JP2015075382A - Object detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an object detection device capable of simplifying the adjustment of an attachment posture on a manufacturing line and reducing the influence of the change of the posture of a vehicle with respect to a road surface.SOLUTION: From a plurality of range-finding points obtained by sweep and irradiation of a laser radar 1, an attachment posture estimating unit 24 extracts a road surface candidate point as a range-finding point at which a road surface 6 is detected (step S3). The attachment posture is estimated on the basis of the angle or distance between a road flat surface S calculated from a plurality of road surface candidate points and a reference surface (z=0) corresponding to the road surface 6. The attachment posture update unit updates the attachment posture stored in a storage unit. A position calculation unit 22 calculates the relative position of the range-finding point using the updated attachment posture.

Description

  The present invention relates to an object detection device that is mounted on a vehicle and calculates a relative position of an object detected by a laser radar based on a mounting posture of the laser radar and an irradiation angle of laser light emitted from the laser radar.

  2. Description of the Related Art Conventionally, there is an object detection device that detects a pedestrian or a preceding vehicle existing around a vehicle using a scan type laser radar. In this type of object detection device, the distance from the laser radar to the ranging point is measured by measuring the time required for the laser beam emitted by the laser radar to be reflected and returned by the object (this is the ranging point). Measure the distance to.

  The laser radar used in the object detection device sweeps (i.e., scans) pulsed laser light in a predetermined angle range, for example, every fixed angle in the vehicle width direction and the vehicle height direction.

  The object detection device calculates the relative position of the distance measuring point with respect to the laser radar from the distance to the distance measuring point, the irradiation angle of the laser beam that detected the distance measuring point, and the mounting posture of the laser radar. The mounting posture here refers to the height from the road surface where the laser radar is mounted, the pitch angle, the roll angle, and the yaw angle with respect to the road surface of the laser radar. The irradiation angle refers to an angle formed by the irradiated laser light with the front direction of the laser radar.

  By the way, in general, an object detection device calculates a relative position on the assumption that a laser radar has a pre-designed mounting posture. For this reason, when the mounting posture of the laser radar does not match the mounting posture designed in advance, there arises a problem that the relative position of the distance measuring point detected by the object detection device is deviated from the actual relative position. .

  As means for solving this problem, for example, Patent Document 1 discloses a technique for adjusting the mounting posture of a laser radar mounted on a vehicle. In Patent Document 1, in a production line or the like, a reflector is installed at a position (hereinafter referred to as a reference position) that is the front direction when the laser radar is in an accurate mounting posture. Then, the mounting posture is adjusted so that the laser radar detects the reflector at the reference position.

JP-A-8-29536

  However, the technique disclosed in Patent Document 1 also requires the accuracy of the test environment itself, such as the installation position of test equipment such as a reflector for the laser radar. In addition, since it is necessary to adjust the mounting posture in response to the test result, the time and personnel required for adjusting the mounting posture and the cost for adjusting the test environment itself are required on the production line.

  Even when the laser radar is accurately installed in the vehicle, the posture of the vehicle with respect to the road surface (that is, the inclination of the vehicle body) changes due to changes in tire air pressure or summer / winter tire replacement. Furthermore, the posture of the vehicle with respect to the road surface also changes depending on the occupant's boarding position and the load. When the attitude of the vehicle with respect to the road surface changes, the attitude of the laser radar with respect to the road surface (that is, the mounting attitude) also changes, and the relative position of the ranging point cannot be calculated accurately.

  The present invention has been made in view of the above circumstances, and the object of the present invention is to simplify the adjustment of the mounting posture on the production line and reduce the influence of changes in the posture of the vehicle with respect to the road surface. The object is to provide an object detection device capable of performing the above.

  The present invention for achieving the object is input from a laser radar (1) which is mounted on a vehicle and sequentially sweeps and emits laser light in a predetermined angular range in the horizontal direction at a plurality of different angles in the vertical direction. A distance calculation unit (21) that calculates a distance from the laser radar to a distance measuring point that reflects the laser beam based on the signal, a storage unit (20) that stores the mounting posture of the laser radar on the road surface, and a storage A position calculation unit (22) for detecting a relative position of the ranging point with respect to the vehicle based on the mounting posture stored in the unit, the direction in which the laser beam is irradiated, and the distance calculated by the distance calculation unit; And an attachment posture estimation unit (24) for estimating the posture, and a predetermined initial set value is registered in advance in the storage unit as the attachment posture, and the attachment posture estimation unit emits laser light in the vertical direction. For each of multiple ranging points obtained by changing the irradiation azimuth, which is the irradiation direction of the laser beam in the horizontal direction, for each irradiation elevation angle that is the direction, the distance to the ranging point and the irradiation azimuth are associated with each other Based on the frequency component of the one-dimensional signal generated by the one-dimensional signal generator (S303) that generates the one-dimensional signal and the one-dimensional signal generated by the one-dimensional signal generator, A road surface candidate point extraction unit (S3) that extracts a road surface candidate point that is a distance measurement point for detecting a road surface, and a plane that represents the road surface based on a plurality of road surface candidate points extracted by the road surface candidate point extraction unit. Estimate the mounting posture from the road surface plane calculation unit (S4) that calculates a certain road surface plane, the angle formed by the road surface plane calculated by the road surface plane calculation unit, and the reference plane that is the plane that represents the road surface determined from the mounting posture. First mounting posture estimation processing unit (S5) , Based on the mounting orientation estimated by the first attachment orientation estimation processing unit, a first mounting attitude updating unit for updating the mounting attitude stored in the storage unit (S8), characterized in that it comprises a.

  In the above configuration, the road surface candidate point extraction unit extracts a road surface candidate point from which a road surface is detected, from a plurality of distance measurement points obtained by sweep irradiation by the laser radar. The first mounting posture estimation processing unit estimates the mounting posture of the laser radar from the angle formed by the road surface plane calculated from the plurality of road surface candidate points and the reference surface, and the mounting posture update unit is stored in the storage unit. Update the mounting orientation.

  According to the above configuration, since the position calculation unit calculates the relative position using the mounting posture updated by the mounting posture update unit, even if there is an error with the design value, The relative position can be acquired more accurately.

  In other words, it is possible to reduce the influence of the error between the mounting posture stored in advance in the storage unit and the actual mounting posture on the detection accuracy of the relative position of the distance measuring point. For this reason, it is possible to simplify the precise adjustment of the mounting posture on the production line, and it is possible to relax the requirements for the accuracy of the test environment. Therefore, the cost accompanying adjustment of the mounting posture can be reduced.

  Moreover, since the estimation of the mounting posture performed by the mounting posture estimation unit can be performed without using a test facility such as a reflector, it can be performed when the vehicle is traveling. That is, even when the posture of the vehicle with respect to the road surface changes due to a change in tire air pressure, a occupant's boarding position, a load, etc., the mounting posture in the changed vehicle posture can be estimated. Since the position calculation unit calculates the relative position using the estimated mounting posture, it is possible to reduce the influence of the change in the posture of the vehicle on the road surface in the calculation of the relative position.

It is a block diagram which shows the outline of the system configuration | structure which concerns on this embodiment. It is a figure for demonstrating the irradiation direction of the laser beam which the laser radar 1 irradiates in a height direction. It is a figure for demonstrating the detection area of the laser radar. It is a flowchart which shows an example of the flow of an attachment attitude | position estimation process. It is a flowchart which shows an example of the flow of a road surface candidate point extraction process. It is a schematic diagram for demonstrating the operation | movement of a road surface candidate point extraction process. It is a flowchart which shows an example of the flow of a road surface estimation process. It is a conceptual diagram for demonstrating the operation | movement of a road surface estimation process. It is a schematic diagram for demonstrating the process which estimates pitch angle (theta) p. It is a schematic diagram for demonstrating the process which estimates pitch angle (theta) p. It is a schematic diagram for demonstrating the process which estimates roll angle (theta) r. It is a schematic diagram for demonstrating the process which estimates roll angle (theta) r. It is a schematic diagram for demonstrating the process which estimates mounting height. It is a flowchart which shows an example of the flow of a stability evaluation process. It is a conceptual diagram for demonstrating the action | operation of a stability evaluation process. It is a conceptual diagram for demonstrating the process which estimates yaw angle (theta) y. It is a conceptual diagram which shows an example of the front area | region Fr, the back area | region Rr, the right side area | region Rt, and the left side area | region Lf. It is a schematic diagram for demonstrating the variation | change_quantity of the gradient in a vehicle front-back direction. FIG. 10 is a diagram for explaining the effect of application example 2; 12 is a flowchart illustrating an example of a flow of candidate road surface point extraction processing according to Modification 2.

<Overall configuration>
Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows an example of the configuration of an object detection system 100 to which the present invention is applied. An object detection system 100 shown in FIG. 1 is mounted on a vehicle 5 and detects a distance from an object that exists in all directions such as right and left and front and rear of the vehicle 5. The object detection system 100 includes a laser radar 1 and an ECU (Electronic Control Unit) 2.

  The driving support system 3 in FIG. 1 is an example of a function that uses the detection result of the object detection system 100. The driving support system 3 performs various controls that support the driving of the driver based on the detection result of the object detection system 100 (details will be described later). Of course, not only the driving support system 3 but also various systems mounted on the vehicle can use the detection result of the object detection system 100.

  The laser radar 1 includes a rotation mechanism including a motor and the like, and rotates the rotation mechanism to change the irradiation direction of the laser light with respect to the front-rear direction of the vehicle 5 when viewed from the vehicle height direction. That is, the laser radar 1 irradiates laser light toward 360 ° around the vehicle (that is, all directions) by rotating the rotating mechanism. The rotation axis of the rotation mechanism is also referred to as the center axis of the laser radar 1. Further, the rotation axis direction coincides with the height direction. Here, the term “match” includes not only perfect match but also approximate match. This is because deviation occurs due to the attitude of the laser radar 1 with respect to the road surface 6 as described later.

  The irradiation direction when the rotation mechanism is in the neutral position (that is, the rotation angle is 0 °) is the front direction of the laser radar 1 and is also referred to as the front-rear direction of the laser radar 1. Further, a direction orthogonal to the front-rear direction and the rotation axis direction of the laser radar 1 is defined as the left-right direction of the laser radar 1. That is, the left-right direction of the laser radar 1 coincides with the irradiation direction when the rotation angle of the rotation mechanism is 90 ° (and 270 °).

  As an example, in the present embodiment, the front direction of the laser radar 1 and the front direction of the vehicle 5 are matched (including substantially the same as described above), but of course, as another aspect, the rotation angle of the rotation mechanism The direction in which the angle becomes 180 ° may coincide with the front direction of the vehicle 5.

  Here, before describing the configuration and operation of the laser radar 1, the road surface determined based on the posture of the vehicle 5 with respect to the road surface 6 with respect to its mounting position and the posture of the laser radar 1 with respect to the road surface 6 (this is referred to as the mounting posture). Description will be made using a coordinate system as appropriate.

  The road surface coordinate system has the center of the mounting position of the laser radar 1 as an origin and a point orthogonally projected onto the road surface 6, the X axis in the vehicle width direction, the Y axis in the vehicle longitudinal direction, the Z axis in the height direction, They are arranged so as to be orthogonal to each other. Therefore, the XY plane of the road surface coordinate system corresponds to the road surface 6.

  As shown in FIGS. 2 and 3, the laser radar 1 is installed, for example, on the upper surface of the vehicle roof so as to have a predetermined mounting posture. The mounting posture here refers to the pitch angle θp, the roll angle θr, the yaw angle θy, and the mounting height h with respect to the road surface 6 of the laser radar 1. The mounting height h refers to the height h from the road surface 6 to the laser radar 1 (strictly, the light emitting unit 11).

  The pitch angle θp refers to the angle that the central axis of the laser radar 1 makes with the Z axis in the YZ plane of the road surface coordinate system, and the roll angle θr refers to the rotation axis of the laser radar 1 with respect to the Z axis in the XZ plane. Refers to the corner to make. The yaw angle θy is an angle formed by the front direction of the laser radar 1 with the Y axis in the XY plane.

  The pitch angle θp takes a positive value in the direction of rotation from the Z-axis direction to the Y-axis direction, and the roll angle θr takes a positive value in the direction of rotation from the Z-axis direction to the X-axis direction. Further, the yaw angle θy has a positive value in the same rotational direction as the irradiation azimuth angle β. Note that the position of the laser radar 1 in the road surface coordinate system is represented by (0, 0, h).

  Returning to FIG. 1 again, the configuration and operation of the laser radar 1 will be described. The laser radar 1 includes a light emitting unit 11, a light receiving unit 12, and a TOF (Time of Flight) calculating unit 13.

  The light emitting unit 11 includes a laser diode, and irradiates laser light discontinuously in a range of a predetermined angle in the rotation axis direction (this is an irradiation range in the rotation axis direction) according to a drive signal from the ECU 2. More specifically, pulsed laser light is irradiated for each layer formed by dividing the irradiation range in the rotation axis direction into L (L is a positive integer) irradiation directions. The rotation axis direction corresponds to the vertical direction described in the claims, and the direction perpendicular to the rotation axis direction corresponds to the horizontal direction described in the claims.

  In the present embodiment, the rotation axis direction irradiation range of the light emitting unit 11 is a range from −10 ° to + 30 ° with respect to the direction perpendicular to the rotation axis (that is, the direction of the optical axis), and the rotation axis direction irradiation range is L. = Laser light is divided into 32 layers. That is, the laser beam is irradiated with a shift of 1.25 ° in the height direction for each layer. The numbers of the layers are 0, 1, 2,..., 31 in ascending order of the angle α formed with the optical axis in the rotation axis direction.

  Note that the angle formed by the laser beam with the optical axis in the rotation axis direction (this is the irradiation elevation angle) represents the rotation angle from the rotation axis direction to the y-axis direction as a positive value. The irradiation elevation angle α is determined for each layer. For example, α = −10 ° for layer 0, and α = 30 ° for layer 31.

  The light emitting unit 11 sequentially irradiates the laser light of each layer every time the rotation mechanism rotates by a certain angle (for example, 0.1 °), so that the laser radar 1 is directed to the vehicle in all directions for each layer. The laser beam is swept and irradiated. The concept of the irradiation angle in the rotation axis direction of each layer is shown in FIG. 2, and the trajectory of the distance measuring point when the laser beam of each layer detects a point on the road surface 6 is scanned at 360 °. 3 shows. The distance measuring point is a point where the laser beam is reflected and refers to a point detected by the laser radar 1.

  The light receiving unit 12 receives, with a light receiving lens, reflected light obtained by reflecting the laser light emitted from the light emitting unit 11 at a distance measuring point, and gives the light to a light receiving element (photodiode). The light receiving element outputs a voltage corresponding to the intensity of the reflected light. The output voltage of the light receiving element is amplified by an amplifier and then output to a comparator. The comparator compares the output voltage of the amplifier with the reference voltage, and outputs a predetermined light reception signal to the TOF calculation unit 13 when the output voltage becomes larger than the reference voltage.

  A drive signal output from the ECU 2 to the light emitting unit 11 is also input to the TOF calculation unit 13. The TOF calculation unit 13 measures the time from when the drive signal is output to when the light reception signal is generated, that is, the time difference between the time when the laser light is emitted and the time when the reflected light is received (meaning the existing TOF). Then, a signal indicating this time difference (hereinafter, TOF signal) is input to the ECU 2.

  When the light emitting unit 11 irradiates the laser light, a drive signal is output from the ECU 2 to the light emitting unit 11 so that the laser light is emitted in a different pattern for each irradiation angle. In other words, the laser light pattern is determined for each irradiation angle, so when the reflected light is received, the irradiation angle of the laser light that is the source of the reflected light is specified from the reflected light pattern. can do.

  The irradiation angle of the laser beam is represented by an irradiation elevation angle α and an irradiation azimuth angle β. As described above, the irradiation elevation angle α is an angle formed by the laser beam with the optical axis in the rotation axis direction, and is determined for each layer. The irradiation azimuth angle β is a rotation angle from the neutral position of the rotation mechanism. As shown in FIG. 3, the irradiation azimuth angle β represents a clockwise angle as a positive value.

  The ECU 2 is composed mainly of a microcomputer composed of a CPU, ROM, RAM, backup RAM, etc., and executes various processes by executing various control programs stored in the ROM based on input information. To do. This ECU 2 corresponds to the object detection device recited in the claims. More specifically, the ECU 2 includes a storage unit 20, a distance calculation unit 21, a position calculation unit 22, a vehicle information acquisition unit 23, and an attachment posture estimation unit as functional blocks for performing various processes as shown in FIG. 24.

  The storage unit 20 is configured by a rewritable nonvolatile storage medium. The storage unit 20 stores an initial value of the mounting posture of the laser radar 1 (this is referred to as an initial mounting posture), and stores various data such as a mounting posture estimated by a mounting posture estimation unit 24 described later. To do. Among the mounting postures stored in the storage unit 20, the mounting posture used by the position calculation unit 22 described later is updated by the first mounting posture update process (step S8 in FIG. 4) and the second mounting posture update process (S12). It will be done.

  For the initial mounting posture, for example, a result measured in a state where there is no occupant and the air pressure of each tire is a specified value, such as a production line, may be used. In the present embodiment, the pitch angle θp, the roll angle θr, and the yaw angle θy are all set as the initial mounting posture so that no deviation occurs (that is, 0 °). The mounting height h as the initial mounting posture is configured to use a value designed in advance.

  As for the mounting height h, the pitch angle θp and the roll angle θr are both 0 °, the irradiation elevation angle α of an arbitrary laser beam, and the distance r between the distance measuring point detected by the laser beam and the laser radar 1 Therefore, a value calculated as h = r × sin α may be used. Furthermore, as the initial mounting posture, a result measured by a relatively rough hand may be used. The reason why the initial mounting posture may be set to be rougher than before is that the mounting posture used when calculating the relative position of the distance measuring point is updated by the mounting posture estimation process described later.

  When the TOF signal (that is, the time difference between the emission time of the laser beam and the reception time of the reflected beam) is input from the TOF calculation unit 13, the distance calculation unit 21 reflects the reflection point of the laser beam (this is calculated based on the time difference). A distance r (refer to FIG. 3) is calculated.

  The position calculation unit 22 is based on the distance r to the distance measuring point calculated by the distance calculation unit 21, the irradiation angle of the laser beam from which the reflected light is obtained, and the mounting posture stored in the storage unit 20. To create position data for the AF points. The position data here is expressed in the sensor coordinate system described below.

  The sensor coordinate system has an x-axis, a y-axis, and a z-axis so as to be orthogonal to each other, and has an origin so that the position of the laser radar 1 in the sensor coordinate system is (0, 0, h). The ECU 2 recognizes that the y-axis direction corresponds to the vehicle longitudinal direction, the x-axis direction corresponds to the vehicle width direction, and the z-axis direction corresponds to the height direction.

  The position calculation unit 22 represents the relative position of the distance measuring point obtained in consideration of the current mounting posture with coordinates (x, y, z) of the sensor coordinate system. In other words, each axis and origin of the sensor coordinate system are set based on the mounting posture adopted by the current position calculation unit 22.

  This sensor coordinate system represents a relative position of the ranging point (ie, object) detected by the laser radar 1 and recognized by the ECU 2 based on the detection result of the laser radar 1 with respect to the vehicle 5. On the other hand, the road surface coordinate system represents the actual relative position of the object detected by the laser radar 1 with respect to the vehicle.

  The sensor coordinate system and the road surface coordinate system coincide with each other when the laser radar 1 is accurately installed (as designed) on the vehicle 5 and the attitude of the vehicle 5 with respect to the road surface 6 is also in accordance with the specified value. To do. That is, the center axis (that is, the z axis) of the laser radar 1 coincides with the height direction of the vehicle 5, and the front direction (that is, the y axis) of the laser radar 1 coincides with the front in the front-rear direction of the vehicle 5. Of course, the x-axis and the x-axis also coincide.

  However, as mentioned in the problem column, in order to accurately install the laser radar 1 with respect to the vehicle 5, strict optical axis adjustment using a reflector or the like is required. In addition, since the accuracy of the test environment for adjusting the optical axis (the installation position of the reflector or the like) itself is also required, the production line requires time, manpower, and test environment costs.

  Even when the laser radar 1 is accurately installed on the vehicle 5, the vehicle 5 is inclined with respect to the road surface 6 due to, for example, a change in tire air pressure or a change in summer / winter tires. The posture (this is the vehicle posture) changes. Furthermore, the vehicle posture also changes depending on the boarding position of the occupant and the load. When the vehicle posture changes, the posture of the laser radar 1 with respect to the road surface 6 (that is, the mounting posture) also changes.

  In the actual use environment of the vehicle 5, the number of passengers and the load are largely changed, and it is assumed that the tire state also fluctuates. Therefore, the road surface coordinate system and the sensor coordinate system often do not match. It is possible. Accordingly, the detection result of the laser radar 1 includes an error due to the mounting posture of the laser radar 1. In other words, the error due to the mounting posture of the laser radar 1 corresponds to a deviation between the sensor coordinate system and the road surface coordinate system.

  Hereinafter, in order to simplify the description, it is assumed that the sensor coordinate system is set based on the initial mounting posture. That is, in the initial mounting posture of the present embodiment, since the pitch angle θp, the roll angle θr, and the yaw angle θy are all 0 °, the y-axis is the front direction of the laser radar 1 and the x-axis is the left and right of the laser radar 1. The z-axis is provided in the direction of the rotation axis of the laser radar 1. The origin is a point where the position of the laser radar 1 is (0, 0, h) in the sensor coordinate system.

  The vehicle information acquisition unit 23 is connected to the vehicle information detection unit 4 including a sensor group that detects information of the host vehicle 5 via an in-vehicle LAN (not shown). The vehicle information detector 4 includes, for example, a vehicle speed sensor that detects the traveling speed of the host vehicle, a yaw rate sensor that detects the magnitude of the yaw rate that acts on the host vehicle, and a steering angle sensor that detects the steering angle of the steering wheel (all illustrated) Abbreviation). The vehicle information acquisition unit 23 calculates the traveling speed of the host vehicle 5, the magnitude of the yaw rate, the amount of movement, and the like using sensor signals from these sensors.

  The mounting posture estimation unit 24 estimates the mounting posture of the laser radar 1 from the detection result of the laser radar 1. The operation of the mounting posture estimation unit 24 will be described later in detail using the flowchart shown in FIG.

  Based on the detection result of the object detection system 100, the driving support system 3 performs, for example, a known ACC (Adaptive Cruise Control), a collision damage reduction brake function, a lane keep assist (LKA), and the like. For example, in the case of ACC, the driving system and the braking system are controlled so as to keep the inter-vehicle distance constant based on the distance from the preceding vehicle detected by the object detection system 100.

  In addition to the functions described above, the driving support system 3 in the present embodiment implements an automatic driving function for automatically driving the vehicle 5 and an automatic parking function for automatically parking the vehicle 5 at the parking position.

<Mounting posture estimation process>
Next, the flow of the mounting posture estimation process performed by the mounting posture estimation unit 24 will be described using the flowchart shown in FIG. The flowchart shown in FIG. 4 is performed sequentially (for example, every 100 milliseconds) when, for example, an ignition switch (not shown) is turned on and power is supplied to the object detection system 100.

  First, in step S1, the mounting posture reading process is performed, and the process proceeds to step S2. In the mounting posture reading process in step S <b> 1, among the mounting postures stored in the storage unit 20, the mounting posture to be used by the position calculation unit 22 is read out. When this flow is performed for the first time after the ignition switch is turned on, the initial mounting posture is read as an example. If the mounting posture is updated in steps S8 and S12 to be described later, the updated mounting posture is read out.

  In step S2, a scan process is performed and the process proceeds to step S3. In the scan process of step S2, a drive signal is output to the laser radar 1, and laser light is scanned 360 ° in all layers. Then, the distance calculation unit 21 calculates the distance r to each distance measuring point based on the time difference data input from the TOF calculation unit 13, and associates the irradiation azimuth angle β with the distance r between the distance measuring points. Further, irradiation azimuth-distance data is generated. For each distance measuring point, the position calculation unit 22 calculates a relative position from the distance r calculated by the distance calculation unit 21, the irradiation angle of the laser light, and the mounting posture read in step S1.

  In addition, since the laser radar 1 in this embodiment irradiates the entire circumference of the vehicle with the laser beam by 0.1 ° from 0 ° to 360 °, the laser beam is irradiated 3601 times in one layer. Further, since the number L of layers is 32, the ECU 2 obtains a maximum of 115232 distance measuring points when the laser radar 1 scans 360 °. The reason why the maximum value is expressed here is that, among the laser beams emitted from the light emitting unit 11, the light receiving unit 12 cannot receive the reflected light and the distance measuring point may not be acquired.

  In step S3, road surface candidate point extraction processing is performed, and the process proceeds to step S4. The operation of the road surface candidate point extraction process in step S3 will be described later using a flowchart shown in FIG. 5 and a conceptual diagram shown in FIG. By performing this road surface candidate point extraction process, a distance measuring point (which is assumed to be a road surface candidate point) estimated to have detected the road surface 6 is extracted from all the distance measuring points acquired in step S2. The ECU 2 that executes step S3 corresponds to a road surface candidate point extraction unit.

  In step S4, a road surface estimation process is performed and the process proceeds to step S5. The operation of the road surface estimation process in step S4 will be described later using a flowchart shown in FIG. 7 and a conceptual diagram shown in FIG. By performing this road surface estimation process, a plane S (which is referred to as a road surface plane) S corresponding to the road surface 6 in the sensor coordinate system is estimated. The road surface plane S is represented by S = ax + by + cz + e = 0 (a, b, c, e are coefficients determined by the road surface estimation process) in the sensor coordinate system. Of course, when the sensor coordinate system and the road surface coordinate system coincide with each other, the road surface plane S coincides with the plane with z = 0. ECU2 which implements this step S4 is equivalent to the road surface plane calculation part as described in a claim.

  In step S5, a first mounting posture estimation process is performed, and the process proceeds to step S6. In the first mounting posture estimation process in step S5, the road surface plane S estimated in the previous step S4 is used to estimate the pitch angle θp, roll angle θr, and mounting height h among the mounting postures by the following method. .

  First, a method for estimating the pitch angle θp will be described with reference to FIGS. 9 and 10. FIG. 9 is a conceptual diagram of the yz plane of the sensor coordinate system. FIG. 9 shows that the surface where z = 0 (this is the reference surface) coincides with the road surface 6 (Z = 0 in the road surface coordinate system) for convenience, and the alternate long and short dash line in the figure is , A straight line (referred to as Syz) formed by intersecting the road surface plane S and the yz plane. The straight line Syz may be x = 0 on the road surface plane S, and Syz = by + cz + e = 0.

式２のように求めることができる。 Here, since the road surface plane S is a plane calculated from the road surface candidate points, the road surface plane S and the road surface 6 should essentially match. FIG. 10 shows a diagram in which the sensor coordinate system is rotated so that the road surface plane S coincides with the road surface 6. As shown in FIG. 10, the angle formed between the y-axis and the road surface plane S, that is, the angle formed between the y-axis and the straight line Syz corresponds to the pitch angle θp of the laser radar 1. The pitch angle θp (−90 ° ≦ θp ≦ 90 °) is obtained by transforming the straight line Syz = 0 with respect to z.

It can obtain | require like Formula 2.

  Next, a method for estimating the roll angle θr will be described with reference to FIGS. 11 and 12. Since the roll angle θr can be derived based on the same concept as the method for deriving the pitch angle θp described above, a part of the description will be omitted.

  FIG. 11 is a conceptual diagram of the xz plane of the sensor coordinate system. The reference plane z = 0 and the road surface 6 (Z = 0) are shown to coincide with each other, and the alternate long and short dash line in the drawing represents a straight line (Sxz) in which the road surface plane S and the xz plane intersect. . The straight line Sxz may be y = 0 on the road surface plane S, and Sxz = ax + cz + e = 0.

式４のように求めることができる。 FIG. 12 shows a diagram in which the sensor coordinate system is rotated so that the road surface 6 and the road surface plane S coincide with each other. As shown in FIG. 12, the angle formed by the x-axis and the road surface plane S, that is, the angle formed by the x-axis and the straight line Sxz corresponds to the roll angle θr of the laser radar 1. Here, the roll angle θr (−90 ° ≦ θr ≦ 90 °) is obtained by transforming the straight line Sxz = 0 with respect to z.

Equation 4 can be obtained.

The mounting height h is calculated by using, for example, a distance r to an arbitrary road surface candidate point at the irradiation azimuth angle β = 0, an irradiation elevation angle α of the laser beam that detected the road surface candidate point, and a pitch angle θp. To do. Specifically, when the irradiation azimuth angle β = 0, as shown in FIG. 13, the size of the angle formed by the laser beam that detects the road surface candidate point and the road surface 6 is determined by the irradiation of the laser beam. It is represented by the sum of the elevation angle α and the pitch angle θp. Therefore, the mounting height h can be obtained by the following formula 5.

  The pitch angle θp, roll angle θr, and mounting height h obtained as described above are stored in the storage unit 20 with a time stamp indicating the time at which they are obtained. Note that the time stamp is not limited to the one representing the time, but may be any number as long as the time series of each estimation result can be identified. The ECU 2 that performs the first attachment posture estimation process corresponds to a first attachment posture estimation processing unit described in the claims.

  In step S6, a stability evaluation process is performed. The detailed operation of the stability evaluation process in step S6 will be described later using a flowchart shown in FIG. 14 and a conceptual diagram shown in FIG. Here, the outline of the stability evaluation process will be described. In the stability evaluation process, a variation (that is, variance) σ of the estimation result is calculated from time-series data in which the estimation results of the previous mounting postures are arranged in time series, and the likelihood of the estimation result obtained in step S5 (reliability) Can also be called gender).

  In the stability evaluation process, when the variance σ is smaller than a predetermined threshold Σ, it is determined that the estimation result is stable. When the variance σ is equal to or greater than the threshold Σ, it is determined that the estimation result is not stable.

  In general, the detection result of the laser radar 1 is affected by the vehicle state and the traveling environment. Here, the vehicle state refers to whether or not a user operation such as a sudden acceleration / deceleration or a sudden turning operation is performed, and the traveling environment refers to the presence or absence of unevenness or a step on the road surface 6. For example, as an example in which the vehicle state is not stable, when a sudden braking operation is performed on the front wheels by a sudden braking operation, the rear wheels are lifted according to inertia and the vehicle posture changes. Further, as an example where the traveling environment is not stable, the vehicle posture vibrates before and after the vehicle climbs over a step.

  As described above, when the vehicle state and the traveling environment are not stable, the vehicle posture fluctuates with time, and the detection result of the laser radar 1 includes an error due to the vehicle posture. Along with this, the estimation result in step S5 also varies over time. On the other hand, when the vehicle state and the traveling environment are stable, the change in the vehicle posture is relatively small, so the estimation result is also stable.

  That is, in step S6, the fact that the estimation result is stable means that the mounting posture of the laser radar 1 is estimated in a state where both the vehicle state and the traveling environment are stable. Moreover, the fact that the estimation result is not stable means that the mounting posture of the laser radar 1 is estimated in a state where at least one of the vehicle state and the traveling environment is not stable.

  If it is determined in step S6 that the estimation result is stable, step S7 becomes YES and the process proceeds to step S8. On the other hand, if it is determined in step S6 that the estimation result is not stable, step S7 is NO and this flow is terminated.

  In step S8, a first attachment posture update process is performed, and the process proceeds to step S9. In step S8, the mounting posture used in the position calculation unit 22 is updated using the pitch angle θp, roll angle θr, and mounting height h estimated in step S5. The ECU 2 that executes step S8 corresponds to a first mounting posture update unit described in the claims.

  In step S9, straightness evaluation processing is performed. In this step S9, it is determined whether or not the vehicle 5 is traveling straight for a certain period of time (for example, 5 seconds) from the past time point to the present time. Whether or not the vehicle 5 is traveling straight may be determined from the vehicle information that the vehicle information acquisition unit 23 acquires from the vehicle information detection unit 4. For example, if the vehicle speed is greater than 0 and the yaw rate is 0, it is determined that the vehicle is traveling straight. On the other hand, in other cases, it is determined that the vehicle is not traveling straight. The case where the yaw rate is 0 includes not only the case where the yaw rate is completely 0 but also the case where the yaw rate is regarded as 0 if it is equal to or less than a predetermined threshold.

  If it is determined in step S9 that the vehicle is traveling straight, step S10 is YES, and the process proceeds to step S11. On the other hand, if it is determined in step S9 that the vehicle is not traveling straight, step S10 is NO and this flow is terminated.

  In step S11, the second mounting posture estimation process is performed, and the process proceeds to step S12. In the second mounting posture estimation process, the yaw angle θy in the mounting posture is estimated. The second mounting posture estimation process will be described with reference to FIG. The ECU 2 that performs the second mounting posture estimation processing corresponds to a second mounting posture estimation processing unit described in claims.

  The left column of Table 8 shown in FIG. 16 represents the positional relationship between the position of the laser radar 1 on the XY plane of the road surface coordinate system (that is, the origin of the road surface coordinate system) and an arbitrary distance measuring point Py. The one-dot chain line virtually represents the y-axis position of the sensor coordinate system. The angle formed by the Y axis and the y axis corresponds to the yaw angle θy. The right column of Table 8 represents the positional relationship between the position of the laser radar 1 on the xy plane of the sensor coordinate system (that is, the origin of the sensor coordinate system) and the distance measuring point Py.

  16 shows the relative position of the ranging point Py with respect to the laser radar 1 at time T-Tc, and the lower stage shows the relative position of the ranging point Py with respect to the laser radar 1 at time T. Represents. Tc is a pre-designed time, for example, 500 milliseconds.

  Originally, when the front-rear direction of the host vehicle 5 and the front direction of the laser radar 1 (that is, the y-axis direction) match, the drawings in the left column and the right column match. However, when the yaw angle θy is not 0 °, the distance measuring point Py that is actually in the position shown in the left column of Table 8 for the vehicle 5 is recognized as being in the right column of Table 8 for the ECU 2. Is done.

  Therefore, the relative position of the distance measuring point Py recognized by the ECU 2 is recognized so as to move along a locus γ as shown on the outside of Table 8 in FIG. Here, the magnitude of the angle θv between the y-axis direction of the sensor coordinate system and the relative distance measuring point locus γ is the angle between the traveling direction of the vehicle 5 (that is, the Z-axis direction of the road surface coordinate system) and the laser radar 1. That is, it is equal to the magnitude of the yaw angle θy. Also, the angle θv and the yaw angle θy are opposite in sign.

  Therefore, the mounting posture estimation unit 24 tracks an arbitrary distance measuring point Py using a known tracking (tracking) method, and calculates an angle θv formed by the locus γ of the distance measuring point Py and the y-axis direction. Thus, the yaw angle θy (= −θv) can be calculated. Here, the distance measuring point Py may be obtained from a stationary object, and a known method may be used as a method for determining whether or not the object is a stationary object. The yaw angle θy determined as described above is stored in the storage unit 20 with a time stamp indicating the time when the yaw angle θp determined in step S5 is determined.

  In step S12, the yaw angle θy used by the position calculation unit 22 is updated using the yaw angle θy estimated in step S11, and this flow ends. ECU2 which implements this step S12 is equivalent to the 2nd attachment posture update part given in a claim.

  In the above description, the case where the sensor coordinate system is set based on the initial mounting posture has been described as an example. Of course, the above-described mounting posture estimation process is repeatedly performed using the mounting posture updated in step S8 or step S12. You may implement.

  In the present embodiment, the position calculated by the position calculation unit 22 in consideration of the mounting posture of the laser radar 1 is used as the position of the road surface candidate point used when calculating the road surface plane S. For this reason, the mounting posture (for example, the pitch angle θp) newly calculated by performing the above-described mounting posture estimation process represents a difference from the currently employed mounting posture.

  Therefore, when updating the mounting posture in step S8 or step S12, it is necessary to update the value obtained by adding the newly calculated mounting posture to the currently employed mounting posture as the mounting posture with respect to the road surface 6. Similarly, when generating time series data of the mounting posture, the mounting posture at each time point needs to use a value obtained by adding the newly calculated mounting posture to the mounting posture adopted at that time. There is.

  Of course, as another configuration, when the position calculation unit 22 uses values calculated by setting the pitch angle θp, the roll angle θr, and the yaw angle θy to 0 as the position of the road surface candidate point used when calculating the road surface plane S. In (this is referred to as a first modification), the newly calculated mounting posture represents the current mounting posture as it is. Therefore, in the case of the first modification, the calculated value as it is may be used for the time series data.

<Road surface candidate point extraction processing>
The flow of road surface candidate point extraction processing performed by the mounting posture estimation unit 24 will be described using the flowchart shown in FIG. The flowchart shown in FIG. 5 is performed when the process proceeds to step S3 in the attachment posture estimation process shown in FIG.

  L used in the description of this process is the number of layers included in the laser radar 1 as described above, and is 32 in this embodiment. Moreover, lyr is an integer and is a variable used when performing the following processing. The lyr is for specifying the layer that is the target of the current processing among all the layers included in the laser radar 1, and represents the number of the layer that is the target of the current processing.

  First, in step S301, lyr is set to 0 (that is, lyr is initialized), and the process proceeds to step S302. In step S302, it is determined whether or not the variable lyr is smaller than the number L of layers. If lyr is smaller than the number L of layers, step S302 is YES, and the process proceeds to step S303. On the other hand, if lyr is greater than or equal to the number of layers L, step S302 is NO and this flow ends.

  In step S303, a distance measurement point detected by the layer having the layer number lyr is extracted from the irradiation azimuth angle-distance data generated by the distance calculation unit 21 as a result of the scanning process, and a one-dimensional signal is generated. . That is, the one-dimensional signal indicates the correspondence between the irradiation azimuth angle β and the distance r between the distance measuring points in the layer whose layer number is lyr. The ECU 2 that executes step S303 corresponds to a one-dimensional signal generation unit described in the claims.

  In step S304, a frequency analysis is performed by applying a known Fast Fourier Transform (FFT) to the one-dimensional signal generated in step S303. In the present embodiment, FFT is used as a method for decomposing a one-dimensional signal for each frequency component, but other methods may be used. For example, discrete wavelet transform (DWT) is used. There may be.

  In step S305, the low-frequency component is extracted by passing the one-dimensional signal frequency-resolved in step S304 through a known low-pass filter (LPF). The maximum value of the frequency that the low-pass filter passes (so-called cutoff frequency) may be designed as appropriate. In step S306, a known inverse Fourier transform (IFFT: Inverse FFT) is performed on the low-frequency component obtained in step S305.

  By performing the processes in steps S304 to S306, a low frequency component is extracted from the one-dimensional signal generated in step S302 (that is, the original). Here, the significance of extracting the low-frequency component of the one-dimensional signal will be described with reference to FIG.

  FIG. 6 is a graph showing the correspondence between the irradiation azimuth angle β and the distance r in a one-dimensional signal. In the graph shown in FIG. 6, the horizontal axis represents the irradiation azimuth angle β, and the vertical axis represents the distance r from the laser radar 1 to the distance measuring point. Further, the solid line portion in FIG. 6 (strictly, a row of points separated by 0.1 °) represents a distance measuring point for each irradiation azimuth angle, and the dotted line represents a low-frequency component of a one-dimensional signal obtained through the LPF. Represents.

  Normally, the laser light emitted from the light emitting unit 11 reaches the road surface 6 when there is no three-dimensional object such as a preceding vehicle, a person, or a wheel stopper in the traveling direction of the laser light. On the other hand, when the three-dimensional object exists in the traveling direction of the laser beam, the laser beam does not reach the road surface 6 and is reflected and returned by the three-dimensional object. Therefore, the distance r when the three-dimensional object is detected is smaller than the distance r when the road surface 6 is detected.

  Further, when the road surface 6 is detected, the detection result (that is, the distance r) is substantially the same distance r even in the adjacent irradiation azimuth angle, while the irradiation azimuth from the state in which the road surface 6 is detected. When the three-dimensional object is detected by changing the angle β, the distance r changes rapidly. That is, in the graph shown in FIG. 6, the portion where the distance r changes abruptly means that a three-dimensional object is detected. Further, the portion where the distance r changes abruptly appears as a high frequency component in the frequency component of the one-dimensional signal.

  Therefore, the high frequency component in the frequency component of the one-dimensional signal corresponds to a portion where a three-dimensional object is detected, and the relatively low frequency component corresponds to a portion where a road surface 6 is detected. That is, extracting the low frequency component from the one-dimensional signal generated in step S302 (ie, the original) by performing the processing in steps S304 to S306 means that the distance measurement point detecting the road surface 6, that is, the road surface This corresponds to extracting candidate points.

  Returning to FIG. 5 again, the description of the road surface candidate point extraction process will be continued. In step S307, the road surface candidate points obtained in step S306 are stored in the storage unit 20, and the process proceeds to step S308. For example, an array for storing road surface candidate points may be prepared in the storage unit 20, and the road surface candidate points may be stored in the array in the order of calculation.

  In step S308, 1 is added to lyr, and the process returns to step S302. That is, a layer whose layer number is one higher than the layer that has been the target of processing so far (that is, the next layer) is set as a target of subsequent processing, and steps S303 to S307 are performed. .

  Then, by repeating the above processing until lyr becomes L or more, road surface candidate points in all layers can be extracted.

<Road surface estimation process>
The flow of the road surface estimation process performed by the mounting posture estimation unit 24 will be described using the flowchart shown in FIG. The flowchart shown in FIG. 7 is performed when the process proceeds to step S4 in the attachment posture estimation process shown in FIG. In the following, N represents the number of all road surface candidate points acquired by the road surface candidate point extraction process, and the integer M (M ≧ 1) is a preset number of repetitions, for example, 10. Moreover, m and n are integers and are variables used in performing the following processing.

  First, in step S401, m = 0 (that is, m is initialized), and the process proceeds to step S402. In step S402, it is determined whether the variable m is smaller than the number of repetitions M. If m is smaller than the number of repetitions M, step S402 becomes YES and the process moves to step S403. On the other hand, if m is equal to or greater than the number of repetitions M, step S402 is NO and the process moves to step S430.

In step S403, three points are randomly selected from all the road surface candidate points, and the process proceeds to step S404. The road surface candidate points selected in step S403 are assumed to be Pa (x _a , y _a , z _a ), Pb (x _b , y _b , z _b ), Pc (x _c , y _c , z _c ). As the coordinates representing the positions of the points Pa, Pb, and Pc, values calculated by the position calculation unit 22 for the respective road surface candidate points may be used.

  In step S404, a temporary road surface plane St is calculated from the three road surface candidate points Pa, Pb, and Pc selected in step S403. ECU2 which implements this step S404 is equivalent to the temporary road surface plane calculation part as described in a claim. The temporary road surface plane St obtained when m = 0 is expressed as St (0), and the temporary road surface surface St calculated when m = 1 is expressed as St (1). That is, the plane St (m) represents the temporary road surface plane St calculated at an arbitrary m.

Since the temporary road surface plane St is a plane that passes through the three road surface candidate points Pa, Pb, and Pc, as shown in FIG. 8, from the outer product of the vector T that goes from Pa to Pb and the vector U that goes from Pa to Pc. It can be calculated. More specifically, assuming that each component of the vector T is (T _x , T _y , T _z ) and each component of the vector U is (U _x , U _y , U _z ), Become.

Then, when the expression representing the temporary road surface plane St in the sensor coordinate system is expressed as ax + by + cz + e = 0 using the coefficients a, b, c, e, the coefficients a, b, c calculated the outer product of the vector T and the vector U. It can obtain | require by Formula 8 obtained as a result. The coefficient e is a formula 9 using the a, b, c calculated in Equation 8, Pa coordinates _{(x a, y a, z} a) a.

  Returning to FIG. 7 again, the description of the road surface estimation process will be continued. In step S405, n = 0 and the process proceeds to step S410. In step S410, it is determined whether the variable n is smaller than the number N of road surface candidate points. If n is smaller than the number N of road surface candidate points, step S410 becomes YES and the process moves to step S411. On the other hand, if n is equal to or greater than the number N of road surface candidate points, step S410 is NO and the process proceeds to step S420.

  In step S411, a distance d between the road surface candidate point stored in the storage unit 20 and the temporary road surface plane St (m) calculated in step S404 is calculated. Note that the n + 1th road surface candidate point stored from the top is represented as P (n + 1), and the distance d between the road surface candidate point P (n) and the temporary road surface plane St (m) is represented as d (n). For example, P (0) indicates a road surface candidate point stored at the head in the array in which the storage unit 20 is provided, and d (0) is a relationship between P (0) and the temporary road surface plane St (m). Represents the distance.

The distance d (n) between the road surface candidate point P (n) and the temporary road surface plane St (m) calculated in step S404 is changed from P (n) to the temporary road surface surface St (m) as shown in FIG. The length of the dropped perpendicular, which can then be determined by Equation 10. Here, for convenience, the coordinates of the road surface candidate point P (n) are represented by (x _n , y _n , z _n ), but the coordinates calculated by the position calculation unit 22 may be used.

  In step S412, the distance d (n) calculated in step S411 is stored in the storage unit 20, and the process proceeds to step S413. In step S413, 1 is added to n, and the process returns to step S410. Thereby, in the next step S411, the distance d (n + 1) between the road surface candidate point P (n + 1) stored next and the temporary road surface plane St (m) calculated in step S404 is calculated.

  That is, by repeating steps S410 to S413 until n becomes N or more (NO in step S410), all the road surface candidate points P (0) to P (N-1), the temporary road surface plane St (m), and Distances d (0) to d (N-1) are calculated.

  In step S420, the median value d_med of d (0) to d (N-1) obtained by the above processing is obtained, and the process proceeds to step S421. The median value d_med obtained for the temporary road surface plane St (m) is represented by d_med (m). As a method for obtaining the median value d_med (m), a known method may be obtained. In the present embodiment, an array in which d (0) to d (N-1) are sorted in ascending order is generated, and a value stored in the center of the array is set as a median value d_med (m). For the sorting, a known method such as bubble sorting or quick sorting may be applied.

  If all the road surface candidate points P (0) to P (N-1) detect points on the same plane without error, the distances d (0) to d (N-1) Is also 0, and the median d_med (m) is also 0. However, in reality, in addition to being affected by the detection accuracy of the laser radar 1 itself, in the road surface candidate point extraction process in step S3, some of the distance measurement points that are detected other than the road surface 6 are also road surface candidate points. Will choose as. For this reason, the distances d (0) to d (N−1) are not necessarily zero.

  If the three points Pa, Pb, Pc selected to calculate the temporary road surface plane St (m) are road surface candidate points with relatively large errors, the road surface candidate points satisfying the distance d ≠ 0. And the distance d to each road surface candidate point becomes a relatively large value.

  Since all the road surface candidate points are selected as points detecting the road surface 6, the plane representing all the road surface candidate points is a plane corresponding to the road surface 6 in the sensor coordinate system, that is, the road surface plane S. .

  Further, since the median value d_med (m) obtained in step S420 is a value representative of the distance between the temporary road surface plane St (m) and all the road surface candidate points, the median value d_med (m) is the temporary road surface plane. This represents the distance (ie, error) between St (m) and all the road surface planes S. Therefore, the smaller the median value d_med (m), the closer to the road surface plane S. ECU2 which implements this step S420 is equivalent to the road surface likelihood evaluation part as described in a claim.

  In step S421, the median value d_med (m) obtained in step S420 is stored in association with the temporary road surface plane St (m), and the process proceeds to step S422. In step S422, 1 is added to m, and the process returns to step S402. Then, by repeating steps S402 to S422, M temporary road surface planes St and a median value d_med for each temporary road surface plane St are obtained.

  In step S430, among the M temporary road surface planes St obtained by the above process, the temporary road surface surface St having the minimum median value d_med associated with the temporary road surface surface St is determined as the road surface surface S. Set this to end this flow.

  In the present embodiment, M = 10, but other values may be used, for example, M = 50. As M is larger, the number of temporary road surface planes St generated is increased, so that a more appropriate temporary road surface surface St can be set as the road surface surface S. As another configuration, the median value d_med (m) is obtained in step S420. However, the present invention is not limited to this, and an average value of d (0) to d (N-1) may be obtained and used. Furthermore, in this embodiment, the likelihood of the temporary road surface plane as the road surface plane S is evaluated by the median value d_med of the distance d, but may be evaluated by the sum of the squares of the distance d.

<Stability evaluation process>
The flow of stability evaluation processing performed by the mounting posture estimation unit 24 will be described using the flowchart shown in FIG. The flowchart shown in FIG. 14 is implemented when it moves to step S6 in the attachment attitude | position estimation process shown in FIG. Hereinafter, the integer I is a preset number of read frames, and is set to 10, for example. Here, the number of read frames refers to the number of estimation results of the mounting posture read back from the present time to the past. Further, i is an integer and is a variable used in performing the following processing.

  As the estimation result of the mounting posture, there are at least the pitch angle θp, the roll angle θr, and the mounting height h, but here, for simplicity, the estimation result of the mounting posture used for evaluating the stability is the pitch angle. Only θp is used. Of course, as another configuration, the following stability evaluation process may be performed using another roll angle θr and the mounting height h.

  First, in step S601, i = 0 is set, and the process proceeds to step S602. In step S602, it is determined whether the variable i is smaller than the number I of read frames. If the variable i is smaller than the number I of read frames, step S602 becomes YES and the process moves to step S603. On the other hand, if the variable i is greater than or equal to the number I of read frames, step S602 is NO and the process moves to step S610.

  In step S603, the pitch angle θp calculated i frames before the current time is read, and the process proceeds to step S604. In the case of i = 0, the pitch angle θp calculated 0 frames before the current time is read out, and this indicates the pitch angle θp calculated in step S5 of the current mounting posture estimation process. When i = 1, the pitch angle θp calculated in the previous mounting posture estimation process is read.

  In step S604, 1 is added to i, the process returns to step S602, and steps S603 and S604 are repeated until the variable i becomes equal to or greater than the number I of read frames. That is, if it is determined NO in step S602, it means that a total of I = 10 pitch angles θp are read including the pitch angles θp calculated in the current mounting posture estimation process.

  In step S610, the variance σ of the latest I pitch angles θp is calculated, and the process proceeds to step S611. In step S611, it is determined whether or not the variance σ calculated in step S610 is smaller than a preset threshold value Σ. If the variance σ is smaller than the preset threshold value Σ, step S611 becomes YES and the process moves to step S612. On the other hand, if the variance σ is greater than or equal to the preset threshold Σ, step S611 becomes NO and the process moves to step S613.

  In step S612, it is determined that the estimation result is stable, and this flow ends. In step S613, it is determined that the estimation result is unstable, and this flow is terminated. In the present embodiment, whether or not the estimation result is stable is determined using the variance σ, but may be determined by a standard deviation or the like. Further, as another configuration, it may be determined that the difference is stable when the difference from the estimation result at the previous time is equal to or less than a predetermined threshold.

  Here, a state where the estimation result is stable and a state where the estimation result is not stable will be described with reference to FIG. The pitch angle θp will be described here as an estimation result, but the same applies to other elements such as the roll angle θr. FIG. 15 is a graph showing time-series data of the pitch angle θp. The vertical axis represents the pitch angle θp and the horizontal axis represents time. Note that θp0 and Δθp in FIG.

  When the current time is T, the pitch angle θp one frame before is the pitch angle estimated at time T-1. That is, since the number of read frames I = 10, the pitch angle θp from time T to time T-9 is read in steps S601 to S604 of the stability evaluation process shown in FIG. Then, the variance σ is calculated using the pitch angle θp as a population to evaluate the stability (steps S610 and S611).

  Here, since the dispersion σ is obtained by evaluating the variation of the pitch angle θp, the case where the dispersion σ is smaller than the threshold Σ means that the change in the vertical axis in FIG. Indicates that it is continuing. For example, the section from time T to T-10 shows a stable state.

  On the other hand, when the vehicle 5 climbs a step or the like, the pitch angle θp varies greatly, for example, at times T-13 and T-14 in FIG. In a case where the population includes elements with large fluctuation amounts (time fluctuation ranges described in claims), there is a high possibility that the variance σ is also equal to or greater than the threshold Σ. In other words, when the variance σ is equal to or greater than the threshold value Σ, it means that there is a change in the attitude of the vehicle body relative to the road surface 6 such as the vehicle 5 climbing a step in the time series data that is the population of the variance σ. To do.

  Further, when the vehicle 5 is traveling on an unpaved road such as a gravel road, the vehicle 5 constantly vibrates, so that the amplitude of the time-series data of the pitch angle θp increases, and the variance σ increases.

  Therefore, by performing the above-described stability evaluation process, the result estimated in the current mounting posture estimation process is estimated in a state where the relative position of the vehicle 5 (laser radar 1) to the road surface 6 is stable. It can be determined whether or not.

(Summary of embodiment)
In the above configuration, road surface candidate points are extracted from a plurality of distance measuring points obtained by the laser radar 1 performing a scanning process (step S3 in FIG. 4). The pitch angle θp, the roll angle θr, and the mounting angle of the laser radar 1 are determined from the angles formed by the road surface plane S calculated from the randomly selected road surface candidate points Pa, Pb, and Pc and the reference plane with z = 0. Estimate the height h.

  In step S11, a relative distance measurement point locus γ of an arbitrary distance measurement point Py (however, a stationary object) is calculated, and a yaw angle θy is also obtained from an angle formed by the relative distance measurement point locus y and the y axis. be able to.

  If the time variation of the mounting posture estimated by the mounting posture estimation unit is smaller than a predetermined threshold, the estimated mounting posture is adopted as the actual mounting posture (steps S8 and S12) and used for calculating the relative position. In other words, the relative position of the distance measuring point is calculated using a value obtained by estimating the actual mounting posture, not the design value or the mounting posture at the time of factory shipment.

  Therefore, even when there is an error between the initial mounting posture stored in the storage unit 20 and the design value, the relative position of the distance measuring point can be acquired more accurately. That is, the relative position error caused by the initial mounting posture error can be reduced. For this reason, it is possible to simplify the precise adjustment of the mounting posture on the production line, and it is possible to relax the requirements for the accuracy of the test environment. Therefore, the cost accompanying adjustment of the mounting posture can be reduced.

  Moreover, since the above installation attitude | position estimation process can be implemented without using test facilities, such as a reflector, it can be implemented during driving | running | working, for example. For this reason, even when the vehicle posture changes due to the change in tire air pressure, the position of an occupant or a load in the vehicle, etc., the mounting posture in the state where the vehicle posture has changed can be estimated.

  Since the position calculation unit calculates the relative position using the estimated mounting posture, it is possible to reduce the influence of the change in the vehicle posture in the calculation of the relative position. Therefore, according to the present embodiment, the relative position can be calculated more accurately even when the vehicle posture is different from the reference posture at the time of factory shipment. It can be said that the present embodiment works more effectively on a rental car or a vehicle used for car sharing that is likely to cause changes in passengers and loading capacity.

  Further, the error in the detection position of the distance measuring point due to the error between the design value and the actual mounting posture increases as the distance from the laser radar 1 increases. For this reason, this embodiment has a further effect at the time of detecting a distant object (for example, 40 m to 70 m).

  Furthermore, according to the above embodiment, since the mounting posture of the laser radar 1 can be estimated with high accuracy, the relative position of the ranging point can be calculated with higher accuracy. As a result, the road surface 6 and the non-road surface can be calculated. Separation with high accuracy is possible.

  Conventionally, application to ACC, PCS, etc. is assumed, and since the focus has been on the detection of relatively large objects such as vehicles, errors due to dynamic changes in vehicle attitude as described above are It has never been a problem.

  However, assuming application to applications such as automatic driving and automatic parking in the future, it is preferable that even relatively small objects such as curbs on both sides of the roadway and parking lots can be detected.

  According to the present embodiment, since the road surface 6 and the non-road surface can be separated with high accuracy, the present invention is applied not only to ACC and PCS, but also to applications that require higher detection accuracy such as automatic driving and automatic parking. be able to. Moreover, this embodiment can show | form a further effect by applying like the application examples 1-3 shown below.

(Application 1)
According to the embodiment described above, the amount of change in the gradient in the front-rear direction of the road surface 6 from the distance measurement point obtained by performing the update of the mounting posture at least once in the state in which the mounting posture is not deviated and obtained by the scanning process of S2. The amount of change in the gradient in the vehicle width direction can be estimated. Hereinafter, Application Example 1 will be described with reference to FIG. The state in which the mounting posture is not deviated refers to a state in which the laser radar 1 is sufficiently fixed to the vehicle 5 and the posture of the laser radar 1 with respect to the vehicle 5 does not change.

  FIG. 17 is a diagram in which the detection range by the object detection system 100 is divided into four regions, a front region Fr, a rear region Rr, a right side region Rt, and a left side region Lf. The front area Fr points to a detection area that extends in front of the vehicle 5 (including not only directly in front but also diagonally forward). Here, as an example, the irradiation azimuth angle β is in the range of −45 ° to + 45 °. Further, the rear region Rr points to a detection area extending behind the vehicle 5 (including not only the back but also an oblique rear), and the irradiation azimuth angle β is in the range of 135 ° to 225 °.

  The right side region Rt points to a detection area that extends to the right side of the vehicle 5, and the irradiation azimuth angle β is in the range of 45 ° to 135 °. The left side region Lf points to a detection area that extends to the left side of the vehicle 5, and the irradiation azimuth β is in the range of 225 ° to 315 °.

  First, a method for estimating the amount of change in the gradient in the front-rear direction using the region obtained by dividing the vehicle periphery into a plurality of parts will be described. If the processing unit that estimates the amount of change in the gradient in the ECU 2 is a gradient estimation unit, the gradient estimation unit is a road surface candidate belonging to the front area Fr among the road surface candidate points obtained by the road surface candidate point extraction process (step S3 in FIG. 4). Extract points. Then, a road surface estimation process (step S4) using the road surface candidate points belonging to the front area Fr as a population is performed to estimate the road surface plane (which is referred to as the front plane) Sfr of the front area Fr.

  In addition, the gradient estimation unit extracts road surface candidate points belonging to the rear region Rr from the road surface candidate points obtained by the road surface candidate point extraction process, and performs a similar process to the road surface plane of the rear region Rr (this is referred to as a rear plane). Estimate Srr.

  Here, the amount of change in the gradient in the vehicle front-rear direction can be estimated by obtaining the angle formed by the front plane Sfr and the rear plane Srr in the yz plane. For example, when the inclination of the front plane Sfr on the yz plane with respect to the rear plane Srr is large, it means that an upward gradient starts, as shown in FIG. 18, and the rear plane Srr and the front plane Sfr are reached. Is the amount of change in the slope.

  In addition, when the vehicle 5 is in the middle of climbing uphill, road surface candidate points can be obtained for both the front and rear of the vehicle 5, while the laser irradiated to the front region Fr as it approaches the peak of the uphill. Reflected light from the road surface 6 with respect to light is difficult to obtain. That is, the number of road surface candidate points belonging to the front area Fr decreases. Therefore, for example, when the rear plane can be calculated in the rear region Rr, but the front plane cannot be calculated due to a lack of road surface candidate points in the front region Fr, it is estimated that the vehicle has approached the top of the uphill slope. it can.

  In addition, in the vicinity of the peak of the upward gradient, the smaller the laser beam irradiation elevation angle α (the smaller the layer number), the more difficult it is to obtain the reflected light. Therefore, the number of road surface candidate points for each layer in the forward region Fr is Decrease. That is, it is possible to determine whether or not the vehicle 5 is approaching the top of the ascending slope from the time transition of the number of road surface candidate points for each layer in the front area Fr.

  Further, at the point where the uphill slope has been climbed, the incident angle of the laser beam with respect to the road surface 6 also becomes large in the rear region Rr, so that it becomes difficult to obtain the reflected light from the road surface 6. That is, the number of road surface candidate points belonging to the rear region Rr decreases.

  Therefore, while it is possible to calculate the front plane Sfr, when the rear plane Srr cannot be calculated due to the lack of road surface candidate points in the rear region Rr, it can be estimated that the peak of the uphill has been reached. If both the front plane Sfr and the rear plane Srr cannot be calculated due to a lack of road surface candidate points, it may be estimated that the vehicle is located between an uphill road and a downhill road.

  As described above, the gradient in the front-rear direction of the road surface 6 can be estimated by performing the plane estimation process using the road surface candidate points belonging to each region as the denominator. Furthermore, the position of the vehicle 5 with respect to the road surface 6 in which a gradient in the vehicle longitudinal direction exists from the number of road surface candidate points in each of the front region Fr and the rear region Rr and the number of road surface candidate points for each layer in each region. Can be estimated. The road gradient information estimated by the gradient estimation unit can be used for various applications.

  For example, when the object detection system 100 detects that the vehicle 5 has reached an upward slope, the driving support system 3 may receive the detection result and perform control for assisting the driving force.

  Moreover, in the conventional object detection system, a road having a steep uphill existing in front of the vehicle may be erroneously recognized as a three-dimensional object. When the object detection system misrecognizes a road with a steep uphill that exists in front of the vehicle as a three-dimensional object, the driving support system 3 performs a braking process (that is, a collision damage reduction brake) for avoiding a collision. Function). On the other hand, according to the object detection system 100 of the first application example, since the three-dimensional object on the road surface and the road gradient can be distinguished, the erroneous recognition and erroneous braking as described above can be suppressed.

  Furthermore, by using the gradient of the start point as a reference, for example, by integrating the amount of change in the gradient in the front-rear direction, it is possible to estimate the gradient of the currently traveling road with respect to the start point. In particular, when the starting point is a horizontal plane, the actual road gradient can be estimated.

  The method for estimating the gradient in the vehicle width direction is the same as the method for estimating the gradient in the front-rear direction. That is, the gradient estimation unit performs the road surface estimation process using the road surface candidate points belonging to the right side area Rt among the road surface candidate points obtained by the road surface candidate point extraction process, thereby performing the road surface of the right side area Rt. Estimate the plane (this is the right side plane) Srt.

  Further, the gradient estimation unit performs the road surface estimation process using the road surface candidate points belonging to the left side area Lf among the road surface candidate points obtained by the road surface candidate point extraction process, thereby obtaining the road surface of the left side area Lf. A plane (this is the left side plane) Slf is estimated.

  Then, the amount of change in the gradient in the vehicle width direction is estimated by obtaining the angle formed by the right side plane Srt and the left side plane Slf in the xz plane. For example, when the clockwise rotation angle in the xz plane is a positive value and the right side plane Srt forms a positive angle with respect to the left side plane Slf, the road surface 6 viewed from the vehicle front-rear direction is shown. It can be presumed that the cross section is convex or the road surface 6 has a right shoulder. Further, when the right side plane Srt has a negative angle with respect to the left side plane Slf, it can be estimated that the road surface 6 is concave in the vehicle width direction or is raised to the right.

  Note that the range indicated by each region described above is an example, and may be designed as appropriate. For example, as another form, a predetermined gap may be generated between the regions. That is, when the irradiation azimuth angle β is in the range of −20 ° to + 20 °, the front region Fr, the range of 160 ° to 200 ° is the rear region Rr, the range of 60 ° to 120 ° is the right region Rt, 240 ° to 300 °. May be the left side region Lf.

(Application example 2)
Furthermore, a rapid change (referred to as pitching) of the vehicle attitude can be estimated from the time series data of the mounting attitude of the laser radar 1 obtained in the above-described embodiment or the slope of the road surface 6 obtained in the above-described application example 1. . Pitching occurs when the vehicle 5 gets over a step on the road surface 6 or when the driver suddenly performs a braking operation. Hereinafter, Application Example 2 will be described with reference to FIG. Here, for the sake of simplicity, a description will be given of a configuration in which pitching in the vehicle front-rear direction is detected from time-series data of the pitch angle θp of the laser radar 1. .

  FIG. 15 is a graph showing the time-series data of the pitch angle θp as described above, and the pitch angle θp fluctuates rapidly at times T-14 and T-13. That is, it indicates that pitching has occurred at these times. Further, θp0 represents an average value of the pitch angles θp in a state determined to be stable in the stability evaluation process.

  As a method for determining whether or not pitching has occurred, the variation amount of the pitch angle θp may exceed a predetermined threshold. For example, if it is determined that pitching has occurred when the pitch angle θp estimated at a certain time (here, T-14 and T-13) is greater than or equal to a certain multiple (for example, twice) of the pitch angle θp0. Good. The threshold value used for determining whether or not pitching has occurred is not limited to the number obtained by multiplying θp0 by a predetermined multiple, and may be a constant value (for example, 1 °) added to θp0. Δθp in FIG. 15 indicates the difference between the pitch angle θp0 and the pitch angle θp at time T-14.

  As described above, it is possible to provide such information to various applications by detecting the occurrence of pitching and detecting the pitching degree Δθp.

  For example, as an existing driving support system 3, there is a system that detects a pedestrian by performing image processing on an image captured by an in-vehicle camera (not shown) and detects a distance to the pedestrian. As an example, it is assumed that the in-vehicle camera is attached in the vicinity of the roof mirror so as to capture an appropriately designed range (referred to as an imaging range) in front of the vehicle.

  In general, in such a driving support system 3, in order to reduce the processing load of the CPU that performs image processing, a range in which image processing is performed in the shooting range (this processing range is set) is set in advance. . When a pedestrian is detected in the processing range, walking based on the pre-designed camera mounting posture (installation position and shooting direction) and the position of the pedestrian's foot in the shot image. Estimate the distance to the person.

  More specifically, as shown on the left side of FIG. 19, from the distance in the photographed image from the end portion closer to the vehicle 5 (referred to as the vehicle side end portion) to the feet of the pedestrian in the photographing range of the camera. The distance from the front end of the vehicle 5 to the pedestrian is estimated. The example shown on the left side of FIG. 19 represents a case in which no pitching has occurred in the vehicle 5 and the in-vehicle camera is in a mounting posture designed in advance with respect to the road surface 6. That is, the case where the distance from the front end of the vehicle 5 to the pedestrian can be detected with high accuracy is shown, and the distance to the pedestrian at this time is assumed to be 10 m.

  However, when pitching occurs, the pedestrian may be out of the processing range and cannot detect the pedestrian, or the estimated distance to the pedestrian may deviate from the actual distance. There is.

  For example, it is shown on the right side of FIG. 19 when the pitching has occurred so that the vehicle 5 assumes a forward leaning posture. As shown on the right side of FIG. 19, when the vehicle is in the forward tilt posture, the angle formed by the center line of the viewing angle of the in-vehicle camera with the horizontal plane is larger than a predesigned value (that is, the center of the viewing angle). The line is directed toward the road surface 6), and the photographing range is closer to the front side in the traveling direction than the design value.

  At this time, the distance in the photographed image from the vehicle side end in the photographing range to the foot of the pedestrian is longer than in the case where no pitching occurs, so the distance from the vehicle 5 to the pedestrian is It is estimated to be longer than the value (10 m) (for example, 30 m).

  If such a false detection is made, there is a possibility that the collision damage reduction brake function will not operate even though the vehicle is actually approaching a pedestrian.

  According to this application example 2, since the pitching of the vehicle 5 can be detected, when pitching occurs, the pedestrian and the pedestrian can be detected by canceling the pedestrian detection process based on the captured image of the in-vehicle camera. It is possible to reduce the possibility of erroneous detection of the distance. Alternatively, by correcting the mounting posture of the in-vehicle camera with the pitching angle Δθp, it is possible to improve the accuracy of the detection distance when pitching occurs.

(Application 3)
Further, by observing the time change of the estimated road gradient and the vehicle information simultaneously from the time series data of the gradient of the road surface 6 obtained in the application example 1 described above, the vehicle posture during sudden acceleration / deceleration or corner driving is observed. Can be estimated.

  For example, the road surface gradient at the time when the vehicle is traveling stably is stored, and when the vehicle speed input from the vehicle information detection unit 4 suddenly changes, the road surface gradient estimated at the time and the storage unit 20 are stored. The difference from the road slope at the previous time is calculated. Note that this calculated difference corresponds to the inclination of the vehicle 5 due to pitching, as described above. In other words, the difference from the calculated previous time of the road surface gradient corresponds to the amount of change in the vehicle attitude with respect to the road surface 6.

  As described above, by estimating the vehicle posture at the time of sudden acceleration / deceleration or cornering, for example, a suspension (not shown) is controlled according to the vehicle posture with respect to the road surface 6, and the ride comfort and operation stability of the occupant are controlled. Etc. can be improved.

(Comparison with Patent Document 1)
Patent Document 1 discloses the following method as a method for adjusting the vertical shift (that is, the pitch angle) of the optical axis. That is, in Patent Document 1, the distance between distance measuring points of a plurality of layers irradiated at a predetermined irradiation azimuth angle is registered using a laser radar that has been confirmed to have an accurate mounting posture in advance. deep. In addition, all of these layers are layers that have an irradiation elevation angle for detecting a road surface.

  At the time of adjustment, the degree of deviation is evaluated by evaluating the degree of deviation in the vertical direction of the optical axis of the laser radar by comparing the distance between distance measuring points of the corresponding layer with a value registered in advance. The pitch angle is adjusted so that becomes zero. The yaw angle is adjusted using test equipment such as a reflector as described in the background section.

  This Patent Document 1 only mentions the shift between the pitch angle and the yaw angle, and does not consider the shift in mounting height or roll angle. However, the distance between distance measuring points of a plurality of layers actually changes not only by the pitch angle but also by the mounting height and the roll angle deviation. The deviations of the pitch angle, the mounting height, and the roll angle are sums that interact with each other, and appear as distances between distance measuring points of each layer.

  Therefore, in the method of Patent Document 1, when the mounting height or the roll angle is deviated, it cannot be adjusted to the correct mounting posture. In addition, the yaw angle cannot be adjusted without using a reflector. That is, in order to evaluate the yaw angle error, it is necessary to bring the vehicle to a repair shop or the like.

  On the other hand, in this embodiment, an error can be evaluated for each of the pitch angle θp, the roll angle θr, and the mounting height h. Further, since the error can be evaluated for the yaw angle θy without using a test facility such as a reflector, it is not necessary to bring it to a repair shop or the like when evaluating the error of the yaw angle θy.

(Modification 2)
In the road surface candidate point extraction process in the above-described embodiment, an example in which road surface candidate points are extracted from distance measurement points using FFT, LPF, and IFFT has been described, but the present invention is not limited thereto. The road surface candidate points may be obtained by performing the processing shown in FIG. 20 using all the distance measurement points acquired in step S2 of FIG. 4 as a population. The flowchart shown in FIG. 20 is performed when the process proceeds to step S3 in the attachment posture estimation process shown in FIG.

  K used in the description of the present processing is the number of angles that the irradiation azimuth angle β can take, and is 3601 because the laser beam is irradiated by 360 ° by 0.1 ° here. The integer k (0 ≦ k <K) is a variable representing the number of laser beams irradiated when the number of laser beams irradiated when the irradiation azimuth angle β = 0 is zero. That is, the variable k and the irradiation azimuth angle β satisfy the relationship β = 0.1 × k. Here, the distance r to the distance measuring point detected by the laser beam irradiated at an arbitrary k (that is, β = 0.1 × k) is represented as r (k). For example, the distance r to the distance measuring point detected by the laser beam irradiated when k = 0 (that is, β = 0) is expressed as r (0).

  First, in step T301, lyr = 0 is set, and the process proceeds to step T302. In step T302, it is determined whether or not the variable lyr is smaller than the number L of layers. When lyr is smaller than the number L of layers, step T302 becomes YES and the process moves to step T303. On the other hand, if lyr is greater than or equal to the number L of layers, step T302 is NO and this flow ends.

  In step T303, k = 1 is set, and the process proceeds to step T310. In Step T310, it is determined whether or not the variable k is smaller than the irradiation azimuth number K. If k is smaller than K, step T310 is YES and the process moves to step T311. On the other hand, if k is greater than or equal to K, step T310 is NO and the process moves to step T320.

  In step T311, the distance r (k) from the kth ranging point and the distance r (k) from the k-1 ranging point in the layer currently being processed (that is, the layer number is lyr). k-1) is calculated (this is the distance change amount) f, and the process proceeds to step T312.

  In Step T312, it is determined whether or not the distance change amount f is smaller than a preset threshold value (for example, 0.2 m) Thres. If the distance change amount f is smaller than the threshold value Thres, step T312 is YES, and the process proceeds to step T313. On the other hand, if the distance change amount f is greater than or equal to the threshold value Thres, step T312 is NO and the process moves to step T314.

  In step T313, the kth ranging point is stored in the storage unit 20 as a temporary road surface candidate point, and the process proceeds to step T314. In step T314, 1 is added to K, and the process returns to step T310. By repeating Step T310 to Step T314 until NO is determined in Step T310, a temporary road surface candidate point can be extracted from the distance measurement points in the layer having the layer number lyr.

  In step T320, by evaluating the distribution of the distance r using a plurality of temporary road surface candidate points as a population, the temporary road surface candidate points detecting the road surface 6 are extracted, and the process proceeds to step T321. More specifically, as shown in FIG. 6, the temporary road surface candidate points detecting the road surface 6 have relatively close values, and therefore can be grouped into one group in the distribution of the distance r. Moreover, since the temporary road surface candidate point which has detected the same solid object also takes a comparatively close value, it can group.

  Here, as described in FIG. 5, the distance measurement point detecting the road surface 6 has a distance r larger than the distance measurement point detecting the three-dimensional object. Therefore, it can be estimated that among the plurality of groups formed in the distribution of the distance r, the group having the largest median value is a group composed of temporary road surface candidate points detecting the road surface 6. it can. In Step T320, a temporary road surface candidate point belonging to the group having the largest median value of the distance r is extracted from the plurality of groups.

  In step T321, the temporary road surface candidate point extracted in previous step T320 is stored as a road surface candidate point, and the process proceeds to step T322. In step T322, 1 is added to lyr and the process returns to step T302. By repeating steps T302 to T320 until NO is determined in step T302, road surface candidate points in all layers are extracted.

  In the above, the laser radar 1 that sweeps and emits laser light toward 360 ° in all directions is used. Of course, the laser radar 1 has a certain angular range in the horizontal direction (for example, only the front region Fr shown in FIG. 17). Alternatively, the laser beam may be swept and irradiated. Further, the laser radar that irradiates 32 lines in the vertical direction is used.

DESCRIPTION OF SYMBOLS 100 Object detection system, 1 Laser radar, 2 ECU (object detection apparatus), 3 Driving assistance system, 4 Vehicle information detection part, 5 Vehicle, 6 Road surface, 20 Storage part, 21 Distance calculation part, 22 Position detection part, 23 Vehicle Information acquisition unit, 24 mounting posture estimation unit, S3 road surface candidate point extraction unit, S303 one-dimensional signal generation unit, S4 road surface plane calculation unit, S404 temporary road surface plane calculation unit, S420 road surface likelihood evaluation unit, S5 first mounting posture estimation Processing unit, S8 first mounting posture update unit, S11 second mounting posture estimation processing unit, S12 second mounting posture update unit

Claims (11)

Mounted on the vehicle,
Based on a signal input from a laser radar (1) that sequentially sweeps and irradiates laser light in a predetermined angle range in the horizontal direction at a plurality of different angles in the vertical direction, measurement is performed by reflecting the laser light from the laser radar. A distance calculation unit (21) for calculating the distance to the distance point;
A storage unit (20) for storing the mounting posture of the laser radar with respect to the road surface;
A position calculation unit that detects a relative position of the ranging point with respect to the vehicle based on the mounting posture stored in the storage unit, the direction in which the laser light is irradiated, and the distance calculated by the distance calculation unit. (22)
An attachment posture estimation unit (24) for estimating the attachment posture;
In the storage unit, a predetermined initial setting value is registered in advance as the mounting posture,
The mounting posture estimation unit is
The plurality of distance measuring points obtained by changing the irradiation azimuth angle that is the irradiation direction of the laser beam in the horizontal direction for each irradiation elevation angle that is the irradiation direction of the laser beam in the vertical direction. A one-dimensional signal generation unit (S303) that generates a one-dimensional signal in which the distance to the distance point and the irradiation azimuth are associated with each other;
Based on the frequency component of the one-dimensional signal generated by the one-dimensional signal generation unit, the distance measuring point detecting the road surface among the distance measuring points that are elements of the one-dimensional signal. A road surface candidate point extraction unit (S3) for extracting road surface candidate points;
A road surface plane calculating unit (S4) that calculates a road surface plane, which is a plane representing the road surface, based on the plurality of road surface candidate points extracted by the road surface candidate point extracting unit;
A first mounting posture estimation processing unit (S5) that estimates the mounting posture from an angle formed by the road surface plane calculated by the road surface plane calculation unit and a reference plane that is a plane representing the road surface determined from the mounting posture. )When,
A first mounting posture update unit (S8) that updates the mounting posture stored in the storage unit based on the mounting posture estimated by the first mounting posture estimation processing unit; Object detection device. In claim 1,
The mounting posture includes at least one of a mounting height that is a height from the road surface to the laser radar, a pitch angle of the laser radar with respect to the road surface, and a roll angle of the laser radar with respect to the road surface. An object detection apparatus characterized by that. In claim 1 or 2,
The mounting posture includes a yaw angle with respect to the road surface of the laser radar,
The object detection device performs sequential sweep irradiation,
A second attachment posture estimation processing unit (S11) for estimating the yaw angle by calculating a locus of a relative position of the same distance measurement point from the data of the distance measurement point at a plurality of time points;
A second mounting posture update unit (S12) that updates the yaw angle stored in the storage unit using the yaw angle estimated by the second mounting posture estimation processing unit; Object detection device. In any one of Claims 1-3,
The said road surface candidate point extraction part extracts the said road surface candidate point by frequency-resolving the said one-dimensional signal, and extracting the frequency component lower than the preset frequency. In any one of Claims 1-4,
The road surface plane calculating unit
A temporary road surface plane calculating unit (S404) that selects at least three points from the plurality of road surface candidate points and calculates a temporary road surface plane that is a candidate of the road surface plane from the selected road surface candidate points;
A road surface likelihood evaluation unit (S420) that evaluates the likelihood of the temporary road surface plane as a road surface plane by calculating the distance between the selected road surface candidate point other than the selected road surface candidate point and the temporary road surface plane; With
The temporary road surface plane calculating unit calculates a plurality of different temporary road surface planes by using different combinations of the road surface candidate points,
The road surface likelihood evaluation unit evaluates the likelihood for each of the plurality of temporary road surface planes,
The said road surface calculation part sets the said temporary road surface plane with the said highest likelihood to the said road surface plane, The object detection apparatus characterized by the above-mentioned. In claim 5,
The road surface likelihood evaluation unit is more likely as the median of a plurality of distances obtained by calculating the distance from the temporary road surface plane for each of the road surface candidate points other than the selected road surface candidate point is smaller. An object detection device characterized by high evaluation. In any one of Claim 1 to 6,
The storage unit stores time series data in which the mounting postures estimated by the mounting posture estimation unit are arranged in time series,
The mounting posture update unit
In the time series data in which the mounting postures estimated by the mounting posture estimation unit are arranged in time series, when the time variation width of the mounting posture estimated by the mounting posture estimation unit is smaller than a preset threshold value, An object detection apparatus that updates the mounting posture. In any one of Claims 1-7,
The laser radar sweeps and irradiates the laser light in all directions around the vehicle,
From an angle formed by a front plane that is a road plane calculated from the road surface candidate point located in front of the vehicle and a rear plane that is a road plane plane calculated from the road surface candidate point located in the rear of the vehicle, An object detection apparatus that estimates a change amount of a slope of the road surface in a vehicle front-rear direction. In claim 8,
A left side plane that is a road surface plane calculated from the road surface candidate point located on the left side of the vehicle and a right side plane that is a road surface plane calculated from the road surface candidate point located on the right side of the vehicle. An object detection apparatus that estimates an amount of change in the slope of the road surface in the vehicle width direction from an angle formed. In any one of Claim 1 to 9,
The storage unit stores time series data in which the mounting postures estimated by the mounting posture estimation unit are arranged in time series,
An object detection apparatus that estimates a change amount of a vehicle posture relative to the road surface from a change amount of the attachment posture in time-series data in which the attachment postures estimated by the attachment posture estimation unit are arranged in time series. Mounted on the vehicle,
Based on a signal input from a laser radar (1) that sequentially sweeps and irradiates laser light in a predetermined angle range in the horizontal direction at a plurality of different angles in the vertical direction, measurement is performed by reflecting the laser light from the laser radar. A distance calculation unit (21) for calculating the distance to the distance point;
A storage unit (20) for storing a yaw angle of the laser radar with respect to a road surface;
Position calculation for detecting the relative position of the distance measuring point with respect to the vehicle based on the yaw angle stored in the storage unit, the direction in which the laser beam is irradiated, and the distance calculated by the distance calculation unit Part (22);
An attachment posture estimation unit (24) for estimating the yaw angle,
In the storage unit, a predetermined initial setting value is registered in advance as the yaw angle,
The mounting posture estimation unit is
A second attachment posture estimation processing unit (S11) for estimating the yaw angle by calculating a locus of a relative position of the same distance measurement point from the data of the distance measurement point at a plurality of time points;
A second mounting posture update unit (S12) that updates the yaw angle stored in the storage unit using the yaw angle estimated by the second mounting posture estimation processing unit; Object detection device.

Images