Marking line detection device and marking line detection program

Abstract

To provide a compartment line detecting device capable of properly extracting a compartment line even if the compartment line contains a blur.SOLUTION: A compartment line detecting device 1 acquires point-group data from a point-group data generating device 2 to generate three-dimensional coordinates of a road as point-group data based on reflected light of a laser beam emitting toward a road where a dolly moves, and detects a road compartment line based on the acquired point-group data. The compartment line detecting device comprises direction vector estimation means 103 to estimate the direction vector of the compartment line based on the reflection intensity of the reflection light in the acquired point-group data, center point calculation means 104 to project the point-group data included in a prescribed interval along the direction vector estimated by the direction vector estimation means 103 and to calculate the center point of the compartment line based on the reflection intensity of the projected point-group data, and compartment line determination means 105 to determine the compartment line based on the direction vector and the center point.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a lane marking detection device and a lane marking detection program that appropriately detect lane marks. [Background technology]

[0002] In recent years, the use of point clouds obtained by MMS (Mobile Mapping System) has been increasing in order to develop high-precision road maps such as maps for autonomous driving, and to improve the efficiency and sophistication of road maintenance work.

[0003] For example, lane markings painted on the road surface, including center lines and outer edges of the road, are extremely important information for drivers and environmental recognition sensors for autonomous driving to recognize the boundaries of lanes.

[0004] Therefore, it is necessary to properly detect the marking lines painted on the road surface.

[0005] In view of this, Non-Patent Document 1 discloses a technique for extracting white lines by using the reflection intensity of a point cloud acquired by an MMS and generating a white line model by grouping measurement points on the white lines. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] Ryuji Miyazaki, Makoto Yamamoto, “Road model generation using high-density mobile measurement data (5th report) - Generation of road centerlines based on road lane markings -”, 2018 Japan Society for Precision Engineering Spring Meeting, pp.553-554 Summary of the Invention [Problem to be solved by the invention]

[0007] However, in the technology described in Patent Document 1, a white line model is generated by grouping measurement points on the white line, so the extraction target is a white line consisting of only a continuous solid line along the route. Therefore, for example, extraction is difficult in a section where the lane markings are faded.

[0008] Similarly, in the process of extracting the lane marking centerline, the maximum and minimum laser irradiation angles are used, so the positional accuracy of the centerline may decrease in sections where the lane markings are faint.

[0009] In addition, the presence of highly reflective markings, such as speed reduction marks, close to lane markings significantly affects the extraction rate, and it also becomes difficult to respond to branching and merging lane markings.

[0010] The present invention has been made in consideration of the above-mentioned problems, and aims to provide a lane line detection device and a lane line detection program that can properly extract lane lines even when the lane lines are blurred. [Means for solving the problem]

[0011] In order to achieve the above object, a first feature of the lane marking detection device according to the present invention is to A lane marking detection device that acquires point cloud data from a point cloud data generating device that generates three-dimensional coordinates of a road as point cloud data based on reflected light of a laser light irradiated onto the road along which a mobile object travels, and detects lane marks of the road based on the acquired point cloud data, a direction vector estimation means for estimating a direction vector of the lane marking based on a reflection intensity of the reflected light in the acquired point cloud data; a center point calculation means for projecting point cloud data included in a predetermined interval along the direction vector estimated by the direction vector estimation means, and calculating a center point of the lane marking based on the reflection intensity of the projected point cloud data; A demarcation line determination means for determining the demarcation line based on the direction vector and the center point; The reason is that it is equipped with the following features.

[0012] A second feature of the lane marking detection device according to the present invention is that The direction vector estimation means A threshold value that maximizes the degree of separation is calculated based on the point cloud data using a discriminant analysis method, binarized data is generated by binarizing the reflection intensity of the reflected light based on the calculated threshold value, and the direction vector is estimated based on the generated binarized data. The present invention is characterized in that:

[0013] A third feature of the lane marking detection device according to the present invention is that The center point calculation means Point cloud data included in a predetermined interval is projected along the direction vector estimated by the direction vector estimation means, a rate of change in reflection intensity of the projected point cloud data is calculated, and the center between two points where the calculated rate of change is at a peak is calculated as the center point of the division line. The point is...

[0014] A fourth feature of the lane marking detection device according to the present invention is that The center point calculation means point cloud data included in a predetermined interval is projected along the direction vector estimated by the direction vector estimation means, a value of the reflection intensity of the projected point cloud data that is equal to or greater than the threshold value calculated by the direction vector estimation means is changed to the threshold value, a rate of change of the changed reflection intensity is calculated, and the center between two points at which the calculated rate of change is at a peak is calculated as the center point of the division line. The point is...

[0015] A first feature of the lane marking detection program according to the present invention is to A lane marking detection program executed by a lane marking detection device that acquires point cloud data from a point cloud data generating device that generates three-dimensional coordinates of a road as point cloud data based on reflected light of a laser light irradiated onto the road along which a mobile object travels, and detects lane marks of the road based on the acquired point cloud data, a direction vector estimation step of estimating a direction vector of the lane marking based on a reflection intensity of the reflected light in the acquired point cloud data; a center point calculation step of projecting point cloud data included in a predetermined interval along the direction vector estimated in the direction vector estimation step, and calculating a center point of the lane marking based on the reflection intensity of the projected point cloud data; a demarcation line determination step of determining the demarcation line based on the direction vector and the center point; The object of the present invention is to have the lane marking detection device execute the above-mentioned. Effect of the Invention

[0016] According to the lane marking detection device and lane marking detection program of the present invention, lane markings can be appropriately detected. [Brief description of the drawings]

[0017] [Figure 1] 1 is a schematic diagram showing a schematic configuration of a lane marking detection system according to an embodiment of the present invention; [Diagram 2] FIG. 2 is an explanatory diagram that illustrates the generation of point cloud data by a point cloud data generating device 2 included in the lane marking detection system according to one embodiment of the present invention. [Diagram 3] 1 is an explanatory diagram showing a schematic diagram of estimation of a direction vector by a direction vector estimation means 103 provided in a lane marking detection device 1 according to an embodiment of the present invention. FIG. [Figure 4] 1 is an explanatory diagram showing a schematic diagram of a projection by a center point calculation means provided in the lane marking detection device 1 according to an embodiment of the present invention. FIG. [Diagram 5] 2 is an explanatory diagram for explaining a model of calculation of a center point by a center point calculation means provided in the lane marking detection device 1 according to one embodiment of the present invention. FIG. [Figure 6] 1 is an explanatory diagram illustrating the processing contents of the detection processing of the end points by the demarcation line determination means provided in the demarcation line detection device 1 according to one embodiment of the present invention. FIG. [Figure 7] 4 is a flowchart showing a processing procedure of the lane marking detection system according to an embodiment of the present invention. [Figure 8]8 is a flowchart showing the processes in steps S103, S107, and S115 of the flowchart showing the processing procedure of the lane marking detection system according to one embodiment of the present invention shown in FIG. 7. [Figure 9] This is a flowchart executed when failure determination occurs multiple times consecutively in step S217 of the flowchart shown in FIG. 8. [Figure 10] 1 is an explanatory diagram showing a schematic diagram of detection of a marking line when a dashed line is painted on a road surface in a marking line detection device according to an embodiment of the present invention; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. The same or equivalent parts and components are designated by the same or equivalent reference numerals throughout the drawings. However, it should be noted that the drawings are schematic and may differ from the actual product. In addition, the drawings may include parts with different dimensional relationships and ratios.

[0019] The embodiments described below are merely examples of devices for embodying the technical idea of ​​the present invention, and the technical idea of ​​the present invention does not limit the arrangement of each component to that described below. Various modifications can be made to the technical idea of ​​the present invention within the scope of the claims.

[0020] <Configuration of lane marking detection system> Hereinafter, a description will be given of a lane marking detection system including a lane marking detection device according to one embodiment of the present invention.

[0021] FIG. 1 is a schematic diagram showing the schematic configuration of a lane marking detection system according to an embodiment of the present invention.

[0022] As shown in FIG. 1, the lane marking detection system 100 is mounted on a cart (mobile body) 3 that moves along a road surface R, and includes a lane marking detection device 1, a point cloud data generation device 2, and a monitor 5.

[0023] While the dolly 3 equipped with the lane marking detection system 100 is moving, the point cloud data generating device 2 irradiates a laser light onto a road surface R of the road on which the dolly 3 is traveling. Then, the point cloud data generating device 2 generates three-dimensional coordinates of the road surface R as point cloud data based on the reflected light of the irradiated laser light. Details will be described later.

[0024] The lane marking detection device 1 acquires point cloud data from the point cloud data generation device 2, and detects lane marks based on the acquired point cloud data. The lane marks include solid lines and dashed lines painted on the road surface in, for example, white or yellow. The lane marks are used as center lines and outer edges of the roadway.

[0025] The monitor 5 displays various screens, such as a screen displaying the lane markings detected by the lane marking detection device 1.

[0026] <Generation of point cloud data by point cloud data generator 2> FIG. 2 is an explanatory diagram that illustrates the generation of point cloud data by the point cloud data generating device 2 included in the lane marking detection system 100 according to one embodiment of the present invention.

[0027] As shown in FIGS. 1 and 2, the point cloud data generating device 2 includes a laser scanner 21, a Global Navigation Satellite System (GNSS) 22, an Inertial Measurement Unit (IMU) 23, a data collecting means 24, and a temporary data storage unit 25.

[0028] The laser scanner 21 is provided behind the dolly 3, and irradiates a laser beam toward the road surface R. The reach of the irradiated laser beam is in the range of 80 m to 100 m, and for example, rotates at a cycle of 200 Hz and emits a laser beam at 1000 kHz. The laser scanner 21 receives reflected light of the emitted laser beam, and obtains laser data (Time, θ, L, I) indicating the measurement time of reception, the angle and distance to the road surface R, and reflection intensity based on the received reflected light.

[0029] The GNSS 22 acquires GNSS data (position coordinates) using signals emitted from artificial satellites.

[0030] The IMU 23 can detect the attitude of the point cloud data generating device 2. The attitudes acquired by the IMU 23 are provided to the data collecting means 24 as POS (Position and Orientation System) data in association with the GNSS data. That is, the POS data includes data (Time, xi, yi, zi, Roll, Pitch, Yaw) indicating the reference position and attitude (measurement time, position coordinates, inclination) of the IMU 23, which is a combination of the GNSS data (position coordinates) from the GNSS 22 and the detected attitude. Zi is the altitude value from the ground level at which the IMU 23 is attached.

[0031] The position coordinates of the laser scanner 21 based on the IMU 23 can be expressed as (Δx, Δy, Δz).

[0032] The data collection means 24 generates a collection of three-dimensional coordinates indicating the positions of features as point cloud data (X, Y, Z, I) based on the POS data acquired from the IMU 23 and the laser data acquired from the laser scanner 21.

[0033] The data temporary storage unit 25 temporarily stores the point cloud data (X, Y, Z, I) generated by the data collection means 24. The stored point cloud data is supplied to the lane line detection device 1. Here, the point cloud data stored in the data temporary storage unit 25 is supplied to the lane line detection device 1 in real time, but this is not limited thereto. Only the point cloud data generation device 2 may be mounted on the trolley 3, and after the point cloud data is temporarily stored, the point cloud data may be supplied from the data temporary storage unit 25 to the lane line detection device 1 when the point cloud data generation device 2 and the lane line detection device 1 are connected.

[0034] <Configuration of lane marking detection device 1> Returning to Figure 1, the lane marking detection device 1 has a point cloud data acquisition means 101, a point cloud data memory unit 102, a direction vector estimation means 103, a center point calculation means 104, a lane marking determination means 105, a lane marking data memory unit 106, and a display control means 107.

[0035] The point cloud data acquisition means 101 acquires the point cloud data supplied from the temporary data storage unit 25 of the point cloud data generation device 2, and stores the acquired point cloud data in the point cloud data storage unit .

[0036] The point cloud data storage unit 102 is configured, for example, by a hard disk drive, and stores point cloud data (X, Y, Z, I).

[0037] The direction vector estimation means 103 estimates the direction vector of the lane marking based on the reflection intensity of the reflected light in the acquired point cloud data.

[0038] 3(a) and (b) are explanatory diagrams that show a schematic diagram of the estimation of a direction vector by the direction vector estimation means 103 provided in the lane marking detection device 1 according to one embodiment of the present invention.

[0039] As shown in FIG. 3(a), the direction vector estimation means 103 sets an arbitrary point in the point cloud data as a seed point S1.

[0040] The direction vector estimation means 103 acquires point cloud data P included in a rectangle T1 centered on the seed point S1. When there are less than a predetermined number of points in the point cloud P around the seed point S1, i.e., within the rectangle T1, the direction vector estimation means 103 estimates that the estimation has failed, that is, that the seed point S1 is outside the lane line.

[0041] When a predetermined number or more of points P exist within the rectangle T1, the direction vector estimation means 103 estimates that the seed point S1 is within the lane marking.

[0042] Therefore, the direction vector estimation means 103 uses the reflection intensity values ​​of the point cloud P to classify point cloud data inside the lane lines and point cloud data outside the lane lines, and calculates a threshold value Th1 that maximizes the degree of separation using a discriminant analysis method (Otsu's binarization).

[0043] Then, the direction vector estimation means 103 performs binarization using the threshold value Th1, and sets a set of point cloud data whose reflection intensity is greater than the threshold value Th1 as point cloud data P'.

[0044] Next, the direction vector estimation means 103 calculates the distribution direction angle θ of the lane markings for the point cloud data P′ using the following (Formula 1). This allows the direction vector estimation means 103 to estimate a direction vector Q having the distribution direction angle θ.

[0045]

number

[0046] In this case, M(2,0), M(0,2), and M(1,1) are expressed as shown in Table 1.

[0047] [Table 1]

[0048] When the direction vector Q is estimated by the direction vector estimation means 103, a point moved a distance d from the seed point S1 along the direction vector Q is set as a seed point S2, as shown in Fig. 3(b). Similarly, the direction vector estimation means 103 estimates the direction vector of the lane marking based on the reflection intensity of the reflected light in the acquired point cloud data.

[0049] In this way, by moving the seed point along the direction vector Q and calculating a new direction vector Q, it is possible to estimate the direction in which the lane lines continue.

[0050] The center point calculation means 104 projects the point cloud data included in the predetermined interval Th2 along the direction vector Q estimated by the direction vector estimation means 103, and calculates the center point of the lane marking based on the reflection intensity of the projected point cloud data.

[0051] 4(a) and (b) are explanatory diagrams that typically show projection by the center point calculation means 104 provided in the lane marking detection device 1 according to one embodiment of the present invention.

[0052] 4(a), the center point calculation means 104 sets a projection target region 41 including point cloud data P' along a direction vector Q. The projection target region 41 is divided into a 10×12 mesh (10 meshes in the direction of vector Q, and 12 meshes in the direction perpendicular to vector Q) or the like.

[0053] The center point calculation means 104 generates reflection intensity projection data 42 by projecting the point cloud data P along the direction vector Q. For example, the point cloud data P for 10 meshes included in one row of the projection target area 41a along the direction vector Q of the projection target area 41 is projected, and reflection intensity projection data 42a is generated as a representative value. At this time, the representative value may be the maximum value, median value, or average value of the projection target area 41a. The center point calculation means 104 then generates reflection intensity projection data 42 by similarly making the projection target area 41 one-dimensional.

[0054] At this time, there may be cases where the point cloud data is missing.

[0055] FIG. 4(b) is an explanatory diagram illustrating a process of correcting defects in point cloud data in the lane marking detection device 1 according to an embodiment of the present invention.

[0056] Assume that the center point calculation means 104 generates reflected intensity projection data 42A as shown in FIG. 4(b).

[0057] At this time, it is assumed that the reflected intensity projection data 42A2, 42A3, and 42A6 included in the reflected intensity projection data 42A are missing data.

[0058] At this time, the center point calculation means 104 performs the correction by replacing the average values ​​of the reflected intensity projection data on both sides of the missing data.

[0059] For example, the average value of the reflection intensity projection data 42A1 and the reflection intensity projection data 42A4 sandwiched in a direction perpendicular to the directional vector Q of the missing data 42A2 and 42A3 is calculated, and the calculated average value is set as the value of the missing data 42A2 and 42A3.

[0060] Similarly, the average value of the reflected intensity projection data 42A5 and the reflected intensity projection data 42A7 sandwiched in a direction perpendicular to the directional vector Q of the missing data 42A6 is calculated, and the calculated average value is set as the value of the missing data 42A6.

[0061] In this manner, the center point calculation means 104 corrects the reflected intensity projection data 42A that includes missing data, and generates the reflected intensity projection data 42B that does not include missing data.

[0062] Then, the center point calculation means 104 changes values ​​of the generated reflection intensity projection data 42B that are equal to or greater than the threshold value Th1 to the threshold value Th1.

[0063] 5(a) to 5(c) are explanatory diagrams that typically explain the calculation of a center point by the center point calculation means 104 included in the lane line detection device 1 according to one embodiment of the present invention. FIG. 5(a) is a graph of the reflection intensity of the reflection intensity projection data 42B. FIG. 5(b) is a diagram showing values ​​obtained by changing the reflection intensity to a threshold value Th1. FIG. 5(c) is a diagram showing values ​​obtained by differentiating FIG. 5(c). In FIG. 5(a) to FIG. 5(c), the x-axis represents the position coordinate in a direction perpendicular to the direction vector Q, and the y-axis represents the reflection intensity.

[0064] As shown in FIG. 5(a), the reflection intensity 201 of the reflection intensity projection data 42B has a high value in a central portion 201b and low values ​​at both ends 201a and 201c.

[0065] Since the value of the central portion 201b may vary, it is preferable to change the value of the central portion 201b so as to make it easier to classify the point cloud data inside the lane lines and the point cloud data outside the lane lines.

[0066] Therefore, the center point calculation means 104 changes the value of the reflection intensity 201 of the reflection intensity projection data 42B that is equal to or greater than the threshold value Th1 calculated by the direction vector estimation means 103 to the threshold value Th1.

[0067] This makes it possible to obtain an altered reflection intensity 204 in which a central portion 201b of the reflection intensity 201 is altered to the threshold value Th1, as shown in FIG. 5(b).

[0068] Then, the center point calculation means 104 differentiates the changed reflection intensity 204 to calculate a reflection intensity change rate 206 as shown in FIG. 5(c).

[0069] 5(c), the reflection intensity change rate 206 reaches its peak at A1 and A2. The distance from peak A1 to peak A2 is the width of the demarcation line.

[0070] Therefore, the center point calculation means 104 calculates the center point between the peak A1 and the peak A2 of the calculated rate of change as the lane marking center point q.

[0071] At this time, if only one end point of the lane line exists in the direction perpendicular to the direction vector Q, if the width of the lane line that can be calculated from the lane line end point is greater than the threshold value Th4 or less than the threshold value Th5, or if the angle difference between the direction vector consisting of the calculated lane line center point q and the previous lane line center point q and the direction vector Q calculated by the direction vector estimation means 103 is equal to or greater than a predetermined threshold value Th3, it is determined that the extraction of the lane line has failed. Note that, as a result of experiments, it has been found that when this process is applied to a section where no lane line exists, the extraction is determined to have failed due to the determination of the lane line width or direction vector Q. If it is determined that the extraction is successful, the lane line center coordinate q is obtained by converting the lane line center point on the projection data to the original spatial coordinate system. At this time, the height (elevation value) of the lane line center point can be selected from the average height value and median value of the lane line point group P'.

[0072] The lane marking determination means 105 determines the lane marking based on the direction vector Q estimated by the direction vector estimation means 103 and the lane marking center point q calculated by the center point calculation means 104. Specifically, the lane marking is extracted based on the calculated direction vector Q and lane marking center point q until an end point appears.

[0073] 6(a) and 6(b) are explanatory diagrams illustrating the processing contents of the endpoint detection processing performed by the lane line determination means 105 included in the lane line detection device 1 according to one embodiment of the present invention.

[0074] As shown in FIG. 6(a), the lane line determination means 105 sets a seed point S at a position a distance e (eg, 3 cm) in the direction of the direction vector Q from the last successful extraction point ESP, which is the last extracted center point.

[0075] Then, as shown in Fig. 6(b), the demarcation line determination means 105 acquires point cloud data P contained within a rectangle T centered on the seed point S, and determines whether or not a predetermined number of points P exist within the rectangle T. If less than the predetermined number of points P exist within the rectangle T, the demarcation line determination means 105 determines that it is an end point.

[0076] Then, the lane line determination means 105 detects the width of the lane line by detecting peaks based on the reflection intensity projection data, calculates the center of the width of this lane line as the lane line center point q, and replaces the final extraction success point ESP with the lane line center point q.

[0077] The lane marking data storage unit 106 stores the lane markings determined by the lane marking determination means 105. The lane markings are stored as lane marking data, for example, by associating a lane marking center point q with a direction vector Q.

[0078] The display control means 107 draws the lane markings on the monitor 5 based on the lane marking data stored in the lane marking data storage unit 106 .

[0079] <Function of lane marking detection system> Next, the operation of the lane marking detection system 100 including the lane marking detection device according to one embodiment of the present invention will be described.

[0080] FIG. 7 is a flowchart showing a processing procedure of the lane marking detection system 100 according to an embodiment of the present invention.

[0081] As shown in FIG. 7, in step S101, the direction vector estimation means 103 sets an arbitrary point as a seed point S1, which is an extraction start point, in the point cloud data.

[0082] In step S103, the direction vector estimation means 103 and the center point calculation means 104 perform a process of extracting the lane line. As a result, the direction vector Q and the lane line center point q are calculated. Details are shown in FIG.

[0083] When the lane markings are extracted, in step S105, the seed point S1 is moved forward in the image by a distance d (for example, 30 cm) to set a new seed point.

[0084] Furthermore, in step S107, similarly to step S103, the direction vector estimation means 103 and the center point calculation means 104 perform the process of extracting the lane markings.

[0085] In step S109, the lane line determination means 105 determines whether or not the lane line has reached an end point in the forward direction.

[0086] In step S111, the lane marking determination means 105 extracts an end point in the forward direction. Details are shown in FIG.

[0087] Up to this point, the lane lines from the seed point S1 set in step S101 to the end points in the forward direction have been extracted.

[0088] Therefore, in step S113, the seed point S1 is moved in the image in the direction opposite to the forward direction by a distance d (for example, 30 cm) to set a new seed point.

[0089] Furthermore, in step S115, similarly to step S111, the direction vector estimation means 103 and the center point calculation means 104 perform the process of extracting the lane markings.

[0090] In step S117, the lane marking determination means 105 determines whether or not the lane marking has reached an end point in the opposite direction.

[0091] In step S119, the lane marking determination means 105 extracts the end points in the opposite direction. Details are shown in FIG.

[0092] In step S121, the lane marking determination means 105 integrates the center points from the forward end point extracted in step S111 to the reverse end point extracted in step S119 to form a lane marking.

[0093] <Extraction of lane lines> FIG. 8 is a flowchart showing the processes in steps S103, S107, and S115 of the flowchart showing the processing procedure of the lane marking detection system 100 according to one embodiment of the present invention shown in FIG.

[0094] As shown in FIG. 8, in step S201, the direction vector estimation means 103 acquires point cloud data P included in a rectangle T1 centered on a seed point S1.

[0095] In step S203, the direction vector estimation means 103 determines whether or not a predetermined number of points P exist within the rectangle T1.

[0096] When there are less than a predetermined number of points P in the rectangle T1 (step S203; NO), the direction vector estimation means 103 determines that the estimation has failed (step S217) and ends the process.

[0097] On the other hand, if a predetermined number or more of point group P exists within rectangle T1 (step S203; YES), the direction vector estimation means 103 uses the reflection intensity values ​​of point group P to calculate a threshold value Th1 that maximizes the degree of separation using a discriminant analysis method, and binarizes the points using threshold value Th1, thereby setting the set of point cloud data whose reflection intensity is greater than threshold value Th1 as point cloud data P' (step S205).

[0098] Next, the direction vector estimation means 103 calculates the distribution direction angle θ of the lane markings for the point cloud data P′, thereby estimating a direction vector Q having the distribution direction angle θ (step S207).

[0099] In step S209, the center point calculation means 104 projects the point cloud data included in the predetermined interval Th2 along the direction vector Q estimated by the direction vector estimation means 103.

[0100] In step S211, the center point calculation means 104 detects the width of the lane marking by detecting a peak based on the reflected intensity projection data.

[0101] In step S213, the center point calculation means 104 judges the likelihood of the line being a white line.

[0102] If the lane line has only one end point in the direction (width direction) perpendicular to the direction vector Q, or if the width of the lane line that can be calculated from the lane line end points is greater than threshold value Th4 or smaller than threshold value Th5, it is determined that the lane line does not appear to be a white line. In addition, the direction vector formed by the calculated lane line center point q and the previous lane line center point q will significantly deviate from the direction vector Q calculated by the direction vector estimation means 103.

[0103] Therefore, the center point calculation means 104 also determines that the line does not appear to be a white line when the angle difference is equal to or greater than a predetermined threshold value Th3. Experiments have shown that when this process is applied to a section where no lane lines exist, the width and directional vector Q of the lane lines are judged to be unsuccessful in extraction. If it is determined that the extraction is successful, the lane line center coordinate q is obtained by converting the lane line center point on the projection data into the original spatial coordinate system. At this time, the height (elevation value) of the lane line center point can be selected from the average height value or median value of the lane line point group P'.

[0104] If it is determined in step S213 that the marking is likely to be a white line (YES), the center point calculation means 104 calculates the marking line center point q based on the calculated peak of the rate of change (step S215).

[0105] <End point extraction process> FIG. 9 is a flowchart that is executed when failure determination occurs multiple times consecutively in step S217 of the flowchart shown in FIG.

[0106] As shown in FIG. 9, in step S301, the lane line determination means 105 sets the last successfully extracted point ESP, which is the center point extracted last, as a seed point.

[0107] In step S303, the lane line determination means 105 calculates a direction vector Q based on the previous seed point BS and the last successful extraction point ESP.

[0108] In step S305, the lane line determination means 105 sets a seed point S at a position away from the last successful extraction point ESP in the direction of the direction vector Q by a distance e (for example, 3 cm) that is sufficiently shorter than the distance d.

[0109] In step S307, the demarcation line determination means 105 acquires point cloud data P contained within a rectangle T having the newly set seed point S as its center.

[0110] In step S309, the demarcation line determination means 105 determines whether or not a predetermined number of point cloud data P exist within the rectangle.

[0111] If there is a predetermined number or more of point cloud data P within the rectangle (step S309; ​​YES), the demarcation line determination means 105 uses the reflection intensity values ​​of the point cloud P to calculate a threshold value Th1 that maximizes the degree of separation using a discriminant analysis method, and by binarizing using threshold value Th1, the set of point cloud data whose reflection intensity is greater than threshold value Th1 is set as point cloud data P' (step S311).

[0112] In step S313, the lane line determination means 105 generates reflected intensity projection data by projecting the point cloud data included in the predetermined interval Th2 along the direction vector Q estimated in step S303.

[0113] In step S315, the lane line determination means 105 detects the width of the lane line by detecting peaks based on the reflected intensity projection data.

[0114] In step S316, the lane marking determining means 105 determines whether the lane marking is a white line.

[0115] If it is determined in step S316 that the point is likely to be a white line (YES), in step S317, the lane line determination means 105 replaces the last successful extraction point ESP with the lane line center point q.

[0116] In step S319, the lane line determination means 105 determines whether the movement of the seed point exceeds a prescribed distance d (for example, 30 cm).

[0117] If it is determined that the movement of the seed point exceeds a prescribed amount (for example, 30 cm) (step S319; YES), the seed point is determined to be an end point and the process ends.

[0118] In this manner, the seed point is determined to be an end point until the movement of the seed point exceeds a prescribed amount (for example, 30 cm), and the process is terminated.

[0119] As described above, the lane line detection device 1, one embodiment of the present invention, is equipped with a direction vector estimation means 103 that estimates the direction vector of the lane line based on the reflection intensity of reflected light in the acquired point cloud data, a center point calculation means 104 that projects point cloud data included at a predetermined interval along the direction vector estimated by the direction vector estimation means 103 and calculates the center point of the lane line based on the reflection intensity of the projected point cloud data, and a lane line determination means 105 that determines the lane line based on the direction vector and the center point.

[0120] This prevents the positional accuracy of the center line from decreasing even in sections where the lane markings are faint, and also allows the lane markings to be properly extracted without being affected by markings with high reflective intensity, such as speed reduction marks, even when they are close to the lane markings.

[0121] In one embodiment of the present invention, a solid white line painted on the road surface is used as an example of a dividing line, but this is not limited to this, and dashed white lines and solid and dashed lines painted in yellow or other colors can also be detected in the same manner.

[0122] Figures 10(a) and (b) are explanatory diagrams that show a schematic diagram of detection of a lane marking when a broken line is painted on the road surface in the lane marking detection device 1 according to one embodiment of the present invention. Figure 10(a) is an explanatory diagram showing detection of the middle of a lane marking drawn in broken lines, and Figure 10(b) is an explanatory diagram showing detection of an end point of a lane marking drawn in broken lines. In the lane marking detection device 1, a solid line mode that detects lane marks drawn in solid lines and a broken line mode that detects lane marks drawn in broken lines can be selected, and here it is assumed that the broken line mode has been selected.

[0123] When the broken line mode is selected, as shown in FIG. 10(a), the direction vector estimation means 103 sets the previous inner end point of the broken line in the point cloud data as a seed point S3.

[0124] The direction vector estimation means 103 determines, as a seed point S4, a point that is moved a distance f from the seed point S3 along the estimated direction vector Q. The distance f is used when detecting a broken line, and is preset as a value greater than the distance d.

[0125] The direction vector estimation means 103 acquires point cloud data P contained within a rectangle T4 centered on the seed point S4, and if a predetermined number or more of the point cloud P exist within the rectangle T4, the direction vector estimation means 103 estimates that the seed point S4 is within the division line.

[0126] Therefore, the direction vector estimation means 103 uses the reflection intensity values ​​of the point cloud P to calculate a threshold value Th1 at which the degree of separation is maximized using a discriminant analysis method, and by binarizing using the threshold value Th1, the set of point cloud data whose reflection intensity is greater than the threshold value Th1 is set as point cloud data P'.

[0127] Next, the direction vector estimation means 103 calculates the distribution direction angle θ of the lane markings for the point cloud data P′, and estimates a direction vector Q having the distribution direction angle θ from the seed point S4.

[0128] On the other hand, as shown in FIG. 10(b), even if the seed point S5 is an end point, the direction vector estimation means 103 sets it as the seed point S5 detected as the previous end point within the dashed line in the point cloud data.

[0129] The direction vector estimation means 103 determines a point moved by a distance f along the estimated direction vector Q from the seed point S3 as the seed point S6.

[0130] The direction vector estimation means 103 acquires point cloud data P included in a rectangle T6 centered on the seed point S6. Here, since the seed point S5 is an end point, no broken line appears beyond that.

[0131] Therefore, since the predetermined number or more of the points P do not exist within the rectangle T6, the direction vector estimation means 103 estimates that the seed point S5 is an end point of the partition line, which is a dashed line.

[0132] In this manner, by moving the seed point along the direction vector Q and calculating a new direction vector Q, it is possible to detect a lane marking that is a broken line.

[0133] The above-described embodiment can also be realized by executing a lane marking detection program installed in a computer. [Explanation of symbols]

[0134] 1. Marking line detection device 2. Point cloud data generator 3. Cart (mobile) 5 Monitor 21 Laser Scanner 24 Data collection methods 25 Temporary Data Storage Unit 100 Lane Line Detection System 101 Point cloud data acquisition means 102 Point cloud data storage unit 103 Direction Vector Estimation Means 104 Center point calculation means 105 Means of determining lane lines 106 Lane line data storage unit 107 Display control means

Claims

1. A lane marking detection device that acquires point cloud data from a point cloud data generating device that generates three-dimensional coordinates of a road as point cloud data based on reflected light of a laser light irradiated onto the road along which a mobile object travels, and detects lane marks of the road based on the acquired point cloud data, a direction vector estimation means for estimating a direction vector of the lane marking based on a reflection intensity of the reflected light in the acquired point cloud data; a center point calculation means for converting a projection target area into one dimension by extracting a representative value of the mesh with respect to an axis perpendicular to the direction vector estimated by the direction vector estimation means based on point cloud data included in the projection target area divided into a plurality of meshes, and calculating a center point of the division line based on the reflection intensity of the one-dimensional point cloud data; A demarcation line determination means for determining the demarcation line based on the direction vector and the center point; A lane marking detection device comprising:

2. The direction vector estimation means A threshold value that maximizes the degree of separation is calculated using a discriminant analysis method based on the point cloud data, binarized data is generated by binarizing the reflection intensity of the reflected light based on the calculated threshold value, a distribution direction angle of the lane markings is calculated from the variance in the x-axis direction, the variance in the y-axis direction, and the covariance based on the generated binarized data, and the direction vector having an angle of the distribution direction angle is estimated from a predetermined seed point.

2. The lane marking detection device according to claim 1, wherein:

3. The demarcation line determination means When the solid line mode is set, if a predetermined number or more of points are not present within a predetermined rectangular range centered on a second seed point that is moved from the first seed point by the solid line distance along the estimated direction vector, the point is estimated to be an end point of the lane marking line; When the dashed line mode is set, if a predetermined number or more of points do not exist within a predetermined rectangular range centered on a second seed point that is moved from the first seed point along the estimated direction vector by a dashed line distance that is longer than the solid line distance, the point is estimated to be an end point of the division line.

2. The lane marking detection device according to claim 1, wherein:

4. The center point calculation means A rate of change in reflection intensity of the one-dimensional point cloud data is calculated, and the center between two points at which the calculated rate of change is at a peak is calculated as the center point of the division line.

2. The lane marking detection device according to claim 1,

5. The center point calculation means Among the reflection intensities of the one-dimensional point cloud data, values ​​equal to or greater than the threshold value calculated by the direction vector estimation means are changed to the threshold value, a rate of change of the changed reflection intensity is calculated, and the center between two points at which the calculated rate of change is at a peak is calculated as the center point of the division line.

3. The lane marking detection device according to claim 2.

6. A lane marking detection program executed by a lane marking detection device that acquires point cloud data from a point cloud data generating device that generates three-dimensional coordinates of a road as point cloud data based on reflected light of a laser light irradiated onto the road along which a mobile object travels, and detects lane marks of the road based on the acquired point cloud data, a direction vector estimation step of estimating a direction vector of the lane marking based on a reflection intensity of the reflected light in the acquired point cloud data; a center point calculation step of converting point cloud data included in a projection target area divided into a plurality of meshes into one-dimensional data by extracting a representative value of the mesh with respect to an axis perpendicular to the direction vector estimated by the direction vector estimation means, and calculating a center point of the division line based on the reflection intensity of the one-dimensional point cloud data; a demarcation line determination step of determining the demarcation line based on the direction vector and the center point; A lane marking detection program that causes the lane marking detection device to execute the above-mentioned lane marking detection program.

Images