Road line recognition apparatus

• US 20060239509A1
• Filed: 04/25/2006
• Published: 10/26/2006
• Est. Priority Date: 04/26/2005
• Status: Active Grant

• Current Assignee: Subaru Corp.
• Original Assignee: Fuji Heavy Industries Limited (Subaru Corp.)
• Inventors: Saito, Toru
• Granted Patent: US 8,224,031 B2
• Time in Patent Office: Days
• US Class Current: 382/104
• CPC Class Codes: G06V 10/753 Transform-based matching, e.g. Hough transform
• CPC Class Codes: G06V 20/588 Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road

Abstract

A road line recognition apparatus, including: an imaging section imaging a progress path of a own vehicle including a road to output a couple of images; an image processing section calculating a distance in a real space in a set region of at least an image on one side based on the imaged couple of images; and a detection section detecting a road line; wherein the detection section includes: a road line candidate point detection and conversion processing unit detecting a pixel on a road surface as a road line candidate point based on luminance and the distance with regard to the image on one side, and performing Hough conversion of the road line candidate point; a road line straight line detection processing unit detecting one straight line proper to the road line on each of a right side and a left side of the own vehicle based on at least a position or a behavior of the own vehicle between straight lines obtained by the Hough conversion; and a road line detection processing unit detecting the road line of a shape of a straight line or a curved line by recording a road line position which is a road line candidate point indicating a road line among the road line candidate points based on the detected straight line.

Claims

1. A road line recognition apparatus, comprising:

an imaging section imaging a progress path of a own vehicle including a road to output a couple of images;

an image processing section calculating a distance in a real space in a set region of at least an image on one side based on the imaged couple of images; and

a detection section detecting a road line;

wherein the detection section includes;

a road line candidate point detection and conversion processing unit detecting a pixel on a road surface as a road line candidate point based on luminance and the distance with regard to the image on one side, and performing Hough conversion of the road line candidate point;

a road line straight line detection processing unit detecting one straight line proper to the road line on each of a right side and a left side of the own vehicle based on at least a position or a behavior of the own vehicle out of straight lines obtained by the Hough conversion; and

a road line detection processing unit detecting the road line of a shape of a straight line or a curved line by recording a road line position which is a road line candidate point indicating a road line among the road line candidate points based on the detected straight line.

2. The road line recognition apparatus according to claim 1,

wherein the road line candidate point detection and conversion processing unit of the detection section sets a search region on the image on one side in order to search for a road line candidate point only in a circumference of a detection position of the road line or the straight line, detected last time, on the image on one side in this time detection when the road line has been detected in a last time detection or when the road line has not been detected but the straight line has been detected in the last time detection.

3. The road line recognition apparatus according to claim 1,

wherein the road line candidate point detection and conversion processing unit of the detection section sets a search region on the image on one side in order to search for a road line candidate point only in a circumference of a position distant from a presumption locus presumed by a behavior of the own vehicle by a predetermined distance in this time detection when both of the road line and the straight line have not been detected for predetermined times.

4. The road line recognition apparatus according to claim 3,

wherein the road line candidate point detection and conversion processing unit of the detection section sets the search region of this time detection to be expanded in comparison with the search region in a last time detection.

5. The road line recognition apparatus according to claim 2,

wherein the road line candidate point detection and conversion processing unit of the detection section enlarges the search region of a road line part or a straight line part distant more than the first separation distance threshold value from a presumption locus presumed based on the behavior of the own vehicle in this time detection to a neighborhood of an end of the own vehicle presumed from the presumption locus when the road line or the straight line, detected in the last time detection, are distant more than the first separation distance threshold value from the presumption locus.

6. The road line recognition apparatus according to claim 2,

wherein the road line candidate point detection and conversion processing unit of the detection section enlarges the search region of a road line part or a straight line part, close to be within a second separation distance threshold value to a presumption locus presumed based on the behavior of the own vehicle in this time detection to a direction of being distant from the own vehicle by a predetermined quantity when the road line or the straight line, detected in the last time detection, is close to be within a second separation distance threshold value to the presumption locus.

7. The road line recognition apparatus according to claim 2,

wherein the road line candidate point detection and conversion processing unit of the detection section, when setting the search region based on the road line or the straight line detected in the last time detection, in this time detection, enlarges the search region from a road line or a straight line which is in an opposite side of the road line or the straight line detected in the last time detection, to a position distant to a direction of the road line or the straight line by a predetermined quantity.

8. The road line recognition apparatus according to claim 2,

wherein the road line candidate point detection and conversion processing unit of the detection section expands the search region in a direction of a line segment when setting the search region based on the road line or the straight line detected in the last time detection, and searching for a road line candidate point while shifting a horizontal line of the image on one side upward by one pixel in the search region, in this time detection, and when the line segment connecting a road line candidate point detected on a present horizontal line with a road line candidate point detected immediately before deviates from the search region.

9. The road line recognition apparatus according to claim 1,

wherein the road line candidate point detection and conversion processing unit of the detection section does not search a region with a width of the own vehicle from a presumption locus presumed based a behavior of the own vehicle, the presumption locus being set as a center of the width, in a search of a road line candidate point.

10. The road line recognition apparatus according to claim 9,

wherein the road line candidate point detection and conversion processing unit of the detection section searches up to a position distant from the road line or the straight line, detected in the last detection, by a predetermined distance, in the region with the width of the own vehicle from the presumption locus presumed based on the behavior of the own vehicle, in the search of a road line candidate point.

11. The road line recognition apparatus according to claim 1,

wherein the road line candidate point detection and conversion processing unit of the detection section sets a pixel portion other than a pixel portion indicating a solid body extracted on the image on one side based on information of the distance calculated by the image processing section, as a search region and perform a search for a road line candidate point in the set search region.

12. The road line recognition apparatus according to claim 1,

wherein the road line straight line detection processing unit of the detection section detects the straight line proper to the road line based on a distance between road lines calculated from distance data detected by the image detection section and the number of pixels between the road lines.

13. The road line recognition apparatus according to claim 1,

wherein the road line straight line detection processing unit of the detection section detects the straight line proper to the road line each of right and left sides of the own vehicle out of the straight lines obtained by the Hough conversion, based on parallelism with a presumption locus of the own vehicle calculated based on distance data detected by the image detection section, a distance from a road line detected last time, and a distance from a road line presumed from an opposite road line detected last time.

14. The road line recognition apparatus according to claim 13,

wherein the road line straight line detection processing unit of the detection section detects only one of straight lines, each detected as the straight line proper to the road line on the right side and the left side of the own vehicle, as the straight line proper to the road line based on each counted value on the Hough plane of each straight line, the parallelism with the presumption locus of the own vehicle calculated based on the distance data detected by the image detection section, and a displacement quantity from a road line position detected last time when a road width in the real space of the straight lines detected on the right side and the left side of the own vehicle is not within a predetermined range, or the parallelism of the straight lines is worse than a set threshold value.

15. The road line recognition apparatus according to claim 1,

wherein the road line detection processing unit of the detection section converts the number of pixels between the road line positions on the right side and the left side to a road width in the real space based on distance data, and calculates a maximum value and a minimum value of the road width based on the converted road width when the road line positions on both of the right and the left sides are detected at time of recording the road line positions of the road line candidate points indicating the road line, out of the road line candidate points using the straight line detected on each of the right and the left sides of the own vehicle as a reference.

16. The road line recognition apparatus according to claim 1,

wherein the road line detection processing unit of the detection section is shifting a horizontal line of the image on one side upward by one pixel while recording a road line candidate point as the road line position when displacement in a vertical direction and a horizontal direction between a road line candidate point detected on an upper horizontal line after the road line positions detected on the horizontal line based on the detected straight lines have reached a predetermined number and the last detected road line position is severally within a predetermined value.

17. The road line recognition apparatus according to claim 15,

wherein the road line detection processing unit of the detection section is shifting a horizontal line of the image on one side upward by one pixel while recording a road line candidate point as the road line position when displacement in a vertical direction and a horizontal direction between a road line candidate point detected on an upper horizontal line after the road line positions detected on the horizontal line based on the detected straight lines have reached a predetermined number and the last detected road line position is severally within a predetermined value, and

when the road width in the real space converted from the number of pixels between the road line positions on the right side and the left side is out of a range from the minimum value to the maximum value, the road line detection processing unit detects a road line on which a road line position having higher parallelism with the presumption locus of the own vehicle as a road line to be set as the reference.

18. The road line recognition apparatus according to claim 17,

wherein the road line detection processing unit of the detection section detects a road line on which a road line position having lower parallelism with the presumption locus of the own vehicle as a road line at a road line branching when the road width in the real space converted form the number of pixels between the road line positions on the right side and the left side exceeds the maximum value, and calculates a position in the real space of a road line position at which the calculated road width first exceeds the maximum value, and further calculates a distance from the own vehicle to the road line position to detect the calculated distance as a distance where the road line branching starts.

19. The road line recognition apparatus according to claim 17,

wherein the road line detection processing unit of the detection section detects a road line on which a road line position having lower parallelism with the presumption locus of the own vehicle as a road line at a road line joining when the road width in the real space converted form the number of pixels between the road line positions on the right side and the left side is lower than the minimum value, and calculates a position in the real space of a road line position at which the calculated road width first becomes lower than the minimum value, and further calculates a distance from the own vehicle to the road line position to detect the calculated distance as a distance where the road line joining starts.

20. The road line recognition apparatus according to claim 1,

wherein the road line detection processing unit of the detection section calculates distances between road line positions belonging to a detected road line to detect whether the road line is a continuous line or a broken line, and characterizes information of the detected road line or information of the road line positions as the continuous line or the broken line to output the information.

21. The road line recognition apparatus according to claim 1,

wherein the road line candidate point detection and conversion processing unit detects a pixel on the image on one side as a road line candidate point based on the luminance and the distance of each pixel with regard to the image on one side, and performs the Hough conversion of the road line candidate point, the pixel detected as the road line candidate having the luminance and a derivative value of luminance being larger than a predetermined first start point luminance threshold value and first start point edge strength threshold value, respectively.

22. The road line recognition apparatus according to claim 21,

wherein the road line candidate point detection and conversion processing unit of the detection section searches a horizontal line of the image on one side, and

when a pixel portion is found in which luminance values are equal to or more than a second start point luminance threshold value which is larger than the first start point luminance threshold value and derivative values of luminance is equal to or more than a second start point edge strength threshold value and further average luminance of a high luminance pixel portion corresponding to a road line is equal to or more than a predetermined first road line average luminance threshold value, the road line candidate point detection and conversion processing unit deletes road line candidate points corresponding to a pixel portion in which average luminance of an already detected high luminance pixel portion is smaller than the first road line average luminance threshold value, and saves road line candidate points corresponding to the newly found pixel portion to set as an object of the Hough conversion.

23. The road line recognition apparatus according to claim 21,

wherein the road line candidate point detection and conversion processing unit of the detection section searches a horizontal line of the image on one side from a start point capable of being the road line candidate point, and

when pixels, at each of which a derivative value of luminance is not more than a predetermined end point edged strength threshold value or a luminance value is smaller than a luminance value of the start point, are not detected, despite searching, up to the number of the pixels, the number corresponding to a road line width threshold value set according to a distance from the own vehicle on the real space of the horizontal line and a road line width, the road line candidate point detection and conversion processing unit does not set the start point as a road line candidate point and does not set it as an object of the Hough conversion.

24. The road line recognition apparatus according to claim 21,

wherein the road line candidate point detection and conversion processing unit of the detection section searches a horizontal line of the image on one side from a start point capable of being the road line candidate point, and detects a pixel at which a derivative value of luminance is not more than a predetermined end point edged strength threshold value or a luminance value is smaller than a luminance value of the start point as an end point, and further

when a difference between average luminance of pixels from the start point to the end point and road surface luminance is smaller than a predetermined second road line average threshold value, the road line candidate point detection and conversion processing unit does not set the start point as a road line candidate point, and does not set it as an object of the Hough conversion.

25. The road line recognition apparatus according to claim 24,

wherein, when the derivative value of luminance at the start point is smaller than a third start point edge strength threshold value which is set to a value larger than the first start point edge strength threshold value, and the number of the pixels from the start point to the end point is larger than a second road line width threshold value set correspondingly to an ordinary road line width based on a distance from the own vehicle on the real space of the horizontal line and a road line width, the road line candidate point detection and conversion processing unit of the detection section sets the second road line width average luminance threshold value to a higher value, and compares the difference between the average luminance of the pixels from the start point to the end point and the road surface luminance with the second road line average luminance threshold value.

26. The road lien recognition apparatus according to claim 24,

wherein when the derivative value of luminance at the start point is not smaller than a third start point edge strength value which is set to a value larger than the first start point edge strength threshold value, and a number of the pixels from the start point to the end point is not larger than a second road line width threshold value set correspondingly to an ordinal road line width based on a distance from the own vehicle on the real space of the horizontal line and a road line width, the road line candidate point detection and conversion processing unit of the detection section sets the second road line width average luminance threshold value to a lower value, and compares the difference between the average luminance of the pixels from the start point to the end point and the road surface luminance with the second road line average luminance threshold value.