Image analysis – Pattern recognition – Feature extraction
Reexamination Certificate
2001-02-16
2004-11-02
Ahmed, Samir (Department: 2623)
Image analysis
Pattern recognition
Feature extraction
C382S104000, C382S106000, C382S107000, C382S108000, C382S199000
Reexamination Certificate
active
06813383
ABSTRACT:
TECHNICAL FIELD
The present invention relates to a detecting method of a linear mark and a detector. The present invention particularly relates to a detecting method of a linear mark and a detector, which allows a linear mark characteristic point to be detected without confusion with surrounding structure and noise, for example, which is suitable for detecting a white line in an image of a road.
BACKGROUND ART
A conventional linear mark detector is disclosed in Japanese Patent Laid-Open No. HEI 10-47923.
FIG. 6
is a block diagram showing a configuration of the conventional linear mark detector. The linear mark detector is constituted by an image pickup means
501
, an edge detecting means
502
, an edge memory means
503
, an image processing means
504
, a curvature center specifying means
505
, an edge column extracting means
506
, a tangent line specifying means
507
, a center of gravity specifying means
508
, a normal line specifying means
509
, a segment setting means
510
and so on.
The image pickup means
501
picks up an image of a road surface and produces image information including boundaries of both sides of a lane, which is provided on the road surface to be detected.
The edge detecting means
502
detects edges based on intensities of the image information and successively outputs edges representing the boundaries of both sides.
The edge memory means
503
successively stores the edges representing the boundaries of both sides outputted by the edge detecting means
502
and forms continuous edge groups in two columns representing the boundaries of both sides.
The image processing means
504
specifies the boundaries of both sides on a three-dimensional plain surface at least based on the edge groups arranged in two columns.
The curvature center specifying means
505
specifies a curvature center relative to the boundaries of both sides based on the previous image, in which the boundaries of both sides are specified.
The edge column extracting means
506
extracts a pair of edge columns respectively by predetermined lengths along a concentric circle whose center conforms to the curvature center. The extraction is made from the continuous edge groups arranged in two columns. The edge groups have been additionally stored by the edge memory means
503
and represent the boundaries of both sides.
The tangent specifying means
507
specifies a pair of tangents respectively relative to the edge columns in the case where a pair of the edge columns extracted by the edge column extracting means
506
are in parallel with each other.
The center of gravity specifying means
508
specifies a center of gravity of the edge groups constituting a pair of the edge columns extracted by the edge column extracting means
506
.
The normal line specifying means
509
specifies a normal line passing through the center of gravity relative to a pair of the tangents.
The segment setting means
510
specifies intersection points of the normal line and a pair of the tangents. The segment setting means
510
sets a line segment between a pair of the intersection points as a segment of the lane and outputs the segment to the image processing means
504
.
The following will be described an operation as below.
The edge detecting means
502
detects edges representing the boundaries of both sides based on intensities of the image information including the boundaries of both sides of the lane, which is provided on the road surface to be detected. The edge memory means
503
successively stores the edges into an edge memory.
Further, the continuous edge groups in two columns are formed, which represent the boundaries of both sides. The image processing means
504
specifies the boundaries of both sides on the three-dimensional plain surface at least based on the edge groups arranged in two columns. The curvature center specifying means
505
specifies a curvature center relative to the boundaries of both sides based on the previous image, in which the boundaries of both sides are specified.
Next, the edge column extracting means
506
extracts a pair of the edge columns respectively by predetermined lengths along a concentric circle whose center conforms to the curvature center. The extraction is made from the continuous edge groups arranged in two columns. The edge groups have been additionally stored into the edge memory and represent the boundaries of both sides.
And then, the tangent specifying means
507
specifies a pair of tangents respectively for the edge columns in the case where a pair of the extracted edge columns is in parallel with each other. Moreover, the center of gravity specifying means
508
specifies a center of gravity of the edge groups constituting a pair of the extracted edge columns.
Further, the normal line specifying means
509
specifies a normal line which is passing through the center of gravity, relative to a pair of the tangents.
Finally, intersection points of the normal line and a pair of the tangents are specified. And then, the segment setting means
510
sets a line segment between a pair of the intersection points as a segment of the lane, and the image for specifying the boundaries of both sides is updated according to the segment.
Problems to be Solved by the Invention
The conventional linear mark detector has the above-mentioned configuration, in which only edges are detected as a boundary. However, the edges exist due to a variety of surrounding configurations and noise as well as linear mark boundaries. It is therefore difficult to clearly distinguish the linear mark boundaries from such configurations and noise. Consequently, the linear mark boundaries cannot be detected with reliability.
An object of the present invention is to provide a detecting method of a linear mark and a detector, by which a linear mark can be detected without the influence of the surrounding configurations and noise.
Means for Solving the Problems
A linear mark detecting method according to the present invention is characterized by including a computing step of computing characteristic values such as a difference in luminance value between the inside and the outside of a linear mark region, edge intensity at an end of a linear mark, and uniformity of luminance in the linear mark, for each point in an inputted target image; and an evaluation value giving step of giving to each point in the target image an evaluation value, which is indicative of a likelihood of a linear mark characteristic point for detecting the linear mark, based on each of the characteristic values computed in the computing step.
A linear mark detector according to the present invention is characterized by including an image input means for inputting a target image, an image processing means for computing characteristic values such as a difference in luminance value between the inside and the outside of a linear mark region, edge intensity at an end of the linear mark, and uniformity of luminance in the linear mark, for each point in the target image inputted by the image input means; a linear mark characteristic integrating means for giving to each point in the target image an evaluation value, which is indicative of a likelihood of a linear mark characteristic point for detecting a linear mark, based on each of the characteristic values computed by the image processing means; and a linear mark characteristic point data output means for outputting the evaluation value indicative of a likelihood of a linear mark characteristic point, the value being added to each of the points by the linear mark characteristic integrating means.
According to the linear mark detecting method and the detector of the present invention, computation is conducted on characteristic values such as a difference in luminance value between the inside and the outside of a linear mark region, edge intensity at an end of the linear mark, and uniformity of luminance in the linear mark, for each point in the above target image. Based on each of the computed characteristic values, an evaluation value indicative of a likelihood of a line
Ahmed Samir
Le Brian
NEC Corporation
LandOfFree
Linear mark detecting method and device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Linear mark detecting method and device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Linear mark detecting method and device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3308566