JP2018075866A - Attitude estimation method for parking control apparatus, and attitude estimation apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an attitude estimation method for a parking control apparatus and an attitude estimation apparatus, which are capable of accurately estimating a vehicular attitude.SOLUTION: An attitude estimation apparatus includes: a feature point-feature article selection part 151 for selecting a number of feature points, the number increasing proportionately as a distance from a parking target position nears, by referring to an environment map 12; a feature point-feature article detection part 152 for detecting a feature point in surroundings of a vehicle; and an attitude estimation part 153 for estimating an attitude of the vehicle by collating the feature point selected by the feature point-feature article selection part 151 with the feature point detected by the feature point-feature article detection part 152.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a posture estimation method for a parking control device and a posture estimation device.

  As described in Patent Document 1 from the past, based on the position and posture of the person estimated based on the rotation speed of the wheel and the steering angle of the vehicle, and the position and posture of the person estimated based on the environment map, Methods for determining the final self position and posture are known.

  An environment map (for example, a three-dimensional map) is created from data acquired by a vehicle that has traveled in the past, and stores a plurality of feature points and feature objects such as road signs and structures. Then, by comparing environmental information such as the shape and texture of an object acquired by a surrounding detection sensor such as a camera or a laser radar mounted on the vehicle with the feature points and features of the environment map, the position of the vehicle in the environment map and Estimate posture.

JP 2009-149194 A

  However, since the conventional example disclosed in Patent Document 1 described above stores feature points without considering the distance from the vehicle, the posture of the vehicle in the environment map may not be estimated with high accuracy.

  The present invention has been made to solve such a conventional problem, and an object of the present invention is to provide a posture estimation method and posture of a parking control device capable of estimating the posture of a vehicle with high accuracy. It is to provide an estimation device.

  In order to achieve the above object, the closer the distance from the parking target position is, the more feature points are stored and the posture of the vehicle is estimated.

  According to the present invention, the posture of the vehicle can be estimated with high accuracy, and the vehicle can be parked at the correct position and posture with respect to the parking target position.

FIG. 1 is a block diagram illustrating a configuration of an attitude estimation apparatus to which an attitude estimation method according to an embodiment of the present invention is applied and peripheral devices thereof. FIG. 2 is a flowchart showing a processing procedure of the posture estimation apparatus according to the first embodiment of the present invention. FIG. 3 is an explanatory diagram showing a target route set by the posture estimation apparatus according to the first embodiment of the present invention. FIG. 4A is an explanatory diagram illustrating a state in which feature points existing outside the neighborhood region are selected according to the first embodiment of the present invention. FIG. 4B is an explanatory diagram illustrating a state in which two feature points existing inside the neighborhood region are selected according to the first embodiment of the present invention. FIG. 5 is an explanatory diagram illustrating a change in an image captured by the camera when the posture of the vehicle changes. FIG. 6 is a flowchart showing a processing procedure of the posture estimation apparatus according to the second embodiment of the present invention. FIG. 7 is an explanatory diagram showing a state of selecting a linear feature existing in the three-dimensional data of the environment map according to the second embodiment of the present invention. FIG. 8 is an explanatory diagram showing a state of selecting two feature points existing in the three-dimensional data of the environment map according to the second embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[Description of First Embodiment]
FIG. 1 is a block diagram showing the configuration of the posture estimation apparatus and its peripheral devices according to the first embodiment of the present invention. As shown in FIG. 1, the posture estimation apparatus 100 according to this embodiment includes a camera 11 (ambient detection unit) that captures an environment around a vehicle, three-dimensional data around the vehicle, and feature points included in the three-dimensional data, An environment map 12 for storing feature objects, an odometry sensor 13 for detecting vehicle odometry, a storage unit 14 (feature point storage circuit) for storing feature point information, and a control signal corresponding to the attitude of the vehicle is output. A control unit 15 is provided. The control unit 15 is connected to an actuator 16 that controls steering, accelerator, brake, shift, parking, and the like mounted on the vehicle, and outputs a control signal to the actuator 16. Note that the feature point information may be stored in the environment map 12 without providing the storage unit 14.

  The environment map 12 may store two-dimensional data around the vehicle. Moreover, the structure connected via a network other than the case where it mounts in a vehicle may be sufficient. Although the case where the camera 11 is used as a sensor for detecting the situation around the vehicle will be described as an example, it is also possible to detect the situation around the vehicle using laser radar, sonar, beacon, and WiFi communication. .

  The control unit 15 includes a feature point / feature selection unit 151 that selects a feature point or a feature included in the environment map 12 based on conditions described later with reference to the environment map 12. Further, a feature point / feature object detection unit 152 that detects a feature point around the vehicle or a feature object from an image captured by the camera 11 is provided.

  Also, a posture estimation unit that estimates the posture of the vehicle by collating the feature point or feature selected by the feature point / feature selection unit 151 with the feature point or feature detected by the feature point / feature detection unit 152 153.

  The posture of the host vehicle on the environment map 12 is estimated by detecting the distance and angle with the feature point. Generally, as the distance between the vehicle and the feature point increases, the posture of the vehicle becomes easier to estimate. For example, when there is a feature point in front of the vehicle, if the angle of the vehicle posture changes by 1 degree, the position of the feature point is detected with a larger deviation as the distance between the vehicle and the feature point increases. . On the other hand, when the distance between the vehicle and the feature point is small, even if the angle of the posture of the vehicle changes by 1 degree, the displacement of the feature point is small and difficult to detect. Thus, since it is difficult to estimate the posture of the vehicle V1 when the distance between the vehicle V1 and the feature point is small, in this embodiment, the vicinity of the parking target position (parking target position) where the accuracy of the posture of the vehicle V1 is required. (The smaller the distance from), the more feature points are memorized. Thereby, since a plurality of feature points can be referred to, the posture of the vehicle can be estimated with high accuracy. The reason why the accuracy of the posture of the vehicle V1 is required as the distance from the parking target position is smaller is that the occupant tends to be more sensitive to the posture of the vehicle as it approaches the parking target position. For example, when the vehicle is parked at the parking target position, if the posture of the vehicle V1 is deviated from the parking target position, the occupant may feel uncomfortable, and it may be possible to park again. Therefore, the accuracy of the posture of the vehicle V1 is required as the distance from the parking target position is smaller.

  Each function shown in the present embodiment can be implemented by one or a plurality of processing circuits. The processing circuit includes a processing device including an electric circuit. Processing devices also include devices such as application specific integrated circuits (ASICs) and conventional circuit components arranged to perform the functions described in the embodiments.

  Then, the posture estimation device 100 according to the present embodiment detects the posture of the vehicle with high accuracy when the vehicle approaches a parking target position (for example, a parking space in a parking lot), and correct posture with respect to the parking target position. The vehicle is controlled so that the vehicle stops.

  Hereinafter, the processing procedure of the posture estimation apparatus 100 according to the present embodiment will be described with reference to the flowchart shown in FIG. This control is executed by the control unit 15 shown in FIG.

  First, in step S <b> 11, the control unit 15 determines whether or not the vehicle has approached the parking target position 21. In this process, for example, when an occupant of the vehicle determines that the vehicle has approached the parking target position 21 and an operation switch (not shown) is pressed, it can be determined that the vehicle has approached. Alternatively, in a vehicle equipped with a navigation device (not shown), the parking target position 21 is registered in advance in the navigation device, and the vehicle is within a predetermined distance from the parking target position 21 based on the GPS vehicle position. It is also possible to determine that the vehicle has approached. The reference point of the parking target position 21 is, for example, the center position (for example, the position indicated by the symbol Q1 in FIG. 3) when the vehicle is correctly parked in the parking target position 21. Further, the predetermined distance can be set to 20 m, for example. That is, it can be determined that the vehicle has approached the parking target position 21 when the vehicle has reached a radius of 20 m from the point indicated by the symbol Q1. Further, when the vehicle reaches a radius of 20 m, the passenger may be notified by emitting a sound or light so as to prompt the passenger to operate the switch. In addition, it can replace with said center position Q1 and can also set the position of the camera 11 when a vehicle stops correctly in the parking target position 21 to a reference position.

  In step S <b> 12, the control unit 15 collates the feature points, features, and obstacles included in the image captured by the camera 11 with the feature points, features, and obstacles stored in the environment map 12. The position and posture of the vehicle with respect to the parking target position are estimated. The feature point is an arbitrary point such as a road sign or a building. The characteristic object is, for example, a white line laid on a road surface, a wall of a building, or the like. The obstacle is an object or the like that is installed on the road surface and hinders vehicle travel.

  In this embodiment, the camera 11 is used as the surrounding detection unit, and an example in which feature points, features, and obstacles around the vehicle are detected using an image captured by the camera 11 will be described. As another example, a laser radar or the like can be used. At the same time, it is better to acquire odometry information such as a wheel speed pulse and a yaw rate sensor by the odometry sensor 13.

  In step S <b> 13, the control unit 15 refers to information on the position and orientation of the vehicle and information on obstacles included in the image captured by the camera 11, and sets the target until the vehicle reaches the parking target position from the current position. Generate a route. For example, as shown in FIG. 3, when the vehicle V1 approaches the parking target position 21, a suitable target route L1 for parking the vehicle V1 at the parking target position 21 is generated. At this time, for example, the vehicle V1 recognizes the position and posture with reference to the roadside structure 31.

  In step S14, the control unit 15 outputs a control signal to the actuator 16 so that the vehicle V1 travels along the target route L1, and executes parking control.

  In step S <b> 15, the control unit 15 determines whether or not the distance from the vehicle V <b> 1 to the parking target position 21 has reached an area that is less than a preset threshold distance (this is referred to as “neighboring area”). The threshold distance is set to 10 m, for example. For example, as shown in FIG. 4A, a range with a radius of 10 m is set in the vicinity region R1 with the center point Q1 of the parking target position 21 as a reference. If the vehicle V1 enters the vicinity region R1 shown in FIG. 4A, the process proceeds to step S16.

  In step S <b> 16, the control unit 15 acquires an image captured by the camera 11, analyzes the image, and detects a feature point included in an object or a building that exists around the vehicle. Further, the position and posture of the host vehicle on the environment map 12 are estimated from the correspondence with the feature points of the environment map 12. The feature points to be used at this time are selected according to the following steps S17 to S19.

  In step S <b> 17, the control unit 15 determines whether or not a feature point exists outside the vicinity region R <b> 1 of the environment map 12. As shown in FIG. 4A, when the feature point p1 exists outside the vicinity region R1 and inside the detection range R2 of the image captured by the camera 11 (YES in step S17), the process is performed in step S18. Proceed. On the other hand, as shown in FIG. 4B, if there is no feature point outside the neighboring region R1 (NO in step S17), the process proceeds to step S19.

  In step S18, the control unit 15 selects one feature point (unique feature point) existing outside the neighboring region R1. Specifically, as shown in FIG. 4A, when a feature point p1 exists outside the neighboring region R1, only the feature point p1 is selected. Thereafter, the process proceeds to step S20.

  In step S19, the control unit 15 selects two feature points existing inside the neighborhood region R1. Specifically, as shown in FIG. 4B, feature points p2 and p3 existing inside the neighboring region R1 are selected. Thereafter, the process proceeds to step S20.

  In step S <b> 20, the control unit 15 stores information on the feature points selected in the process of step S <b> 18 or S <b> 19 in the storage unit 14. The feature point information stored in the storage unit 14 can be used when the vehicle V1 is parked at the parking target position 21 from the next time onward.

  In step S <b> 21, the control unit 15 estimates the attitude of the vehicle V <b> 1 on the environment map 12 based on the feature points selected on the environment map 12. Specifically, as shown in FIG. 4A, when the feature point p1 outside the region R1 is selected, the posture of the vehicle V1 is estimated based on the feature point p1. As shown in FIG. 4B, when the feature points p1 and p2 inside the region R1 are selected, the posture of the vehicle V1 is estimated based on these feature points p2 and p3. By doing so, when the vehicle V1 approaches the parking target position 21, it is possible to improve the estimation accuracy of the posture of the vehicle V1 with respect to the parking target position 21.

  Hereinafter, when estimating the posture of the vehicle, if a feature point exists outside the neighborhood region R1, one feature point outside is used, and if no feature point exists outside the neighborhood region R1, the neighborhood region The reason why the posture of the vehicle is estimated using two feature points inside R1 will be described in detail.

  FIG. 5 is an explanatory diagram schematically showing a surrounding image captured by the camera 11 mounted on the vehicle. An image D1 indicates an image captured when the vehicle is in the posture F1, and an image D2 indicates an image captured when the vehicle is in a posture F2 in which the posture angle of the vehicle is slightly changed with respect to the posture F1. In the posture F1, the imaging region is f1, and in the posture F2, the imaging region is f2. The image D1 captured by the camera 11 in the posture F1 includes an object B1 (for example, a road sign) that exists in the vicinity of the vehicle and an object B2 (for example, a mountain) that exists in the distance. Similarly, the image D2 captured by the camera 11 in the posture F2 includes the objects B1 and B2.

  When the images D1 and D2 are compared, it is understood that an object that is farther away from the vehicle is displayed with a greater movement with respect to the posture change. Specifically, the change of the object B2 of the image D1 and the object B2 of the image D2 is larger than the change of the object B1 of the image D1 and the object B1 of the image D2. That is, an object farther from the vehicle is greatly affected by the posture change.

  From the above, it can be seen that the posture of the vehicle can be estimated with higher accuracy by using feature points that are located farther from the vehicle. Accordingly, in the processing of steps S17 to S19, a vehicle vicinity region R1 is set, and when a feature point exists outside (ie, far from) the vicinity region R1, it is possible to estimate the posture with high accuracy. Therefore, the posture of the vehicle is estimated using this one feature point. On the other hand, when there is no feature point outside the neighboring region R1, the posture of the vehicle is estimated using two feature points in order to improve posture estimation accuracy.

  By doing this, when there are feature points outside the neighboring region R1, it is possible to avoid selecting feature points redundantly by setting the number of feature points to be one, and outside the neighboring region R1. When there are no feature points, the number of feature points to be selected is set to two, so that the posture of the vehicle is estimated with high accuracy. As a result, when the vehicle enters the vicinity region R1, the posture of the vehicle with respect to the parking target position 21 can be estimated with higher accuracy.

  Next, in step S <b> 22 of FIG. 2, the control unit 15 controls the vehicle to travel along the target route L <b> 1 (see FIG. 3), and moves the vehicle V <b> 1 to the parking target position 21.

  In step S23, the control unit 15 determines whether or not the vehicle V1 has reached the parking target position 21 and the parking operation has been completed. If the parking operation has been completed, the process ends. When the parking is completed, it is possible to recognize that the vehicle V1 is parked at the parking target position 21 by detecting that the occupant puts the shift position into parking or operated the parking brake.

  In the above-described embodiment, the neighborhood region R1 is set around the point of the parking target position 21 (Q1 in FIG. 3), and one feature point is selected when the feature point exists outside the neighborhood region R1. In the case where no feature point exists outside the neighboring region R1, an example in which two feature points are selected inside the neighboring region R1 has been described. The present invention is not limited to this, and the same effect can be achieved by increasing the number of feature points to be selected as the distance from the parking target position 21 to the feature points approaches. For example, the distance from the parking target position 21 can be divided into three stages, and the number of feature points to be selected gradually increases as the parking target position 21 is approached in each region.

  Thus, in the posture estimation apparatus 100 according to the first embodiment, when the vehicle V1 is parked in the parking target position 21, the feature points of the environment map 12 and the feature points of the image captured by the camera 11 are collated. Then, the attitude of the vehicle V1 in the environment map is estimated. At this time, as the distance from the parking target position 21 is shorter, more feature points are stored. As a result, when the vehicle V1 is parked at the parking target position 21 using the environment map, a plurality of feature points can be used as the vehicle V1 approaches the parking target position, so that the attitude of the vehicle V1 can be estimated with high accuracy. The vehicle can be parked at a correct position with respect to the parking target position 21.

  Further, when the parking target position 21 is used as a reference and the range less than the threshold distance is set as the neighboring region R1, and the feature point exists outside the neighboring region R1, the posture of the vehicle is estimated using one feature point, If there is no feature point outside the neighboring region R1, the posture of the vehicle is estimated using two (plural) feature points existing inside the neighboring region R1.

  Therefore, when a feature point exists outside the neighborhood region R1, the posture is estimated using only one feature point, so that the calculation is easier and the calculation load is reduced as compared with the case of using a plurality of feature points. it can. Further, when no feature point exists outside the neighborhood region R1, the posture of the vehicle is estimated using two feature points existing inside. Therefore, even if the feature point is close to the parking target position 21, it is possible to estimate the posture with high accuracy by using two (plural) feature points.

  Furthermore, since the information of the selected feature point is stored in the storage unit 14, when the vehicle V1 is parked at the same parking target position 21 after the next time, the feature point stored in the storage unit 14 is adopted. It ’s fine. Therefore, since the feature point to be used is specified at the time of parking after the next time, it is not necessary to select a feature point suitable for posture estimation, and the calculation load can be reduced.

  Further, since the distance to the feature point is calculated using the attachment position of the camera 11 when the vehicle V1 is parked at the parking target position 21 as a reference point, the distance to the feature point can be determined more accurately. The accuracy of estimating the posture angle can be improved.

[Description of Second Embodiment]
Next, a second embodiment of the present invention will be described. The apparatus configuration is the same as the block diagram shown in FIG.

  In the second embodiment, it is determined whether or not a linear feature by a linear object exists in the environment map 12, and if there is a linear feature, parking is performed using this feature. The attitude of the vehicle V1 with respect to the target position 21 is estimated. Hereinafter, the processing procedure of the posture estimation apparatus according to the second embodiment will be described with reference to the flowchart shown in FIG.

  Since the process of step S31-35 shown in FIG. 6 is the same as the process of step S11-S15 shown in FIG. 2, description is abbreviate | omitted.

  In step S <b> 36, the control unit 15 acquires an image captured by the camera 11 and detects a feature point such as an object, a white line, or the like existing around the vehicle. Further, the position and posture of the host vehicle on the environment map 12 are estimated from the correspondence relationship with the characteristic objects of the environment map 12. A “linear feature” is a target or object that forms a straight line on a plane, such as a white line laid on the road, a straight curb, a straight wall of a building, or a building. A linear pattern, etc.

  In step S <b> 37, the control unit 15 determines whether or not a linear feature exists in the environment map 12. If a linear feature exists (YES in step S37), the process proceeds to step S38. If a linear feature does not exist (NO in step S37), the process proceeds to step S39.

  In step S <b> 38, the control unit 15 selects a linear feature existing in the environment map 12. For example, as shown in FIG. 7, a linear white line 22 for dividing the parking space is laid around the parking target position 21, and this white line 22 exists in the environment map 12 as a feature object. The white line 22 is preferentially selected as a feature. Thereafter, the process proceeds to step S40.

  In step S39, the control unit 15 selects two feature points of the environment map 12. For example, as shown in FIG. 8, two feature points p4 and p5 existing around the parking target position 21 are selected within the detection range R2 of the image captured by the camera 11. Thereafter, the process proceeds to step S40.

  In step S <b> 40, the control unit 15 stores in the storage unit 14 information about the feature object and feature point selected in the process of step S <b> 38 or S <b> 39. Information on the feature and feature points stored in the storage unit 14 can be used when the vehicle is parked at the parking target position 21 from the next time onward.

  In step S <b> 41, the control unit 15 estimates the posture of the vehicle V <b> 1 based on the feature or feature point selected on the environment map 12.

  When there is a linear feature, the orientation of the vehicle V1 with respect to the parking target position 21 is highly accurate by checking the direction of the straight line in the environment map 12 and the direction of the straight line of the feature captured by the camera 11. Can be estimated. On the other hand, when there is no linear feature, the posture of the vehicle V1 is estimated by collating the two feature points of the environment map 12.

  Thereafter, in step S42, the control unit 15 controls the vehicle V1 to travel along the target route, and moves the vehicle V1 to the parking target position 21.

  In step S43, the control unit 15 determines whether or not the vehicle V1 has reached the parking target position 21 and the parking operation has been completed. If the parking operation has been completed, the process ends. When the parking is completed, it is possible to recognize that the vehicle V1 is parked at the parking target position 21 by detecting that the occupant puts the shift position into parking or operated the parking brake.

  As described above, in the posture estimation apparatus according to the second embodiment, when a linear feature exists in the environment map 12, this feature is collated with the feature included in the image captured by the camera 11. Then, the attitude of the vehicle V1 in the environment map is estimated. Then, this feature is stored. As a result, when the vehicle V1 is parked at the parking target position 21 using the environmental map, it is possible to use a linear feature that is easy to specify the position and orientation, so that the orientation of the vehicle V1 is highly accurate. The vehicle can be parked at a correct position with respect to the parking target position 21.

  If there is no linear feature in the environment map 12, two feature points are selected and collated with the feature points included in the image captured by the camera 11 to estimate the posture of the vehicle V 1. Further, this feature point is stored. As a result, when the vehicle V1 is parked at the parking target position 21 using the environment map, two (plurality) feature points can be used, so that the attitude of the vehicle V1 can be estimated with high accuracy, and parking is performed. The vehicle can be parked at a correct position with respect to the target position 21.

  As mentioned above, although the attitude | position estimation apparatus of this invention and the attitude | position estimation method of the parking control apparatus were demonstrated based on embodiment of illustration, this invention is not limited to this, The structure of each part has the same function. It can be replaced with any configuration.

11 Camera 12 Environmental Map 13 Odometry Sensor 14 Storage Unit (Feature Point Storage Circuit)
DESCRIPTION OF SYMBOLS 15 Control part 16 Actuator 21 Parking target position 22 White line 31 Roadside structure 100 Attitude estimation apparatus 151 Feature point / feature object selection part 152 Feature point / feature object detection part 153 Attitude estimation part p1-p5 Feature point R1 Neighborhood area R2 Detection range V1 vehicle

Claims (9)

When parking at a parking target position, an environment map indicating a surrounding situation of the parking target position is stored, a feature point in the environment map is stored, and a posture of the vehicle in the environment map based on the feature point In a posture estimation method for estimating
The attitude estimation method for a parking control device, wherein more feature points are stored as the distance from the parking target position is shorter. Set a threshold distance based on the parking target position,
The posture estimation method for a parking control device according to claim 1, wherein a plurality of feature points are stored in an area where the distance from the parking target position is less than the threshold distance. Set a threshold distance based on the parking target position,
When there is no feature point in an area where the distance from the parking target position is equal to or greater than the threshold distance, a plurality of feature points are stored in an area where the distance from the parking target position is less than the threshold distance. The attitude | position estimation method of the parking control apparatus of Claim 1 or 2 to do. Set a threshold distance based on the parking target position,
3. The feature point according to claim 1, wherein in the area where the distance from the parking target position is less than the threshold distance, more feature points are stored than in the area where the distance from the parking target position is equal to or greater than the threshold distance. The attitude | position estimation method of the parking control apparatus of description. Storing the parking target position in the environment map;
The said parking target position is set based on the attachment position of the circumference | surroundings detection part. The attitude | position estimation method of the parking control apparatus as described in any one of Claim 1 thru | or 4 characterized by these.   6. The method for estimating a posture of a parking control device according to claim 1, wherein when a linear feature is detected, the feature is stored as the feature. 7. The posture of the vehicle is estimated by preferentially using the linear feature object when a plurality of the feature points are stored and the linear feature object is stored. Attitude estimation method for the parking control device of the present invention. 8. The posture of the vehicle in the environment map is estimated using more feature points as it is closer to the parking target position when a plurality of the feature points are stored. 8. The attitude | position estimation method of the parking control apparatus of description. When parking at a parking target position, an environment map indicating a surrounding situation of the parking target position is stored, a feature point in the environment map is stored, and a posture of the vehicle in the environment map based on the feature point In the posture estimation device for estimating
A posture estimation device comprising: a feature point storage circuit that stores more feature points as the distance from the parking target position is shorter.

Images