updated radius of curvature R is smaller than a predetermined threshold value Rth. If the curvature radius R is smaller than the threshold value Rth, it is determined that the surface to be detected of the curved object B2 is curved, and the process proceeds to step S15.

In step S15, curved object B2 is regarded as two objects and divided, and the process proceeds to step S16. Specifically, of the surfaces to be detected of the curved object B2, the curved surface on which the detection point group DA is detected is divided into two curved surfaces. As a result, the detection point group DA is divided into a first detection point group DA1 corresponding to the divided first curved surface and a second detection point group DA2 corresponding to the second curved surface. It should be noted that in the present embodiment, the process of step S15 corresponds to the "mode changing section".

On the other hand, if the curvature radius R is larger than the threshold value Rth in step S14, it is determined that the curved object B2 does not have a curved surface, and the process proceeds to step S16 without dividing the detection point group DA. In addition, in this embodiment, the process of step S14 corresponds to a "judgment part."

In step S16, information on a predetermined number of detection points is stored in the memory 11 for each object. In the subsequent step S17, the shape of the object is recognized based on the predetermined number of detection points stored in the memory 11. FIG. Therefore, when the curved object B2 is regarded as two objects and divided in step S15, the shape of each divided object is recognized based on the predetermined number of detection points.

In step S18, based on the shape of the curved object B2 recognized in step S17, the driving support control of the own vehicle 50 is performed, and this process is once terminated.

Next, with reference to FIG. 8, the effect of dividing an object will be described. As shown in FIG. 8A, when the curved object B2 is not divided even though the surface to be detected of the curved object B2 is curved, the three detection points D1 extracted from the detection point group DA , D6 and D11 recognize the shape of the curved object B2. Therefore, the missing portion KB becomes relatively large.

In the present embodiment, as indicated by the dividing line LB in FIG. 8B, it is determined that the surface to be detected of the curved object B2 is curved, and the curved object B2 is regarded as two objects and divided. In this case, the detection point group DA is divided into a first detection point group DA1 and a second detection point group DA2. In the example shown in FIG. 8B, the detection point group DA including the detection points D1 to D11 includes a first detection point group DA1 including the detection points D1 to D6 and a second detection point group including the detection points D7 to D11. divided into DA2. Then, three detection points D1, D3, and D6 extracted from the first detection point group DA1 and three detection points D7, D9, and D11 extracted from the second detection point group DA2 are used to detect the curved object B2. Shapes are recognized. Therefore, the shape of the curved object B2 can be recognized from more detection points than when the object is not divided, and the missing portion KB can be reduced. As a result, even when driving support control is performed based on the recognized shape of the curved object B2, it is possible to appropriately prevent the host vehicle 50 from getting too close to the curved object B2.

According to this embodiment detailed above, the following effects can be obtained.

When the own vehicle 50 is running, it is determined that the surface to be detected of the object is curved based on a plurality of detection points detected in time series. The implementation of object recognition performed has been changed. In this case, object recognition can be performed in different manners depending on whether the surface to be detected of the object is curved or not curved, and the shape of the object can be properly recognized. As a result, it is possible to properly recognize the shape of the object while reducing the amount of memory used.

Based on the detected plurality of detection points, the radius of curvature R of the surface to be detected of the object is calculated, and if the calculated radius of curvature R is smaller than a threshold value Rth, it is determined that the surface to be detected is curved. I made it The radius of curvature R of a curved object is smaller than the radius of curvature R of a non-curved object. Therefore, by determining whether or not the calculated radius of curvature R is smaller than the threshold Rth, it is possible to properly determine that the object has a curved surface.

The curvature radius R is calculated using the line segment distance W and the separation distance H. A circle CA that passes through these three detection points is determined by the start point and end point used to calculate the line segment distance W and the detection points used to calculate the separation distance H. FIG. In the circle CA, the line segment distance W indicates the length of the chord with respect to the arc between the start point and the end point, and the separation distance H indicates the distance from the chord to the arc. Therefore, using the line segment distance W and the separation distance H, it is possible to calculate the radius of curvature R corresponding to the radius of the circle CA.

When the surface to be detected of the object is curved, the radius of curvature R of the surface to be detected may not be uniform. In this case, the smaller the curvature radius R, the more difficult it is to properly recognize the shape of the object. In this regard, the smallest radius of curvature R among the calculated multiple radiuses of curvature R is selected as the radius of curvature R of the surface to be detected of the object. Therefore, it is possible to appropriately recognize the shape of the object even in a portion having the smallest radius of curvature in the surface to be detected of the object.

When the surface to be detected of the object is determined to be curved, the number of detection points, which is a predetermined number, is increased compared to when the surface to be detected is determined not to be curved. As a result, it is possible to increase the number of detection points required for recognizing an object having a curved surface to be detected, and to properly recognize the shape of the object.

When it is determined that the surface to be detected of the object is curved, the object to be detected is regarded as a plurality of objects and divided into multiple objects. Made to recognize the shape. As a result, when an object is divided into N pieces (N is a natural number of 2 or more), the shape of the object before division is recognized by N times the predetermined number of detection points, and an object having a curved surface to be detected can be detected. can be properly recognized.

The driving support control of the present embodiment is useful when the own vehicle 50 is parked in the parking area PA near the parked vehicle (another vehicle) 60, as described below. In a vehicle, it is conceivable that the corners between the front or rear surface and the side surfaces are curved. In other words, it is conceivable that the outer peripheral surface of the vehicle switches between a flat surface, a curved surface, and a flat surface near the corners of the vehicle, and the radius of curvature R changes accordingly. When the radius of curvature R is changing, it is more difficult to properly recognize the shape of the object in a portion where the radius of curvature R is smaller.

In this regard, according to the driving support control of the present embodiment, by selecting the smallest curvature radius R among the plurality of curvature radii R calculated near the corner as the curvature radius R near the corner, the parking vehicle The corner shape of 60 can be correctly recognized. As a result, the possibility of contact with the parked vehicle 60 can be reduced when the own vehicle 50 is parked in the parking area PA.

Specifically, as shown in FIG. 9 , there are cases where own vehicle 50 parks in parking area PA located in front of parked vehicle 60 after passing by the side of parked vehicle 60 . In this case, the ECU 10 detects the front right portion 60A of the parked vehicle 60 when the front surface of the own vehicle 50 passes near the front right portion 60A of the parked vehicle 60 as indicated by an arrow YB in FIG. is curved. As a result, when the rear surface of the own vehicle 50 passes near the right front portion 60A of the parked vehicle 60, or when the own vehicle 50 moves backward toward the parked vehicle 60 as indicated by the arrow YC in FIG. In addition, the ECU 10 can appropriately prevent the host vehicle 50 from approaching the right front portion 60A of the parked vehicle 60 too much.

Further, as shown in FIG. 10 , there are cases where own vehicle 50 parks in parking area PA located on the side of parked vehicle 60 after passing in front of parked vehicle 60 . In this case, the ECU 10 detects the left front portion 60B of the parked vehicle 60 when the own vehicle 50 passes in front of the parked vehicle 60 as indicated by an arrow YD in FIG. 10, and the left front portion 60B is curved. to judge. As a result, as shown by arrow YE in FIG. 10, when own vehicle 50 turns right while traveling backward and passes near left front portion 60B of parked vehicle 60, ECU 10 detects that own vehicle 50 is parked vehicle 60. It is possible to appropriately prevent the vehicle from approaching too much to the left front portion 60B.

(Other embodiments)
Each of the above embodiments may be modified and implemented as follows.

- In the above-described embodiment, an example is shown in which an object is divided into two parts based on the determination that the surface to be detected of the object is curved, but the number of divisions may be three or more. In this case, the number of divisions may be variably set according to the radius of curvature R of the surface to be detected of the object. For example, the smaller the radius of curvature R, the greater the number of divisions.

- When the object is divided, the ECU 10 may recalculate the radius of curvature R for each divided object. Then, when the recalculated radius of curvature R is smaller than the threshold value Rth, the divided object may be divided again.

- In the above-described embodiment, the formula (1) is used to calculate the radius of curvature R, but the radius of curvature R may be calculated using a map or the like.

- In the above-described embodiment, the smallest radius of curvature R is selected from among the plurality of radiuses of curvature R calculated on the surface to be detected of the object, but the present invention is not limited to this. For example, an average value of a plurality of curvature radii R including the smallest curvature radius R may be selected. For example, when the starting point is set to the detection point D2 and the end point is set to the detection point D10 and the curvature radius R is the minimum, the curvature radius R and the start point are set to the detection point D1, and the end point is detected. The average value of the curvature radius R when the point D9 is set and the curvature radius R when the start point is set at the detection point D3 and the end point is set at the detection point D11 is calculated, and this average value is selected. can be

- The imaging device is not limited to a monocular camera, and may be a stereo camera. A device using transmitted waves and reflected waves is not limited to the radar sensor 22, and may be a laser sensor.

- The object recognition apparatus and method described in the present disclosure can be performed by a dedicated computer provided by configuring a processor and memory programmed to perform one or more functions embodied by a computer program. , may be implemented. Alternatively, the object recognition apparatus and techniques described in this disclosure may be implemented by a dedicated computer provided by configuring a processor with one or more dedicated hardware logic circuits. Alternatively, the object recognition apparatus and techniques described in this disclosure are a combination of a processor and memory programmed to perform one or more functions and a processor configured by one or more hardware logic circuits. may be implemented by one or more dedicated computers configured by The computer program may also be stored as computer-executable instructions on a computer-readable non-transitional tangible recording medium.

10 -- ECU, 22 -- radar sensor.

Claims (7)

Applied to a vehicle equipped with an object detection device (22) for detecting an object, extracting a predetermined number of detection points per object from among a plurality of detection points detected in time series by the object detection device (22) while the vehicle is running. , an object recognition device (10) for recognizing an object shape based on the predetermined number of detection points,
a determination unit that determines whether a surface to be detected of an object is curved based on the plurality of detection points detected in time series by the object detection device while the vehicle is running;
a mode changing unit that changes a mode of object recognition performed by the predetermined number of detection points based on the determination result of the determining unit;
An object recognition device comprising: a calculation unit that calculates a radius of curvature of the surface to be detected based on the plurality of detection points detected by the object detection device;
2. The object recognition device according to claim 1, wherein the determination unit determines that the surface to be detected is curved when the radius of curvature calculated by the calculation unit is smaller than a predetermined threshold. The calculator draws a straight line passing through two predetermined points among the plurality of detection points, and calculates a distance between the predetermined two points and a distance from another detection point existing between the two points to the straight line. 3. The object recognition device according to claim 2, wherein said radius of curvature is calculated based on . The calculation unit draws a straight line passing through two points in a plurality of combinations of the plurality of detection points, calculates a plurality of the curvature radii corresponding to each of the straight lines, and 4. The object recognition device according to claim 3, wherein the small radius of curvature is selected as the radius of curvature of the surface to be detected. An object recognition device that, when passing near another vehicle (60), uses the vicinity of the front and rear corners of the other vehicle as the surface to be detected,
The calculating unit draws a straight line passing through two points in a plurality of combinations of the plurality of detection points near the corner, calculates a plurality of the curvature radii corresponding to each of the straight lines, and calculates the plurality of the calculated radii of curvature. 4. The object recognition device according to claim 3, wherein the smallest radius of curvature is selected as the radius of curvature near the corner. The mode changing unit is configured such that when the determining unit determines that the surface to be detected has a curved surface, the detection surface having the predetermined number of points is higher than a case where the surface to be detected is determined not to have a curved surface. The object recognition device according to any one of claims 1 to 5, wherein the number of points is increased. When the determination unit determines that the surface to be detected has a curved surface, the aspect changing unit divides the object to be detected by considering the object to be detected as a plurality of objects, and for each object after division, 6. The object recognition device according to claim 1, wherein the object shape is recognized based on the predetermined number of detection points.

Images