Lane detection device and method thereof, curve starting point detection device and method thereof, and steering assistance device and method thereof
A more accurate lane is estimated by using a lane at a previous time point after coordinate conversion that is coordinate-converted based on traveling information (a vehicle speed and a yaw rate) at a current time point, as well as a lane at the current time point.
1. A lane detection method, comprising:
an image capturing step of capturing an image of a road surface in a forward view of a vehicle traveling on a lane to provide a road surface image;
a lane detection step of detecting a lane mark in the road surface image and defining an imaginary lane based on the detected lane mark to provide lane information of the imaginary lane;
a lane storage step of storing current-time lane information detected and defined in the lane detection step and a previous-time lane information detected and defined in the lane detection step a predetermined time before the current time point;
a traveling information detection step of detecting traveling information of the vehicle when the lane is detected; and
a lane estimation step of coordinate-converting the previous-time lane information based on the traveling information at the current time point and generating reliable current-time lane information through curve approximation on the basis of the coordinate-converted previous-time lane information and the current-time lane information,
wherein the reliable current-time lane information generated in the lane estimation step indicates an approximate curve lane extending forward from an subject vehicle position, the method further comprises a curve starting point detection method which comprises:
a straight-lane estimation step of coordinate-converting a previous-time straight lane information of an imaginary lane extending backward from the subject vehicle position based on the traveling information at the current time point and generating reliable current-time straight lane information of an imaginary line extending backward from the subject vehicle position through linear approximation on the basis of the coordinate-converted previous-time straight lane information and the current-time straight lane information; and
a curve starting point estimation step comprising the steps of
superimposing the estimated reliable current-time approximate straight lane on the approximate curve lane extending forward from the subject vehicle position estimated in the lane estimation step;
detecting, as a current-time curve starting point candidate point in the current-time lane, a point at which a distance in a vehicle width direction between the straight line extended forward from the subject vehicle position along the approximate straight lane and the approximate curve lane reaches or exceeds a threshold distance and detecting a curve starting point candidate point in the coordinate-converted previous-time lane;
matching a maximum value of a first Gaussian distribution function that is defined along a direction of the straight line extended forward from the subject vehicle position to the curve starting point candidate point detected at the current time point;
matching a maximum value of a second Gaussian distribution function that is defined along the direction of the straight line extended forward from the subject vehicle position to a curve starting point candidate point in the coordinate-converted previous-time lane, the maximum value of the second Gaussian distribution function being made smaller than the maximum value of the first Gaussian distribution function; and
estimating, as a more reliable curve starting point, a position of a maximum value of a composite distribution function obtained by adding the first Gaussian distribution function having the maximum value and the second Gaussian distribution function having the smaller maximum value.
2. The lane detection method according to claim 1 , wherein
the traveling information at the current time point includes a vehicle speed as well as any one of a yaw rate and a steering angle at the current time point, and
in the lane estimation step, the reliable current-time lane is generated by superimposing the current-time lane information and the coordinate-converted previous-time lane information.
3. The lane detection method according to claim 2 , wherein
in the lane estimation step,
when the reliable current-time lane information is generated by superimposing the current-time lane information and the coordinate-converted previous-time lane information, in which a weighting factor of the coordinate-converted previous-time lane is made smaller than a weighting factor of the current-time lane.
4. The lane detection method according to claim 3 , wherein
in the lane estimation step,
when the reliable current-time lane information is generated by superimposing the current-time lane information and the coordinate-converted previous-time lane information, if the current-time lane information has a missing lane part, the weighting factor of the current-time lane information is used for a weighting factor of the coordinate-converted previous-time lane information, without making the weighting factor of the coordinate-converted previous-time lane information smaller than the weighting factor of the current-time lane information with respect to a position corresponding to the missing lane part in the current-time lane information.