METHOD FOR DETECTING LANE DEPARTURE AND APPARATUS THEREOF

• US 20100002911A1
• Filed: 07/06/2008
• Published: 01/07/2010
• Est. Priority Date: 07/06/2008
• Status: Active Grant

• Current Assignee: Automotive Research & Testing Center
• Original Assignee: Automotive Research & Testing Center
• Inventors: Su, Yi-Feng, Wu, Jui-Hung, Liu, Jing-Fu
• Granted Patent: US 8,311,283 B2
• Time in Patent Office: Days
• US Class Current: 382/104
• CPC Class Codes: B60W 30/12 Lane keeping
• 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 method for detecting the lane departure of a vehicle includes an image recognition process and a deviation estimation process. The image recognition process includes the following steps: an image capturing step for capturing image frame data by using an image capturing unit; and a lane line recognition for analyzing the image frame data for determining the lane lines. By using a quadratic curve fitting equation, a plurality of lane line being detected so as to establish a road geometry estimation model. The road geometry estimation model is inputted into the deviation estimation process to detect the lane departure of the vehicle so as to alert the driver. Furthermore, an apparatus for detecting the deviation of the vehicle, comprising: an image capturing unit, a processing unit and a signal output unit.

Claims

1. A method for detecting lane departure of a vehicle, comprising:

an image recognition process which includes the following steps;

an image capturing step for capturing an image frame data of the road ahead of the vehicle by using an image capturing unit; and

a lane line recognition step for analyzing the image frame data and establishing a road geometry estimation model by using a quadratic curve fitting equation, a plurality of parameters solved by substituting a plurality of position data calculated by recognizing a plurality of lane lines in the image frame data to correct the road geometry estimation model and make the corrected road geometry estimation model approximating a practical lane line track.

2. The method for detecting lane departure of the vehicle as claimed in claim 1, wherein the lane line recognition step further includes a lane line feature recognition method which includes the following methods:

a high grayscale recognition for discriminating the lane line from road surface by the character that the lane lines have higher grayscale than the road surface;

a lane line edge feature recognition for calculating a lane line width in the image frame data by the edge feature that the lane line connects to the road surface; and

a lane width recognition for solving a standard lane width in the image frame data by the equation which a practical standard lane width multiplied by a ratio of a lane line width in the image frame data to a practical standard lane line width equals to the standard lane width in the image frame data.

3. The method for detecting lane departure of the vehicle as claimed in claim 2, wherein the lane line recognition step includes the following steps:

A. establishing the road geometry estimation model;

a position information of a predefined disappear point is substituted in a quadratic curve fitting equation to establish a initial road geometry estimation model, whereby the disappear point is an intersection point which is formed by two lane lines of the image frame data intersecting in a distant place, the predefined disappear point is a center point of the image frame data as the initial road geometry estimation model is established;

B. initial point search;

the image frame data is divided into a plurality of sections, a initial point of the lane line in a predetermined region is detected based on at least one side of the estimated lane line of the initial road geometry estimation model and by using the lane line feature recognition method to determine which section the initial point of practical lane line is located in;

C. road geometry estimation model correction;

in the section which the initial point is located in, the lane line feature recognition method is continuously used to recognize the lane lines in each section, the position information of the lane lines is substituted in the quadratic curve fitting equation to correct the road geometry estimation model; and

D. disappear point detection;

check whether the image intersection of the two side of the estimated lane lines calculated by the corrected road geometry estimation model is located in a predetermined region to determine whether the road geometry estimation model approximates to the practical lane line track;

if not, repeat the lane line recognition step to find the most approximate road geometry estimation model.

4. The method for detecting lane departure of the vehicle as claimed in claim 3, wherein the lane line recognition step further includes the following step:

C′

. compensation for line point of image;

as in the step C a lane line point is not detected the compensation for line point of image is used by the equation that the practical standard lane width multiplied by the ratio of the lane line width in the image frame data to the practical standard lane line width equals to the standard lane width in the image frame data;

the standard lane width in the image frame data is used to compensate the lane line point;

the compensation equation is listed as the following;

5. The method for detecting lane departure of the vehicle as claimed in claim 3, wherein a single section in step C is divided into a plurality of rows;

the ROI (region of interest) detection is performed;

the ROI detection function is listed as the following;

ROI_n=[u_i-1−

λ

_n·

Mark_I,u_i-1+λ

_n·

Mark_I]

ROI_d=[u_i-1−

λ

_d·

Mark_I,u_i-1+λ

_d·

Mark_I]

where u_i-1 is a transversal coordinate of the lane line of a former row;

λ

_n is a parameter showing that no lane line is detected in the former row;

λ

_d is a parameter showing that a lane line is detected in the former row; and

Mark_I;

is a width of a lane line on an image plane of the row processed now.

6. The method for detecting lane departure of the vehicle as claimed in claim 1, wherein the quadratic curve fitting equation is y=k1·

x²+m1·

x+b1, where x and y are the longitudinal axis and transversal axis of a two dimensional space;

k1, m1, b1 are the predefined parameters, and

7. The method for detecting lane departure of the vehicle as claimed in claim 1, wherein the lane line recognition step further includes the following equation to determine daytime or nighttime so as to switch the lane line detection modes:

8. The method for detecting lane departure of the vehicle as claimed in claim 1 further comprising a deviation estimation process which includes the following steps:

an operation processing step;

use the corrected road geometry estimation model and a quadratic curve fitting equation to calculate a transversal displacement, a traveling slope, and a lane curvature; and

a pre-alert step;

compare the transversal displacement and the corrected road geometry estimation model to determinate a distance between the vehicle and the lane line;

if the vehicle is close to the lane line, a warning is issued.

9. The method for detecting lane departure of the vehicle as claimed in claim 8, wherein the transversal displacement is calculated by the equation:

Δ

=y_L−

L×

ε

_L

where Δ

is the transversal displacement of the vehicle;

y_L is a displacement of the vehicle from a reference point;

L is a preview distance; and

ε

_L is the traveling slope.

10. The method for detecting lane departure of the vehicle as claimed in claim 8, wherein the distance between the vehicle and the lane line is calculated by the equation:

11. The method for detecting lane departure of the vehicle as claimed in claim 8, wherein the quadratic curve fitting equation is y=k2·

x²+m2·

x+b2, where x and y are longitudinal axis and transversal axis of a real plane; and

k2, m2, and b2 are parameters which are solved by the corrected road geometry estimation model.

12. The method for detecting lane departure of the vehicle as claimed in claim 11, wherein the traveling slope is calculated by the corrected road geometry estimation model, the equation of the traveling slope is ε

_L=2·

k2·

x+m2.

13. The method for detecting lane departure of the vehicle as claimed in claim 11, wherein the lane curvature is derived from a lane line track estimated by the corrected road geometry estimation model, the equation of the lane curvature is

14. An apparatus for detecting lane departure of a vehicle comprising:

an image capturing unit for capturing a image frame data and transferring the captured image frame data to a processing unit;

the processing unit analyzing the image frame data to establish a road geometry estimation model and recognizing a position data of a plurality of lane lines to correct the road geometry estimation model, the correct the road geometry estimation model used to determine whether the vehicle is deviating from the lane, the processing unit emitting signals to a signal output unit as achieving a predefined alert range; and

the signal output unit for providing light source, images or audio signals as the signals from the processing unit are transferred to the signal output unit.

15. The apparatus for detecting lane departure of the vehicle as claimed in claim 14, wherein the image capturing unit could be a CCD, a CMOS camera.