JP5402828B2 - Lane boundary detection device, lane boundary detection program - Google Patents

Description

  The present invention relates to a lane boundary detection device and a lane boundary detection program for detecting a lane boundary.

  In order to be used for a lane departure warning device that generates a warning when the vehicle is likely to deviate from the lane, the front landscape is photographed with a camera mounted toward the front of the vehicle, and the boundary of the lane in which the vehicle travels is shown. Lane boundary detection that detects lane boundary lines (white lines painted on the road, etc.), which are road markings, and calculates parameters related to the lane boundary lines (relative position, angle, curvature of the lane, etc.) The device is known.

  By the way, as shown in FIG. 11, the lane boundary line may form a composite line with one or a plurality of auxiliary lines. Even in such a case, the lane boundary detection device needs to correctly obtain the position of the lane boundary (for example, the position of the lane center side edge of the lane boundary line).

  As one of the methods, when a composite line is detected, whether or not it is a composite line and the type of the composite line are determined from the characteristics of the marking pattern of the composite line, and the composite line is determined. In this case, a method of recognizing the position of the lane boundary by removing a marking pattern other than the lane boundary line (that is, an auxiliary marking or the like) as noise has been proposed (for example, see Patent Document 1).

Japanese Patent No. 3748811

  By the way, road markings such as lane boundary lines are painted on the road, so they are partially worn by being stepped on by the vehicle, or the road surface part and the paint part due to the influence of shadows, rain, etc. The contrast between the two may be significantly reduced, and it may be partially difficult to distinguish between the two.

  Therefore, even if the lane boundary line can be identified from the composite line, there is a problem that the shape of the identified lane boundary line (and thus the position of the lane boundary) cannot be detected with high accuracy. There was a problem that reliability of various controls using the position of the boundary was lowered.

  In order to solve the above problems, an object of the present invention is to improve the detection accuracy of a lane boundary by using information on road markings other than the lane boundary obtained from a captured image.

  In the lane boundary detection device of the present invention made to achieve the above object, the candidate extraction means can detect a lane boundary line that is a road marking for indicating the lane boundary from the captured image obtained by imaging the road surface in front of the vehicle. A candidate for a boundary line is extracted, and a candidate position parameter, which is a parameter for defining an approximate line indicating the position of the boundary line candidate, is obtained. The approximate line may be a straight line or a curve defined by an n-order function, an elliptic function, a hyperbolic function, or the like.

  When a plurality of boundary line candidates are detected by the candidate extraction unit, the determination unit determines that the plurality of boundary line candidates are a lane boundary line and one or a plurality of boundary lines based on the arrangement pattern of the plurality of boundary line candidates. It is determined whether or not a composite line composed of auxiliary lines is formed.

  When it is determined by the determination means that a plurality of boundary line candidates form a composite line, the inter-line interval detection means determines that each of the boundary line candidates is obtained from the arrangement pattern of the boundary line candidates that form the composite line. In addition to identifying the line type that corresponds to the lane boundary line or the auxiliary line, a line interval that is an interval between the boundary line candidate identified as the lane boundary line is determined for each boundary line candidate identified as the auxiliary line. Ask.

  In the lane boundary detection device of the present invention, the reference line setting means sets, as a reference line, the candidate position parameter with the highest reliability among the boundary line candidates forming the composite line. Then, when the set reference line is a lane boundary line, the boundary parameter setting means sets the position indicated by the candidate position parameter calculated for the reference line as a boundary parameter indicating the position of the lane boundary, When the set reference line is an auxiliary line, the position indicated by the candidate position parameter calculated for the reference line is shifted to the lane boundary side by the line interval calculated for the reference line. Set as a boundary parameter indicating the position of the lane boundary.

  Thus, in the lane boundary detection device of the present invention, when a plurality of boundary line candidates form a composite line, the auxiliary line information is not removed as noise, but is the most reliable among the boundary line candidates. The boundary parameter indicating the position of the lane boundary is obtained by using the information on the candidate position parameter having a high value and correcting the information as necessary using the information on the arrangement pattern of the composite line.

  Therefore, according to the lane boundary detection device of the present invention, it is impossible to accurately detect the shape of the lane boundary line (and hence the candidate position parameter) due to the influence of lane boundary wear, roadside shadows, rain, and the like. Even in a stable environment, when the lane boundary line forms a composite line, the boundary parameter can be obtained with high accuracy by using the auxiliary line information (candidate position parameter, line interval).

  That is, the auxiliary line is painted along the lane boundary line, and the distance between the two lines (line interval) is constant, so if the position of the auxiliary line can be accurately detected, the detection result is shifted by the line interval. Thus, the boundary parameter of the lane boundary can be accurately detected.

  In the lane boundary detection device of the present invention, the candidate extraction means, for example, as described in claim 2, the value of the pixel values of the pixels arranged along the vehicle width direction in the captured image is greatly changed. The position of the edge point to be detected may be set as a detection value, and the candidate position parameter may be obtained by performing Hough transform on the detection value.

  In this case, as described in claim 3, the reference line setting means may be configured to set the reliability of the candidate position parameter to be higher as the number of detected values that match the candidate position parameter is larger. Good.

  Note that the number of detected values that match the candidate position parameter is the number of votes in the Hough transform, and the number of detected values that are on the approximate line defined by the candidate position parameter obtained by the Hough transform. Represent.

  Further, as described in claim 4, when the candidate extracting unit is configured to obtain the line width of the boundary line candidate together with the candidate position parameter, the reference line setting unit determines the reliability of the candidate position parameter. The line width of the boundary line candidate to which the detection value matching the candidate position parameter belongs is configured to be set higher as the line type identified by the line interval detection means is closer to the preset pattern width. May be.

Further, as described in claim 5, the candidate extraction unit uses an edge extraction filter that processes a pixel value in a region of a specified size set around the target pixel to be processed, thereby The degree of change of the pixel value at, and the size of the region to be processed by the edge extraction filter may be changed based on the past determination result by the determination means.

  That is, it is desirable that the region size of the edge extraction filter used in the candidate extraction unit is optimized for the detection of the edge of the lane boundary line (so that the recognition performance is maximized), for example, the width of the lane boundary line It is conceivable to double it.

  However, in this setting, if the distance between the lane boundary line and the auxiliary line constituting the composite line is narrower than the width of the lane boundary line, the recognition performance may be deteriorated at the portion where the composite line is painted. Therefore, when a composite line is detected, it is desirable to change the area size of the edge extraction filter to a size optimized for the composite line, for example, twice the interval between the lane boundary line and the auxiliary line. It is possible to do.

By doing so, the recognition performance of the edge extraction filter can always be maximized regardless of whether it is a composite line or not.
By the way, it is conceivable that the candidate position parameter is set so as to represent the position of the edge of the boundary line candidate positioned on the center side of the lane, for example, as described in claim 6. However, the present invention is not limited to this, and it may be the center position of the boundary line candidate, or the edge position of the boundary line candidate positioned outside the lane.

  The boundary parameter setting means is configured to obtain a boundary parameter for each region obtained by dividing the region into left and right with reference to a position corresponding to the front of the vehicle in the captured image, as defined in claim 7. May be.

  Further, as described in claim 8, the determination means performs determination in each of the regions divided into left and right with reference to a position corresponding to the front of the vehicle in the captured image, and the determination result of both is When they are different from each other, they may be configured not to be composite lines.

  In other words, composite lines are installed on the left and right sides of the lane, and the same pattern is used. By using this feature, it is possible to prevent a non-composite line from being erroneously determined as a composite line. be able to.

  By the way, although each means which comprises the lane boundary detection apparatus of any one of Claims 1-8 may be implement | achieved by hardware, as described in Claim 9, these computers are these. You may implement | achieve by the lane detection program functioned as each means.

(A) is a block diagram which shows schematic structure of the lane departure warning system to which this invention was applied, (B) is a block diagram which shows the structure of the lane boundary detection apparatus used as the principal part. The flowchart which shows the content of the main process. The flowchart which shows the content of the boundary line candidate extraction process in the main process. The flowchart which shows the content of the composite line determination process in the main process. The flowchart which shows the content of the reference line selection process in the main process. The flowchart which shows the content of the boundary parameter calculation process in the main process. Explanatory drawing which shows the structure and operation | movement of an edge extraction filter. Explanatory drawing which illustrated the calculation method of a boundary parameter. (A) is a map showing the relationship between the number of votes by Hough transform and the probability (PFi), and (B) is a map showing the relationship between the candidate width parameter and the probability (PPi). Explanatory drawing which shows the modification of an edge extraction filter. Explanatory drawing which illustrated the compound line.

Embodiments of the present invention will be described below with reference to the drawings.
[overall structure]
FIG. 1 is a block diagram showing a configuration of a lane departure warning system 1 mounted on a vehicle. FIG. 1A is an overall configuration, and FIG. 1B is a detailed configuration of a lane boundary detection device as a main part.

  As shown in FIG. 1 (A), the lane departure warning system 1 generates a lane boundary detection device 10 that generates boundary parameters indicating positions and the like that are boundaries of a lane in which the host vehicle is traveling (hereinafter referred to as “lane boundary”). The lane boundary detection device 10 is connected to the lane boundary detection device 10 via the in-vehicle LAN. Based on the boundary parameters acquired from the lane boundary detection device 10 through the in-vehicle LAN, the possibility that the host vehicle deviates from the traveling lane is obtained. The vehicle control ECU 20 executes various processes for notifying the driver of the fact.

  As shown in FIG. 1B, the lane boundary detection device 10 includes a CCD camera 11 arranged to image a road surface ahead of the vehicle, and an analog image signal output from the CCD camera 11 as digital image data. An analog / digital converter (ADC) 12 for converting to image data, an image memory 13 for storing image data obtained by the ADC 12, processing for generating boundary parameters based on the image data stored in the image memory 13, etc. , A ROM 15 that stores a program for processing executed by the CPU 14, a RAM 16 that functions as a work area for the CPU 14, and a communication for outputting processing results in the CPU 14 to the outside via the in-vehicle LAN. IC17 etc. are provided.

  Like the lane boundary detection device 10, the vehicle control ECU 20 includes a CPU, a ROM, a RAM, and a communication IC. In addition, the vehicle control ECU 20 receives a detection signal from a sensor directly connected to the ECU 20, and outputs a control signal for a control target. An IO port or the like is provided. Here, at least a speaker that generates an alarm sound is connected as a control target.

  The vehicle control ECU 20 may be configured to function as a device for preventing a traffic accident such as a steering torque control device that controls the steering torque so that the vehicle does not depart from the travel lane.

[Main processing]
Next, main processing (contents of processing by the lane boundary detection program) executed by the CPU 14 will be described with reference to the flowchart shown in FIG. This process is started when the power source of the vehicle such as an ignition switch is turned on, and then repeatedly executed at regular time intervals (for example, every 100 ms).

When this process is started, first, the CCD camera 11 and ADC 12 are operated, and image data for one screen is stored in the image memory 13 (S110).
Thereafter, boundary line candidate extraction processing (S120), composite line determination processing (S130), reference line selection processing (S140), and boundary parameter calculation processing (150), which will be described in detail later, are executed, and this processing is terminated.

  In addition, about S130-S140, a process shall be performed about each area | region which divided | segmented the area | region into right and left on the basis of the position corresponded to the front of the said vehicle in a captured image. That is, the boundary parameter is obtained for the lane boundaries at the left and right ends of the lane in which the host vehicle is traveling.

[Boundary line extraction process]
Here, the boundary line candidate extraction process executed in the previous S120 will be described with reference to the flowchart shown in FIG. In addition, this processing is a road marking (a white line painted on the road surface, etc.) that may be a lane boundary indicating the boundary (left and right edges) of the lane in which the host vehicle is traveling (hereinafter referred to as “traveling lane”) ) Are extracted as boundary line candidates, and candidate position parameters indicating the position of the boundary line candidates are generated.

  In this process, first, in S210, the image data stored in the image memory 13 is processed using an edge extraction filter of a prescribed size, so that for each pixel, the horizontal line direction of the image (the vehicle width in the image). A differential value representing the degree of change in luminance value of pixels adjacent along (direction) is calculated.

  As shown in FIG. 7A, the edge extraction filter has, for example, a filter size of 1 row and 2n columns, and one row includes “1, 1,..., −1, −1,. 1 ”, n filter parameters“ 1 ”and n filter parameters“ −1 ”.

  As shown in FIG. 7B, this edge extraction filter is applied while shifting pixel by pixel on the image to obtain a filter output for the entire image. The output value calculation method when a filter is applied to a certain target pixel is as follows.

  The elliptical region shown in FIG. 7C is an enlarged part of the image shown in FIG. 7B, and the squares in the drawing represent each pixel, and each pixel has a pixel value ( (Luminance) Gi. The number in the pixel indicates the number for identifying the pixel column, and indicates a case where the edge extraction filter is applied to the pixel in the kth column as the target pixel.

  In this case, the output Hk of the edge extraction filter is calculated as in equation (1) using pixel values near the target pixel.

The filter size is set to a different size depending on the row to which the edge extraction filter is applied so as to be twice the width of the lane boundary line. Such an edge extraction filter is a known filter described in detail in, for example, Japanese Patent Application Laid-Open No. 2000-306110.

  In subsequent S220, based on the differential value obtained in S210, a point where the absolute value of the differential value is greater than or equal to a preset threshold value, that is, a point where the luminance value of an adjacent pixel greatly changes is extracted as an edge point. The coordinate value of the extracted edge point is registered in the RAM 16.

  In subsequent S230, a well-known Hough conversion process is performed using the registered edge points, thereby extending in the traveling direction of the vehicle (an angle (yaw angle) formed with the traveling direction of the vehicle is constant). Detects linear patterns (less than values).

  In the Hough transform process, the position in the image of each pixel constituting the image data is represented by coordinates (x, y), and the linear pattern is approximated by a linear function y = ax + b. The variables (a, b) that define the approximate line to be obtained are obtained.

  In subsequent S240, a pair of adjacent linear patterns representing both edges of the same marking pattern is extracted from the detected linear pattern as a boundary line candidate and labeled. This extraction is performed by a known method using, for example, the polarity of the differential value at the edge point.

Note that labeling refers to a process of numbering each boundary line candidate such as i = 0, 1,.
In the subsequent S250, candidate parameters are obtained and registered in the RAM 16 for each boundary line candidate, and this process is terminated.

  As candidate parameters, a candidate position parameter indicating the position of the boundary line candidate and a candidate width parameter indicating the line width of the boundary line candidate are obtained. However, among the pair of straight line patterns constituting the boundary line candidate, the straight line pattern located on the lane center side is used as a representative pattern, and as the candidate position parameter, a variable obtained by the Hough transform process for the representative pattern is used, and the candidate parameter is set. Uses a calculated value of the distance between a pair of straight line patterns constituting the boundary line candidate.

Further, the candidate parameters include the number of votes in the Hough transform process for the representative pattern (that is, the total number of edge points existing on the representative pattern).
[Compound line judgment processing]
Next, the composite line determination process executed in S130 will be described with reference to the flowchart shown in FIG.

In this process, first, in S310, intervals between adjacent boundary line candidates are obtained.
Here, the distance between the representative patterns of each boundary line candidate is used as the interval between the boundary line candidates.

  In subsequent S320, based on the candidate width parameter obtained in S250 and the interval between the boundary line candidates obtained in S310, it is determined whether or not the arrangement pattern of the boundary line candidates matches the pre-defined composite line arrangement pattern. If it is determined and they do not match, the process is terminated as it is.

  If the layout of the boundary line candidate matches the composite line pattern, the line type (lane boundary line, auxiliary line) of the boundary line candidate is identified and registered in the RAM 16 in S330 according to the configuration of the matched composite line pattern. To do.

  In S340, the calculation result in S310 represents an interval (hereinafter referred to as “line interval”) between the auxiliary line (identified boundary line) and the lane boundary line (identified boundary line). Then, each auxiliary line is registered in the RAM 16 and this process is terminated.

  That is, as shown in FIG. 8A, when there are auxiliary lines on both sides of the lane boundary line, in S310, D1 and D2 are calculated as the distance between the boundary line candidates, and in S320 and S340, The line type of the boundary line candidate is identified, and in S340, D1 is registered as the line interval of the auxiliary line positioned outside the lane boundary line, and D2 is registered as the line interval of the auxiliary line positioned inside the lane boundary line. Become.

[Reference line selection processing]
Next, the reference line selection process executed in S140 will be described with reference to the flowchart shown in FIG.

In this process, first, in S410, one of the boundary line candidates extracted in the boundary line candidate extraction process executed in the previous S120 (i = 0) is selected.
In S420, a first probability (PFi) representing the reliability of the candidate position parameter calculated for the selected boundary line candidate is calculated.

Here, the first probability (PFi) is set to a higher value as the number of votes by the Hough transform recorded in the process of S250 is larger. Specifically, as shown in FIG. 9A, if the number of votes by Hough transform is determined, a first probability (PFi) is calculated using a map whose probability is uniquely determined. In the map shown in FIG. 9A , when the number of votes is 50 or less, the first probability (PFi) is set to 0, and when the number of votes is 150 or more, the first probability (PFi). ) Is set to 0.5.

  When the number of votes is 50 or more and less than 150, the first probability (PFi) is set to increase proportionally as the number of votes increases. The number of votes by the Hough transform represents the number of detection values that match (position on the straight line) the straight line indicated by the candidate position parameter.

In S430, the second probability (PPi) representing the reliability of the candidate width parameter calculated for the selected boundary line candidate is calculated.
Here, the second probability (PPi) is calculated using a map whose probability is uniquely determined according to the candidate width parameter calculated in S250. In this map, the second probability is set to be higher as the boundary line candidate width is closer to the width that can be taken by the lane boundary line or the auxiliary line.

  Specifically, as shown in FIG. 9B, if the boundary line candidate width is within a range of 0.15 m to 0.3 m, the second probability (PPi) is 0.5 which is the maximum value. When the boundary line candidate width is 0.45 m or more, the second probability (PPi) is set to 0 which is the minimum value. Also, if the boundary line candidate width is within the range of 0 to 0.15 m, the longer the boundary line candidate width, the larger the value of the second probability (PPi) is set proportionally. In the range of 0.3 to 0.45 m, the value of the second probability (PPi) is set proportionally smaller as the boundary line candidate width is longer.

  In S440, a total likelihood (Pi) that is the sum of the respective likelihoods (PFi and PPi) is calculated, and this is registered in the RAM 16 in association with the boundary line candidate selected in S410.

  In S450, it is determined whether or not the processing of S410 to S440 has been executed for all lane boundary line candidates. If there are unprocessed boundary line candidates, the process returns to S410, and unprocessed boundary line candidates (here, i is larger by 1), and the processing from S420 onward is repeated.

  On the other hand, if it is determined in S450 that all the boundary line candidates have been processed, the process proceeds to S460, and the position corresponding to the front of the vehicle in the captured image (for example, the center of the image in the left-right direction) is used as a reference. In each of the regions divided into the left and right regions, the boundary line candidate having the highest overall probability (Pi) is selected as the reference line, which is registered in the RAM 16 and the present process is terminated.

[Boundary parameter calculation processing]
Next, the boundary parameter calculation process executed in S150 will be described with reference to the flowchart shown in FIG.

  In this process, first, in S510, based on the reference line selected in S140, the lateral position and angle of the reference line at a position away from the host vehicle by a predetermined distance set in advance are calculated as boundary parameters.

  In S520, it is determined whether or not the reference line is an auxiliary line. If the reference line is not an auxiliary line, that is, if it is a lane boundary line, the present process ends. That is, in this case, the boundary parameter calculated in S510 becomes the boundary parameter representing the lane boundary as it is.

  If the reference line is an auxiliary line, the process proceeds to S530, the lateral position calculated in S510 is corrected by the line interval calculated in S340 with respect to the reference line, and the present process ends. That is, in this case, the boundary parameter calculated in S510 and corrected by the processing in S530 is the boundary parameter representing the lane boundary.

  Specifically, when the auxiliary line located inside the lane boundary line is set as the reference line, in S510, D3 is calculated as the lateral position of the reference line (see FIG. 8B), and in S530, The position D2 + D3 corrected to shift the lateral position D3 of the reference line to the lane boundary side by the distance (line interval) D2 between the auxiliary line that becomes the reference line and the lane boundary line becomes the boundary parameter (FIG. 8C )reference).

  Note that the boundary parameter calculated in this way (corrected in some cases) is provided to the vehicle control ECU 20 via the in-vehicle LAN, and detects, for example, the approach speed at which the vehicle approaches the lane boundary line. This is used for a well-known departure determination process for determining a departure possibility that indicates a possibility of departure from a traveling lane detected based on the speed.

[effect]
As described above, in the lane departure warning system 1, the lane boundary detection device 10 detects the lane boundary line and the auxiliary line of the lane boundary line and the auxiliary line when the composite line composed of the lane boundary line and the auxiliary line is detected. Boundary parameters that represent the position of the lane boundary using all of the candidate position parameters obtained from these boundary line candidates and the most reliable candidate position parameters and the layout pattern (line spacing) of the composite line Seeking.

  Therefore, according to the lane boundary detection device 10, the shape of the lane boundary line (and hence the candidate position parameter) cannot be accurately detected due to wear of the lane boundary line, the shadow of the roadside object, rain, or the like. Even in the environment, when the lane boundary line forms a composite line, the boundary parameter representing the lane boundary can be obtained with high accuracy by using the auxiliary line information (candidate position parameter, line interval). Furthermore, it is possible to improve control accuracy such as deviation determination processing using this boundary parameter.

[Correspondence with Invention]
In the above embodiment, S120 corresponds to candidate extracting means, S310 to S320 as determining means, S330 to S340 as line interval detecting means, S140 as reference line setting means, and S150 as boundary parameter setting means.

[Other Embodiments]
Embodiments of the present invention are not limited to the above-described embodiments, and can take various forms as long as they belong to the technical scope of the present invention.

  For example, in the above embodiment, in S210, the change in the luminance value of the pixel is obtained as the differential value of the image data. However, when the CCD camera 11 is a color camera, the RGB signal output from the color camera, The change of the color difference signal when the RGB signal is converted into the luminance signal and the color difference signal may be obtained.

  In the above embodiment, the size of the region to be processed (hereinafter referred to as “filter size”) is changed according to the position where the edge extraction filter is applied in the image plane. You may comprise so that a filter size may be changed by whether it is a compound line. In this case, as shown in FIG. 10, when the detection target is not a composite line, the filter size is set to twice the width of the lane boundary line, and when the detection target is a composite line, the lane boundary line and the auxiliary line are set. It is desirable to set it to twice the width of the exposed portion of the road surface between the lines.

  In the above embodiment, the edge extraction filter uses a region that is set one-dimensionally with the target pixel as the center, but is set two-dimensionally with the target pixel as the center. Alternatively, a region that is a target for processing may be used, or a combination of the two may be used.

  In the above embodiment, the probability (Pi) of the boundary line candidate is obtained from the candidate position parameter and the candidate width parameter. For example, as the change in the pixel value with the detection value that matches the candidate position parameter increases, The probability (Pi) of the boundary line candidate may be increased. In addition, when information about the width of the traveling lane or the like can be acquired from a navigation device (not shown), a region where a lane boundary line or an auxiliary line is detected is estimated according to the information, and the boundary existing in the estimated region You may make it raise the probability of a line candidate.

  In the above embodiment, it is determined whether or not the lane is a composite line individually on both the left and right sides of the lane, but only when it is determined that both sides are composite lines, the process for the composite line is executed. It may be configured. However, in this case, it cannot be determined from the composite line pattern which of the plurality of boundary line candidates is a lane boundary line. For example, the boundary line candidate selected as the reference line is left as it is. It can be considered to be configured to handle as a lane boundary line.

  DESCRIPTION OF SYMBOLS 1 ... Lane departure warning system 10 ... Lane boundary detection apparatus 11 ... CCD camera 12 ... Analog-digital converter (ADC) 13 ... Image memory 14 ... CPU 15 ... ROM 16 ... RAM 17 ... Communication IC 20 ... Vehicle control ECU

Claims (9)

A lane boundary detection device that is mounted on a vehicle and detects a lane boundary indicating a boundary of a traveling lane that is a traveling lane,
Boundary line candidates that may be lane boundary lines, which are road markings to indicate lane boundaries, are extracted from captured images obtained by capturing the road surface ahead of the vehicle, and an approximate line indicating the position of the boundary line candidates is defined. Candidate extracting means for obtaining a candidate position parameter which is a parameter to be
When a plurality of boundary line candidates are detected by the candidate extracting unit, a plurality of boundary line candidates are combined from lane boundary lines and one or a plurality of auxiliary lines from the arrangement pattern of the plurality of boundary line candidates. A determination means for determining whether or not it is formed;
When it is determined by the determining means that the plurality of boundary line candidates form a composite line, each of the boundary line candidates is determined from the arrangement pattern of the boundary line candidates forming the composite line. And a line interval that identifies a line type corresponding to one of the auxiliary lines, and is an interval between the boundary line candidates identified for the lane boundary line for each of the boundary line candidates identified for the auxiliary line A line interval detecting means for obtaining
Among the boundary line candidates forming the composite line, a reference line setting unit that sets a candidate line parameter having the highest reliability as a reference line;
When the reference line is the lane boundary line, the position indicated by the candidate position parameter calculated for the reference line is set as a boundary parameter indicating the position of the lane boundary, and the reference line is the auxiliary line. In some cases, the position indicated by the candidate position parameter calculated for the reference line is shifted to the lane boundary side by the line interval calculated for the reference line to indicate the position of the lane boundary. Boundary parameter setting means for setting as a boundary parameter;
A lane boundary detection device comprising:   The candidate extracting means uses the position of an edge point at which the pixel value of a pixel arranged in the vehicle width direction in the image as a detected value is detected as the detected value, and a Hough transform is performed on the detected value. The lane boundary detection device according to claim 1, wherein the candidate position parameter is obtained by performing the following.   The lane boundary detection device according to claim 2, wherein the reference line setting unit sets the reliability of the candidate position parameter to be higher as the number of detection values that match the candidate position parameter is larger. The candidate extraction means obtains a line width of the boundary line candidate together with the candidate position parameter,
The reference line setting means sets the reliability of the candidate position parameter, the line width to which the detection value that matches the candidate position parameter belongs in advance for the line type identified by the line interval detection means. The lane boundary detection device according to claim 2, wherein the lane boundary detection device is set to be higher as the pattern width is closer. The candidate extraction means uses an edge extraction filter that processes pixel values in a region of a prescribed size set around the target pixel to be processed, thereby determining the degree of change in the pixel value at the target pixel. 5. The lane according to claim 2, wherein the lane size is changed, and a region size to be processed by the edge extraction filter is changed based on a past determination result by the determination unit. Boundary detection device.   The lane boundary detection device according to any one of claims 1 to 5, wherein the candidate position parameter represents a position of an edge of the boundary line candidate positioned on a center side of the lane.   The boundary parameter setting means obtains the boundary parameter for each region obtained by dividing the region into left and right with reference to a position corresponding to the front of the vehicle in the captured image. The lane boundary detection device according to claim 6.   The determination means performs determination in each of the regions divided into left and right with reference to a position corresponding to the front of the vehicle in the captured image, and when both determination results are different, both are combined lines The lane boundary detection device according to any one of claims 1 to 7, wherein the lane boundary detection device is determined to be not.   A lane boundary detection program for causing a computer to function as each means constituting the lane boundary detection device according to any one of claims 1 to 8.