JP2021135191A - Object detection device - Google Patents

Abstract

To provide an object detection device that can more accurately perform object detection.SOLUTION: A first object recognition unit 31a prepares a first grid map showing a position where an object exists around an own vehicle by setting the prescribed number of votes with respect to a principal grid serving as a grid corresponding to a detection coordinate showing a position where an object included in sensing information on a range-finding sensor 10 is detected. A second object recognition unit 31b prepares a second grid map showing the position where the object exists around the own vehicle by setting the prescribed number of votes with respect to a principal grid serving as a grid corresponding to a detection coordinate showing a position where an object included in sensing information on an on-vehicle camera 20 is detected. An object integration unit 32 prepares an integration grid map obtained by integrating the first grid map and second grid map. An object identification unit 33 that identifies a grid in which a total value of the number of votes in the integration grid map exceeds a prescribed threshold as a grid where the object exists.SELECTED DRAWING: Figure 1

Description

The present invention relates to an object detection device that detects an object existing in the vicinity of its own vehicle (hereinafter referred to as its own vehicle).

Conventionally, an object existing around the own vehicle is detected by using sonar or the like, and the detection result is used for vehicle control such as automatic parking. For example, as an object detection device, there is one proposed in Patent Document 1.

This device includes a vote number calculation unit, a cluster setting unit, a grid discrimination unit, and a moving object discrimination unit. The vote number calculation unit calculates the number of votes, which is the number including the detection points of the distance measurement sensor, for each grid of the grid map based on the detection result of the detection points by the distance measurement sensor such as the radar sensor. The cluster setting unit sets the cluster, which is the object to be detected, on the grid map by clustering the detection points based on the detection result of the detection points by the distance measuring sensor. The grid discriminating unit discriminates between the static grid and the moving grid from the grid occupied by the cluster based on the total value of the number of votes for each grid calculated a plurality of times. The moving object determination unit determines whether the cluster is a moving object or a stationary object based on the arrangement of the stationary grid and the moving grid included in the grid occupied by the cluster. By using these, it is possible to detect an object existing around the own vehicle and to determine whether the cluster, which is the object to be detected, is a moving object or a stationary object.

JP-A-2017-187422

However, accurate object detection cannot be performed by object detection using only a ranging sensor such as a radar sensor. Specifically, the distance measuring sensor is good at measuring the linear distance from the distance measuring sensor to the object (hereinafter, simply referred to as the distance to the object), but is not good at measuring the width of the object. Therefore, the object width cannot be measured accurately, and the object can not be detected accurately.

In view of the above points, it is an object of the present invention to provide an object detection device capable of performing more accurate object detection.

In order to achieve the above object, the invention according to claim 1 of the present application detects an object (100) existing around the own vehicle (V) by using a grid map in which the periphery of the own vehicle (V) is divided into a plurality of grids. It is an object detection device, and the sensing information of the distance measuring sensor (10) that measures the position of an object by outputting the exploration wave and acquiring the reflected wave reflected by the object is input to the grid map. By setting a predetermined number of votes in the main grid, which is the grid corresponding to the detection coordinates indicating the detected position of the object included in the sensing information of the ranging sensor, the object existing around the own vehicle The sensing information of the first object recognition unit (31a) that creates the first grid map (M1) showing the position of the vehicle and the in-vehicle camera (20) that images the surroundings of the own vehicle is input, and the grid map is used. By setting a predetermined number of votes in the main grid, which is the grid corresponding to the detection coordinates indicating the position where the object included in the sensing information of the in-vehicle camera was detected, the position of the object existing around the own vehicle was shown. An integrated grid map (Mc) that integrates the second object recognition unit (31b) that creates the second grid map (M2), the number of votes for each grid in the first grid map, and the number of votes for each grid in the second grid map. ), And an object identification unit (33) that specifies a grid in which the total number of votes in the integrated grid map exceeds a predetermined threshold as a grid on which an object exists. There is.

In this way, the first object recognition unit and the second object recognition unit create the first grid map and the second grid map based on the sensing information of the distance measuring sensor and the vehicle-mounted camera. Then, the object integration unit creates an integrated grid map by integrating the first grid map and the second grid map. That is, the first grid map, which may not be sufficient for measuring the object width, which the distance measuring sensor is not good at, and the second grid map, which is based on the sensing information of the in-vehicle camera, which is good at measuring the object width, are combined. We are doing sensor fusion. In this way, by creating an integrated grid map that considers the characteristics of various types of sensors, it is possible to improve the detection accuracy of objects, and false detections such as when object detection is performed with only one ranging sensor. Can be suppressed.

The reference reference numerals in parentheses attached to each component or the like indicate an example of the correspondence between the component or the like and the specific component or the like described in the embodiment described later.

It is a block diagram of the vehicle control system which concerns on 1st Embodiment. It is the figure which showed the image of the object detection by the sonar and the in-vehicle camera when the cylindrical object exists in front of the own vehicle. It is a schematic diagram of the 1st grid map which converted the detection coordinates of an object by a sonar into a grid map. It is a schematic diagram of the 2nd grid map which converted the detection coordinates of an object by an in-vehicle camera into a grid map. It is a schematic diagram of the integrated grid map which integrated the 1st grid map and the 2nd grid map. It is a flowchart of an object detection process.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In each of the following embodiments, parts that are the same or equal to each other will be described with the same reference numerals.

(First Embodiment)
A vehicle control system provided with the object detection device according to the present embodiment will be described. Here, as the vehicle control, the one that executes the parking support control that performs automatic parking will be described as an example, but if the vehicle control is performed based on the object detection result by the object detection device, other control is used. It does not matter if it does.

As shown in FIG. 1, the vehicle control system 1 includes a sonar 10, an in-vehicle camera 20, an electronic control device (hereinafter referred to as an ECU) 30 constituting an object detection device, various actuators 40, and a display 50. ing.

The sonar 10 corresponds to a distance measuring sensor. The sonar 10 outputs ultrasonic waves as exploration waves and acquires the reflected waves at predetermined sampling periods, such as relative velocity and relative distance to the target, and the azimuth angle at which the target exists. The position measurement result is sequentially output to the ECU 30 as sensing information. When the sonar 10 detects an object, it outputs the detection coordinates, which are the coordinates of the detected position, in the sensing information. The detection coordinates of the object are specified using the moving triangulation method, and since the distance to the object changes as the vehicle moves, it is specified based on the change in the measurement result for each sampling cycle. There is.

Although only one sonar 10 is shown here, it is actually provided at a plurality of locations with respect to the vehicle. Examples of the sonar 10 include front sonars arranged side by side on the front and rear bumpers in the left-right direction of the vehicle, and side sonars arranged at lateral positions of the vehicle. The sampling cycle of the sonar 10 is set according to the arrangement location of the sonar 10. For example, the front sonar has a first cycle, and the side sonar has a second cycle equal to or less than the first cycle.

Here, the sonar 10 is taken as an example of the distance measuring sensor, but examples of the distance measuring sensor include a millimeter wave radar and LIDAR (Light Detection and Ranging). The millimeter wave radar performs measurement using millimeter waves as exploration waves, and LIDAR performs measurements using laser light as exploration waves, both of which output exploration waves within a predetermined range such as in front of a vehicle. , The measurement is performed within the output range.

The in-vehicle camera 20 is a peripheral monitoring camera that captures a predetermined range around the own vehicle, captures a peripheral image of the own vehicle, analyzes the captured data, and when an object is detected, senses the detection coordinates. Is included in and output to the ECU 30. The in-vehicle camera 20 includes, but is limited to, a front camera that captures the front of the vehicle, a rear camera that captures images of the rear and left and right sides of the vehicle, a left-side camera, and a right-side camera. is not it. The sampling cycle, which is the cycle for capturing images of the vehicle-mounted camera 20, is, for example, a third cycle equal to or higher than the first cycle. Also in the vehicle-mounted camera 20, the detection coordinates of the object are specified by using the moving triangulation method.

The ECU 30 performs various controls for detecting an object and realizing parking support control for automatic parking, and is composed of a microcomputer equipped with a CPU, ROM, RAM, I / O, and the like. There is. Specifically, the ECU 30 detects an object existing around the own vehicle by creating a grid map based on the sensing information of the sonar 10 and the vehicle-mounted camera 20, and selects the detection coordinates of the object. Further, in the case of the present embodiment, the ECU 30 performs parking support control based on the detection result of the object, and controls various actuators 40 based on the parking support control to perform automatic parking. Specifically, the ECU 30 is configured to include a first object recognition unit 31a, a second object recognition unit 31b, an object integration unit 32, an output selection unit 33, and a control unit 34 as functional units that execute various controls. ing.

Hereinafter, the details of each functional unit constituting the ECU 30 will be described. However, the first object recognition unit 31a, the second object recognition unit 31b, and the object integration unit 32 are created by each unit assuming the existence of an object. This will be explained with reference to the grid map. As an example of the case where an object exists in the vicinity of the own vehicle, it is assumed that the cylindrical object 100 exists in front of the own vehicle V as shown in FIG. The O mark in FIG. 2 shows the state of detection of the object 100 by the sonar 10, and the X mark in FIG. 2 shows the state of detection of the object 100 by the in-vehicle camera 20. 3A to 3C show a part of the grid map created by each of the first object recognition unit 31a, the second object recognition unit 31b, and the object integration unit 32 in the situation shown in FIG.

Based on the sensing information input from the sonar 10, the first object recognition unit 31a uses a grid map Map in which the periphery of the vehicle V is divided into a plurality of grids G, as shown in FIG. 3A, to surround the vehicle V. A grid map M1 showing the positions of the objects 100 of the above is created. The grid map M1 is a map in which the periphery of the own vehicle V is divided into a plurality of grids G in a grid pattern in a plan view. Although only a part thereof is shown in FIG. 3A, the first object recognition unit 31a creates, for example, a grid map M1 having a range of several m × several m larger than the own vehicle V centered on the own vehicle V. The first object recognition unit 31a uses, for example, a square of several to several tens of centimeters as one grid G, that is, one unit area, and the grid map M1 is divided into a grid in the front-rear direction of the own vehicle V and the left-right direction of the own vehicle V. Is being created. Then, when the sensing information is input from the sonar 10, the first object recognition unit 31a sets the number of votes in each grid G on the grid map Map, and the grid reflects the detection coordinates included in the sensing information of the sonar 10. Map M1 is being created. The number of votes is a value that serves as an index indicating the existence probability of the object 100, and here, the larger the number of votes, the higher the existence probability of the object 100. In the following description, the grid map M1 created by the first object recognition unit 31a is referred to as the first grid map M1.

Specifically, as shown in FIG. 3A, the grid G corresponding to the detection coordinates included in the sensing information from the sonar 10 is the main grid Gm, and the grid G located adjacent to the main grid Gm in the width direction of the object 100. Is the width direction grid Gw. Further, the grid G located behind the main grid Gm and the width direction grid Gw, that is, on the side far from the sonar 10, is referred to as a rear grid Gb, and other than the main grid Gm, the width direction grid Gw, and the rear grid Gb is referred to as another grid Go. The main grid Gm has the first vote, the width direction grid Gw and the rear grid Gb have the second vote less than the first vote, and the other grid Go has the third vote less than the second vote. Set the number of votes for each grid G. For example, the number of positive votes is set as the number of first votes, which is +1 here. 0 is set for the number of second votes. The number of negative votes is set to -1 for the third number of votes.

The reason why the width direction grid Gw is set to have a larger number of votes than the other grid Go is that the sonar 10 is not good at measuring the object width. There is a possibility that the object 100 exists deviating from the detection coordinates of the sonar 10 in the width direction. Assuming this, a width direction grid Gw located in the width direction with respect to the main grid Gm is provided, and the number of votes for the width direction grid Gw is set to be larger than that of other grid Gos having a low existence probability of the object 100. The reason why the rear grid Gb is set to have a larger number of votes than the other grid Go is that the sonar 10 cannot measure the position behind the main grid Gm and the object 100 may exist. Also in this case, assuming that the object 100 may exist in the rear grid Gb, the existence probability of the object 100 is low for the rear grid Gb located behind the main grid Gm. The number of votes is higher than that of Go.

Based on the sensing information input from the vehicle-mounted camera 20, the second object recognition unit 31b uses a grid map Map in which the periphery of the vehicle V is divided into a plurality of grids G to determine the position of the object 100 around the vehicle V. The shown grid map M2 is created. Although only a part of the grid map M2 is shown in FIG. 3B, the overall size of the grid map M2 and the size of each grid G are the same as those of the first grid map M1 created by the first object recognition unit 31a. When the sensing information is input from the vehicle-mounted camera 20, the second object recognition unit 31b also sets the number of votes for each grid G on the grid map Map, and the grid reflects the detection coordinates included in the sensing information of the vehicle-mounted camera 20. Map M2 is being created. In the following description, the grid map M2 created by the second object recognition unit 31b is referred to as the second grid map M2.

Specifically, as in the case of the sonar 10, the grid G corresponding to the detection coordinates included in the sensing information of the vehicle-mounted camera 20 is the main grid Gm, and the grid located behind the main grid Gm, that is, on the side far from the vehicle-mounted camera 20. G is the rear grid Gb. In addition, the grid G located in front of the main grid Gm, that is, on the side closer to the vehicle-mounted camera 20, is referred to as the front grid Gf. In addition, other grids other than the main grid Gm, the front grid Gf, and the rear grid Gb are used as other grid Go. The main grid Gm has the 4th vote, the front grid Gf and the rear grid Gb have the 5th vote less than the 4th vote, and the other grid Go has the 6th vote less than the 5th vote. Set the number of votes for grid G. For example, the number of positive votes for the fourth number of votes is set to +1 here. The number of fifth votes is set to 0 here. The number of votes for the sixth vote is negative, which is set to -1 here.

The reason why the front grid Gf is set to have a larger number of votes than the other grid Go is that the vehicle-mounted camera 20 is not good at measuring the distance to the object 100. There is a possibility that the object 100 is present at a position shifted forward from the coordinates detected by the vehicle-mounted camera 20. Assuming this, a front grid Gf located in front of the main grid Gm is provided, and the number of votes for the front grid Gf is set to be larger than that of other grid Gos having a low probability of existence of the object 100. The reason why the rear grid Gb is set to have a larger number of votes than the other grid Go is that the vehicle-mounted camera 20 cannot photograph the position behind the main grid Gm and the object 100 may exist. Also in this case, assuming that the object 100 may exist in the rear grid Gb, the existence probability of the object 100 is low for the rear grid Gb located behind the main grid Gm. The number of votes is higher than that of Go.

The resolution of the sonar 10 and the in-vehicle camera 20 is high, and the grid map Map can be configured with a finer grid G. However, in order to reduce the memory capacity and the processing load, the data can be increased by increasing the grid size. May be thinned out to make it rough.

As shown in FIG. 3C, the object integration unit 32 integrates the first grid map M1 created by the first object recognition unit 31a and the second grid map M2 created by the second object recognition unit 31b. Hereafter, Mc is created (hereinafter referred to as an integrated grid map). Since the overall size and the number of grids of the first grid map M1 and the second grid map M2 are the same, the number of votes of the grid G at the same position is added to obtain one integrated grid map Mc. The number of votes of each grid G in the integrated grid map Mc is the total number of votes of the grid G at the same position in each of the first grid map M1 and the second grid map M2.

The output selection unit 33 corresponds to the object identification unit, identifies the grid Ge in which the object exists from the integrated grid map Mc created by the object integration unit 32, and detects the detection coordinates of the object 100 detected to exist in the grid Ge. Is output to the control unit 34. For example, the output selection unit 33 stores the threshold value of the number of votes, and if the number of votes of each grid G in the integrated grid map Mc is less than the threshold value, the object 100 is determined to be a non-existent grid Gn, and the value is equal to or higher than the threshold value. If there is, it is determined that the grid Ge in which the object 100 exists. Here, the threshold value is set to 1, the grid G in which the total value of the number of votes is 1 or more is specified as the grid Ge in which the object 100 exists, and the grid G having the total value less than 1 is specified as the object 100. Is specified as the existing grid Gn. Then, of the first grid map M1 or the second grid map M2, the detection coordinates of the object 100 included in the grid Ge determined to have the object 100 here are output to the control unit 34 as the detection coordinates of the actual object 100. do.

At this time, the coordinates of the grid Ge in which the object 100 exists can be output to the control unit 34. However, here, as for the detection coordinates of the object 100 detected to exist in the grid Ge, the detection coordinates indicated by the sensing information of the sonar 10 and the vehicle-mounted camera 20 are output as they are. By doing so, it is not necessary to create the coordinates of the grid Ge in which the object 100 exists, and the memory capacity and the processing load of the ECU 30 can be reduced. Further, since the resolution of the sonar 10 and the in-vehicle camera 20 is high and the data of the grid map Map is coarser, it is more detailed data to output the detection coordinates indicated by the sensing information of the sonar 10 and the in-vehicle camera 20 as they are. ..

The control unit 34 performs parking support control for automatic parking. In the parking support control, parking route generation from the current position to the planned parking position, route tracking control for moving the own vehicle V along the generated parking route, and display control related to parking support control are executed. The parking route is generated based on the detection coordinates of the object 100 transmitted from the output selection unit 33. For example, the control unit 34 generates a parking route so as to avoid the detection coordinates of the object 100. The route follow-up control is a control that performs vehicle motion control such as acceleration / deceleration control and steering control in order to move the own vehicle V along the parking route generated by the parking route generation. Further, when the object 100 is newly detected in the parking route during the parking support control, the control unit 34 regenerates the parking route or the own vehicle in the route follow-up control until the object 100 moves. By stopping the movement of V, contact with the object 100 can be avoided.

The display control informs the driver that the parking support control is in progress by displaying the parking support control, the state of the parking route and the own vehicle V, for example, which position in the parking route is being moved. It controls the display of. The control unit 34 controls the display 50 as display control.

In the present embodiment, the ECU 30 is configured by one, and each functional unit is provided in the ECU 30, but a plurality of ECUs may be combined. For example, the first object recognition unit 31a and the second object recognition unit 31b are separate ECUs, the first object recognition unit 31a is unitized with the sonar 10, and the second object recognition unit 31b is an in-vehicle camera 20. It may be unitized as. Further, only the control unit 34 may be composed of one or a plurality of ECUs. Examples of various ECUs in which the control unit 34 is composed of a plurality of ECUs include a steering ECU that performs steering control, a power unit control ECU that performs acceleration / deceleration control, and a brake ECU.

Specifically, although not shown, the control unit 34 acquires detection signals output from each sensor such as an accelerator position sensor, a brake pedal force sensor, a steering angle sensor, and a shift position sensor mounted on the vehicle V. There is. Then, the control unit 34 detects the state of each unit from the acquired detection signal, and outputs control signals to various actuators 40 in order to move the own vehicle V following the parking path.

The various actuators 40 are various travel control devices related to the travel and stop of the own vehicle V, and include an electronically controlled throttle 41, a brake actuator 42, an EPS (Electric Power Steering) motor 43, a transmission 44, and the like. These various actuators 40 are controlled based on the control signals from the control unit 34, and the traveling direction, steering angle, and control drive torque of the own vehicle V are controlled. As a result, parking support control including route tracking control in which the own vehicle V is moved according to the parking route and parked at the planned parking position is realized.

The display 50 is composed of a display (hereinafter, referred to as a navigation display) 51 in a navigation system, a display (hereinafter, referred to as a meter display) 52 provided in a meter, and the like. Here, during the parking support control, the navigation display 51 and the meter display 52 are set to display the parking route or the position in the parking route where the vehicle V is moving.

As described above, the vehicle control system 1 according to the present embodiment is configured. Next, the operation of the vehicle control system 1 according to the present embodiment will be described. The parking support control performed by the ECU 30 of the vehicle control system 1 includes parking route generation, route tracking control, and display control. However, since these are as described above, the description thereof will be omitted. Here, object detection including grid map creation used for parking route generation and output of object detection coordinates will be described with reference to FIG. FIG. 4 is a flowchart of the object detection process executed by the ECU 30. This object detection process is executed every predetermined control cycle, for example, every cycle shorter than the sampling cycle of the sonar 10 or the vehicle-mounted camera 20 when the start switch of the own vehicle V such as the ignition switch is turned on. The first object recognition unit 31a, the second object recognition unit 31b, the object integration unit 32, and the output selection unit 33 of the ECU 30 function as processing units to perform this object detection processing.

First, the process of step S100 and the process of step S105 are performed in parallel for each predetermined control cycle. In step S100, it is determined whether or not sonar recognition has been performed, that is, whether or not the sensing information output from the sonar 10 includes information indicating that the object 100 has been recognized. Then, if the sensing information of the sonar 10 does not include information indicating that the object 100 has been recognized, the process of step S100 is repeated, and if it is included, the process proceeds to step S110.

On the other hand, in step S105, it is determined whether or not the vehicle-mounted camera recognition has been performed, that is, whether or not the sensing information output from the vehicle-mounted camera 20 includes information indicating that the object 100 has been recognized. Then, if the sensing information of the vehicle-mounted camera 20 includes information indicating that the object 100 has been recognized, the process proceeds to step S125, and if not, the process of step S105 is repeated.

When the object 100 is detected by the sonar 10 or the vehicle-mounted camera 20, the first grid map M1 or the second grid map M2 is created based on the object 100, the integrated grid map Mc is updated, and the detection coordinates of the object 100 are controlled by the control unit 34. Need to tell. Therefore, in steps S100 and S105, it is determined that the object 100 is detected by either the sonar 10 or the vehicle-mounted camera 20.

Then, when the object 100 is detected by the sonar 10, each process of steps S110 to S120 is executed. In step S110, the grid G corresponding to the detection coordinates included in the sensing information of the sonar 10 is searched. This grid G becomes the main grid Gm in this control cycle. In step S115, the number of votes of the main grid Gm searched in step S110 is added. That is, the number of votes of the main grid Gm in the current control cycle is added to the first grid map M1 created up to the previous control cycle. Here, the number of votes is added by +1. In step S120, the number of votes is added to other grid Go including the front of the main grid Gm, that is, other grid Go other than the main grid Gm, the width direction grid Gw, and the rear grid Gb in the first grid map. Here, the number of votes is added by -1. As described above, since the sonar 10 is not good at measuring the object width, 0 is added to the width direction grid Gw in step S120, that is, it is not changed. In this way, the first grid map M1 is updated.

On the other hand, when an object is detected by the vehicle-mounted camera 20, each process of steps S125 to S135 is executed. In steps S125 to S135, the same processing as in steps S110 to S120 is performed based on the detection coordinates included in the sensing information of the vehicle-mounted camera 20. However, in step S135, the number of votes is added by -1 to the other grid Go including the side of the main grid Gm, that is, to the other grid Go other than the main grid Gm, the front grid Gf, and the rear grid Gb with respect to the second grid map M2. do. As described above, since the vehicle-mounted camera 20 is not good at measuring the distance to the object 100, in step S135, 0 is added to the front grid Gf, that is, it is not changed. In this way, the second grid map M2 is updated.

Subsequently, the process proceeds to step S140, and the first grid map M1 and the second grid map M2 are integrated to create an integrated grid map Mc. As a result, the integrated grid map Mc is updated. Then, the process proceeds to step S145.

In step S145, the output selection process is performed. That is, the grid Ge in which the object 100 exists is specified from the integrated grid map Mc created in step S140, and the detection coordinates of the object 100 detected to exist in the grid Ge are output to the control unit 34. Specifically, the grid G in which the total number of votes is equal to or greater than the threshold value, and here is 1 or more, is specified as the grid Ge in which the object 100 exists from the integrated grid map Mc. Then, among the detection coordinates included in the sensing information of the sonar 10 and the vehicle-mounted camera 20, the detection coordinates corresponding to the grid Ge in which the object 100 exists are output to the control unit 34.

In this way, the object detection including the grid map creation used for the parking route generation and the output of the detection coordinates of the object 100 is completed, and the detection coordinates of the object 100 are transmitted to the control unit 34. Of the processes described here, the processes of steps S100 and S110 to S120 are executed by the first object recognition unit 31a, and the processes of steps S105 and S125 to S135 are executed by the second object recognition unit 31b. Further, the processing of step S140 is executed by the object integration unit 32, and the processing of step S145 is executed by the output selection unit 33.

Then, when the detection coordinates of the object 100 are transmitted to the control unit 34, the control unit 34 executes parking route generation, route tracking control, and display control as parking support control. As a result, after the parking route is created so as to avoid the object 100, various actuators 40 are controlled to perform automatic parking following the parking route, and the display 50 displays which of the parking route and the parking route. It is displayed at the position whether or not the own vehicle V is moving.

As described above, in the vehicle control system 1 of the present embodiment, the first object recognition unit 31a and the second object recognition unit 31b have the first grid map M1 and the first grid map M1 based on the sensing information of the sonar 10 and the vehicle-mounted camera 20. The second grid map M2 is created. Then, the object integration unit 32 integrates the first grid map M1 and the second grid map M2 to create an integrated grid map Mc. That is, the sensor is a combination of the first grid map M1 based on the sensing information of the sonar 10 which is not good at measuring the object width and the second grid map M2 based on the sensing information of the in-vehicle camera 20 which is good at measuring the object width. We are doing fusion. In this way, by creating an integrated grid map Mc that considers the characteristics of various types of sensors, it is possible to improve the detection accuracy of the object 100, as in the case of performing object detection with only one distance measuring sensor. It is possible to suppress false positives.

Specifically, assume the situation shown in FIG. 2 described above, that is, the case where the cylindrical object 100 exists in front of the own vehicle V.

As shown by the O mark in FIG. 2, when the cylindrical object 100 is detected by the sonar 10, the sonar 10 is not good at measuring the object width, so that the detection coordinates of the sonar 10 are the object 100. It can be something that does not correspond to the actual width. FIG. 2 shows a case where the detection coordinates of the sonar 10 deviate in the width direction of the object 100 and each detection coordinate spreads over a wider range than the actual width of the object 100. However, the accuracy of detecting the distance to the object of the sonar 10 is high. Therefore, the detection coordinates of the sonar 10 are almost correct with respect to the distance to the object.

Further, as shown by the X mark in FIG. 2, when the in-vehicle camera 20 detects the cylindrical object 100, the in-vehicle camera 20 is good at measuring the object width, so that the detection coordinates of the in-vehicle camera 20 Is approximately corresponding to the actual width of the object 100. However, the vehicle-mounted camera 20 is not good at measuring the distance to the object 100. Therefore, the detection coordinates of the vehicle-mounted camera 20 may not correspond to the distance to the object 100. FIG. 2 shows a case where the detection coordinates of the vehicle-mounted camera 20 are shorter than the distance to the actual object 100.

Therefore, when the first object recognition unit 31a creates the first grid map M1 based on the detection coordinates included in the sensing information of the sonar 10, the map becomes as shown in FIG. 3A. Looking at the portion where the number of votes is +1 in the first grid map M1, the distance to the object 100 is accurate, but the width direction of the object 100 is not so accurate. Further, when the second object recognition unit 31b creates the second grid map based on the detection coordinates included in the sensing information of the vehicle-mounted camera 20, the map becomes as shown in FIG. 3B. Looking at the portion where the number of votes is +1 in the second grid map M2, the map is accurate in the width direction of the object 100, but the distance to the object 100 is not accurate.

On the other hand, when the first grid map M1 and the second grid map M2 created in this way are integrated by the object integration unit 32, the integrated grid map Mc as shown in FIG. 3C is obtained. Looking at the portion where the number of votes is +1 in the integrated grid map Mc, both the distance to the object 100 and the width direction of the object 100 are accurate maps. That is, the first grid map M1 having an accurate distance to the object 100, which the sonar 10 is good at, and the second grid map M2 having an accurate width direction of the object 100, which the vehicle-mounted camera 20 is good at, are integrated.

In this way, by creating an integrated grid map Mc that considers the characteristics of various types of sensors, it is possible to improve the detection accuracy of an object, which is an error such as when object detection is performed with only one distance measuring sensor. It is possible to suppress the detection.

(Other embodiments)
Although the present disclosure has been described in accordance with the above-described embodiment, the present disclosure is not limited to the embodiment, and includes various modifications and modifications within an equal range. In addition, various combinations and forms, as well as other combinations and forms that include only one element, more, or less, are also within the scope of the present disclosure.

(1) For example, the number of votes and the numerical value of the predetermined threshold value used when detecting the object 100 as described in the first embodiment are only shown as an example, and can be set arbitrarily. For example, the number of 1st and 4th votes was +1 and the number of 2nd and 5th votes was 0, and the number of 3rd and 6th votes was -1, but the number of 1st and 4th votes was set. The number may be +5, the number of 2nd and 5th votes may be +1 and the number of 3rd and 6th votes may be 0. Further, the threshold value for determining the grid on which the object exists may be set to a value corresponding to the number of votes. For example, when the first number of votes is +5, the threshold value may be set to +20. ..

(2) Further, in the first embodiment, the number of votes applied to the first grid map M1 created based on the sensing information of the sonar 10 and the second grid map M2 created based on the sensing information of the vehicle-mounted camera 20. Equalized the number of votes applied. That is, the number of the first vote and the number of the fourth vote are the same, the number of the second vote and the number of the fifth vote are the same, and the number of the third vote and the number of the sixth vote are the same. This means that the sonar 10 and the vehicle-mounted camera 20 have the same weight, and the weights can be different. That is, the number of the first vote and the number of the fourth vote may be different, the number of the second vote and the number of the fifth vote may be different, and the number of the third vote and the number of the sixth vote may be different. In that case, for example, the weighting should be changed according to the high reliability, that is, the error between the detection coordinates and the actual position of the object 100, and the weighting of the higher reliability should be higher. Can be done. Further, for example, the weighting of the shorter sensing cycle may be lowered so that the weighting is changed according to the sensing cycle. In that case, at least the number of the first vote and the number of the fourth vote may be set to different values according to the weight.

(3) Further, in the first embodiment, other than the main grid Gm, the width direction grid Gw, the front grid Gf and the rear grid Gb is set as the other grid Go, but the range of the other grid Go may be limited. For example, if the detectable range of a distance measuring sensor such as a sonar 10 or an in-vehicle camera 20 is a predetermined angle range in a plan view, the range is limited to a range of predetermined angles on both sides in the horizontal direction, and the other grid Go is used. do. In that case, for the grid G corresponding to the detectable range of the distance measuring sensor and the vehicle-mounted camera 20 and the positions exceeding the predetermined angles on both sides thereof, the number of votes may be set to 0 so that the total value of the number of votes is not changed.

(3) In the first embodiment, the object detection process shown in FIG. 3 is executed when the start switch of the own vehicle V is turned on, but the start switch is always turned on. It does not have to be executed. For example, the object detection process may be executed when a switch (not shown) instructing the execution of parking support control is turned on. That is, the object detection process may be executed when the execution of the control using the result of the object detection process is instructed.

(4) The processing unit and its method described in the present disclosure are provided by configuring a processor and a memory programmed to execute one or more functions embodied by a computer program. It may be realized by a dedicated computer. Alternatively, the processing unit and methods thereof described in the present disclosure may be realized by a dedicated computer provided by configuring a processor with one or more dedicated hardware logic circuits. Alternatively, the processing unit and method thereof described in the present disclosure may be a combination of a processor and memory programmed to perform one or more functions and a processor composed of one or more hardware logic circuits. It may be realized by one or more dedicated computers configured. Further, the computer program may be stored in a computer-readable non-transitional tangible recording medium as an instruction executed by the computer.

1 Vehicle control system 10 Sonar 20 In-vehicle camera 30 ECU
31a 1st object recognition unit 31b 2nd object recognition unit 32 Object integration unit 33 Output selection unit 34 Control unit

Claims (5)

An object detection device that detects an object (100) existing around the own vehicle by using a grid map (Map) in which the periphery of the own vehicle (V) is divided into a plurality of grids (G).
The sensing information of the distance measuring sensor (10) that measures the position of the object by outputting the exploration wave and acquiring the reflected wave reflected by the exploration wave by the object is input, and the grid map is used. By setting a predetermined number of votes in the main grid (Gm), which is a grid corresponding to the detection coordinates indicating the position where the object is detected included in the sensing information of the distance measuring sensor, the object exists around the own vehicle. A first object recognition unit (31a) that creates a first grid map (M1) showing the positions of the objects to be used.
The sensing information of the vehicle-mounted camera (20) that captures the surroundings of the own vehicle is input, and the grid map is used to correspond to the detection coordinates indicating the position where the object included in the sensing information of the vehicle-mounted camera is detected. With the second object recognition unit (31b) that creates a second grid map (M2) showing the positions of the objects existing around the own vehicle by setting a predetermined number of votes in the main grid which is a grid. ,
An object integration unit (32) that creates an integrated grid map (Mc) that integrates the number of votes for each grid in the first grid map and the number of votes for each grid in the second grid map.
An object detection device having an object identification unit (33) that identifies a grid in the integrated grid map in which the total value of the number of votes exceeds a predetermined threshold value as a grid on which the object exists. The first object recognition unit has a first vote number on the main grid corresponding to the detection coordinates by the distance measuring sensor, and a grid located in the width direction of the object adjacent to the main grid in the width direction grid ( Gw), the grid located behind the main grid and the width direction grid is defined as a rear grid (Gb), and the width direction grid and the rear grid have a second vote number smaller than the first vote number, the main grid. And the grid in the width direction and the grid different from the rear grid are set as other grids (Go), and the number of third votes, which is smaller than the number of second votes, is set in the other grid, respectively.
The second object recognition unit has a fourth vote number on the main grid corresponding to the detection coordinates by the vehicle-mounted camera, and a grid located in front of the object adjacent to the main grid is a front grid (Gf). With the grid located behind the main grid and the front grid as the rear grid, the front grid and the rear grid have a fifth vote number less than the fourth vote, the main grid, the front grid, and the rear grid. The grid different from the above is set as another grid, and the number of sixth votes, which is less than the number of fifth votes, is set in the other grid.
The object detection device according to claim 1, wherein the object integration unit creates the integrated grid map by adding the number of votes for each grid in one grid map and the number of votes for each grid in the second grid map. .. The number of the first vote and the number of the fourth vote are the same, the number of the second vote and the number of the fifth vote are the same, the number of the third vote and the number of the sixth vote are the same, claim 2. The object detection device described. The second claim, wherein the first vote and the fourth vote are different, the second vote and the fifth vote are different, and the third vote and the sixth vote are different. Object detection device. The first object recognition unit uses a grid different from the main grid, the width direction grid, and the rear grid as the other grid for a grid in a range of predetermined angles on both sides in the horizontal direction of the detectable range of the distance measuring sensor. ,
The second object recognition unit uses a grid different from the main grid, the front grid, and the rear grid as the other grids in a range of predetermined angles on both sides in the horizontal direction of the detectable range of the vehicle-mounted camera. The object detection device according to any one of claims 2 to 4.

Images