Road curvature estimation device

The road curvature estimation device uses a stereo camera and machine learning to accurately determine road curvature by identifying lane boundaries and adjusting for potential obstructions, ensuring stable curvature estimation and smooth vehicle control.

JP2024164418APending Publication Date: 2024-11-27SUBARU CORP
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023079874
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-15
Publication Date
2024-11-27

AI Technical Summary

Technical Problem

Existing lane keeping control systems face challenges in accurately estimating road curvature at high speeds due to obstruction by preceding or oncoming vehicles, leading to erroneous recognition of lane boundaries and inaccurate curvature calculations.

Method used

A road curvature estimation device using a stereo camera system with machine learning to identify lane boundaries, combined with a curvature calculation unit that detects the maximum curvature position and sets an attention position based on vehicle speed, ensuring accurate curvature estimation by comparing and correcting for potential misrecognitions.

Benefits of technology

Enables stable and accurate road curvature setting even when lane markings are obscured, allowing for smooth vehicle control through curves by setting curvature based on the maximum detected curvature and an attention position adjusted for vehicle speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024164418000001_ABST
    Figure 2024164418000001_ABST
Patent Text Reader

Abstract

To make it possible to estimate a curvature of a road with stable accuracy even when a distant area is unclear, when the curvature of the road is estimated using image data.SOLUTION: A road curvature estimation device comprises: a camera device provided with an imaging part, and a borderline estimation part for estimating a borderline between a compartment line and a travel lane of a curve from image data for every at least one frame; and a curvature calculation device. The curvature calculation device comprises a maximum curvature position detection part, an attention position setting part, and an attention position curvature calculation part. The attention position curvature calculation part compares a maximum curvature position of the borderline detected by the maximum curvature position detection part with an attention position of the borderline set by the attention position setting part; sets a curvature of the attention position as an attention position curvature when the attention position is located on one's own vehicle side relative to the maximum curvature position; and sets a curvature of the maximum curvature position as an attention position curvature when the attention position is ahead of the maximum curvature position.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a road curvature estimation device. [Background technology]

[0002] Conventionally, when a vehicle travels straight on an expressway or a general road, a lane keeping (ALK) control is known that reduces the burden on the driver by assisting the driver in steering. This type of lane keeping control system obtains information on the driving environment in front of the vehicle using a sensing device such as a camera (on-board camera) mounted on the vehicle.

[0003] Then, based on the acquired driving environment information, the lane keeping control device recognizes the dividing lines (mainly white lines) that divide the left and right sides of the lane (driving lane) in which the vehicle is traveling. Next, the lane width is calculated from the distance between the left and right dividing lines, and a target travel path is set based on this lane width so that the vehicle will travel in the center of the driving lane. The lane keeping control device performs steering control so that the vehicle travels along the target travel path.

[0004] When the vehicle is traveling around a curve, the lane keeping control device measures (estimates) the curvature of the recognized left and right dividing lines, and sets a target course along the curve based on the curvature.

[0005] Recently, image data has been created that divides the area of ​​the vehicle's lane from the area of ​​the road other than the vehicle's lane, using machine learning such as a neural network based on an image of the road ahead captured by a camera along which the vehicle is traveling. A technology is also known that estimates the curvature of the lane along which the vehicle is traveling (the vehicle's lane) from this image data (for example, JP 2021-128612 A).

[0006] For example, as shown in FIG. 10A, image data captured by a camera is used to estimate a boundary line between the vehicle's lane and the marking lines dividing the lane to the left and right, as shown in FIG. 10B, using machine learning. Then, the curvature of this boundary line is calculated. This makes it possible to estimate the curvature with high accuracy even if the marking lines dividing the vehicle's lane to the left and right are worn or part of the marking line is blocked by an obstacle. The blank area in FIG. 10B is the space above the horizon. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Patent Publication No. 2021-128612 Summary of the Invention [Problem to be solved by the invention]

[0008] Incidentally, when the vehicle is traveling at high speed, in lane keeping control, when the vehicle is entering a curve or traveling around a curve, it is important to accurately obtain the curvature of the road over a long distance in order to ensure driving stability.

[0009] In the captured image shown in FIG. 11A, a preceding vehicle and an oncoming vehicle are captured. In this captured image, a part of a distant lane marking is blocked by the side of the preceding vehicle or the oncoming vehicle. When an estimated image of the lane marking is generated from image data of at least one frame and an attempt is made to estimate the left and right boundary lines from this estimated image, there is a possibility that the left and right boundary lines are erroneously recognized as shown in FIG. 11B. That is, in the estimated image of the boundary line shown in FIG. 11B, the part surrounded by the dashed dotted ellipse is the part of the lane marking in the captured image of FIG. 11A that is blocked by the preceding vehicle or the oncoming vehicle. For example, when an attempt is made to estimate the left and right boundary lines from image data of one frame, or even when the lane marking is unclear, there is a possibility that the outside of the preceding vehicle or the oncoming vehicle is erroneously recognized as the boundary line. Note that the blank part in FIG. 11B is the space on the horizon.

[0010] In this case, since the contours of the sides of the preceding vehicle and the oncoming vehicle extend toward the top of the image, the more distant the vehicle is, the smaller the value of the curvature that is obtained will be compared to the actual value.

[0011] An object of the present invention is to provide a road curvature estimation device that can set the curvature of a road with stable accuracy when setting the curvature of the road from image data even when the curvature is unclear in the distance. [Means for solving the problem]

[0012] The present invention relates to a road curvature estimation device that includes a camera device having an imaging unit that images the area ahead of the vehicle, and a boundary estimation unit that estimates a boundary line between a lane on a curve ahead of the vehicle and a dividing line that divides the lane on the left and right based on image data for at least one frame captured by the imaging unit, sets a plurality of boundary line positions on the boundary line, and calculates a curvature at each of the boundary line positions, and a curvature calculation device, the curvature calculation device detects a maximum curvature position that exhibits a maximum curvature from among the curvatures at the plurality of boundary line positions calculated by the boundary line estimation unit. The vehicle control system includes a maximum curvature position detection unit, a focus position setting unit that sets a focus position in front of the vehicle on the boundary line, and a focus position curvature calculation unit, wherein the focus position curvature calculation unit compares the maximum curvature position detected by the maximum curvature position detection unit with the focus position set by the focus position setting unit, and if the focus position is on the vehicle side of the maximum curvature position, sets the curvature of the focus position as the focus position curvature, and if the focus position is ahead of the maximum curvature position, sets the curvature of the maximum curvature position as the focus position curvature. Effect of the Invention

[0013] According to the present invention, the focus position curvature calculation unit compares the maximum curvature position of the boundary line with the focus position, and if the focus position is closer to the vehicle than the maximum curvature position, sets the curvature of the focus position as the focus position curvature, and if the focus position is further away than the maximum curvature position, sets the curvature of the maximum curvature position as the focus position curvature.Therefore, when setting the curvature of the road from image data for at least one frame, the curvature of the road can be set with stable accuracy even when the distant object is unclear. [Brief description of the drawings]

[0014] [Figure 1] Functional block diagram of a road curvature estimation device according to a first embodiment [Diagram 2] A flowchart showing the maximum curvature position detection routine. [Diagram 3] A flowchart showing the attention position setting routine. [Figure 4] A flowchart showing the routine for calculating the curvature at the target position. [Diagram 5] FIG. 1 is an overhead view showing the state in which the maximum curvature position is detected when entering a curved road. [Figure 6] Conceptual diagram of a focus position setting table [Figure 7] Functional block diagram of a road curvature estimation device according to a second embodiment [Figure 8] Functional block diagram of a road curvature estimation device according to a third embodiment [Figure 9] A flowchart showing the maximum curvature position detection routine. [Figure 10A] FIG. 1 shows a captured image of the area in front of a vehicle. [Figure 10B] A diagram showing an image in which the boundary line is estimated by machine learning from FIG. 8A. [Figure 11A] FIG. 1 shows a captured image of the area in front of a vehicle. [Figure 11B] FIG. 10B shows an image in which the boundary line is estimated by machine learning from FIG. 10A. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, one embodiment of the present invention will be described with reference to the drawings. [First embodiment] A first embodiment of the present invention is shown in Fig. 1 to Fig. 5. A road curvature estimation device 1 shown in Fig. 1 is mounted on a host vehicle M shown in Fig. 5. This road curvature estimation device 1 includes a camera device 2 and a curvature calculation device 3. The camera device 2 and the curvature calculation device 3 are connected so as to be able to communicate with each other.

[0016] The camera device 2 includes an imaging unit 11 and a road shape estimation unit 12. The road shape estimation unit 12 and the curvature calculation device 3 are configured with a microcontroller or the like including a CPU, a RAM, a ROM, a rewritable non-volatile memory (flash memory or EEPROM), and peripheral devices. The RAM is provided as a work area for the CPU, and various data in the CPU are temporarily stored. The ROM stores programs and fixed data required for executing each process in the CPU. The CPU is also called an MPU (Microprocessor) or a processor. Instead of the CPU, a GPU (Graphics Processing Unit) or a GSP (Graph Streaming Processor) may be used. Alternatively, a CPU, a GPU, and a GSP may be selectively combined.

[0017] The imaging unit 11 is configured with a stereo camera consisting of a main camera 11a and a sub camera 11b. The cameras 11a and 11b are installed horizontally with a predetermined baseline length, for example, above the rearview mirror, in the center of the vehicle width direction at a position close to the windshield. Each of the cameras 11a and 11b is provided with an imaging element such as a CCD or CMOS, and the driving environment in front of the vehicle M, including the lane in which the vehicle M is traveling, is captured by the imaging elements.

[0018] The imaging unit 11 captures reference image data with the main camera 11a and captures comparison image data with the sub-camera 11b. Both cameras 11a and 11b perform imaging operations at a preset frame rate (for example, 30 to 60 [fps]).

[0019] The road shape estimation unit 12 includes an image processing unit 12a and a boundary line estimation unit 12b. The image processing unit 12a performs image processing such as well-known stereo matching processing and filtering processing on the captured images of the road shape and the like captured by both cameras 11a and 11b of the imaging unit 11. Then, image data including distance information is generated from the captured images that have been subjected to predetermined image processing.

[0020] The boundary estimation unit 12b uses machine learning such as a neural network to identify the vehicle's own lane and the dividing line from the image data for each frame generated by the road shape estimation unit 12, and generates image data (estimated image) that estimates the boundary area between the vehicle's own lane and the dividing line (see Fig. 8B and Fig. 10B).Then, it calculates the curvature ρ1, ρ2...ρn[1 / R] on the boundary line at each boundary line position (X1, X2...Xn[m]) from the entrance of the curve to a predetermined point (see Fig. 5).

[0021] Then, each calculated curvature data is transmitted to the curvature calculation device 3. In Fig. 5, four boundary line positions are simply set at X1 = 30, X2 = 40, X3 = 50, and X4 = 60 [m] forward from the curve entrance, and four points are set at X1' = 30, X2' = 40, X3' = 50, and X4' = 60 [m] in the direction continuing from the constant curvature section through the transition curve section to the curve exit, but this is merely an example, and in reality the curvature is calculated for each finer boundary line position.

[0022] The curvature calculation device 3 includes a maximum curvature position detection unit 3a, a focus position setting unit 3b, and a focus position curvature calculation unit 3c. A running state detection unit 21 is connected to the maximum curvature position detection unit 3a and the focus position setting unit 3b. A vehicle speed detection unit 22 is connected to the focus position setting unit 3b.

[0023] On the other hand, the output side of the focus position curvature calculation unit 3c is connected to a driving control unit 31. This driving control unit 31 is a collective name for an output control unit that controls the output of a power unit such as an engine or an electric motor, a steering control unit that controls the steering operation, and a brake control unit that forcibly decelerates the vehicle speed.

[0024] The running state detection unit 21 detects the running state of the host vehicle M. The running state detection unit 21 detects the running state of the host vehicle M, whether the host vehicle M is running on a straight road or running around a curve, by collating the host vehicle position detected based on a positioning signal from a global navigation satellite system (GNSS), which is an external signal, with road map information of a car navigation system. The host vehicle speed detection unit 22 detects the current vehicle speed of the host vehicle M. The host vehicle speed detection unit 22 is, for example, a vehicle speed sensor. <Processing in maximum curvature position detection unit 3a> The maximum curvature position detection unit 3a detects the position showing the maximum curvature from the curvature calculated for each predetermined position on the boundary line calculated by the boundary line estimation unit 12b of the camera device 2.

[0025] Specifically, the detection of the maximum curvature position is executed according to a maximum curvature position detection routine shown in Fig. 2. In this routine, first, it is checked whether the vehicle M is about to enter the entrance of the curve (step S1). Whether the vehicle M is about to enter the entrance of the curve is determined based on map information of a car navigation system (not shown). Alternatively, it may be determined based on a captured image processed by the image processing unit 12a of the camera device 2. Incidentally, "about to enter the entrance of the curve" is set to about 10 to 30 m before the entrance of the curve. However, the distance about to enter the entrance of the curve is not limited to this.

[0026] If the current position of the vehicle M is not immediately before the entrance of the curve (NO), the process branches to step S7. If the vehicle M is immediately before entering the entrance of the curve (YES), the process proceeds to step S2. When the process proceeds to step S2, the curvatures ρ1, ρ2...ρn[1 / R] of the boundary line position (X1, X2...Xn[m]) calculated by the boundary line estimation unit 12b are read. Note that a predetermined weighting may be applied to each of these curvatures ρ1, ρ2...ρn[1 / R]. By weighting each of the curvatures ρ1, ρ2...ρn[1 / R], it becomes easier to detect the curvature as far away as possible as the maximum value in step S3 described later. Alternatively, by weighting each of the curvatures ρ1, ρ2...ρn[1 / R], it becomes easier to detect the curvature of the near boundary line position as the maximum value in step S3 described later.

[0027] Next, the maximum curvature among the curvatures is detected (step S3). As shown in Fig. 10A, many curves are connected from a straight line through a transition curve section to a constant curvature section showing the maximum curvature. When the boundary line set on this curve is estimated from the image data for each frame using machine learning, the boundary line has a trajectory similar to that of the lane markings, as shown in Fig. 10B.

[0028] On the other hand, as shown in Fig. 11A, the lane markings from the transition curve section to the constant curvature section of the curve may be blocked by obstacles such as a preceding vehicle or an oncoming vehicle. When the boundary line of such a curve is estimated from image data for each frame using machine learning, the contour of the side of the preceding vehicle or the oncoming vehicle may be mistakenly recognized as the boundary line, as shown in the area surrounded by the dashed-dotted ellipse in Fig. 11B. When the curvature is calculated based on the incorrectly recognized boundary line, the curvature of this area is small, and the curvature in the transition curve section is maximized.

[0029] Then, the calculated maximum curvature is set as the maximum curvature ρm (step S4). Furthermore, the boundary position indicating the maximum curvature ρm is set as the maximum curvature position Xm (step S5). After that, the maximum curvature ρm and the boundary position Xm are stored in memory, and the routine ends.

[0030] On the other hand, when the process branches from step S1 to step S7, it is checked whether the host vehicle M has just passed the curve exit. Whether the host vehicle M has just passed the curve exit is determined based on the running state detected by the running state detection unit 21. Alternatively, the curvature of the boundary line position X1' gradually decreases, and after finally detecting X1'=0[1 / R], it may be determined that the host vehicle M has just passed the curve exit.

[0031] If it is determined that the vehicle has passed the curved road exit (YES), the process proceeds to step S8. If it is determined that the vehicle is traveling on a curved road or on a road that has passed the curved road exit (NO), the process exits the routine. When the process proceeds to step S8, the maximum curvature ρm and the boundary line position Xm stored in the memory are reset, and the process exits the routine. <Processing in the attention position setting unit 3b> The attention position setting unit 3b sets the attention position of the boundary line based on the vehicle speed of the host vehicle M just before the curve entrance (vehicle speed when entering the curve). This attention position is, for example, a attention point required for performing feedforward control in steering control. Steering control has a specific response delay. Therefore, the faster the host vehicle speed when entering the curve, the more distant a position needs to be paid attention to.

[0032] Specifically, the focus position set by the focus position setting unit 3b is obtained according to a focus position setting routine shown in Fig. 3. When this routine is started, it is checked whether or not the vehicle has just entered the entrance of a curve (step S11). Whether or not the vehicle has just entered the entrance of a curve is determined based on the running state detected by the running state detection unit 21.

[0033] If it is determined in step S11 that the vehicle has just entered the entrance of a curve (YES), the process proceeds to step S12. If it is determined that the vehicle has not just entered a curve (NO), the process exits the routine.

[0034] In step S12, the vehicle speed when entering the curve (entry vehicle speed) detected by the vehicle speed detection unit 22 is read. In step S13, a focus position Xo on the boundary line ahead of the vehicle M is set based on the entry vehicle speed. This focus position Xo is set by referring to a focus position setting table. FIG. 6 shows the concept of the focus position setting table. The focus position Xo is, for example, a focus point required for performing feedforward control in steering control. Steering control has a specific response delay. Therefore, in the focus position setting table, the faster the entry vehicle speed is, the farther the focus position is set. This relationship between the entry vehicle speed and the focus position is previously determined and set through experiments or the like. Thereafter, the set focus position Xo is stored in memory and the routine ends (step S14). <Processing in the target position curvature calculation unit 3c> The focus position curvature calculation unit 3c calculates the curvature at the focus position Xo on the boundary line set by the focus position setting unit 3b. The curvature at the focus position Xo calculated by the focus position curvature calculation unit 3c is calculated according to a focus position curvature calculation routine shown in FIG.

[0035] When this routine is started, the boundary line position Xm indicating the maximum curvature ρm detected by the maximum curvature position detection unit 3a is read (step S21), and the focus position Xo set by the focus position setting unit 3b is read (step S22).

[0036] Then, the attention position curvature calculation unit 3c compares the attention position Xo with the boundary line position Xm (step S23). If Xo≦Xm (YES), the process proceeds to step S24. If Xo>Xm (NO), the process branches to step S25.

[0037] In step S24, the focus position curvature calculation unit 3c calculates the curvature of the focus position Xo, and the process proceeds to step S26. For example, as shown in Fig. 5, if the boundary position Xm showing the maximum curvature ρm is X3 and the focus position Xo is X2 or X3, the process proceeds to step S24. On the other hand, as described above, if the boundary position Xm showing the maximum curvature ρm is set to X3 due to erroneous recognition of the boundary, and the focus position Xo is the boundary position X4, the process branches to step S25.

[0038] Furthermore, for example, when the boundary line position Xm showing the maximum curvature ρm is the boundary line position X4 in Figure 5 and the focus position Xo is one of the boundary line positions X1 to X4, the focus position curvature calculation unit 3c applies the corresponding curvatures ρ1 to ρ4 calculated by the boundary line estimation unit 12b as they are as the curvature of the focus position Xo.

[0039] On the other hand, for example, if the boundary position Xm showing the maximum curvature ρm is X3 in Fig. 5 and the focus position Xo is between the boundary positions X3-X4, then the boundary position Xm showing the maximum curvature ρm is X3. Therefore, the curvature of the boundary position X3 is set as the curvature of the focus position Xo. Also, if the boundary position Xm showing the maximum curvature ρm is X3 and the focus position Xo is between X2-X3, then the curvature of the focus position Xo is calculated by linearly interpolating between the curvature ρ2 of the boundary position X2 and the curvature ρ3 of the boundary position X3.

[0040] In step S26, the curvature of the attention position Xo calculated in step S24 is expressed as the attention position curvature ρo Then, the process proceeds to step S28.

[0041] Moreover, when the process branches from step S23 to step S25, the maximum curvature ρm is read in. Next, this maximum curvature ρm is set as the target position curvature ρo (step S27), and the process proceeds to step S28.

[0042] When the process proceeds from step S26 or step S27 to step S28, the curvature ρo at the target position is stored in memory and the routine ends.

[0043] This attention position curvature ρo is read by the cruise control unit 31. When entering the entrance of the curve, the cruise control unit 31 sets a target vehicle speed when passing through the attention position Xo based on the vehicle speed at the time of entering and the attention position curvature ρo. Then, the cruise control unit 31 is operated to control the host vehicle M to pass through the curve smoothly.

[0044] In this way, in this embodiment, when calculating the curvature of a boundary line on a curve, first, the boundary line position Xm showing the maximum curvature ρm is obtained. Then, the curvature (focus position curvature) ρo of the focus position Xo is set with the maximum curvature ρm of this boundary line position Xm as the upper limit. As a result, when setting the curvature of a road from image data for each frame using machine learning, even if a distant lane marking is obstructed by an obstacle such as a preceding vehicle or an oncoming vehicle, making the lane marking unclear due to this obstacle and causing the boundary line to be erroneously recognized, the curvature of the boundary line as far away as possible can be set with stable accuracy. [Second embodiment] A second embodiment of the present invention is shown in Fig. 7. In this embodiment, a driving state determination unit 24 determines whether the host vehicle M is traveling on a straight road or traveling on a curve, based on the vehicle state detected by a vehicle state detection unit 23. Note that the camera device 2, the curvature calculation unit 3, and the host vehicle speed detection unit 22 have the same configuration as in the first embodiment, and therefore their explanation will be omitted.

[0045] The vehicle state detection unit 23 is pre-installed in the host vehicle M. The vehicle state detection unit 23 detects the behavior of the host vehicle M, and includes a steering angle sensor, a yaw rate sensor, a lateral acceleration sensor, and the like.

[0046] When the vehicle state detection unit 23 is a steering wheel angle sensor, the driving state determination unit 24 determines that the vehicle has entered the entrance of a curve when the steering wheel angle detected by the steering wheel angle sensor increases from the neutral position. The driving state determination unit 24 determines that the vehicle has passed the exit of a curve when the steering wheel angle detected by the steering wheel angle sensor is returned to the neutral position.

[0047] When the vehicle state detection unit 23 is a yaw rate sensor, the driving state determination unit 24 determines that the vehicle has entered the entrance of a curve when the yaw rate detected by the yaw rate sensor increases from 0. The driving state determination unit 24 determines that the vehicle has passed the exit of a curve when the yaw rate detected by the yaw rate sensor decreases and returns to 0.

[0048] When the vehicle state detection unit 23 is a lateral acceleration sensor, the driving state determination unit 24 determines that the vehicle has entered the entrance of a curve when the lateral acceleration detected by the lateral acceleration sensor increases from 0. The driving state determination unit 24 determines that the vehicle has passed the exit of a curve when the lateral acceleration detected by the lateral acceleration sensor is returned to 0.

[0049] The maximum curvature position detection unit 3a of the curvature calculation device 3 determines whether or not the vehicle has just passed the curve exit based on the determination result of the running state determination unit 24. In addition, the focus position setting unit 3b of the curvature calculation device 3 determines whether or not the vehicle has just entered the curve entrance based on the determination result of the running state determination unit 24. [Third embodiment] 8 and 9 show a third embodiment of the present invention. In this embodiment, a constant curvature zone traveling determination unit 25 is added to the configuration of the second embodiment. Note that the camera device 2, the vehicle speed detection unit 22, the focus position setting unit 3b of the curvature calculation device 3, and the focus position curvature calculation unit 3c have the same configurations as in the first embodiment, so their explanations are omitted. Also, the vehicle state detection unit 23 and the traveling state determination unit 24 have the same configurations as in the second embodiment, so their explanations are omitted.

[0050] The constant curvature section travel determination unit 25 determines whether the host vehicle M is traveling in a constant curvature section indicating the maximum curvature of the curve, based on the vehicle state detected by the vehicle state detection unit 23. That is, in the case where the vehicle state detection unit 23 is a steering angle sensor, the constant curvature section travel determination unit 25 determines that the host vehicle M is traveling in a constant curvature section of the curve when the steering angle detected by the steering angle sensor increases and becomes constant. In the case where the vehicle state detection unit 23 is a yaw rate sensor, the constant curvature section travel determination unit 25 determines that the host vehicle M is traveling in a constant curvature section of the curve when the yaw rate detected by the yaw rate sensor gradually increases and indicates a constant value.

[0051] When the vehicle state detection unit 23 is a lateral acceleration sensor, the constant curvature section driving determination unit 25 determines that the vehicle M is driving through a constant curvature section of a curve when the lateral acceleration detected by the lateral acceleration sensor increases and indicates a constant value.

[0052] The result of the determination by the constant curvature section traveling determination unit 25 is read by the maximum curvature position detection unit 3a of the curvature calculation device 3. The maximum curvature position detection unit 3a executes a maximum curvature position detection routine shown in Fig. 9, instead of the routine shown in Fig. 2 of the first embodiment. Steps S1 to S6 and S8 of this routine are the same as those of the first embodiment.

[0053] In this embodiment, the process of step S7' is executed instead of the process of step S7. That is, in step S7', it is determined whether or not the host vehicle M is traveling in a constant curvature section of a curve, based on the determination result of the constant curvature section traveling determination unit 25.

[0054] If it is determined that the vehicle M is not traveling in a constant curvature section of a curve (NO), the routine is terminated. On the other hand, if it is determined that the vehicle M is traveling in a constant curvature section of a curve (YES), the process proceeds to step S8, where the maximum curvature ρm and the boundary line position Xm stored in the memory are reset, and the routine is terminated.

[0055] The maximum curvature position detection unit 3a detects the position indicating the maximum curvature of the curve. As the vehicle M passes through a certain curve, the curvature of the curve through which the vehicle M passes gradually becomes smaller. Therefore, the position indicating the maximum curvature of the curve is the section where the curvature of the curve is constant. By resetting the maximum curvature ρm and the boundary line position Xm stored in the memory when the vehicle M is traveling through a certain curvature section of the curve, the timing of this reset can be made earlier than in the first and second embodiments. This reduces the burden on the memory.

[0056] Although not described in the above-described embodiments, the boundary line estimation unit 12b of the camera device 2 also estimates the boundary line (right boundary line) between the right-side dividing line and the driving lane using machine learning from image data for each frame captured by the imaging unit 11.

[0057] Furthermore, the boundary line estimation unit 12b may identify the vehicle's own lane and the lane markings from image data for every two or more frames. [Explanation of symbols]

[0058] 1...Road curvature estimation device, 2...Camera equipment, 3...curvature calculation device, 3a…Maximum curvature position detector, 3b...Attention position setting section, 3c... Curvature calculation unit at target position, 11...imaging unit, 11a…Main camera, 11b…Sub camera, 12...Road shape estimation section, 12a...image processing unit, 12b...boundary line estimation section, 21...running condition detection unit, 22... Vehicle speed detection unit, 23... Vehicle state detection unit, 24...Running state determination unit, 25...Constant curvature section travel determination unit, 31...cruise control unit, X1~X4,X1'~X4'...Border line position, Xm...Maximum curvature position, Xo…Position of interest, M: Vehicle, ρ1~ρ4…Curvature, ρm...maximum curvature, ρo…curvature at the target position

Claims

1. An imaging unit that images an image ahead of the vehicle; a boundary estimating unit that estimates a boundary line between a lane in a curve ahead of the vehicle and a dividing line that divides the lane into left and right sides based on image data of at least one frame captured by the imaging unit, sets a plurality of boundary line positions on the boundary line, and calculates a curvature of each of the boundary line positions; A camera device having Curvature calculation device A road curvature estimation device comprising: The curvature calculation device includes: a maximum curvature position detection unit that detects a maximum curvature position that indicates a maximum curvature from among the curvatures of the plurality of boundary line positions calculated by the boundary line estimation unit; a focus position setting unit that sets a focus position ahead of the host vehicle on the boundary line; A curvature calculation unit for a position of interest having The attention position curvature calculation unit compares the maximum curvature position detected by the maximum curvature position detection unit with the attention position set by the attention position setting unit, and if the attention position is closer to the vehicle than the maximum curvature position, sets the curvature of the attention position as the attention position curvature, and if the attention position is ahead of the maximum curvature position, sets the curvature of the maximum curvature position as the attention position curvature. A road curvature estimation device characterized by:

2. When the attention position is between the boundary line positions set by the boundary line estimation unit, the attention position curvature calculation unit calculates the curvature of the attention position by linear interpolation based on the curvatures of the boundary line positions before and after the attention position.

2. The road curvature estimating device according to claim 1.

3. The maximum curvature position detection unit stores the maximum curvature detected from among the curvatures of the multiple boundary line positions calculated by the boundary line estimation unit and a maximum curvature position indicating the maximum curvature in a memory, and resets the maximum curvature and the maximum curvature position stored in the memory when the host vehicle passes an exit of the curve or when the host vehicle travels through a certain curvature section of the curve.

2. The road curvature estimating device according to claim 1.

4. Whether or not the vehicle has passed the exit of the curve, or whether or not the vehicle is traveling in a certain curvature section of the curve is determined based on the behavior of the vehicle detected by a vehicle state detection unit.

4. The road curvature estimating device according to claim 3.

5. A vehicle speed detection unit detects a vehicle speed of the vehicle, The focus position set by the focus position setting unit is set farther away as the vehicle speed increases based on the vehicle speed detected by the vehicle speed detection unit when the vehicle enters the curve.

5. The road curvature estimation device according to claim 1, wherein the road curvature estimation device is a road curvature estimation device.

Citation Information

Patent Citations

  • Road shape estimation device

    JP2021128612A