JP5538098B2 - Vehicle white line recognition device - Google Patents

Description

  The present invention relates to a vehicle white line recognition device that recognizes a white line based on an image captured by an in-vehicle camera.

  In recent years, in order to improve the safety of vehicles, driving support devices that actively support driving operations of drivers have been developed. In general, in order to realize a lane departure prevention function and the like in this driving support device, a white line is recognized based on a captured image or the like ahead of the host vehicle, and the own vehicle traveling lane is estimated based on the white line.

  By the way, the white line on the actual road is a double white line in which a sight line (auxiliary line) made up of a broken line is provided inside the lane line in addition to the white line composed of a single lane line. There are various variations.

  Therefore, in this type of white line recognition, it is necessary to fully consider the presence of auxiliary lines as well as lane markings. As a technique for such white line recognition, for example, in Patent Document 1, a plurality of search lines extending in the horizontal direction from the inner side to the outer side in the vehicle width direction with respect to the search area set on the captured image. The first edge point where the luminance changes more than a predetermined value from dark to bright is detected, and noise detection using the Hough transform is performed on the detected point group to obtain the white line start point. A technique for improving detection accuracy is disclosed.

  According to such a technique, when the white line is a single line composed only of a lane line, the white line is recognized based on the lane line, and when the white line is a double white line or the like, it is mainly located inside. A white line is recognized based on an auxiliary line or the like.

JP 2007-264955 A

  By the way, when the white line recognition based on the inner auxiliary line is performed on the double white line or the like as in the technique disclosed in Patent Document 1 described above, for example, in the lane departure prevention function, the own vehicle deviates from the auxiliary line. There is a risk that control will be performed at a timing slightly different from the actual driver's feeling, such as an alarm being issued just by doing.

  To cope with this, if there is a second edge point on each search line, the second edge point may also be detected to recognize the lane marking and the auxiliary line.

  However, when edge points are detected for a double white line or the like, the first edge point does not necessarily correspond to the auxiliary line, and the second edge point does not necessarily correspond to the lane marking. That is, for example, when an edge point is detected for a double white line or the like in which the auxiliary line is composed of a broken line or the like, the first edge point is detected from both the auxiliary line and the lane marking. Therefore, when the second edge point is detected for each search line, it is necessary to classify with high accuracy whether each edge point belongs to a double white line.

  For such edge point classification, it is conceivable to use a known curve detection technique using the Hough transform. However, such a technique uses a curvature parameter in addition to the position and inclination parameters that define a straight line. Therefore, the processing time becomes enormous, and it becomes difficult to apply to a driving support device or the like that requires real-time performance.

  The present invention has been made in view of the above circumstances, and even when an auxiliary line exists inside a lane marking line, a white line recognition for a vehicle that can classify edge points corresponding to each line in a short time and with high accuracy. An object is to provide an apparatus.

A vehicle white line recognition device according to an aspect of the present invention changes luminance from the inside to the outside in the vehicle width direction for each search line extending in the horizontal direction within a white line detection region on an image obtained by capturing an image of the own vehicle traveling path. The first edge point whose luminance changes from dark to bright over a predetermined level is detected as the first edge point. If an edge point exists outside the first edge point in the vehicle width direction, the edge point is determined. Candidate point detection means for detecting as a second edge point, and approximation of a point group consisting of either the first edge point or the second edge point, or both the first edge point and the second edge point A straight line approximating means for calculating a straight line, and a band-like candidate point classification area based on the approximate straight line are set, and when the ratio of the second edge points in the candidate point classification area is less than the setting, the first Determine the pattern, and When the proportion of the second edge point is equal to or greater than a set and determining the pattern determining means of the second pattern, when the first pattern is determined, the respective edge points present the candidate point classification region first When registering as one candidate point and registering each of the edge points existing outside the candidate point classification area as a second candidate point and determining the second pattern, the candidate point Edge point classification means for registering each edge point existing in the classification area as a second candidate point and registering each edge point present inside the vehicle width direction from the candidate point classification area as a first candidate point And .

  According to the vehicular white line recognition device of the present invention, edge points corresponding to each line can be classified in a short time and with high accuracy even when an auxiliary line exists inside the lane marking.

Schematic configuration diagram of a vehicle driving support device Flow chart showing white line recognition routine Explanatory drawing which shows typically an example of the captured image of the environment outside the vehicle Explanatory drawing which shows the 1st edge point and 2nd edge point which are detected from the image of FIG. Explanatory drawing which shows an example of the change of the brightness | luminance on a search line Explanatory drawing showing the calculation method of Hough transform Explanatory drawing showing Hough space Flowchart showing an edge point classification subroutine in the first region Flowchart showing an edge point classification subroutine in the second region Explanatory drawing which shows an example of the relationship between the Hough straight line and edge point in each pattern Explanatory drawing which shows the candidate point classification | category area | region set in the 1st area | region in each pattern Explanatory drawing which shows an example of the process result in the 1st area | region in each pattern. Explanatory drawing which shows an example of the processing result in the 2nd field in each pattern Explanatory drawing which shows the white line detection area | region set based on the recognized white line

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The drawings relate to an embodiment of the present invention, FIG. 1 is a schematic configuration diagram of a vehicle driving support device, FIG. 2 is a flowchart showing a white line recognition routine, and FIG. 3 is a schematic diagram showing an example of a captured image of an environment outside the vehicle. FIG. 4, FIG. 4 is an explanatory diagram showing the first edge point and the second edge point detected from the image of FIG. 3, FIG. 5 is an explanatory diagram showing an example of a change in luminance on the search line, and FIG. FIG. 7 is an explanatory diagram showing a calculation method, FIG. 7 is an explanatory diagram showing a Hough space, FIG. 8 is a flowchart showing an edge point classification subroutine in the first area, and FIG. 9 is a flowchart showing an edge point classification subroutine in the second area, FIG. 10 is an explanatory diagram showing an example of the relationship between the Hough line and edge points in each pattern, FIG. 11 is an explanatory diagram showing candidate point classification areas set in the first area in each pattern, and FIG. FIG. 13 is an explanatory diagram showing an example of the processing result in the second area in each pattern, and FIG. 14 is based on the recognized white line. It is explanatory drawing which shows the white line detection area | region set.

  In FIG. 1, reference numeral 1 denotes a vehicle such as an automobile (own vehicle), and a driving support device 2 is mounted on the vehicle 1. The driving support device 2 includes, for example, a stereo camera 3, a stereo image recognition device 4, a control unit 5, and the like, and a main part is configured.

  The host vehicle 1 is connected to a vehicle speed sensor 11 that detects the host vehicle speed V, a yaw rate sensor 12 that detects the yaw rate γ, a main switch 13 that performs ON / OFF switching of each function of the driving support control, and the like, and a steering wheel. A steering angle sensor 14 that detects the steering angle θst that is opposed to the steering shaft, an accelerator opening sensor 15 that detects an accelerator pedal depression amount (accelerator opening) θacc by the driver, and the like are provided.

  The stereo camera 3 is constituted by a set of CCD cameras using a solid-state imaging device such as a charge coupled device (CCD) as a stereo optical system. These left and right CCD cameras are each mounted at a certain distance in front of the ceiling in the vehicle interior, take a stereo image of an object outside the vehicle from different viewpoints, and output image data to the stereo image recognition device 4. In the following description, one image (for example, the right image) of the stereo images is referred to as a reference image, and the other image (for example, the left image) is referred to as a comparative image.

  First, the stereo image recognition device 4 divides the reference image into small areas of 4 × 4 pixels, for example, compares the luminance or color pattern of each small area with the comparison image, finds a corresponding area, and performs the entire reference image. Find the distance distribution over. Further, the stereo image recognition device 4 examines the luminance difference between each pixel on the reference image and an adjacent pixel, extracts those whose luminance difference exceeds a threshold value as an edge point, and extracts the extracted pixel (edge A distance image (a distribution image of edge points having distance information) is generated by giving distance information to (point). Then, the stereo image recognition device 4 performs a well-known grouping process on the generated distance image and compares it with a pre-stored three-dimensional frame (window), so that a white line and a side wall in front of the vehicle Recognize three-dimensional objects.

  Here, the white line to be recognized in the present embodiment is, for example, a single lane line or a multiple line (double white line or the like) in which a line of sight guide line is provided inside the lane line, Lines that extend on the road and divide the vehicle lane are collectively referred to, and the form of each line includes a solid line, a broken line, and the like, and further includes a yellow line and the like.

  By the way, the white line laid on the actual road has various variations as described above, and the form of the white line changes with the branching or merging of the traveling road. In addition, there are various noises such as road surface repair marks, water pools and snow on the road. Therefore, it is difficult to accurately recognize white lines in all scenes by uniform processing. Therefore, the stereo image recognition device 4 complementarily performs white line recognition using various methods without relying only on the white line recognition based on the above-described distance image, and finally judges these recognition results in a composite manner. Recognize white lines.

  As one example of such white line recognition, the stereo image recognition device 4 performs white line recognition based on a change in luminance in the horizontal direction on one image (for example, a reference image shown in FIG. 3) captured in front of the host vehicle. . Note that the white line recognition described below is particularly effective for recognizing double white lines.

  Specifically, for example, as shown in FIG. 4, the stereo image recognition device 4 sets left and right white line detection areas Al and Ar on the image, and extends horizontally in each of the white line detection areas Al and Ar. For a plurality of existing search lines L, a change in luminance is examined from the inner side to the outer side in the vehicle width direction for each search line L. Then, the stereo image recognition device 4 detects, as the first edge point Pe1, the first edge point at which the luminance changes from dark to bright over a predetermined amount on each search line L in each white line detection area Al, Ar. When an edge point exists outside the first edge point Pe1 in the vehicle width direction, the edge point is detected as the second edge point Pe2.

  Further, the stereo image recognition device 4 calculates an approximate straight line H with respect to a point group (point group of the first edge point Pe1 and the second edge point Pe2) of each edge point Pe detected on the left and right in front of the host vehicle. Then, the stereo image recognition device 4 sets a band-like candidate point classification region α with the approximate straight line H as a reference, and the ratio 2 of the second edge points in the candidate point classification region α is less than the set threshold value. The first pattern is determined, and the second pattern is determined when the rate 2 of the second edge points is equal to or greater than the set threshold value.

  Then, when the first pattern is determined, the stereo image recognition device 4 registers each edge point Pe existing in the candidate point classification region α as the first candidate point Pc1, and from the candidate point classification region α. Also, each edge point Pe existing outside in the vehicle width direction is registered as a second candidate point. On the other hand, when the second pattern is determined, the stereo image recognition device 4 registers each edge point Pe existing in the candidate point classification region α as the second candidate point Pc2, and more than the candidate point classification region. Each edge point Pe existing inside the vehicle width direction is registered as the first candidate point Pc1. Thereby, when the white line is a double white line, basically, the first candidate point Pc is extracted corresponding to an auxiliary line such as a line of sight guide line, and the second candidate point Pc2 corresponds to a lane line or the like. And extracted.

  Here, when the first pattern is determined, the stereo image recognition device 4 deletes each edge point Pe existing in the vehicle width direction inside the candidate point classification region α as noise, and the second pattern When the pattern is determined, each edge point Pe existing outside the candidate point classification region α in the vehicle width direction is erased as noise.

  By the way, when the own vehicle traveling path is curved far away, for example, the auxiliary line inside the double white line and the outside lane line are partially assimilated on the image so that each edge point Pe is accurately determined. It may be difficult to classify. Therefore, the stereo image recognition device 4 of the present embodiment divides the region on the image into a first region located on the own vehicle 1 side and a second region located far from the own vehicle 1, For the first area, the edge points Pe are classified based on the above-described processes, and for the second area, the edge points Pe are classified based on a separate process.

  In this case, for example, the stereo image recognizing device 4 determines the region until the number of edge points Pe occupies a threshold ratio (for example, about 90%) with respect to the total number of edge points Pe existing on the approximate straight line H. The first area is set, and the area farther than that is set as the second area. That is, in the area where the white line (that is, the point group of the edge points Pe) can be linearly approximated with a single straight line on the side of the own vehicle 1 (first area), the above process based on the approximate straight line H is used. Then, each edge point Pe is classified, and in a far region (second region) where approximation is difficult with a single approximate line H, each edge point Pe is classified based on a separate process.

  Then, when the first pattern is determined, the stereo image recognition device 4 selects the first edge point Pe1 that exists alone on the same search line L among the edge points Pe that exist in the second region. While deleting, other 1st edge point Pe1 is set as 1st candidate point Pc1, and 2nd edge point Pe2 is set as 2nd candidate point Pc2. On the other hand, when the second pattern is determined, the stereo image recognition device 4 selects the first edge point Pe1 that exists alone on the same search line L among the edge points Pe that exist in the second region. The second candidate point Pc2 is set, the other first edge point Pe1 is set as the first candidate point Pc1, and the second edge point Pe2 is set as the second candidate point Pc2.

  Then, the stereo image recognition device 4 calculates an approximate line (first approximate line) of the white line using the least square method for the point group of the first candidate points Pc1 classified in this way, and each first An approximation error of the first approximate line with respect to the candidate point Pc1 is calculated. Similarly, the stereo image recognition device 4 calculates a white line approximation line (second approximation line) using the least square method for the point group of the second candidate points Pc2, and obtains the second candidate point Pc2 for each second candidate point Pc2. The approximation error of the approximate line of 2 is calculated. Then, the stereo image recognition device 4 sets the final approximate line parameter based on the parameter of one of the approximate lines having a small approximate error. In this case, when the approximation error of the first approximate line is small, the stereo image recognition device 4 sets the final approximate line parameters by offsetting the first approximate line to the second approximate line. It is desirable to do.

  Thus, in this embodiment, the stereo image recognition device 4 includes candidate point detection means, straight line approximation means, pattern determination means, edge point classification means, region setting means, second edge point classification means, and approximation line. Each function as a calculation means is realized.

  Note that the stereo image recognition device 4 can perform white line recognition by various other methods. Then, for example, the stereo image recognition apparatus 4 has a relationship between an approximate line of a white line recognized in each white line recognition and a white line candidate point such as an edge point used for the recognition (for example, dispersion of white line candidate points with respect to the approximate line, etc.) ) Is selected from the recognized approximate lines, and the approximate line is output to the control unit 5 as the final approximate line of the white line.

  The control unit 5 receives driving environment information in front of the host vehicle 1 recognized by the stereo image recognition device 4. Further, the vehicle speed V from the vehicle speed sensor 11, the yaw rate γ from the yaw rate sensor 12, and the like are input to the control unit 5 as travel information of the host vehicle 1, and operation input information by the driver from the main switch 13. An operation signal, a steering angle θst from the steering angle sensor 14, an accelerator opening θacc from the accelerator opening sensor 15, and the like are input.

  For example, when an execution of an ACC (Adaptive Cruise Control) function which is one of the functions of the driving support control is instructed through the operation of the main switch 13 by the driver, the control unit 5 recognizes the stereo image recognition device 4. The direction of the preceding vehicle is read, and it is identified whether or not the preceding vehicle to be followed is traveling on the own vehicle traveling path.

  As a result, when the preceding vehicle to be tracked is not detected, the constant speed traveling control for maintaining the vehicle speed V of the host vehicle 1 at the set vehicle speed set by the driver through the opening / closing control (engine output control) of the throttle valve 16. Execute.

  On the other hand, when a preceding vehicle that is a tracking target vehicle is detected and the vehicle speed of the preceding vehicle is equal to or lower than the set vehicle speed, the following traveling control is performed in which the following distance is converged to the target inter-vehicle distance. Executed. During this follow-up running control, the control unit 5 basically converges the inter-vehicle distance to the target inter-vehicle distance through the opening / closing control of the throttle valve 16 (engine output control). Furthermore, when it is determined that sufficient deceleration cannot be obtained only by controlling the throttle valve 16 due to sudden deceleration of the preceding vehicle, the control unit 5 controls the output hydraulic pressure from the active booster 17 (automatic braking intervention). Control) to converge the inter-vehicle distance to the target inter-vehicle distance.

  Further, when execution of the lane departure prevention function, which is one of the functions of the driving support control, is instructed through the operation of the main switch 13 by the driver, the control unit 5 will, for example, the left and right white lines that define the own vehicle traveling lane. A warning determination line is set based on (approximate white line recognized by the stereo image recognition device 4), and the own vehicle traveling path is estimated based on the vehicle speed V and the yaw rate γ of the own vehicle 1. Then, for example, when the control unit 5 determines that the host vehicle travel route crosses either the left or right alarm determination line within a set distance (for example, 10 to 16 [m]) ahead of the host vehicle, It is determined that the host vehicle 1 is likely to depart from the current host vehicle lane, and a lane departure warning is issued.

  Next, white line recognition based on the luminance change on the reference image, which is executed in the stereo image recognition device 4, will be described with reference to the white line recognition routine shown in FIG. In this routine, the same processing is performed separately for the left and right white line detection areas Al and Ar. However, to simplify the description, unless otherwise required in the following description. For example, the white line detection areas Al and Ar are collectively referred to as the white line detection area A, and the subscripts “l” and “r” indicating the left and right attributes are omitted as appropriate.

  When this routine is started, the stereo image recognition device 4 first, in step S101, for example, for each search line L in the white line detection area A set in the processing of step S107 for the image of the previous frame, the first edge. The point Pe1 and the second edge point Pe2 are detected. Specifically, for example, as illustrated in FIG. 5, the stereo image recognition device 4 performs edge detection on each search line L from the inner side to the outer side in the vehicle width direction, and detects pixels on the outer side in the vehicle width direction. While detecting a point (edge point Pe (+)) where the luminance is relatively high with respect to the luminance of the inner pixel and the differential value of the luminance indicating the amount of change is equal to or greater than the set threshold on the plus side, A point where the luminance of the pixel on the outer side in the width direction is relatively low with respect to the luminance of the inner pixel, and the differential value of the luminance indicating the amount of change is equal to or smaller than the set threshold value on the negative side (edge point Pe (−)) Is detected. Then, the stereo image recognizing device 4 selects the edge point Pe (+) located at the innermost side in the vehicle width direction in the white line detection area A among the edge points Pe (+) paired with the edge point Pe (−). While extracting as 1 edge point Pe1, next, edge point Pe (+) located inside a vehicle width direction is extracted as 2nd edge point Pe2 (refer FIG. 4). In FIG. 4 and the like, the search line L, each edge point, and the like are thinned out and displayed in a predetermined manner in order to simplify the description.

In subsequent step S102, the stereo image recognition device 4 is based on the extracted edge point Pe (first edge point Pe1 and second edge point Pe2) or (any one of the first edge point Pe1 and the second edge point Pe2). The approximate straight line of the white line is calculated (in the approximate straight line calculation, when only one of the first edge point Pe1 and the second edge point Pe2 is used, the calculation time is increased. Possible). That is, for example, as shown in FIG. 6, the stereo image recognition device 4 sets the slope θ of the straight line h passing through the point Pe (x, z) to 0 ° to 180 ° for each edge point Pe constituting the point group. The distance from the origin O to the straight line h at each θ (perpendicular length) ρ is obtained based on the following equation (1).
ρ = x · cos θ + z · sin θ (1)
Then, the stereo image recognition device 4 votes (projects) the relationship between θ and ρ obtained for each point Pe as a frequency on a corresponding location on the Hough plane (θ, ρ) shown in FIG. Further, the stereo image recognition device 4 extracts a combination of θ and ρ that has the largest frequency on the Hough plane (θ, ρ), and uses the θ and ρ to approximate the straight line defined by the equation (1) ( A Hough straight line) H is set as an approximate expression of the point group (see FIGS. 10A and 10B).

  When the process proceeds from step S102 to step S103, the stereo image recognition device 4 counts the edge points Pe existing on the approximate straight line H, and the number of edge points Pe from the lower side of the image indicates a threshold ratio (for example, about 90%). The area up to occupying is set as the first area, and the area farther than that is set as the second area (see FIGS. 11A and 11B).

  In step S104, the stereo image recognition device 4 performs a classification process of the edge points Pe in the first region. This process is executed, for example, according to the flowchart of the edge point classification subroutine shown in FIG. 8. When the subroutine starts, the stereo image recognition device 4 first has a belt-like shape with the approximate straight line H as a reference in step S201. The region is set as the candidate point classification region α. That is, the stereo image recognition device 4 sets an area formed by translating the approximate straight line H inward and outward in the vehicle width direction by α1 and α2 as the candidate point classification area α (FIGS. 11A and 11B). )reference). Here, as α1 and α2, for example, it is desirable to use a half length of the width of the white line detected in the previous frame converted to the number of pixels on the screen.

In subsequent step S202, the stereo image recognition device 4 calculates the rate Rate2 occupied by the second edge point Pe2 with respect to the total number of edge points Pe existing in the candidate point classification region α based on the following equation (2). To do.
Rate2 = (Num2 / (Num1 + Num2)) × 100 (2)
In Equation (2), Num1 indicates the total number of first edge points Pe1 existing in the candidate point classification region α, and Num2 indicates the total number of second edge points Pe2 existing in the candidate point classification region α.

  Then, when the rate Rate2 is less than a preset threshold, the stereo image recognition device 4 selects the first pattern as the pattern of the edge point Pe classification process, and the rate Rate2 is equal to or greater than the threshold. Selects the second pattern. In the double white line, since the outer lane line is generally a solid line, the first pattern is, for example, when the auxiliary line inside the double white line is a solid line, or the auxiliary line Is selected when the line segment of the broken line is relatively long and the interval between the line segments is relatively short. The first pattern is also rarely selected, for example, when the lane marking outside the double white line is a broken line. On the other hand, the second pattern is often selected, for example, when the auxiliary line is constituted by a broken line and the broken line segment is relatively short and the interval between the line segments is relatively long.

  In subsequent step S203, the stereo image recognition device 4 checks whether or not the pattern determined in step S202 is the first pattern. If the pattern is the first pattern, the process proceeds to step S204, and if the pattern is the second pattern. Then, the process proceeds to step S207.

  When the process proceeds from step S203 to step S204, the stereo image recognition device 4 registers the edge point Pe existing in the candidate point classification region α as the first candidate point Pc1, and in the subsequent step S205, more than the candidate point classification region α. After all the edge points Pe on the inner side in the vehicle width direction are deleted, and in step S206, the edge points Pe on the outer side in the vehicle width direction than the candidate point classification region α are registered as the second candidate points Pc2 (see FIG. 12A). ) Exit the subroutine.

  On the other hand, when the process proceeds from step S203 to step S207, the stereo image recognition device 4 registers the edge point Pe existing in the candidate point classification region α as the second candidate point Pc2, and in the subsequent step S208, the candidate point classification region α. The edge point Pe on the inner side in the vehicle width direction is registered as the first candidate point Pc1, and all the edge points Pe existing outside the candidate point classification region α in the vehicle width direction are deleted in step S209 (FIG. 12). (Refer to (b)), and the subroutine is exited.

  In the main routine of FIG. 2, when the process proceeds from step S104 to step S105, the stereo image recognition device 4 performs a process of classifying the edge points Pe in the second region. This process is executed, for example, according to the flowchart of the edge point classification subroutine shown in FIG. 9. When the subroutine is started, the stereo image recognition device 4 first has a first that exists on the search line L alone in step S301. It is checked whether or not there is one edge point Pe1 (that is, the second edge point Pe2 does not exist on the same search line L and only the first edge point Pe1 exists).

  In step S301, when the stereo image recognition device 4 determines that there is the first edge point Pe1 that exists solely on the search line L, the process proceeds to step S302, and the first image that exists alone on the search line L exists. If it is determined that there is no edge point Pe1 (that is, the first edge point Pe1 does not exist or the first and second edge points Pe1 and Pe2 coexist in pairs), the process proceeds to step S305.

  When the process proceeds from step S301 to step S302, the stereo image recognition device 4 checks whether or not the pattern determined in the process of step S202 described above is the first pattern. If the pattern is the first pattern, the process proceeds to step S303. If it is the second pattern, the process proceeds to step S304.

  Then, when the process proceeds from step S302 to step S303, the stereo image recognition device 4 deletes all the first edge points Pe1 that exist independently on the search line L, and then proceeds to step S305 (see FIG. 13A). . Here, as described above, when the white line is a double white line, it is highly possible that the outer lane line is a solid line. There is a high possibility that the auxiliary line is a solid line or a broken line similar to the solid line. Therefore, when the second edge point Pe2 is not detected in the distance, it is highly likely that the auxiliary line and the lane marking are partially assimilated on the image. In such a case, it is difficult to determine whether the first edge point Pe1 existing alone on the search line L is due to a lane line or an auxiliary line. Therefore, such first edge point Pe1 is deleted as noise.

  On the other hand, when the process proceeds from step S302 to step S304, the stereo image recognition device 4 registers the first edge point Pe1 that exists alone on the search line L as the second candidate point Pc2, and then proceeds to step S305 (FIG. 13). (See (b)). That is, as described above, when the white line is a double white line, there is a high possibility that the inner auxiliary line is a relatively sparse broken line or the like when the second pattern is determined. Therefore, when the second edge point Pe2 is not detected in the distance, even if the auxiliary line and the lane marking are partially assimilated on the image, they exist alone on the search line L. There is a high possibility that the first edge point Pe1 is due to a lane marking. Therefore, such a first edge point Pe1 is registered as a second candidate point Pc2.

  When the process proceeds from step S301, step S303, or step S304 to step S305, the stereo image recognition device 4 selects the first edge point Pe1 that exists in a pair with the second edge point Pe2 on the same search line L as the first candidate point. After registering as Pc1 (see FIGS. 13A and 13B), the process proceeds to step S306.

  When the process proceeds from step S305 to step S306, the stereo image recognition device 4 registers the second edge point Pe2 coexisting with the first edge point Pe1 on the same search line L as the second candidate point Pc2 (FIG. 13 (a) and (b)), the subroutine is exited.

In the main routine of FIG. 2, when the process proceeds from step S105 to step S106, the stereo image recognition device 4 uses, for example, a least square method represented by the equation (3) based on the point group of the first candidate point Pc1. While calculating one approximate line, a second approximate line using the least square method shown in the equation (4) is calculated based on the point group of the second candidate points Pc2.
X1 = a1 · Z ² + b1 · Z + c1 (3)
X2 = a2 · Z ² + b2 · Z + c2 (4)
Here, when the white line is a double white line, basically, the first approximate line shown in the equation (3) indicates the approximate line of the auxiliary line, and the second approximate line shown in the equation (4) indicates the lane. Indicates a lot line. In the expressions (3) and (4), a1, a2, b1, b2, and c1, c2 indicate parameters obtained by the method of least squares.

  Furthermore, the stereo image recognition device 4 calculates the approximate error of the first approximate line for each first candidate point Pc1, and calculates the approximate error of the second approximate line for each second candidate point Pc2. Then, the stereo image recognition device 4 sets the parameters of the final approximate line X based on the parameters of either the first approximate line or the second approximate line and the approximate line with the small approximate error. That is, in the present embodiment, for example, even if the white line is a double white line, a single approximate line is set as the approximate line of the white line. In this case, for example, in order to uniformly perform various types of driving support control (particularly, lane departure warning, etc.), it is desirable that the final approximate line X parameters follow the lane markings. Therefore, when the approximation error of the first approximate line X1 is small, the stereo image recognition device 4 offsets the first approximate line X1 to the second approximate line X2, thereby setting the final approximate line X parameters. Set. On the other hand, when the approximation error of the second approximate line X2 is small, the stereo image recognition device 4 sets the parameters of the final approximate line X using the second approximate line X2 as it is.

  When the process proceeds from step S106 to step S107, the stereo image recognition apparatus 4 sets the white line search area A to be used in the next frame based on the approximate white line calculated in step S106, and then exits the routine. That is, in step S108, as shown in FIG. 14, for example, as shown in FIG. 14, the stereo image recognition apparatus 4 has the left and right white line approximate lines Xl and Xr calculated in step S106 as shown in equations (5) to (8). In each real space, lines Xsl and Xsr offset in the vehicle width direction by lin_ws [m] and lines Xel and Xer offset in the vehicle width direction by lin_wr [m] are set.

That is, the stereo image recognition device 4 performs the white line approximate line X1 on the left side.
Xsl = al · Zl ² + bl · Zl + cl + lin_Wsl (5)
And for the white line approximation line Xr on the right side,
Xsr = ar · Zr ² + br · Zr + cr−lin_Wsr (6)
Set.
In addition, the stereo image recognition device 4 performs the white line approximate line Xl on the left side.
Xel = al · Zl ² + bl · Zl + cl-lin_Wel (7)
And for the white line approximation line Xr on the right side,
Xer = ar · Zr ² + br · Zr + cr + lin_Wer (8)
Set. Then, the stereo image recognition device 4 sets the area surrounded by the equations (5) and (7) in the real space as the white line detection region Al on the left side by performing coordinate conversion to the coordinate system on the image. , (6) and (8) are similarly transformed into the coordinate system on the image to set the right white line detection area Ar.

  According to such an embodiment, the first and first detected based on the luminance change from the inner side toward the outer side in the vehicle width direction for each search line L extending in the horizontal direction in the white line detection region A on the image. An approximate straight line H of a point group consisting of two edge points Pe1, Pe2 or a point group consisting of either the first edge point Pe1, or the second edge point, Pe2 is calculated, and the approximate straight line H is used as a reference. When the ratio of the second edge point Pe2 in the candidate point classification region α is less than the setting, the first pattern is determined, and the edge point Pe existing in the candidate point classification region α is registered as the first candidate point Pc1 At the same time, the edge point Pe existing outside the candidate point classification area α in the vehicle width direction is registered as the second candidate point Pc2, while the ratio of the second edge point Pe2 occupying the candidate point classification area α is greater than or equal to the set value. The second pattern when Each edge point Pe existing in the candidate point classification area α is registered as the second candidate point Pc2, and each edge point Pe existing inside the candidate point classification area α in the vehicle width direction is determined as the first candidate. By registering as the point Pc1, the edge point corresponding to each line can be classified with high accuracy in a short time even when an auxiliary line exists inside the lane marking.

  In this case, with respect to the total number of edge points Pe existing on the approximate straight line H, it is considered that the area until the edge point Pe occupies a set ratio on the approximate straight line H (that is, the edge point Pe follows the approximate straight line H). Area) is set as the first area, and the above process is limited to the first area, so that the classification of the first candidate point Pc1 and the second candidate point Pc2 can be performed with high accuracy. Can be realized.

  For the second area farther than the first area, when the first pattern is determined, the first edge point Pe1 that exists alone is deleted and the other first edge point Pe1 is used as the first candidate point. While setting as Pc1 and setting the second edge point Pe2 as the second candidate point, when the second pattern is determined, the first edge point Pe1 that exists alone is set as the second candidate point Pc2. In addition, the other first edge point Pe1 is set as the first candidate point Pc1, and the second edge point Pe2 is set as the second candidate point Pc2, thereby eliminating the uncertain edge point, Classification into candidate points Pc1 and Pc2 can be performed with high accuracy.

  Although the present embodiment has been described with respect to a stereo camera, the present invention is not limited to this, including a monaural camera, as long as the white line can be recognized.

1 ... Vehicle (own vehicle)
2 ... Driving support device 3 ... Stereo camera 4 ... Stereo image recognition device (candidate point detection means, straight line approximation means, pattern determination means, edge point classification means, region setting means, second edge point classification means, approximate line calculation means )
DESCRIPTION OF SYMBOLS 5 ... Control unit 11 ... Vehicle speed sensor 12 ... Yaw rate sensor 13 ... Main switch 14 ... Steering angle sensor 15 ... Accelerator opening sensor 16 ... Throttle valve 17 ... Active booster

Claims (6)

First, the luminance changes from dark to bright over a predetermined level for each search line that extends in the horizontal direction within the white line detection area on the image of the host vehicle's travel path. Candidate point detecting means for detecting the edge point as a second edge point when an edge point is present on the outer side in the vehicle width direction than the first edge point.
Linear approximation means for calculating an approximate straight line of a point group consisting of either the first edge point or the second edge point, or both the first edge point and the second edge point;
A band-like candidate point classification area based on the approximate straight line is set, the first pattern is determined when the ratio of the second edge points in the candidate point classification area is less than the setting, and the second edge Pattern determination means for determining the second pattern when the ratio of the points is equal to or greater than the setting;
Wherein when the first pattern is determined, and registers the respective edge points present the candidate points classified region as a first candidate point, the present in the vehicle width direction outer side than the candidate points classification region Each edge point is registered as a second candidate point, and when the second pattern is determined, each edge point existing in the candidate point classification area is registered as a second candidate point, and the candidate point An apparatus for recognizing a white line for a vehicle, comprising: edge point classification means for registering each of the edge points existing inside the classification region as a first candidate point.     When the first pattern is determined, the edge point classifying means erases each edge point existing on the inner side in the vehicle width direction from the candidate point classification area as noise, and the second pattern is determined. 2. The vehicle white line recognition device according to claim 1, wherein each edge point existing outside in the vehicle width direction from the candidate point classification region is erased as noise. A region until the edge points occupy a set ratio on the approximate straight line is set as a first region with respect to the total number of the edge points existing on the approximate straight line, and a region beyond the first region is set as the first region. Having an area setting means for setting as the area 2;
3. The vehicle white line recognition device according to claim 1, wherein the edge point classifying unit classifies each of the edge points existing in the first region. Second edge point classification means for classifying each edge point existing in the second region;
The second edge point classifying unit deletes the first edge point that exists independently when the first pattern is determined by the pattern determination unit and sets the other first edge point as a first candidate. Set as a point, set the second edge point as a second candidate point,
When the second pattern is determined by the pattern determination means, the first edge point that exists alone is set as a second candidate point and the other first edge point is set as a first candidate point, 4. The vehicle white line recognition device according to claim 3, wherein the second edge point is set as a second candidate point.     A first approximate line is calculated based on the point group of the first candidate points, an approximation error of the first approximate line with respect to each first candidate point is calculated, and a point group of the second candidate points is calculated. A second approximation line is calculated based on the second approximation line, an approximation error of the second approximation line for each second candidate point is calculated, and a final approximation is performed based on a parameter of one of the approximation lines having a small approximation error. 5. The vehicle white line recognition apparatus according to claim 1, further comprising approximate line calculation means for setting a line parameter.     When the approximation error of the first approximate line is small, the approximate line calculation means sets a final approximate line parameter by offsetting the first approximate line to the second approximate line. The white line recognition device for a vehicle according to claim 5.

Images