Eye position detecting system and method therefor

• US 5,293,427 A
• Filed: 12/11/1991
• Issued: 03/08/1994
• Est. Priority Date: 12/14/1990
• Status: Expired due to Fees

• Current Assignee: Nissan Motor Co., Ltd.
• Original Assignee: Nissan Motor Co., Ltd.
• Inventors: Yoshida, Kazuhiko, Ueno, Hiroshi
• Primary Examiner(s): Mancuso, Joseph
• Assistant Examiner(s): CHANG, JON CARLTON
• Application Number:
  US07/805,143
• Time in Patent Office: 818 Days
• Field of Search: 351/208, 351/209, 351/210, 382/1, 382/2, 382/48, 382/28, 382/6, 340/575, 340/576
• US Class Current: 382/103
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G06V 40/19 Sensors therefor
• CPC Class Codes: G08B 13/19602 Image analysis to detect motion of the intruder, e.g. by frame subtraction
• CPC Class Codes: G08B 21/06 indicating a condition of sleep, e.g. anti-dozing alarms

Abstract

The invention detects the position of an eye in image of a driver'"'"'s face quickly and with high reliability, even if the density of the driver'"'"'s face image is not uniform on the right and left sides due to direct sunlight. Only one eye detection area is obtained on the dark side of the face image when the density of the face image is not uniform on the right and left sides. One eye window is detected in the one eye detection area, and an iris center is detected within the eye window to discriminate an inattentive driver who is dozing.

Claims

1. An eye position detecting system for detecting an eye position in a face image taken by a camera, comprising:

(a) inputting means for inputting face image signals as light density gradation data;

(b) binarizing means for binarizing the inputted face image signals;

(c) deciding means for obtaining a one eye detection area on a dark side in the face image when the inputted light density gradation data are not uniform on the right and left sides of the face image behond a predetermined density level, said deciding means having;

(1) discriminating means for discriminating density change points in a vertical direction on the basis of the binarized image signals to check whether the light density is not uniform on the right and left sides in the face image;

(2) determining means for determining a threshold value only on the dark side of the face image; and

(3) setting means for setting only one eye detection area on the dark side in the face image when the number of the discriminated density change points exceeds the determined threshold value.

2. The eye position detecting system of claim 1, wherein said deciding means comprises:

(a) density detecting means for detecting light density on the basis of the image signals on both sides in the face image, separately;

(b) comparing means for comparing the detected densities on both sides of the face; and

(c) deciding means for obtaining only one eye detection area on the dark side of the face image when a difference in light density between both sides of the face image exceeds a predetermined level.

3. A method of detecting an eye position in a face image, comprising the steps of:

(a) irradiating the face with infrared rays;

(b) taking face images by a camera in synchronism with the infrared rays;

(c) obtaining a one eye detection area on a dark side in the face image when the face image is not uniform in light density between right and left sides beyond a predetermined density level, the step including the steps of

(1) determining a threshold level;

(2) binarizing the face image signal on the basis of the determined threshold level;

(3) detecting density change points in a vertical direction on the basis of the face image signals to check whether face light density is not uniform on the right and left sides in the face image; and

(4) obtaining only one eye detection area on one dark side of the face image when the number of the density change points exceeds a predetermined value;

(d) obtaining one eye window in the one eye detection area in the face image; and

(e) detecting an iris center within the detected eye window to determine an inattentive driver on the basis of change in lightness at the detected iris.

4. The method of detecting an eye position of claim 3, wherein the step of obtaining one eye detection area comprises the steps of:

(a) detecting light density on the basis of the face image signals on both sides in the face image, separately;

(b) comparing the detected both-side light densities; and

(c) obtaining only a one eye detection area on one dark side in the face image when a difference in light density between both sides of the face image exceeds a predetermined level.

5. The method of detecting eye position of claim 4, wherein the steps of detecting light density and obtaining only a one eye detection area comprises the steps of:

(a) retrieving a predetermined number of image signals at pixels in a horizontal direction within a lower half area of the face image beginning from the center to the right and left, respectively;

(b) calculating an average density on the basis of the retrieved image signals on the right and left sides of the face image, respectively;

(c) repeating the above steps (a) and (b) in a predetermined vertical range;

(d) checking whether a difference in average density between the right and left sides of the face image is beyond a predetermined value;

(e) if the checked difference is less than a predetermined value, determining the light density is uniform on the right and left sides of the face image and obtaining two eye detection areas on both sides in the face image;

(f) if the checked difference is the predetermined value or more, determining that the light density is not uniform on the right and left sides of the face image and obtaining a one eye detection area;

(g) comparing the right and left average densities; and

(h) if the right average density is higher than the left average density, determining the eye detection area on the left side of the face image; and

if the left average density is higher than the right average density, determining the eye detection area on the right side of the face image.

6. The method of detecting an eye position of claim 3, wherein the steps of detecting density change points and obtaining the one eye detection area comprises the steps of:

(a) retrieving a predetermined number of binarized image signals at pixels in a horizontal direction within a lower half area of the face image;

(b) calculating an average density on the basis of the retrieved image signals;

(c) checking whether the calculated average density is beyond a predetermined value to determine a continuous bright portion along the horizontal direction;

(d) checking whether the calculated average density is below another predetermined value to detect a density change point in the horizontal direction;

(e) repeating the above steps (a) to (d) along a vertical direction;

(f) counting the number of density change points along the vertical direction;

(g) if the counted number of the density change points is less than a predetermined value, determining that the light density is uniform on the right and left sides of the face image to decide two eye detection areas on both sides in the face image; and

(h) if the counted number of the density change points is the predetermined value or more, determining that the light density is not uniform on the right and left sides of the face image to obtain a one eye detection area on a side of the face image where the continuous bright portion has not been checked.

7. The method of detecting an eye position of claim 3, wherein the step of obtaining an eye window comprises the steps of:

(a) determining a vertical central line of the face image;

(b) retrieving image densities at pixels in a horizontal direction beginning from the determined central line to the left side in the face image;

(c) determining a face left end point on the basis of a continuous bright portion along the horizontal direction;

(d) retrieving image densities at pixels in a horizontal direction beginning from the determined central line to the right side in the face image;

(e) determining a face right end point on the basis of a continuous bright portion along the horizontal direction;

(f) repeating the above steps (a) to (e) along a vertical direction to determine both the left and right face end points;

(g) determining left and right side positions (X₁, X₂) of the left eye window and left and right side positions (XX1, XX2) of the right eye window on the basis of both the determined face end points;

(h) retrieving image densities at pixels in a vertical direction at regular horizontal pixel intervals;

(i) determining an eyeball lowest end point (BY1) and an eyebrow lowest end point (BY2) on the basis of a continuous dark portion along the vertical direction, respectively;

(j) determining an eyeglass detecting vertical position value (BYH) between the two end points BY1 and BY2;

(k) detecting a dark bridge portion of an eyeglass frame along the horizontal direction at the vertical position value (BYH); and

,

(l) if no eyeglass is detected, determining upper and lower end points (YT, YB) on the basis of the eyeball lowest end point (BY1), and if eyeglasses are detected, determining the upper and lower end points (YT, YB) on the basis of the eyebrow lowest end point (BY2).

8. The method of detecting an eye position of claim 3, wherein the step of detecting an iris center comprises the steps of:

(a) setting a circle with a radius R within an eye window and four radial rectangles crossing the circle circumference by p pixels;

(b) calculating a difference in lightness between portions of rectangles outside the circle and portions of rectangles inside the circle for pixels within the circle; and

(c) determining an iris center at a point where a calculated difference is the maximum.

9. The method of detecting an eye position of claim 3, wherein the inattentive driver can be determined by the steps of:

(a) determining a threshold value within the window;

(b) comparing an iris lightness difference with the threshold at regular time intervals to determine a closed eye when the detected iris lightness difference is less than the threshold and an open eye when more than the threshold; and

(c) determining an inattentive driver when the determined closed eye condition continues beyond a predetermined time duration.