Image analysis – Applications – Vehicle or traffic control
Reexamination Certificate
1999-01-27
2002-09-10
Boudreau, Leo (Department: 2621)
Image analysis
Applications
Vehicle or traffic control
C340S907000, C340S933000
Reexamination Certificate
active
06449383
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field of the Invention
The present invention relates generally to a lane mark recognition apparatus for vehicles which is designed to recognize a lane mark printed on a road surface accurately using an image of a frontal view captured by a camera.
2. Background Art
In recent years, automotive lane mark recognition techniques are being developed which recognize white or yellow lane marks printed on a traffic lane using a captured image of a view in front of a vehicle for realizing automatic cruise control of the vehicle. For instance, Japanese Patent First Publication No. 5-289743 teaches binary-coding of a captured image to recognize a lane mark. Japanese Patent First Publication No. 7-239996 teaches processing edges of a captured image before binary-coding of the image. Japanese Patent First Publication No. 6-215292 proposes the use of Hough transform in image processing.
The above prior art systems, however, encounter drawbacks in respect to two points below.
The first relates to a problem in binary coding. A system, as taught in the above publication No. 5-289743 or No. 7-239996, which binary codes an original image or an edge-processed image may erase useful lane mark data or edge data of a lane mark through the binary coding, especially when the image has a low contrast and is unstable, sensitive to a change in environmental condition. Once erased, the data is difficult to recover. A system, as taught in the publication No. 6-215292, using Hough transform requires binary coding of an image and is also sensitive to a change in environmental condition. Usually, Hough transform requires a large number of calculations, which give rise to problems of a processing time and size of hardware. A practical Hough transform technique can only detect a straight line. There is another Hough transform technique capable of detecting curves, but it is unpractical in respect to the quantity of computation. The use of Hough transform in recognizing a lane mark, therefore, requires any additional technique for detecting curved portions of an image of the lane mark.
The second point relates to a problem in dot-to-line image processing. It is necessary for identifying a lane mark in a captured image to trace candidate edges for the lane mark extracted by binary coding the image to define a line. If, however, another edge exists near the lane mark image, it will result in tracing candidate edges. For example, when it is required to identify a white line printed on a road, and when a white car is traveling on the road, the car may form an edge near the white line image, thereby resulting in failure in tracing candidate edges for the line.
Specifically, the prior art systems have the disadvantage that they are influenced adversely by variation in environmental conditions. Lane mark recognition systems that do not require the binary coding and the dot-to-line image processing are sought.
As one of such systems, a contour recognition system called “Snakes” (see “Snakes: Active Contour Models,” International Journal of Computer Vision, pp. 321-331 (1988)) is known. The Snake system works to move contour points of the snake, mathematically represented as v(s)=(x(s), y(s)), to positions of minimum energy within an image. The energy function of a snake system may be represented as:
Esnakes (
v
)=∫[Eint(
v
(
s
))+Eimage(
v
(
s
))+Econ(
v
(
s
))]
ds
Eint(v(s)) is an internal spline energy of the Snakes which is defined by Eint(v(s))=(&agr;(s)|Vs(s)|
2
+&bgr;(s)|Vss(s)|
2
) where Vs=dv/ds, Vss=dv
2
/ds
2
, &agr;(s) is a parameter representing expansion/contraction such as a property of a rubber film, and &bgr;(s) is a parameter representing the degree of smoothness of a closed curve.
Specifically, minimizing the internal spline energy Eint is equivalent to expanding or contracting the closed curve to change it into a smooth curve.
Eimage is an image energy derived by image features such as lines or edges present within the target area. A gray level or an edge strength of the features is used.
Econ is an external constraint force applied to the closed curve and is defined by a force oriented inside a normal of the contour line. It simulates, for example, hydraulic pressure acting on an object submerged in water and works to precisely fit the closed curve to the contour. As described above, the snake is a deformable spline curve whose shape is controlled by the energy derived from an image.
The use of the snakes in lane mark recognition enables a contour to be found autonomously and eliminates the need for binary coding and dot-to-line image processing. This operation, however, is relatively time intensive for a computer. Specifically, each time each contour point of a contour model is moved slightly, the energy of that point must be calculated repeatedly until the energy thereof is minimized, which results in an increase in operation load on the computer. Particularly, the realization of automatic driving of the vehicle requires recognition of rapidly changing lane marks in real-time, thus causing a computer operation load to be increased.
It is, therefore, an object of the invention to provide a lane mark recognition system for vehicles which are capable of recognizing a lane mark at a speed higher than that required for the energy minimizing operation without need for binary coding and dot-to-line image processing.
SUMMARY OF THE INVENTION
It is therefore a principal object of the present invention to avoid the disadvantages of the prior art.
It is another object of the present invention to provide a lane mark recognition apparatus for vehicles which is designed to recognize a lane mark printed on a road accurately using image processing.
According to one aspect of the invention, there is provided a lane mark recognition apparatus for a vehicle which comprises: (a) an image capturing mechanism capturing an image of a road ahead of a controlled vehicle equipped with this apparatus: and (b) a lane mark recognition circuit recognizing a lane mark printed on the road in the image captured by said image capturing mechanism. The lane mark recognition circuit identifies a location in the image at which a string model is converged as a location of the lane mark. The string mode is made up of a plurality of string components and lines connecting adjacent two of the string components and defined to be moved in the image by an attractive force produced based on a measure of a picture property of the lane mark. Each of the connecting lines has a physical property of an elastic member which produces a reactive force against deformation of the string model so as to keep geometry of the whole of the string model.
In the preferred mode of the invention, the measure of the picture property is a brightness per pixel in the image.
The measure of the picture property may alternatively be a chromaticity per pixel in the image.
Each of the string components is allowed to move only in a horizontal direction of the image. The reactive forces produced by the connecting lines are strengthened as approaching a lower end of the image.
The connecting lines have the same physical property. The lengths of the connecting lines are prolonged as approaching the lower end of the image so as to increase the reactive forces for keeping the geometry of the whole of the string model.
The lengths of the connecting lines may be identical with each other. In this case, the connecting lines have elastic moduli which increase as approaching the lower end of the image so as to increase the reactive forces for keeping the geometry of the whole of the string model.
Ranges within which the string components are sensitive to the attractive forces are broadened as approaching a lower end of the image.
The lane mark recognition circuit determines that the string model has been converged at the lane mark in the image if a given condition is met..
The given condition is represented by an evaluation equation in terms of the m
Hibino Kutsuhiko
Imanishi Masayuki
Oike Tatsuya
Boudreau Leo
Choobin M. Barry
Denso Corporation
Harness Dickey & Pierce PLC
LandOfFree
Lane mark recognition system and vehicle traveling control... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Lane mark recognition system and vehicle traveling control..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Lane mark recognition system and vehicle traveling control... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2904190