JP2016057750A - Estimation device and program of own vehicle travel lane - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate an own vehicle travel lane when a lane mark is unclear or not present.SOLUTION: A lane boundary detection part 26 detects a boundary of a lane where an own vehicle is traveling, on the basis of an image created by an imaging device 14. An own vehicle travel lane estimation part 28 estimates a lane parameter of the travel lane where the own vehicle is traveling, on the basis of a vehicle travel locus corresponding to an absolute location of the own vehicle detected by the location detection part 20 and the detected lane boundary, and creates data indicating a virtual travel lane where the own vehicle is traveling, on the basis of the estimated lane parameter.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a host vehicle travel lane estimation apparatus and program.

  2. Description of the Related Art Conventionally, there is known a travel locus estimation device that has means for estimating and recording a vehicle position, and that estimates a travel locus by averaging those in the vicinity of the vehicle position from a past travel locus. (Patent Document 1).

  Further, a traveling lane recognition device that acquires vehicle trajectory information and uses a vehicle trajectory determined to be traveling in the same lane as the host vehicle for lane recognition is known (Patent Document 2).

  Non-Patent Document 1 describes a general method of detecting a lane mark and detecting the inside as a lane.

  A method is known in which a grid map representing the presence or absence of an obstacle is generated, and a route is generated based on the vehicle position and the grid map (Non-Patent Document 2).

  In addition, a method of performing travel support such as travel lane estimation based on a detailed high-accuracy map is known (Non-Patent Document 3).

JP 2008-14870 A JP2013-168016A

JC McCall and MM Trivedi, “Video-based lane estimation and tracking for driver assistance: Survey, system, and evaluation,” IEEE Transactions on Intelligent Transport Systems, vol. 7, no. 1, pp. 20-37, Mar. 2006 . G. Tanzmeister, M. Friedl, A. Lawitzky, D. Wollherr, and M. Buss, “Road course estimation in unknown, structured environments,” in IEEE Intell. Vehicles Symposium, 2013, pp. 630-635. A. Chen, A. Ramanandan, and J. Farrel, High-precision lane-level road map building for vehicle navigation, Proc. Of IEEE / ION Position Location and Navigation Symposium, 2010, pp: 1035-1042.

  In the technique described in Patent Document 1, since the target trajectory is determined by averaging past trajectories, it may be affected by outliers. Also, there is no means for determining the lane boundary.

  In the technique described in Patent Document 2, the traveling lane is recognized using the trajectory information of surrounding vehicles, but image information is not used. If there is a pattern on the road that serves as a clue to lane estimation, it is considered more effective to use it.

  Since the technique described in Non-Patent Document 1 is based on the premise of the presence of lane marks, estimation may be erroneous when a road where no lane mark exists, a road where the lane mark has deteriorated, or an intersection is passed.

  In the technique described in Non-Patent Document 2 above, it is possible to detect a route that allows obstacle avoidance traveling ahead, but it is not suitable for estimation of the own lane on a general road because it does not take into account travel regulations or travel patterns of ordinary vehicles. .

  In the technique described in Non-Patent Document 3, the own lane information can be obtained by using a highly accurate map in lane units, but it requires a costly sensor and manual map creation cost. Renewal also requires enormous costs.

  The present invention has been made in view of the above circumstances, and provides a vehicle lane estimation apparatus and program capable of estimating the vehicle lane even when the lane mark is not clear or does not exist. For the purpose.

  In order to achieve the above object, the host vehicle traveling lane estimation device of the present invention is mounted on the host vehicle, and based on an image generated by an imaging unit that images the periphery of the host vehicle and generates an image, Time-series data of the absolute position corresponding to the absolute position of the own vehicle detected by the lane boundary detecting means for detecting the boundary of the lane in which the own vehicle is traveling and the position detecting means for detecting the absolute position of the own vehicle Lane parameters of the traveling lane in which the host vehicle is traveling are estimated based on the vehicle traveling locus obtained in advance expressed by the following equation and the boundary of the lane detected by the lane boundary detecting means, and the estimated Vehicle lane estimation means for generating data representing a virtual lane in which the vehicle is traveling based on the lane parameters.

  The program according to the present invention is based on an image generated by an imaging unit that is mounted on a host vehicle and images the periphery of the host vehicle to generate an image of the lane in which the host vehicle is traveling. Vehicle obtained in advance represented by time-series data of the absolute position corresponding to the absolute position of the own vehicle detected by the lane boundary detecting means for detecting the boundary and the absolute position of the own vehicle detected by the position detecting means for detecting the absolute position of the own vehicle Based on the travel trajectory and the boundary of the lane detected by the lane boundary detection means, the lane parameter of the travel lane in which the host vehicle is traveling is estimated, and based on the estimated lane parameter, the host vehicle Is a program for functioning as own vehicle traveling lane estimation means for generating data representing a virtual traveling lane in which the vehicle is traveling.

  According to the present invention, the host vehicle is traveling based on the image that is mounted on the host vehicle by the lane boundary detection unit and that is generated by the imaging unit that captures an image of the periphery of the host vehicle and generates an image. Detect lane boundaries. Then, the vehicle obtained in advance represented by the time series data of the absolute position corresponding to the absolute position of the own vehicle detected by the position detecting means for detecting the absolute position of the own vehicle by the own vehicle traveling lane estimating means. Based on the travel trajectory and the boundary of the lane detected by the lane boundary detection means, the lane parameter of the travel lane in which the host vehicle is traveling is estimated, and based on the estimated lane parameter, the host vehicle Data representing a virtual traveling lane in which the vehicle is traveling is generated.

  As described above, based on the vehicle travel locus and the detected lane boundary, the lane parameter of the travel lane in which the host vehicle is traveling is estimated to represent a virtual travel lane in which the host vehicle is traveling. By generating the data, it is possible to estimate the own vehicle traveling lane even when the lane mark is not clear or does not exist.

  The own vehicle travel lane estimation apparatus according to the present invention estimates a boundary of a lane in which the host vehicle is traveling based on a vehicle travel locus corresponding to the absolute position of the host vehicle detected by the position detection unit. The lane boundary detection means further detects the lane boundary based on the image generated by the imaging means and the lane boundary estimated by the lane boundary estimation means. be able to. Thereby, the boundary of the lane can be detected with high accuracy.

  The host vehicle travel lane estimation device according to the present invention is based on a travel route to the destination of the host vehicle, out of the vehicle travel trajectory corresponding to the absolute position of the host vehicle detected by the position detecting means. The vehicle further includes a related trajectory selection means for selecting a vehicle travel trajectory related to the travel, and the host vehicle travel lane estimation means is detected by the vehicle travel trajectory selected by the related trajectory selection means and the lane boundary detection means. Based on the lane boundary, the lane parameter of the travel lane is estimated, and based on the estimated lane parameter, data representing a virtual travel lane in which the host vehicle is traveling is generated. be able to. This makes it possible to estimate the own vehicle traveling lane with higher accuracy.

  The own vehicle travel lane estimation apparatus according to the present invention acquires the vehicle travel trajectory in the travel direction for each of a plurality of travel directions from the vehicle travel trajectory corresponding to the absolute position of the host vehicle detected by the position detection means. A plurality of directional trajectory acquisition means, and the host vehicle travel lane estimation means includes, for each of the plurality of travel directions, the travel direction vehicle travel trajectory acquired by the multi-direction trajectory acquisition means, and the lane boundary. Data representing a virtual traveling lane in which the host vehicle is traveling based on the lane parameter of the traveling lane estimated based on the lane boundary detected by the detecting means, and based on the estimated lane parameter. Can be generated. As a result, even in a scene with a plurality of traveling directions, the vehicle traveling lane can be estimated.

  The own vehicle travel lane estimation device further includes a plurality of lane trajectory acquisition means for acquiring a travel trajectory of the travel lane for each of a plurality of travel lanes, and the host vehicle travel lane estimation means includes the plurality of travel lane estimation means Based on the vehicle travel locus of the travel lane acquired by the plurality of lane track acquisition means and the boundary of the lane detected by the lane boundary detection means for each of the travel lanes, the lane parameter of the travel lane is determined. Based on the estimated lane parameter, data representing a virtual traveling lane in which the host vehicle is traveling can be generated. Thereby, even in a scene with a plurality of traveling lanes, the own vehicle traveling lane can be estimated.

  The own vehicle travel lane estimation means corrects the vehicle travel locus based on the lane boundary detected by the lane boundary detection means, and according to the extended Kalman filter based on the corrected vehicle travel locus. The lane parameter of each control point used in the cubic spline model is estimated, and data representing the virtual traveling lane is generated based on the estimated lane parameter of each control point and the cubic spline model. can do.

  The lane boundary detection means extracts an edge point based on the image generated by the imaging means, and estimates model parameters used in the clothoid spline model according to the extended Kalman filter based on the extracted edge point. Based on the estimated model parameter and the clothoid spline model, the boundary of the lane in which the host vehicle is traveling can be detected.

  The storage medium for storing the program of the present invention is not particularly limited, and may be a hard disk or a ROM. Further, it may be a CD-ROM, a DVD disk, a magneto-optical disk or an IC card. Furthermore, the program may be downloaded from a server or the like connected to the network.

  As described above, according to the own vehicle travel lane estimation apparatus and program of the present invention, the lane parameters of the travel lane in which the host vehicle is traveling are determined based on the vehicle travel trajectory and the detected lane boundary. By estimating and generating data representing a virtual traveling lane in which the host vehicle is traveling, it is possible to estimate the host vehicle traveling lane even if the lane mark is not clear or does not exist An effect is obtained.

1 is a block diagram showing a driving support control device according to a first embodiment of the present invention. (A) The figure which shows the example of the corrected image, (b) It is a figure which shows a lane boundary detection parameter | index. (A) Figure showing edge point extraction results, (b) Figure showing lines within a section, (c) Figure showing lines labeled as consistent edge boundaries, and (d) Lane mark boundaries FIG. 3 is a diagram showing a line segment labeled as. (A) The figure which shows the estimation result of the own vehicle traveling lane by the detection result of a lane mark, (b) The figure which shows the result of estimating the own vehicle traveling lane only using the past traveling locus, (c) Data to be fused (D) Enlarged view of (c), (e) Diagram showing estimation result of own vehicle lane by the method of this embodiment, and (f) Generated by detected road boundary It is a figure which shows the made virtual lane mark. It is a figure which shows the state transition at the time of estimating a virtual own vehicle traveling lane. (A) The figure which shows the estimation result of the virtual own vehicle travel lane using a clothoid spline model, (b) The figure which shows the estimation result of the virtual own vehicle travel lane using a cubic spline model, (c) Clothoid It is a figure which shows the observation result using a spline model, and (d) The figure which shows the observation result using a cubic spline model. It is a flowchart which shows the content of the own vehicle driving | running | working lane estimation process routine in the computer of the driving assistance control apparatus which concerns on the 1st Embodiment of this invention. It is a flowchart which shows the content of the lane boundary detection process in the computer of the driving assistance control apparatus which concerns on the 1st Embodiment of this invention. It is a flowchart which shows the content of the lane estimation process in the computer of the driving assistance control apparatus which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the driving assistance control apparatus which concerns on the 2nd Embodiment of this invention. (A) The figure which shows a vehicle travel locus | trajectory, (b) The figure which visualized the space score, and (c) The figure which shows the voting range. It is a flowchart which shows the content of the own vehicle driving | running | working lane estimation process routine in the computer of the driving assistance control apparatus which concerns on the 2nd Embodiment of this invention. It is a block diagram which shows the driving assistance control apparatus which concerns on the 3rd Embodiment of this invention. It is a flowchart which shows the content of the own vehicle traveling lane estimation process routine in the computer of the driving assistance control apparatus which concerns on the 3rd Embodiment of this invention. It is a block diagram which shows the driving assistance control apparatus which concerns on the 4th Embodiment of this invention. It is a flowchart which shows the content of the own vehicle driving | running | working lane estimation process routine in the computer of the driving assistance control apparatus which concerns on the 4th Embodiment of this invention. It is a block diagram which shows the driving assistance control apparatus which concerns on the 5th Embodiment of this invention. It is a flowchart which shows the content of the own vehicle driving | running | working lane estimation process routine in the computer of the driving assistance control apparatus which concerns on the 5th Embodiment of this invention. It is a figure which shows the detection result of the lane boundary in experiment. It is a figure which shows the estimation result of the virtual own vehicle travel lane in an experiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the first embodiment, a case where the present invention is applied to a driving support control apparatus that estimates the own vehicle traveling lane at the time of performing lane keeping control and outputs it to the driving support unit will be described as an example.

<Overview>
In conventional lane keeping assistance (LKA) systems and adaptive cruise control (ACC) systems, low level lane detection results are used as they are for high level functions.

  This embodiment provides a low-level lane mark for advanced driver assistance systems (ADAS) such as lane keeping assistance (LKA) and adaptive cruise control (ACC). It is related with the method of estimating the own vehicle travel lane from the detection of lane boundaries, such as. Specifically, in the present embodiment, in the tracking method based on the extended Kalman filter (EKF), the lane boundary detection result such as a low-level lane mark is obtained by fusing the online lane boundary detection result with the previous vehicle travel locus. It provides a general framework that uses vehicle travel trajectories to improve detection performance and to generate virtual long-distance vehicle lanes.

  The LKA system and the ACC system, in particular, require estimation of the vehicle lane using an affordable sensor such as a camera. At present, the conventional LKA system and the ACC system assume that the area inside the boundary between the left and right lane marks is the own vehicle traveling lane, and the estimation of the own vehicle traveling lane depends on the detection of the lane mark. ing. However, there are mainly two types of difficulties in performing such functions in everyday traffic.

(1) In many ordinary urban streets, the lane mark is very bad and cannot be reliably detected. Typically, near intersections, or on small roads, there may be no lane mark that defines a driving lane or indicates lane curvature. In this case, there is no possibility that the own vehicle traveling lane will be estimated, resulting in a failure of the LKA function and the ACC function.

(2) Even if there is a clear lane mark, the resolution of an image patch including long-distance lane mark information is very low, so it is difficult to extract lane boundary information at a long distance. Moreover, unlike main road scenes, normal urban streets are usually cluttered and lane marks are often fragmented. As a result, it is difficult to acquire the vehicle lane information at a long distance.

  In the present embodiment, for the purpose of solving the above-described difficulty, a past vehicle travel locus is used to provide additional information. If the detection level is low, the past vehicle trajectory will be reweighted for edge point extraction, reset the segment voting range, re-scoring for segment pairing, and lane boundary tracking For each detection step, additional spatial assistance is provided to improve lane boundary detection. At a high level of understanding, past vehicle trajectories provide curvature information of the own lane at long distances, and this curvature information, in an EKF-based scheme, is used for side lane boundaries, vacant road space boundaries, and past This is used to generate a virtual travel lane of the host vehicle by fusing the vehicle travel trajectory.

  Information about the driving lane is extremely important for a wide range of next-generation driver assistance systems (ADAS). For example, a lane keeping assistance (LKA) system requires information on the own vehicle travel lane in order to keep the vehicle running inside the roadway. The adaptive cruise control (ACC) system needs information on the own vehicle lane to determine the preceding vehicle / object so as to maintain a safe distance.

  At present, the vehicle lane is often estimated by detecting a painted lane mark. A certain number of conventional approaches have focused entirely on the use of only visual sensors, while other conventional approaches have utilized LIDAR sensors, sometimes in combination with vision. LIDAR sensors are useful in non-urban areas to help detect physical road boundaries such as road fences and curbs, but only changes in the intensity of the returned laser data represent painted lane marks, Moreover, the number of LIDAR beams is very limited. Such a system can work very well in certain situations, but vision can provide a large amount of information, so vision can be used to work well in a wide range of situations. There is.

  Apart from the painted lane mark, the estimation of the vehicle lane may also be made by detecting an empty road space or a physical road boundary. The road detection result may give a safe road ahead of the host vehicle, but does not take into account traffic regulations or normal driving patterns. As a result, the estimated roadway is more suitable for a standard mobile robot than a car because the grid map / road boundary includes dynamic obstacles. Mobile robots are free to move as long as space is available, but cars are required to travel within a preset area that follows driving rules and people's expectations even on unsigned roads. Currently, such information is usually provided by a map of the environment that represents the roadway (eg, the center line of each lane) and includes additional information such as curbs, traffic signs, and the like. Generating such a map generally requires traveling the entire road network, recording the vehicle position and road data using dedicated and expensive sensors, and enormous manual post-processing efforts. To do. While large amounts of detail and accuracy can be achieved, the high cost of probe cars and the low efficiency of manual processing can produce such map data or maintain the benefits that come from current ADAS systems. Make updating difficult.

  In this embodiment, the accumulated vehicle travel trajectory provides a lane level cue to the traffic scene and may allow inferences regarding relationships / interactions between traffic objects, so A method for robustly estimating the own vehicle traveling lane based on the vehicle traveling locus is presented.

<First Embodiment>
<System configuration>
As shown in FIG. 1, the driving assistance control apparatus 10 according to the first embodiment of the present invention captures an image of a GPS receiver 12 that receives radio waves from GPS satellites and the front of the host vehicle. Based on the imaging device 14 to be generated, the received signal from the GPS satellite received by the GPS receiver 12, and the image captured by the imaging device 14, a virtual traveling lane of the host vehicle is generated to assist driving. And a computer 16 for outputting to the unit 18.

  The driving support unit 18 performs lane keeping control of the host vehicle, for example, based on the generated virtual traveling lane of the host vehicle.

  The GPS receiver 12 receives radio waves from a plurality of GPS satellites at each time and outputs received signals from all the received GPS satellites to the computer 16.

  The imaging device 14 repeatedly captures the front of the vehicle for each time, repeatedly generates an image, and outputs the image to the computer 16. Note that the imaging device 14 may generate a stereo image.

  When the computer 16 is represented by functional blocks, as shown in FIG. 1, the GPS satellite information is obtained from the GPS receiver 12 for all GPS satellites that have received radio waves, and the absolute position of the vehicle at each time is obtained. The position detector 20 that calculates the time, the vehicle travel locus generator 22 that generates time series data of the absolute position of the host vehicle calculated by the position detector 20 as the vehicle travel locus, and the vehicle travel locus generator 22 A lane boundary detection unit 26 that detects a lane boundary of a traveling lane of the host vehicle and a lane boundary detection unit 26 that detect the lane boundary of the traveling lane of the host vehicle from the trajectory database 24 that stores the generated vehicle travel trajectory and the image generated by the imaging device 14. Lane boundary, the absolute position of the host vehicle detected by the position detection unit 20, and the vehicle travel gauge stored in the trajectory database 24 Based on, and a current lane estimation unit 28 for estimating a virtual lane of the host vehicle.

  The position detection unit 20 acquires GPS satellite information for all GPS satellites that have received radio waves from the GPS reception unit 12, and calculates the absolute position of the host vehicle based on the acquired GPS information. In the present embodiment, the absolute position of the host vehicle is obtained by using the information of GPS satellites and information obtained by an INS (Inertial Navigation System) sensor (not shown), as disclosed in JP-A-2013-130480. It is calculated by the same method as the positioning system described in the reference). In the positioning system method described in Patent Document 3, since the GPS and INS (Inertial Navigation System) sensors are used, the absolute position of the host vehicle has an error. In general, the absolute position error ranges from 1 m to 3 m. As the INS sensor, for example, an acceleration sensor, a geomagnetic sensor, a gyro sensor, or the like may be used.

  The vehicle travel locus generation unit 22 collects time series data of the absolute position of the host vehicle calculated by the position detection unit 20 using the same method as the positioning system described in Patent Document 3, and generates a vehicle travel locus. Stored in the trajectory database 24.

  The lane boundary detection unit 26 detects the lane boundary of the vehicle lane from the image generated by the imaging device 14.

  Here, the principle of detecting the lane boundary by the lane boundary detection unit 26 will be described.

  Road lane boundaries are typically indicated by painted lane marks or other consistent strength / texture changes that define safe and legal driving areas. As shown in FIG. 4 (a) below, the left boundary is indicated by a consistent high-low intensity edge and the right boundary is marked by a dashed lane mark that is a pair of low-high-low intensity edges. It is done. Finding these boundaries, along with locating these types, is crucial for many ADAS systems.

Generally, the lane boundary detection process is classified into four steps: 1) edge point extraction, 2) feature line segment voting, 3) feature line segment grouping, and 4) lane boundary estimation. In this embodiment, the captured original image is first modified by inverse perspective mapping (IPM), and then the lane boundary is divided into four horizontal sections S ₀ , S ₀ , as shown in FIG. . . . , Is divided into S _3, it is detected in the modified image. Usually, clothoid models approximated using Taylor expansion are used to represent side lane boundaries. As a result, the detected lane boundary position is expressed by the following equation.

However, x _{k, t} (z) is a lateral position of the lane boundary at the distance z and the time point t. x _{k, t} (z) is the lateral position offset x _{0, t} of the lane center line with respect to the host vehicle, the yaw angle θ _t of the traveling of the host vehicle with respect to the lane, the curvature c _{0, t,} and the curvature variation c _{1. , T} and These parameters form the estimated state vector S _t , namely S _t = (x _{0, t} , θ _t , c _{0, t} , c _{1, t} ).

FIG. 2 (b) shows a lane boundary detected index, D _L and D _R each represent left and right border detection. In the example of FIG. 2B, S _{0 to} S ₂ mean “boundary detection” in the image, and S ₃ means “boundary non-detection”. S ₀ ^left,. . . , S ₃ ^left and S ₀ ^right,. . . , S ₃ ^right are the left and right horizontal sections, respectively. T _L and T _R represent the type of left and right boundaries (consistent edge boundary, solid lane mark boundary, dashed lane mark boundary, respectively). In the example of FIG. 2 (b), T _L represents a consistent edge boundary, T _R represents the broken-line lane mark boundary.

  Next, the edge point extraction step will be described.

  When an input image is given, first, an edge point is extracted together with each horizontal scanning line in the captured original image by the edge intensity Sep (u, v) calculated by the following equation (2).

However, S _ep (u, v) is the intensity of the point (u, v) as an edge point, S _ep > 0 at the low-high intensity edge point, and S _ep at the high-low intensity edge point. _ep <0. I (•) is the intensity, and F is the size of the differential filter. A point having an edge strength whose absolute value is larger than the threshold value is stored as shown in FIG.

  Next, the feature line segment voting step will be described.

The stored edge points are projected onto the modified image coordinate system. Next, the edge points in sections S ₁ and S ₂ are grouped into a low-high intensity line segment and a high-low intensity line segment, respectively, based on the Hough transform voting scheme. For each section, the parameter voting space (r, α) is defined using the origin O (u ₀ , v ₀ ) at the midpoint of the bottom row of the section, as shown in the embodiment of FIG. . The parameter r represents the distance between the line and O, and α is the angle of the vector from O to this nearest point. Subsequently, for each stored edge point in the section with coordinates (u, v), the line through this edge point is

r (α) = (u−u ₀ ) cos α + (v−v ₀ ) sin α (3)

Is a pair (r, α). Where α extends to the range [−α _max , α _max ] to calculate the corresponding r. The pair (r, α) will be saved for voting if the result r belongs to the range [−r _max , r _max ]. α _max and r _max are positive constants to define the size of the voting space. The pair with the maximum of the voted value indicates a line segment within the section, as shown in FIG. In addition, the average edge strength of each line segment is calculated. Note that strong shadows along the road often generate line segments that can affect lane boundary groupings. In order to remove the effects of shadows, the line segment will be discarded if the average intensity of neighboring points is below the threshold.

  Next, the step of grouping (pairing) feature line segments will be described.

  Based on the acquired line segment, determination processing for finding a line segment pair is performed by the following two methods.

(1) between the section S ₁ and S _2, to find a consistent edge boundary, the line segment of the same type pairs are retrieved. For each pair candidate, between the condition regarding the average edge strength of each line segment, the condition regarding the difference in the direction α between the line segments, and the upper end point of the line segment in the section S ₁ and the lower end point of the line segment in the section S ₂ It is determined whether or not the condition group including the condition relating to the algebraic distance is satisfied. If all of the conditions are met, the two line segments will be paired and labeled as a consistent edge boundary, as shown in FIG. 3 (c).

(2) In the same section S ₁ or S _2, left lower - high strength line and the right high - pairs of low intensity line is searched. For each pair candidate, a condition regarding the average edge strength of each line segment, a condition regarding the direction α difference, a condition regarding the algebraic distance between the line segments, and a condition regarding the average intensity of the points between these two line segments, It is determined whether or not the condition group including the condition is satisfied. If all of the condition groups are met, the inner line segment will be labeled as a lane mark boundary, as shown in FIG.

  For a line that satisfies both of these groups of conditions, this line will have both labels. If the boundary is a newly found boundary, the lane mark boundary label will have a higher priority. Otherwise, the label that matches the tracked boundary will be used.

  Next, steps for estimating the lane boundary will be described.

  Lane boundaries existing on the left and right of the driving lane of the host vehicle are estimated and tracked by using two EKF filters independently. In this embodiment, a vector

Is used instead of _St.

  In addition, the line segment with the consistent edge boundary label or the lane mark boundary label acquired corresponding to the lane boundary existing on the left of the driving lane of the host vehicle is the EKF filter of the left lane boundary. Used as an observation. In addition, a line segment with a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary existing to the right of the driving lane of the host vehicle is the EKF filter of the right lane boundary. Used as an observation.

  As described above, the lane boundary detection unit 26 executes the steps of edge point extraction, feature line segment voting, feature line segment grouping, and lane boundary estimation based on the image generated by the imaging device 14. The left and right lane boundaries of the traveling lane of the host vehicle are detected.

  Next, the principle of estimating the virtual vehicle lane by the vehicle lane estimation unit 28 will be described.

  Understanding nearby obstacles in road conditions along with the road structure itself, such as the vehicle lane, is key to successful ADAS application in an urban environment. This is a very difficult problem especially in a dynamic urban scene where enormous confusion has occurred, or where there is no lane mark indicating a road structure. The conventional system determines the vehicle lane by detecting the lane boundary. However, as shown in FIG. 4 (a), lane boundary detection may lack stability or be more inaccurate due to the aforementioned challenges. In addition, it is difficult to detect a far-range lane boundary in an urban scene because it can hardly be seen with an in-vehicle camera.

  Here, FIG. 4 shows a result of generating a virtual own vehicle traveling lane by fusing the lane boundary and the past vehicle traveling locus. FIG. 4A shows an estimation result of the own vehicle traveling lane based on the detection result of the lane mark as a comparative example. FIG. 4B shows a result of estimating the own vehicle travel lane using only the past travel trajectory as a comparative example. 4C and 4D show the observation results of the data to be merged. The various lines represent splines in which the vehicle travel locus, the left and right boundary observations, the past locus, and the left and right lane boundaries are respectively applied. FIG. 4 (e) represents the estimation result of the host vehicle lane by the method of the present embodiment, and FIG. 4 (f) represents a virtual lane mark generated by the detected road boundary.

  Unlike the conventional ADAS system that uses the lane boundary detection result as it is, assuming that the area inside the detected lane boundary is the own vehicle traveling lane, in the present embodiment, the virtual lane boundary for the own vehicle traveling is used. Estimate and generate the correct driving lane. Here, the vehicle travel trajectory is used to estimate and track the parameters of the long-distance lane model of the virtual own vehicle travel lane. As described above, due to the position positioning error, the own vehicle traveling lane estimated directly from the vehicle traveling locus usually has a lateral position and a yaw angle as shown in FIG. Includes deviations. Accordingly, a virtual own vehicle traveling lane is estimated and generated by fusing the detected lane boundary with the vehicle traveling locus by the tracking process based on EKF modeled by the long-distance lane model based on the cubic spline. FIG. 5 shows the state transition when estimating the virtual vehicle lane. If a lane mark or a consistent edge boundary is not detected, the virtual own vehicle traveling lane is derived as it is from the vehicle traveling locus. If this is not the case, the detected lane mark / road boundary will correct the travel trajectory to compensate for errors caused by vehicle position measurement, as shown in FIGS. 4 (c)-(d). Used for. Next, a virtual own vehicle traveling lane is generated.

  The long-distance lane model based on the cubic spline used in the present embodiment is very important for effective and efficient estimation of the vehicle lane. In general, the vehicle lane is typically modeled based on a clothoid spline model that may obtain good results in close range, such as lane boundary detection. However, this modeling can fail over long distance ranges where the lane curvature is not evenly distributed. As shown in FIG. 6, the lanes in the proximity range are almost straight, but a steep curve exists in the far range. In order to fit curvature in the far range, clothoid splines are also of relatively large curvature, as shown in the observation spline in FIG. 6 (c), and are typically inaccurate in the near range. Bring an angle. Since the actual angle is derived from the detected lane boundary in order to correct the yaw error of the vehicle travel, the vehicle travel trajectory is detected by the yaw angle as shown in the spline in FIG. Further transformations are made to ensure that the lane boundaries are met, producing a large offset error of the past trajectory in the far range. In this example, the curvature in the far range is relatively accurate, but there is an error in position as shown in FIG.

  The control point of the cubic spline is actually on the curve, and the cubic spline has local control, ie modifying only one control point is a curve near this control point. Affects a part of Therefore, in order to deal with the above problem, the present embodiment adopts cubic spline for generating a virtual vehicle lane. These characteristics ensure that steep curve fitting in the far range does not affect the observation of linear lanes in the near range.

  FIG. 6 shows a generation result of a virtual own vehicle traveling lane by merging lane boundaries and past vehicle traveling trajectories. FIG. 6A shows an estimation result of the own vehicle travel lane using the clothoid spline model. FIG. 6B shows the estimation result of the own vehicle travel lane using the cubic spline model. FIG. 6C shows an observation result using a clothoid model. FIG. 6D shows an observation result using a cubic spline model.

Assuming a sequence of positions (P ₀ , P ₁ ,..., P _n ), the Catmull-Rom spline can interpolate (pass) from points P ₁ to P _n−1 . Further, the tangent vector at _{P i is} parallel to a line connecting the _{P i-1} and _{P i + 1.} The Catmull-Rom cubic spline formula for one line segment is:

However, _{P i} is a control point that needs to Catmull-Rom spline passes a t∈ [0,1].

  In the present embodiment, a Catmull-Rom spline having six control points is used to describe the own vehicle travel lane. The first (nearest) control point is set to be at the lower end of the visible road in the modified image. As the last (farthest) control point, the end point of the past vehicle travel locus in the corrected image is set. The remaining control points are distributed at equal intervals between the control points at both ends. The lateral position of the control point is a vector

Is calculated by a tracking process based on EKF. As shown in FIG. 6D, the straight line portion of the lane in the close range does not cause a positioning error of a steep curve in the far range. As shown, it is generated correctly.

  As described above, the host vehicle travel lane estimation unit 28 stores the lane boundary detected by the lane boundary detection unit 26 and the trajectory database 24 corresponding to the absolute position of the host vehicle detected by the position detection unit 20. The lane parameters at each control point of the cubic spline model are estimated based on the vehicle traveling trajectory, and the virtual traveling lane of the host vehicle is represented based on the estimated lane parameters and the cubic spline model. Data is generated and output to the driving support unit 18.

<Operation of the driving support control device>
Next, the operation of the present embodiment will be described.

  When the vehicle equipped with the driving support control device 10 is running, the imaging device 14 captures the front of the host vehicle, sequentially generates images, and the GPS receiver 12 receives signals from GPS satellites. When receiving sequentially, the driving assistance control apparatus 10 performs the own vehicle travel lane estimation processing routine shown in FIG.

  In step 100, an image generated by the imaging device 14 is acquired, and in step 102, a GPS signal received by the GPS receiving unit 12 is acquired.

  In step 104, the absolute position of the host vehicle is detected based on the GPS signal acquired in step 102 and the INS information output from the INS sensor (not shown).

  In the next step 106, a lane boundary such as a lane mark is detected and tracked based on the previously detected lane boundary and the image acquired in step 100.

  In step 108, based on the absolute position of the host vehicle detected in step 104, the vehicle travel locus including the position corresponding to the absolute position of the host vehicle is detected from the track database 24 and the traveling direction of the host vehicle is reached. The vehicle travel locus in the same direction as is acquired. At this time, when a plurality of vehicle traveling tracks are obtained, a vehicle traveling track obtained by averaging the plurality of vehicle traveling tracks is acquired.

  In step 110, based on the lane boundary detected in step 106, the vehicle travel locus acquired in step 108, and the virtual vehicle lane estimated last time, the virtual vehicle travel is performed. The lane is estimated and tracked, output to the driving support unit 18, and the process returns to step 100.

  By repeatedly executing the above-described own vehicle travel lane estimation processing routine, the virtual vehicle travel lanes that are sequentially generated are output to the driving support unit 18, and the driving support unit 18 performs virtual host vehicle travel. Driving assistance based on lanes is provided.

  In addition, the driving support control device 10 generates a vehicle travel locus based on the detected time-series data of the absolute position of the host vehicle, and stores the vehicle travel locus in the locus database 24.

  Step 106 is realized by the lane boundary detection processing routine shown in FIG.

  In step 120, low-high intensity edge points or high-low intensity edge points are extracted from the image acquired in step 100 in accordance with the edge intensity calculated by the above equation (2) on each horizontal scanning line.

In step 122, the image acquired in step 100 is corrected by inverse perspective mapping (IPM), and the edge points extracted in step 120 are projected onto the coordinate system of the corrected image. Next, the edge points in sections S ₁ and S _{2 of} the modified image are grouped into low-high intensity line segments and high-low intensity line segments, respectively, based on the Hough transform voting scheme. In step 124, the system calculates the average edge intensity of each line segment, the line segment average edge strength is below the threshold is discarded, and finally, in the sections S ₁ and S ₂ of the modified image, the low - high strength line Minutes and high-low intensity line segments are acquired.

  In the next step 126, whether or not a condition group for finding a consistent edge boundary is satisfied based on the low-high intensity line segment and the high-low intensity line segment obtained in step 124 above. For each of the low-high intensity line segments and the high-low intensity line segments, group the line segments and label them with consistent edge boundaries. In addition, it is determined whether or not the condition group for finding the lane mark boundary is satisfied, the line segment is grouped for each of the low-high intensity line segment and the high-low intensity line segment, and the label of the lane mark boundary Add.

  Then, in step 128, for each of the left and right lane boundaries, the line segment with the consistent edge boundary label or the line segment with the lane mark boundary label obtained in step 126 above is added. Based on the lane parameters at each point of the lane boundary estimated in step 128 of the above, the lane parameters at each point of the lane boundary are estimated using the EKF filter, and based on the estimated lane parameter and the clothoid model, The lane boundary is estimated and tracked, and the lane boundary detection processing routine is terminated. In step 126, if no consistent edge boundary or lane mark boundary label is attached to any line segment, it is determined that no lane boundary has been detected.

  Step 110 is realized by the lane estimation processing routine shown in FIG.

  In step 130, it is determined whether or not a lane boundary has been detected by the processing in step 106. When the lane boundary is not detected, the routine proceeds to step 134 described later. On the other hand, if a lane boundary is detected, in step 132, the position and yaw angle of the vehicle travel locus acquired in step 108 are corrected based on the lane boundary detected in step 106.

  In step 134, the vehicle travel trajectory acquired in step 108 or the vehicle travel trajectory corrected in step 132, the lane boundary detected by the processing in step 106, and the host vehicle estimated in the previous step 134 Based on the lane parameters of the driving lane, the EKF filter is used to estimate and track the lane parameters of the own vehicle lane, and based on the estimated lane parameters and the cubic spline model, Is generated and output to the driving support unit 18, and the lane estimation processing routine is terminated.

  As described above, according to the driving support control apparatus according to the first embodiment of the present invention, the travel lane in which the host vehicle is traveling is based on the vehicle travel locus and the detected lane boundary. By estimating the lane parameters and generating a virtual vehicle lane in which the vehicle is traveling, the vehicle lane can be estimated even if the lane mark is not clear or does not exist. .

  In addition, by using the vehicle traveling locus, it is possible to estimate the vehicle traveling lane by obtaining the curvature of a distant lane that cannot be estimated only by image information.

<Second Embodiment>
<System configuration>
Next, a second embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The second embodiment is different from the first embodiment in that the lane boundary is predicted using the vehicle travel locus, and the lane boundary is detected using the prediction result.

  The lane boundary detection method by the lane boundary detection unit 26 described in the first embodiment may handle many difficult problems such as complicated patterns, curvature, and noise. However, lane misdetection may occur in situations where visual lane boundary evidence is not discriminating due to low quality lane features. In fact, such false detection occurs in most lane detection based on general appearance. In this embodiment, the vehicle trajectory can reveal the structure of the scene and provide clues to the situation, so as to provide additional support to identify targets that are not noticeable to noise In addition, the vehicle travel locus is adopted.

  Here, the past vehicle travel trajectory is used to provide a premise or prediction for detecting the lane boundary, and improves the lane boundary detection performance in each step of the lane boundary detection unit 26. The lane boundary prediction unit 224 in the present embodiment may be used as an add-on in any of the steps of the lane boundary detection unit 26, and the detection performance is improved.

  As shown in FIG. 10, the computer 216 of the driving support control apparatus 210 according to the second embodiment is detected by the position detection unit 20, the vehicle travel locus generation unit 22, the locus database 24, and the position detection unit 20. A lane boundary prediction unit 224 that predicts a lane boundary based on the absolute position of the subject vehicle that has been recorded and a vehicle travel locus stored in the locus database 24, and an image generated by the imaging device 14 and a lane boundary prediction unit 224. Is provided with a lane boundary detection unit 226 for detecting the lane boundary of the traveling lane of the host vehicle, and a host vehicle traveling lane estimation unit 28. The lane boundary detection unit 226 is an example of a lane boundary estimation unit.

  The lane boundary prediction unit 224 calculates a score for weighting the score used in edge point extraction by the lane boundary detection unit 226 as prediction of the lane boundary.

  Here, the principle of calculating the score by the lane boundary prediction unit 224 will be described.

  Location is one of the five most important classes of contextual relationship between objects and the real world. Assuming that the object is likely to be in the scene, the object is often found at some location and not at any other location. In the example of lane detection, the past vehicle travel trajectory is such that the vehicle travels in the middle of the travel lane most of the time, and as a result reveals the expected position of the lane boundary. There is a possibility to provide.

  Each vehicle trajectory corresponds to the left lane boundary and the right lane boundary, a set of positions that are separated from the left lane boundary by half the lane width, and half the lane width from the right lane boundary. It means a set of positions that are only separated. According to Japanese road construction regulations, the lane width is between 3 and 3.5 meters. As a result, it is assumed that the travel lane has an average width μ of 3.25 meters and a standard deviation σ of 2.0 meters. The spatial score Sp (u, v) is calculated by the following equation (4) in consideration of the dispersion of the distance from the boundary position to the vehicle travel locus.

However, (u, v) is a point in the corrected image coordinates. N is the number of vehicle travel trajectories existing in the current scene, and _uk is the horizontal position of the kth trajectory in the horizontal line v. Here, in order to make use of the statistical value of the past vehicle travel locus, the total is used instead of the maximum score of each vehicle travel locus. In FIG. 11B, the spatial score is visualized as a gradation portion.

Such spatial context information weights edge points to improve boundary feature extraction when local appearance is poor due to a certain number of factors such as confusion, noise and lighting. Will be used to do. That is, the edge strength used in the extraction of the edge points by the lane boundary detection unit 226, it comes to the product of the S _ep and S _p.

As described above, the lane boundary prediction unit 224 calculates the spatial score _Sp (u, v) at each coordinate (u, v) based on the vehicle travel locus acquired corresponding to the absolute position of the host vehicle. To do.

Further, the lane boundary prediction unit 224 calculates a local yaw angle α _tra of the vehicle travel locus acquired corresponding to the absolute position of the _host vehicle.

The lane boundary detection unit 226 calculates the edge strength S _ep (u, v) of each coordinate according to the above equation (1), and the spatial score S _p (u, v) calculated by the lane boundary prediction unit 224 The product is set as the edge strength of each coordinate, and a point having an edge strength whose absolute value is larger than the threshold is extracted as an edge point.

  Next, the feature line segment voting step by the lane boundary detection unit 226 will be described.

In the first embodiment, during the voting process, the voting range of α is set as [−α _max , α _max ] on the assumption that the line segment of the lane boundary is almost parallel to the traveling direction of the vehicle. Is done. Many lane detection systems employ such assumptions because this assumption can eliminate the effects of almost unrelated horizon and noise and reduce the computational load. However, this assumption may not hold on curved roads or while the vehicle is driving. However, past vehicle travel trajectories are accurate with respect to road curvature and may provide local information, and the average trajectory of the plurality of vehicle travel trajectories reduces the influence of vehicle driving. there is a possibility.

Therefore, the lane boundary detection unit 226 sets the voting range as [−α _max + α _tra , α _max + α _tra ] using the local yaw angle α _tra of the vehicle travel locus calculated by the lane boundary prediction unit 224. (See FIG. 11C).

  Next, the step of grouping feature line segments by the lane boundary detection unit 226 will be described.

In addition to resetting the voting range, past vehicle travel trajectories are further used in line segment pairing that gives higher weight to candidates with higher parallelism to local trajectories. More formally, assuming that the direction of the line segment is α _l1 and α _l2 , respectively, the determination of the condition regarding the direction is performed using a score expressed by the following equation (5).

However, the first term is the original score, and the second term is the reweighted score. λ _o and λ _or are weighting constants.

The lane boundary detection unit 226 pairs two line segments depending on whether or not a condition group for finding a consistent edge boundary is satisfied between the sections S ₁ and S _2, and as a consistent edge boundary. Will be labeled. At this time, the condition regarding the difference in the direction α is a condition using the score of the above formula.

Further, lane boundary detecting section 226, in the same section S ₁ or S _2, depending on whether satisfy conditions set for finding the lane mark, the inner segment of the pair candidates, are labeled as lane mark boundary It will be. At this time, the condition relating to the difference in the direction α is a condition using the score of the formula (5).

  In addition, the lane boundary detection unit 226 includes a line segment with a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary existing on the left side of the traveling lane of the host vehicle, and the previous time. Based on the detected left lane boundary, the lane boundary existing on the left of the traveling lane of the host vehicle is estimated and tracked according to the EKF filter. The lane boundary detection unit 226 also obtains a line segment with a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary present on the right of the traveling lane of the host vehicle, and the previous time. Based on the detected right lane boundary, the lane boundary existing on the right of the traveling lane of the host vehicle is estimated and tracked according to the EKF filter.

<Operation of the driving support control device>
Next, the operation of the present embodiment will be described.

  The driving support control device 210 executes a host vehicle travel lane estimation processing routine shown in FIG. In addition, about the process similar to 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step 100, an image generated by the imaging device 14 is acquired, and in step 102, a GPS signal received by the GPS receiving unit 12 is acquired.

  In step 104, the absolute position of the host vehicle is detected based on the GPS signal acquired in step 102 and the INS information output from the INS sensor (not shown).

  In step 108, based on the absolute position of the host vehicle detected in step 104, the vehicle travel locus including the position corresponding to the absolute position of the host vehicle is detected from the track database 24 and the traveling direction of the host vehicle is reached. The vehicle travel locus in the same direction as is acquired. When a plurality of vehicle travel tracks are obtained, a vehicle travel track obtained by averaging the plurality of vehicle travel tracks is acquired.

  In step 250, a spatial score of each coordinate is calculated based on the vehicle travel locus acquired in step 108.

  In step 252, the yaw angle of the local vehicle travel locus is calculated based on the vehicle travel locus acquired in step 108.

  In step 106, using the spatial score calculated in step 250, the local vehicle travel locus yaw angle calculated in step 252 and the lane boundary detected last time, the image acquired in step 100 is used. Detect and track lane boundaries such as lane marks.

  In step 110, based on the lane boundary detected in step 106, the vehicle travel locus acquired in step 108, and the virtual vehicle lane estimated last time, the virtual vehicle travel is performed. The lane is estimated and tracked, output to the driving support unit 18, and the process returns to step 100.

  Step 106 is realized by a processing routine similar to the lane boundary detection processing routine shown in FIG.

  However, in step 120, the low-high intensity is calculated based on the score calculated by the above equation (2) and the spatial score calculated in step 250 on each horizontal scanning line from the image acquired in step 100. Extract edge points or high-low intensity edge points.

  In step 122, line segment voting is performed according to the voting range set by using the yaw angle of the local vehicle travel locus calculated in step 252.

  In step 126, it is determined whether or not a condition group including a condition using the score of the formula (5) as a condition regarding the difference in the direction α and satisfying the condition group for finding a consistent edge boundary is satisfied. For each of the low-high intensity line segments and the high-low intensity line segments, the line segments are grouped and labeled with consistent edge boundaries. Further, as a condition relating to the difference in the direction α, it is determined whether or not a condition group including a condition using the score of the above formula (5) satisfies a condition group for finding a lane mark boundary. For each of the high intensity line segments and the high-low intensity line segments, the line segments are grouped and labeled with lane mark boundaries.

  In addition, about the other structure and effect | action of the driving assistance control apparatus which concern on 2nd Embodiment, since it is the same as that of 1st Embodiment, description is abbreviate | omitted.

  In this way, by using the vehicle travel locus, it is possible to accurately determine the lane boundary even on a road where the boundary line such as the lane mark is deteriorated or does not exist, and the virtual own vehicle traveling lane is accurately estimated. be able to.

<Third Embodiment>
<System configuration>
Next, a third embodiment will be described. In addition, about the part which becomes the same structure as 1st Embodiment and 2nd Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  The third embodiment is different from the first embodiment in that the relevant vehicle travel locus is selected using the route of the host vehicle.

  As described above, the vehicle travel trajectory stored in the trajectory database 24 is collected when the vehicle travels in the road network a plurality of times. The vehicle travel trajectory used for the own vehicle travel lane estimation is a plurality of vehicle travel trajectories associated according to the position of the own vehicle. When the host vehicle is traveling in a non-branching lane, the shapes of a plurality of associated vehicle travel trajectories are basically the same, and the average of the vehicle travel trajectories estimates the own vehicle travel lane, and the lane It will be used to improve the boundary detection performance. However, when the host vehicle is approaching an intersection, the associated multiple vehicle trajectories may be quite different due to these different travel directions, ie, left turn, right turn or straight ahead. An example of a plurality of vehicle travel trajectories is shown in FIG.

  As shown in FIG. 13, the computer 316 of the driving support control apparatus 310 according to the third embodiment includes a position detection unit 20, a vehicle travel locus generation unit 22, a locus database 24, and a lane boundary prediction unit 224. Based on the absolute position of the host vehicle detected by the position detection unit 20 and the route of the host vehicle, the vehicle related to the travel route from the vehicle travel track stored in the track database 24 to the destination of the host vehicle. An associated trajectory selection unit 324 that selects a travel trajectory, a lane boundary detection unit 226, a lane boundary detected by the lane boundary detection unit 226, an absolute position of the host vehicle detected by the position detection unit 20, and an associated trajectory selection unit And a host vehicle travel lane estimation unit 328 that estimates a virtual travel lane of the host vehicle based on the vehicle travel trajectory selected by H.324.

  The related trajectory selection unit 324 selects a vehicle travel trajectory related to the current travel route to the destination of the host vehicle. The travel route of the host vehicle is acquired from a car navigation system (not shown) having a series of relay points having a global GPS position. An overall distance between the travel route and the associated vehicle travel locus is calculated, and a vehicle travel locus having a distance less than a predetermined threshold, i.e., a vehicle travel locus having the same travel direction as the travel route. Will be selected. The average of the selected vehicle travel locus is used to estimate the vehicle travel lane and improve the lane boundary detection performance. Statistically, it should be noted that the averaged vehicle trajectory may provide accurate spatial information of the road despite the use of low cost GPS / INS sensors. .

  The own vehicle travel lane estimation unit 328 calculates the lane boundary detected by the lane boundary detection unit 226, the average of the vehicle travel trajectory selected by the related trajectory selection unit 324, and the lane parameter at each control point estimated last time. Based on the estimated lane parameters at each control point of the cubic spline model, based on the estimated lane parameters and the cubic spline model, a virtual driving lane of the host vehicle is generated and tracked, and the driving support unit 18 is output.

<Operation of the driving support control device>
Next, the operation of the present embodiment will be described.

  The driving support control device 310 executes a host vehicle travel lane estimation processing routine shown in FIG. In addition, about the process similar to 1st Embodiment and 2nd Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step 100, an image generated by the imaging device 14 is acquired, and in step 102, a GPS signal received by the GPS receiving unit 12 is acquired.

  In step 104, the absolute position of the host vehicle is detected based on the GPS signal acquired in step 102 and the INS information output from the INS sensor (not shown).

  In step 300, a travel route to the destination of the host vehicle is acquired from the navigation system. In the next step 302, based on the absolute position of the host vehicle detected in step 104 and the travel route of the host vehicle acquired in step 300, a position corresponding to the absolute position of the host vehicle is included from the trajectory database 24. The vehicle travel locus is a vehicle travel locus in the same direction as the traveling direction of the host vehicle, and a vehicle travel locus related to the travel route of the host vehicle is acquired. When a plurality of vehicle travel tracks are obtained, a vehicle travel track obtained by averaging the plurality of vehicle travel tracks is acquired.

  In step 250, a spatial score of each coordinate is calculated based on the vehicle travel locus acquired in step 302.

  In step 252, the yaw angle of the local vehicle travel locus is calculated based on the vehicle travel locus acquired in step 302.

  In step 106, a lane boundary such as a lane mark is obtained from the image acquired in step 100, using the spatial score calculated in step 250 and the yaw angle of the local vehicle travel locus calculated in step 252. Is detected.

  Then, in step 110, based on the lane boundary detected in step 106, the vehicle travel locus acquired in step 302, and the virtual vehicle lane estimated last time, the virtual vehicle travel is performed. The lane is estimated and tracked, output to the driving support unit 18, and the process returns to step 100.

  In addition, about the other structure and effect | action of the driving assistance control apparatus which concern on 3rd Embodiment, since it is the same as that of 2nd Embodiment, description is abbreviate | omitted.

  In this way, by using the vehicle travel locus related to the route of the host vehicle, the boundary of the lane can be obtained accurately even in a scene having a plurality of travel directions, and the virtual host vehicle travel lane can be accurately determined. Can be estimated.

  In the above-described embodiment, the case where the lane boundary prediction unit is provided has been described as an example. However, the present invention is not limited to this, and the lane boundary prediction unit is not limited to this, as in the first embodiment. It may be used to detect a lane boundary.

<Fourth embodiment>
<System configuration>
Next, a fourth embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment-3rd Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the fourth embodiment, a related vehicle travel locus is acquired for each travel direction of the host vehicle, and tracking in lane boundary detection and host vehicle travel lane estimation is performed for each travel direction of the host vehicle. This is different from the first embodiment.

  The travel route may not be obtained from the car navigation system. For example, a driver typically does not set a destination in a car navigation system when traveling on a familiar road. At this time, the plurality of vehicle travel trajectories associated with the travel of the host vehicle have different shapes when the host vehicle is approaching the intersection, and each of the plurality of vehicle travel trajectories is It can be considered for running the vehicle.

  As shown in FIG. 15, the computer 416 of the driving support control apparatus 410 according to the fourth embodiment is detected by the position detection unit 20, the vehicle travel locus generation unit 22, the locus database 24, and the position detection unit 20. A plurality of trajectory acquisition units 422 for acquiring, from the vehicle travel trajectory stored in the trajectory database 24, a vehicle travel trajectory related to the travel of the own vehicle based on the absolute position of the host vehicle, For each of a plurality of travel directions, a lane boundary prediction unit 424 that predicts a lane boundary based on the vehicle travel trajectory related to the travel of the host vehicle acquired by the plurality of trajectory acquisition units 422, and for each of the plurality of travel directions, A lane boundary that detects the lane boundary of the traveling lane of the host vehicle based on the image generated by the imaging device 14 and the lane boundary predicted by the lane boundary prediction unit 224 For each of a plurality of travel directions, the departure part 426, the lane boundary detected by the lane boundary detection part 426, the absolute position of the host vehicle detected by the position detection part 20, and the vehicle travel acquired by the multiple trajectory acquisition part 422 A host vehicle travel lane estimation unit 428 that estimates a virtual travel lane of the host vehicle based on the trajectory is provided. The multiple trajectory acquisition unit 422 is an example of a multi-directional trajectory acquisition unit.

  The multiple trajectory acquisition unit 422 acquires a vehicle travel trajectory related to the current travel of the host vehicle from the trajectory database 24 based on the absolute position of the host vehicle detected by the position detection unit 20, and performs a plurality of travel directions. For example, the acquired vehicle travel locus is grouped and averaged for each of a left turn, a straight advance, and a right turn.

  The lane boundary prediction unit 424 weights the score used in edge point extraction by the lane boundary detection unit 226 as prediction of the lane boundary based on the vehicle travel trajectory acquired for the travel direction for each of a plurality of travel directions. To calculate the score.

Further, the lane boundary prediction unit 424 calculates a local yaw angle α _tra of the vehicle travel locus acquired for the travel direction for each of a plurality of travel directions.

The lane boundary detection unit 426 calculates the edge strength S _ep (u, v) of each coordinate in accordance with the above equation (1) for each of a plurality of traveling directions, and the lane boundary prediction unit 424 calculates it for the traveling direction The product of the obtained spatial score _Sp (u, v) is used as the edge strength of each coordinate, and a point having an edge strength whose absolute value is larger than the threshold is extracted as an edge point.

The lane boundary detection unit 426 uses the local yaw angle α _tra of the vehicle travel locus calculated for the travel direction by the lane boundary prediction unit 424 for each of a plurality of travel directions, and determines the voting range [−α _max + α _tra , α _max + α _tra ].

The lane boundary detection unit 426 pairs two line segments depending on whether a condition group for finding a consistent edge boundary is satisfied between sections S ₁ and S _{2 for} each of a plurality of traveling directions. And will be labeled as a consistent edge boundary.

Further, the lane boundary detection unit 426 determines whether the line segment inside the pair candidate is a lane depending on whether or not a condition group for finding a lane mark is satisfied in the same section S ₁ or S _{2 for} each of a plurality of traveling directions. It will be labeled as a mark boundary.

  Further, the lane boundary detection unit 426 attaches a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary existing on the left side of the traveling lane of the host vehicle for each of a plurality of traveling directions. Based on the detected line segment and the left lane boundary detected last time, the lane boundary existing on the left of the traveling lane of the host vehicle is estimated and tracked according to the EKF filter. The lane boundary detection unit 426 attaches a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary present on the right side of the traveling lane of the host vehicle for each of a plurality of traveling directions. Based on the detected line segment and the right lane boundary detected last time, the lane boundary existing on the right of the traveling lane of the host vehicle is estimated and tracked according to the EKF filter.

  As described above, the lane boundary detection unit 426 performs tracking for lane boundary detection for each of a plurality of traveling directions. Each of the results holds a detection score to indicate the likelihood or confidence of the results detected for the current run. As soon as the vehicle enters the intersection, detection results that do not match the current driving, i.e. detection results for different driving directions, are immediately discarded and only the detection results in the same direction are saved. It will be.

  The own vehicle travel lane estimation unit 428 includes, for each of a plurality of travel directions, the lane boundary detected by the lane boundary detection unit 426, the average of the vehicle travel trajectory acquired by the multiple trajectory acquisition unit 422, and each previously estimated Based on the lane parameters at the control points, the lane parameters at each control point of the cubic spline model are estimated and tracked, and based on the estimated lane parameters and the cubic spline model, a virtual driving lane of the host vehicle Is output to the driving support unit 18.

  The own vehicle travel lane estimation unit 428 performs tracking for vehicle travel lane estimation for each of a plurality of travel directions. Each of the results holds a detection score to indicate the likelihood or confidence of the estimated result for the current run. However, as soon as the vehicle enters the intersection, results that do not fit the current run, that is, results with a different direction, will be immediately abandoned and only results in the same direction will be saved. Become.

<Operation of the driving support control device>
Next, the operation of the present embodiment will be described.

  The driving support control device 410 executes a host vehicle travel lane estimation processing routine shown in FIG. In addition, about the process similar to 1st Embodiment and 2nd Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step 100, an image generated by the imaging device 14 is acquired, and in step 102, a GPS signal received by the GPS receiving unit 12 is acquired.

  In step 104, the absolute position of the host vehicle is detected based on the GPS signal acquired in step 102 and the INS information output from the INS sensor (not shown).

  In step 400, based on the absolute position of the own vehicle detected in step 104, the vehicle travel locus including the position corresponding to the absolute position of the own vehicle is obtained from the locus database 24 and is the same as the traveling direction of the own vehicle. When a plurality of vehicle travel trajectories are obtained, a plurality of vehicle travel trajectories are grouped in a plurality of travel directions (right turn, left turn, and straight travel). Then, a vehicle travel locus obtained by averaging a plurality of grouped vehicle travel tracks is acquired.

  In step 402, any one of a plurality of travel directions is set as a travel direction to be processed.

  In step 250, the spatial score of each coordinate is calculated based on the vehicle travel trajectory acquired for the travel direction to be processed in step 400.

  In step 252, the yaw angle of the local vehicle travel locus is calculated based on the vehicle travel locus acquired for the travel direction to be processed in step 400.

  In step 106, using the spatial score calculated in step 250, the local vehicle travel locus yaw angle calculated in step 252 and the lane boundary detected last time, the image acquired in step 100 is used. A lane boundary such as a lane mark is detected for the traveling direction to be processed. At this time, when the score obtained from the tracker that tracks the lane boundary is low for the traveling direction of the processing target, the detection result of the lane boundary obtained for the traveling direction of the processing target is discarded.

  In step 110, based on the lane boundary detected in step 106, the vehicle travel locus acquired in step 302, and the virtual vehicle lane estimated last time, the travel direction to be processed is determined. The virtual vehicle lane is estimated and output to the driving support unit 18. At this time, if the score obtained from the tracking device that tracks the virtual vehicle lane is low for the processing direction, the virtual vehicle lane obtained for the processing direction is low. The estimation result is discarded.

  In Step 404, it is determined whether or not the processing of Step 402 to Step 110 has been executed for all of the plurality of traveling directions, and there is a traveling direction in which the processing of Step 402 to Step 110 is not executed. Returns to step 402 and sets the travel direction as the travel direction to be processed. Further, when the processes of Step 402 to Step 110 are executed for all of the plurality of traveling directions, the process returns to Step 100.

  In addition, about the other structure and effect | action of the driving assistance control apparatus which concern on 4th Embodiment, since it is the same as that of 2nd Embodiment, description is abbreviate | omitted.

  In this way, by using related vehicle travel trajectories for each of a plurality of travel directions, a lane boundary can be accurately obtained even in a scene having a plurality of travel directions. It can be estimated with high accuracy.

  In the above-described embodiment, the case where the lane boundary prediction unit is provided has been described as an example. However, the present invention is not limited to this, and the lane boundary prediction unit is not limited to this, as in the first embodiment. It may be used to detect a lane boundary.

<Fifth embodiment>
<System configuration>
Next, a fifth embodiment will be described. In addition, about the part which becomes the structure similar to 1st Embodiment-4th Embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  In the fifth embodiment, for each traveling lane, a related vehicle traveling locus is acquired, and for each traveling lane, tracking in lane boundary detection and own vehicle traveling lane estimation is performed. This is different from the embodiment.

  On a wide road including a plurality of lanes, there may be a number of lanes in the same traveling direction. In order to deal with such problems, a vehicle travel trajectory with the same direction is first selected and multiple trackers will be generated to deal with multiple travel lanes. The own vehicle traveling lane having a higher score is output to the driving support unit.

  As shown in FIG. 17, the computer 516 of the driving support control apparatus 510 according to the fifth embodiment includes a position detection unit 20, a vehicle travel locus generation unit 22, a locus database 24, and an associated locus selection unit 324. Based on the vehicle travel trajectory selected by the related trajectory selection unit 324 and related to the travel of the host vehicle, the vehicle travel trajectory related to the travel of the host vehicle is grouped for each of a plurality of travel lanes. A plurality of lane generation units 522, a lane boundary prediction unit 524 that predicts a lane boundary based on the vehicle travel trajectory acquired by the plurality of lane generation units 522 for each of a plurality of travel lanes, and a plurality of travel lanes In addition, the vehicle that detects the lane boundary of the traveling lane of the host vehicle based on the image generated by the imaging device 14 and the lane boundary predicted by the lane boundary prediction unit 524 For each of a plurality of travel lanes, the boundary detection unit 526, the lane boundary detected by the lane boundary detection unit 526, the absolute position of the host vehicle detected by the position detection unit 20, and the vehicle acquired by the multiple lane generation unit 522 And a host vehicle travel lane estimation unit 528 that estimates a virtual travel lane of the host vehicle based on the travel trajectory. The multiple lane generation unit 522 is an example of a multiple lane track acquisition unit.

  The multiple lane generation unit 522 groups and averages the vehicle travel trajectories for each of the plurality of travel lanes based on the vehicle travel trajectory related to the current travel of the host vehicle acquired by the related trajectory selection unit 324. .

  The lane boundary prediction unit 524 weights, for each of a plurality of travel lanes, a score used in edge point extraction by the lane boundary detection unit 526 as a lane boundary prediction based on the vehicle travel trajectory acquired for the travel lane. To calculate the score.

Further, the lane boundary prediction unit 524 calculates a local yaw angle α _tra of the vehicle travel locus acquired for the travel lane for each of the plurality of travel lanes.

The lane boundary detection unit 526 calculates the edge strength S _ep (u, v) of each coordinate according to the above equation (1) for each of the plurality of traveling lanes, and calculates the lane boundary prediction unit 524 for the traveling lane. The product of the obtained spatial score _Sp (u, v) is used as the edge strength of each coordinate, and a point having an edge strength whose absolute value is larger than the threshold is extracted as an edge point.

In the lane boundary detection unit 526, the voting range is set to [−α for each of a plurality of travel lanes by using the local yaw angle α _tra of the vehicle travel locus calculated for the travel lane by the lane boundary prediction unit 524. _max + α _tra , α _max + α _tra ].

The lane boundary detection unit 526 pairs two line segments depending on whether or not a condition group for finding a consistent edge boundary is satisfied between sections S ₁ and S _{2 for} each of a plurality of traveling lanes. And will be labeled as a consistent edge boundary.

Further, lane boundary detecting section 526, for each of a plurality of traffic lane in the same section S ₁ or S _2, depending on whether satisfy conditions set for finding the lane mark, the inner segment of the pair candidates, lane It will be labeled as a mark boundary.

  The lane boundary detection unit 526 is provided with a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary existing on the left side of the traveling lane of the host vehicle for each of the plurality of traveling lanes. Based on the line segment and the left lane boundary detected last time, the lane boundary existing on the left of the traveling lane of the host vehicle is estimated and tracked according to the EKF filter. Further, the lane boundary detection unit 526 attaches a consistent edge boundary label or lane mark boundary label acquired corresponding to the lane boundary present to the right of the traveling lane of the host vehicle for each of the plurality of traveling lanes. Based on the detected line segment and the right lane boundary detected last time, the lane boundary existing on the right of the traveling lane of the host vehicle is estimated and tracked according to the EKF filter.

  As described above, the lane boundary detection unit 526 performs tracking for lane boundary detection for each of a plurality of traveling lanes. Each of the results holds a detection score to indicate the likelihood or confidence of the results detected for the current run.

  As shown in FIG. 11 (a) above, in a typical lane split scene, when a new lane boundary appears, the tracker typically requires a small number of frames to move to the new lane boundary, The results may be very unstable. However, in this embodiment, the tracker detects the lane boundary from both sides of the traveling lane for each traveling lane. This is in good agreement with the actual situation. The result is switched between trackers instead of moving the lane boundary from the previous lane boundary to the new lane boundary.

  The own vehicle travel lane estimation unit 528 has, for each of a plurality of travel lanes, the lane boundary detected by the lane boundary detection unit 526, the average of the vehicle travel trajectory acquired by the multiple lane generation unit 522, and each previously estimated Based on the lane parameters at the control points, the lane parameters at each control point of the cubic spline model are estimated and tracked, and based on the estimated lane parameters and the cubic spline model, a virtual driving lane of the host vehicle Is output to the driving support unit 18.

  The own vehicle travel lane estimation unit 528 performs tracking for vehicle travel lane estimation for each of a plurality of travel lanes. That is, for each traveling lane, the tracker estimates a virtual traveling lane.

<Operation of the driving support control device>
Next, the operation of the present embodiment will be described.

  The driving support control device 510 executes the own vehicle travel lane estimation processing routine shown in FIG. In addition, about the process similar to 1st Embodiment-3rd Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step 100, an image generated by the imaging device 14 is acquired, and in step 102, a GPS signal received by the GPS receiving unit 12 is acquired.

  In step 104, the absolute position of the host vehicle is detected based on the GPS signal acquired in step 102 and the INS information output from the INS sensor (not shown).

  In step 300, a travel route to the destination of the host vehicle is acquired from the navigation system. In the next step 302, based on the absolute position of the host vehicle detected in step 104 and the travel route of the host vehicle acquired in step 300, a position corresponding to the absolute position of the host vehicle is included from the trajectory database 24. The vehicle travel locus is a vehicle travel locus in the same direction as the traveling direction of the host vehicle, and a vehicle travel locus related to the travel route of the host vehicle is acquired.

  In step 500, the plurality of vehicle travel tracks acquired in step 302 are grouped for each of the plurality of travel lanes, and the vehicle travel track obtained by averaging the plurality of grouped vehicle travel tracks for each of the plurality of travel lanes. get.

  In step 502, any one of the plurality of travel lanes is set as a processing target travel lane.

  In step 250, a spatial score of each coordinate is calculated based on the vehicle travel locus acquired for the travel lane to be processed in step 500.

  In step 252, the yaw angle of the local vehicle travel locus is calculated based on the vehicle travel locus acquired for the travel lane to be processed in step 500.

  In step 106, using the spatial score calculated in step 250, the local vehicle travel locus yaw angle calculated in step 252 and the lane boundary detected last time, the image acquired in step 100 is used. The lane boundary such as the lane mark is detected and tracked for the lane to be processed.

  Then, in step 110, based on the lane boundary detected in step 106, the vehicle travel locus acquired in step 500, and the virtual vehicle lane estimated last time, the travel lane to be processed is determined. Estimate and track virtual vehicle lanes.

  In step 504, it is determined whether or not the processing in steps 502 to 110 has been executed for all of the plurality of traveling lanes, and there is a traveling lane in which the processing in steps 502 to 110 is not performed. Returns to step 502 and sets the travel lane as the travel lane to be processed. In addition, when the processes in steps 502 to 110 are executed for all of the plurality of traveling lanes, the process proceeds to step 506.

  In step 506, the vehicle lane estimated for the plurality of driving lanes in step 110 is selected with the highest score obtained from the tracker that tracks the virtual vehicle lane. As an estimation result of the own vehicle traveling lane, the vehicle driving lane is output to the driving support unit 18 and the process returns to the step 100.

  In addition, about the other structure and effect | action of the driving assistance control apparatus which concern on 5th Embodiment, since it is the same as that of 2nd Embodiment, description is abbreviate | omitted.

  In this way, by using the related vehicle travel trajectory for each of the plurality of travel lanes, the boundary of the lane can be accurately obtained even in a scene with a plurality of travel lanes. It can be estimated with high accuracy.

  Next, experimental results of the method described in the embodiment of the present invention will be described. An example of the result of lane boundary detection is shown in FIG. FIGS. 19A to 19C show the lane boundary detection results in a complicated pattern. FIGS. 19D to 19F show the results of lane boundary detection in a situation where there is a curvature. FIG. 19G shows the result of lane boundary detection in the presence of backlight. FIGS. 19H to 19I show the results of lane boundary detection in a noisy situation caused by the disappearance of the lane mark.

  FIG. 20A shows the own lane estimation result by using only the original lane detection for virtual lane generation, and FIG. 20B shows only the improved lane boundary detection by virtual lane generation. FIG. 20 (c) shows the own lane estimation result by the method described in the embodiment of the present invention using the past vehicle travel locus, and shows a comparison. Yes.

20 (d) to (f), FIGS. 20 (g) to (i), and FIGS. 20 (j) to (l) similarly show comparisons. From these experimental examples, the embodiment of the present invention is shown. It can be seen that the technique described in is much more accurate and robust than the own lane estimation using only lane boundary detection.

  In the fifth embodiment, the case where the lane boundary prediction unit is provided has been described as an example. However, the present invention is not limited to this, and the lane boundary prediction is the same as in the first embodiment. You may make it detect a lane boundary using a prediction part.

  In the first to fifth embodiments, the case where the vehicle travel locus of the host vehicle is stored in the locus database is described as an example. However, the present invention is not limited to this. The vehicle travel locus of other vehicles may be stored. Further, a trajectory database may be provided in another device, and the trajectory database of the other device may be accessed via a network.

  In addition, the case where the lane boundary detection unit detects a lane boundary using a clothoid spline model has been described as an example, but the present invention is not limited to this, and other models may be used to detect a lane boundary. It may be.

  In addition, the case where the vehicle lane estimation unit estimates a virtual vehicle lane using a cubic spline model has been described as an example, but the present invention is not limited to this, and other models are used. Alternatively, a virtual own vehicle traveling lane may be estimated.

10, 210, 310, 410, 510 Driving support control device 12 GPS receiving unit 14 Imaging device 16, 216, 316, 416, 516 Computer 18 Driving support unit 20 Position detection unit 22 Vehicle travel locus generation unit 24 Track database 26, 226 426, 526 Lane boundary detection unit 28, 328, 428, 528 Own vehicle travel lane estimation unit 224, 424, 524 Lane boundary prediction unit 324 Related track selection unit 422 Multiple track acquisition unit 522 Multiple lane generation unit

Claims (8)

Lane boundary detection means for detecting the boundary of the lane in which the host vehicle is traveling, based on an image mounted on the host vehicle and generated by an imaging unit that images the periphery of the host vehicle and generates an image;
The vehicle travel locus determined in advance by the time-series data of the absolute position corresponding to the absolute position of the own vehicle detected by the position detecting means for detecting the absolute position of the own vehicle, and the lane boundary detecting means Based on the detected boundary of the lane, a lane parameter of a traveling lane in which the host vehicle is traveling is estimated, and on the basis of the estimated lane parameter, a virtual traveling lane in which the host vehicle is traveling Vehicle lane estimation means for generating data representing
Vehicle lane estimation device including Lane boundary estimation means for estimating a boundary of a lane in which the host vehicle is traveling based on a vehicle travel locus corresponding to the absolute position of the host vehicle detected by the position detection unit;
The vehicle lane according to claim 1, wherein the lane boundary detection unit detects the lane boundary based on the image generated by the imaging unit and the lane boundary estimated by the lane boundary estimation unit. Estimating device. A related trajectory for selecting a vehicle travel trajectory related to the travel of the host vehicle from the vehicle travel trajectories corresponding to the absolute position of the host vehicle detected by the position detection unit based on the travel route to the destination of the host vehicle. Further comprising a selection means,
The own vehicle travel lane estimation means estimates the lane parameters of the travel lane based on the vehicle travel trajectory selected by the related trajectory selection means and the lane boundary detected by the lane boundary detection means. The host vehicle travel lane estimation device according to claim 1, wherein data representing a virtual travel lane in which the host vehicle is traveling is generated based on the estimated lane parameter. A plurality of direction trajectory acquisition means for acquiring a vehicle travel trajectory in the travel direction for each of a plurality of travel directions from a vehicle travel trajectory corresponding to the absolute position of the host vehicle detected by the position detection means;
The own vehicle travel lane estimation means includes, for each of the plurality of travel directions, the travel direction vehicle travel trajectory acquired by the multi-direction trajectory acquisition means and the lane boundary detected by the lane boundary detection means. The lane parameter of the said driving | running lane is estimated based on these, The data showing the virtual driving | running lane which the own vehicle is drive | working are produced | generated based on the said estimated lane parameter. Self-vehicle travel lane estimation device. For each of a plurality of travel lanes, further comprising a plurality of lane trajectory acquisition means for acquiring a travel trajectory of the travel lane,
The own vehicle travel lane estimation means includes a vehicle travel trajectory of the travel lane acquired by the plurality of lane trajectory acquisition means and a boundary between the lanes detected by the lane boundary detection means for each of the plurality of travel lanes. The lane parameter of the said driving | running lane is estimated based on these, The data showing the virtual driving | running lane which the own vehicle is drive | working are produced | generated based on the said estimated lane parameter. Self-vehicle travel lane estimation device.   The own vehicle travel lane estimating means corrects the vehicle travel locus based on the lane boundary detected by the lane boundary detection means, and based on the corrected vehicle travel locus, cubic Cubic according to an extended Kalman filter. The lane parameter of each control point used in the spline model is estimated, and data representing the virtual traveling lane is generated based on the estimated lane parameter of each control point and the cubic spline model. The own vehicle travel lane estimation device according to any one of claims 5 to 6.   The lane boundary detection means extracts edge points based on the image generated by the imaging means, estimates model parameters used in the clothoid spline model according to the extended Kalman filter based on the extracted edge points, The own vehicle travel lane estimation device according to any one of claims 1 to 6, wherein a boundary of a lane in which the subject vehicle is traveling is detected based on the estimated model parameter and the clothoid spline model. Computer
Lane boundary detection means for detecting a boundary of a lane in which the host vehicle is traveling, based on an image that is mounted on the host vehicle and that is generated by an imaging unit that images the periphery of the host vehicle and generates an image; and The vehicle travel locus determined in advance by the time-series data of the absolute position corresponding to the absolute position of the own vehicle detected by the position detecting means for detecting the absolute position of the own vehicle, and the lane boundary detecting means Based on the detected boundary of the lane, a lane parameter of a traveling lane in which the host vehicle is traveling is estimated, and on the basis of the estimated lane parameter, a virtual traveling lane in which the host vehicle is traveling A program for functioning as a vehicle lane estimation means for generating data representing the vehicle.