Lane line detection method and apparatus

1. A lane line detection method applicable for an in-vehicle device and comprising:

determining a region of interest in an image to be detected;

extracting lane line pixel features in the region of interest;

combining similar lane line pixel features to generate superpixels corresponding to the combined lane line pixel features; and

performing a clustering and fitting process for the superpixels to obtain a target lane line,

wherein the step of determining the region of interest in the image to detected comprises:

setting a lane line procxing region around the in-vehicle device;

determining coordinate values of a midpoint on each of boundary lines of the lane line processing region in a real coordinate system in which the in-vehicle device is located;

converting each of the coordinate values into a corresponding image coordinate value in an image coordinate system corresponding to the image to be detected; and

determining the region of interest in the image to be detected according to the respective image coordinate values.

2. The method according to claim 1 , wherein the step of extracting the lane line pixel features in the region of interest comprises:

selecting a first edge image and a second edge image in the region of interest;

performing a binarization process for each of the rust edge image and the second edge image to obtain a first binarized edge image and a second binarized edge image; and

performing a mw scanning for each of the rust binarized edge image and the second binarized edge image, and obtaining a rust lane line pixel feature point and a second lane line pixel feature point in respective rows.

3. The method according to claim 2 , wherein the step of combining the similar lane line pixel features to generate the superpixel corresponding to the combined lane line pixel features comprises:

copying and saving the first lane line pixel feature point and the second lane line pixel feature point into a new image to obtain a lane line feature map when a distance between the first lane line pixel feature point and the second lane line pixel feature point satisfies a set distance threshold;

searching for a superpixel feature from an edge position of the lane line feature map, and using a first found superpixel feature as a superpixel feature reference point;

finding similar features to the superpixel feature reference point within a candidate range of the superpixel feature reference point; and

combining the superpixel feature reference point with the found similar features to generate the superpixel.

4. The method according to claim 1 , wherein the step of performing the clustering and fitting process for the respective superpixels to obtain the target lane line comprises:

performing a clustering process for the respective superpixels to obtain a plurality of candidate lane lines;

calculating a length value of each of the candidate lane lines; and

performing a quadratic curve fitting for each of the candidate lane lines whose length value is greater than a set threshold to obtain a target lane line.

5. The method according to claim 2 , wherein the binarization process includes comparing pixel values of the first edge image and the second edge image to a pixel threshold which is associated with positions of the pixels in the first edge image and the second edge image.

6. The method according to claim 5 , wherein the pixel threshold is also associated with a vertical gradient of the pixels in the first edge image and the second edge image.

7. The method according to claim 3 , wherein it is determined whether a distance value between the matched first and second lane line pixel feature points is between a first threshold and a second threshold that are associated with positions of the first lane line pixel feature point and the second lane line pixel feature point.

8. The method according to claim 4 , wherein the following sample distance metric formula is defined to perform the clustering and fitting process:

d ( u i ,u j )=α·abs(θ ti −θ tj )+α·abs(θ mi −θ mj )+α·abs(θ bi −θ bj )+β·abs( t i ·u−t j ·u )+β·abs( m i ·u−m j ·u )+β·abs( b i ·u−b j ·u ),

where d(u i , u j )represents a distance between superpixels u i and u j , θ ti represents a gradient direction angle of a top pixel point ti of the superpixel u i , θ mi represents a gradient direction angle of a middle pixel point mi of the superpixel u i , θ bi represents a gradient direction angle of a bottom pixel point bi of the superpixel u i , θ tj represents agradient direction angle of a top pixel point tj of the superpixel u j , θ mj represents a gradient direction angle of a middle pixel point mj of the superpixel u j , θ bj a represents a gradient direction angle of a bottom pixel point bj of the superpixel u j , α represents the weight of angle, β represents the weight of distance, α and β represent a preset fixed value, abs represents calculating an absolute value, t i ·u represents an abscissa of the top pixel point ti, m i ·u represents an abscissa of the middle pixel point mi, b i ·u represents an abscissa of the bottom pixel point bi, t i ·u represents an abscissa of the top pixel point tj, m j ·u represents an abscissa of the middle pixel point mj, and b j ·u represents an abscissa of the bottom pixel point bj.

9. A lane line detection apparatus, comprising:

a region of interest determining component configured to determine a region of interest in an image to be detected;

a pixel feature extracting component configured to extract lane line pixel features in the region of interest;

a superpixel generating component configured to combine similar lane line pixel features to generate superpixels corresponding to the combined lane line pixel features; and

a target lane line obtaining component configured to perform a clustering and fitting process for the superpixels to obtain a target lane line,

wherein the region of interest determining component comprise:

a processing region setting sub-component configured to set a lane line processing region around the in-vehicle device;

a coordinate value determining sub-component configured to determine coordinate values of a midpoint on each of boundary lines of the lane line processing region in a real coordinate system in which the in-vehicle device is located;

an image coordinate value obtaining sub-component configured to convert each of the coordinate values into a corresponding image coordinate value in an image coordinate system corresponding to the image to be detected; and

a region of interest determining sub-component configured to determine the region of interest in the image to be detected according to the respective image coordinate values.

10. The apparatus according to claim 9 , wherein the pixel feature extracting component includes:

an edge image selecting sub-component configured to select a first edge image and a second edge image in the region of interest;

a binarization processing sub-component configured to perform a binarization process for each of the first edge image and the second edge image to obtain a first binarized edge image and a second binarized edge image; and

a scan processing sub-component configured to perform a row scanning for each of the first binarized edge image and the second binarized edge image, and obtain a first lane line pixel feature point and a second lane line pixel feature point in respective rows.

11. The apparatus according to claim 10 , wherein the superpixel generating component comprises:

a lane line feature map obtaining sub-component configured to copy and save the first lane line pixel feature point and the second lane line pixel feature point into a new image to obtain a lane line feature map when a distance between the first lane line pixel feature point and the second lane line pixel feature point satisfies a set distance threshold;

a reference point selecting sub-component configured to search for a superpixel feature from an edge position of the lane line feature map, and use a first found superpixel feature as a superpixel feature reference point;

a finding sub-component configured to find similar features to the superpixel feature reference point within a candidate range of the superpixel feature reference point; and

a superpixel generating sub-component configured to combine the superpixel feature reference point with the found similar features to generate the superpixel.

12. The apparatus according to claim 9 , wherein the target lane line obtaining component comprises:

a clustering and fitting processing sub-component configured to perform a clustering process for the respective superpixels to obtain a plurality of candidate lane lines;

a length value calculating sub-component configured to calculate a length value of each of the candidate lane lines; and

a target lane line obtaining sub-component configured to perform a quadratic curve fitting for each of the candidate lane lines whose length value is greater than a set threshold to obtain a target lane line.

13. The apparatus according to claim 10 , wherein the binarization processing sub-component is configured to compare pixel values of the first edge image and the second edge image to a pixel threshold which is associated with positions of the pixels in the first edge image and the second edge image.

14. The apparatus according to claim 13 , wherein the pixel threshold is also associated with a vertical gradient of the pixels in the first edge image and the second edge image.

15. The apparatus according to claim 11 , wherein the lane line feature map obtaining sub-component is configured to determine whether a distance value between the matched first and second lane line pixel feature points is between a first threshold and a second threshold that are associated with positions of the first lane line pixel feature point and the second lane line pixel feature point.

16. The apparatus according to claim 12 , wherein the clustering and fitting processing sub-component defines the following sample distance metric formula to perform the clustering and fitting process:

d ( u i ,u j )=α·abs(θ ti −θ tj )+α·abs(θ mi −θ mj )+α·abs(θ bi −θ bj )+β·abs( t i ·u−t j ·u )+β·abs( m i ·u−m j ·u )+β·abs( b i ·u−b j ·u ),

where d(u i , u j ) represents a distance between superpixels u i and u j i , θ ti represents a gradient direction angle of a top pixel point ti of the superpixel u i , θ mi represents a gradient direction angle of a middle pixel point mi of the superpixel u i , θ bi represents a gradient direction angle of a bottom pixel point bi of the superpixel u i , θ tj represents a gradient direction angle of a top pixel point tj of the superpixel u j , θ mj represents a gradient direction angle of a middle pixel point mj of the superpixel u j , θ bj a represents a gradient direction angle of a bottom pixel point bj of the superpixel u j , α represents the weight of angle, β represents the weight of distance, α and β represent a preset fixed value, abs represents calculating an absolute value, t i ·u represents an abscissa of the top pixel point ti, m i ·u represents an abscissa of the middle pixel point mi, b i ·u represents an abscissa of the bottom pixel point bi, t j ·u represents an abscissa of the top pixel point tj, m j ·u represents an abscissa of the middle pixel point mj, and b j ·u represents an abscissa of the bottom pixel point bj.