JP2932551B2 - Iris position detecting device and driver state detecting device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial application field) The present invention relates to an iris position detecting device for detecting an iris position of an eye of a driver or the like, and an iris state of a driver's eye by detecting the iris state. The present invention relates to a driver's state detection device that detects a drowsiness or the like.

(Prior Art) A conventional driver state detecting device, for example, a dozing detection device is disclosed in Japanese Patent Application Laid-Open No. 60-158303. This type of drowsiness detector irradiates the driver's face with infrared light using an infrared strobe, detects the reflected light as an image with a two-dimensional image sensor, binarizes the image, and uses this to make the image brighter or darker. The movement of the eye position is detected from the image pattern of the area to determine whether or not the driver is dozing.

(Problems to be Solved by the Invention) However, in such a dozing detection device, since the position of the eye is recognized from the shape of the light and dark area of the binarized face image, the approximate position of the eye can be recognized. It cannot be determined whether the eyes are open or closed. Therefore, there is a problem that if the driver falls asleep without disturbing the posture of the face, the driver cannot fall asleep.

In view of the above, the present invention provides a method for a driver who can process the area of a dark region portion of an area for detecting an iris portion by a statistical method, detect the center position of the iris of the driver, and make a more accurate determination. It is intended to provide a state detection device.

[Structure of the Invention] (Means for Solving the Problems) In order to solve the above problems, the present invention provides an image input means CL1 for inputting a face image including eyes, and an input image transmitted from the image input means. Binarizing means CL2 for binarizing
Area setting means CL3 for setting an area for detecting an iris portion in the binarized image by the binarization means; and detecting and detecting at least the area of a dark area in the area set by the area setting means CL3. An iris center detecting means CL4 for weighting the candidate points by statistical processing according to the size of the area and detecting the iris center position is provided.

Further, a determination means CL5 for determining the state of the driver from the detection result by the iris center detection means CL4 is provided.

(Operation) According to the above configuration, an area for detecting an iris portion is set from the binarized image, at least an area of a dark area is detected in the area, and statistical information is assigned to a candidate point according to the detected area. Iris center position can be detected by performing weighting based on objective processing. Therefore, when the state of the driver is determined by the determining means CL5, more accurate determination can be performed by specifying the iris center position.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

FIG. 2 is a block diagram of a vehicle driver state detecting apparatus according to one embodiment of the present invention, and FIG. 3 is a flowchart based on the block diagram of FIG.

As shown in FIG. 2, this state detection device includes an infrared strobe 1 for irradiating a driver's face portion in a direction in front of the driver in an instrument panel (not shown), and the infrared strobe 1 A TV camera 3 as image input means CL1 for photographing a face portion irradiated with infrared light of the above, and a timing command circuit 5 for synchronizing the emission of the infrared strobe 1 with the image input of the TV camera 3. ing. Then, when a strobe light emission command is output from the timing command circuit 5 to the infrared strobe 1, the infrared strobe 1 emits light to irradiate the driver's face, and at the same time, an image input command is output to the TV camera 3. Thus, a face portion irradiated with infrared light is imaged.

As shown in FIG. 6, the input image of the TV camera 3 is composed of 520 pixels in the horizontal direction and 500 pixels in the vertical direction, and the angle of view is adjusted so that the face portion becomes almost full in the vertical direction.

An image memory 9 is connected to the TV camera 3 via an A / D converter 7 for converting a photographed input image into a digital amount. The image memory 9 stores input image data of the TV camera 3.

The image memory 9 is connected to an eyeball position determining circuit 11 that defines the position of the eyeball based on the input image data stored in the image memory 9. An iris detection circuit 13 for processing image data of the image memory 9 in the area to detect an iris portion of the eye is connected.

In addition, the iris detection circuit 13 is connected to a drowsiness determination circuit 15 that determines the presence or absence of a driver's drowsiness or aside from the detection result of the iris in the iris detection circuit 13. An alarm output circuit 17 that issues an alarm according to a command when it is determined that the driver falls asleep is determined.

The eyeball presence position defining circuit 11 includes a binarizing unit CL2 for binarizing the input image with a certain threshold value, and an area setting unit CL3 for setting an area for detecting an iris portion in the binarized image.
And

Further, the iris detection circuit 13 constitutes the iris center detection means CL4, and the drowsiness determination circuit 15 constitutes the determination means CL5.

Next, the operation will be described with reference to the flowcharts of FIGS. Although the flowchart of FIG. 4 forms the subject of the embodiment of the present invention, the description will be made from the beginning of the flowchart of FIG. 3 as a premise.

First, in step S1, in synchronization with the emission of the infrared strobe light 1,
The image of the driver's face taken by the TV camera 3
A / D conversion circuit 7 converts the signal into a digital signal and converts it into an image memory 9
To be stored. Next, in step S2, the input image data stored in the image memory 9 is fetched into the eyeball position determining circuit 11, and is binarized by a certain threshold value. This is for clarifying the brightness of the face portion, and the binarization threshold is set to a level at which the eyeball portion can be extracted.

That is, the video signal is converted into digital data of 256 gradations (0 to 255), and the white part is “255” and the black part is “0”.
And binarized at a certain threshold level to obtain a binarized image J (x, y).

Next, in step S3, the left and right eyeball existence regions (windows, square portions shown by dotted lines in FIG. 7)
Is determined in the horizontal direction (X direction). When a window is detected, candidate points of the eye are detected within the width of the window in the X direction (step S4), and it is determined whether or not a forehead is detected (step S5). , (B) (steps S6 and S7). Step S
At 8, it is determined whether or not both eyes and a window have been set, and then iris center detection is performed (step S9).
When the center of the iris is detected, the occupant is determined to look aside from the center of the iris (step S10).

Next, FIG. 5 shows a detailed flowchart (SETWIN process) of the process of step S3.

In this process, the width of the face is recognized based on the left and right hair of the face, the contour of the face, and the like, and the width in the X direction of the region where the eyeball exists is determined from the width of the face.

First, a counter A for white dots, a counter B for black dots, a counter C for detecting eyebrows and eyes, which are black portions of the shading of the face, and a counter D used by the frame determination means CL7 for glasses are cleared.

Next, in step S301, the input image is raster-scanned in the X direction (scanning is performed from left to right in FIG. 10, but the actual face is scanned from right to left. However, description will be given). The left side in FIG. 10 is referred to as the left side of the face.), And it is determined whether or not it is a black dot. When the first black dot is detected, for example, when the scanning reaches the end of the head, the process proceeds to step S303, and it is determined whether the value of the counter A for white dots exceeds 0. Assuming that the occupant's back shadow is white by binarization, the white dots are counted before the first black dot is detected, so the black dot counter B is cleared in step S304.

In step S305, the number of black dots is counted, and it is determined whether black dots have been counted five times in a row (step S306). If the count of the black dot is the first one, the process proceeds to step S308. At this point, since the left X coordinate has not been detected, the process proceeds to step S309, and then the process proceeds to step S3.
The process proceeds to step 10, where it is determined whether or not all scanning has been performed in the X direction (X511). However, since all have not been scanned yet, the process proceeds to step S301, where the scanning value in the X direction is added.
If the black dot is continuously detected, steps S302, S
The process proceeds to 303 and S305, and the determination in step S306 is performed again. Here, when the number of black dots B is five times in a row, it is determined that the right side of the face is substantially at the center in the X direction of the hair, and the Y coordinate MEY at that time is stored (step S307). .

Next, in step S308, it is determined whether or not the left X coordinate of the eyeball existing area has been detected. At this time, since the left X coordinate has not been detected yet, the counter A of the white dot is cleared in step S309, and the process proceeds to step S310.

In step S310, it is determined whether or not scanning has been performed in the X direction again. At this time, since all of the scanning in the X direction has not been performed, the process returns to step S301. If the scanning has reached, for example, a white portion of the face, the process proceeds to step S311 via step S302, and the number of white dots is counted. Then, in step S312, it is determined whether or not the first white dot is after the black dot has been counted continuously 10 times or more.

This determination is for determining whether or not all the hair on the left side has been counted at the position of the forehead in the Y direction.

Here, after the black dot has been counted 10 times or more, if 190 or more white dots corresponding to the width of the forehead are detected, the process proceeds from step S313 to step S314, and it is determined whether or not it is the first white dot. You. Here, if it is the first white dot, X = X1 is stored in the memory as a candidate point X1 of the left X coordinate of the window (step S315) (FIG. 7).

At this time, in step S321, it is determined whether or not the counter C of the black dots of the eyebrows and eyes exceeds 0. here,
If the Y direction is the position of the forehead, the white dot should continue, so the black dot is not counted and the process proceeds to step S310.
It is determined whether or not scanning has been performed in all X directions. At this time, since all X directions have not been scanned yet,
Returning to S301, the same processing as described above is repeatedly executed.

When the black dot C is counted in step S321, the Y coordinate at which the candidate point X1 is detected is not a forehead, but, for example, a position at a lower eyebrow position or the like. Clear and cancel the candidate point X1 on the left side of the window in the X direction (step S3
twenty three).

If all the X directions are scanned at the Y coordinate at which the candidate point X1 could not be detected, the counters A, B, and C are cleared, and the Y coordinate is shifted downward by one (steps S310, S328, and S329). Then, the above search for X1 is performed in the same manner, and if all the Y directions are scanned (step S330), it is determined that X1 cannot be detected (step S336), and the process ends. If X1 cannot be detected, it means that the occupant is looking aside or facing down.

By the way, when the occupant wears spectacles, black dots may continue 10 times or more in the X direction even in the lower part of the spectacle frame. If black dots are continuous at the lower portion of the spectacle frame, white dots will not be continuous 190 or more after black dots have been continuous 10 times or more. Then, after detecting the right end (X2) on the left side of the spectacle frame as a candidate point of the window, if white dots do not continue 190 times or more, the process proceeds from step S313 to step S316 (X2-5, Y-axis MEY). ), The upper dot in the Y direction is searched from the point (approximately the center in the X direction of the lower frame on one side of the glasses) and the number of black dots is counted (step S317). Then, it is determined in step S318 whether or not the dot has been counted 20 times or more. If the dot has been counted 20 times or more, it is determined whether or not the black dot has been counted 10 times or more (step S31).
9). This has the eyes from P ₁ position of the frame if wearing glasses as FIG. 9 (b) at the 20 dots or more, where continued black dots than 10 times to detect the eye, without glasses In the case of (a), the eyes and eyebrows are viewed as frames, and FIG.
When counted 20 dots or more from the P ₂ position upwardly as is because black dots from which can without distinction be followed over 10 times. If the number of black dots is 10 or more, X1 is determined as the X coordinate on the left side of the window in consideration of the frame size based on the X coordinate candidate point X2 (step S320). If there are no more than 10 black dots,
The process returns from step S310 to step S301 without determining the X coordinate, and repeats the same processing as described above.

The X coordinate X1 on the left side of the window is shown in FIGS. 8 and 9 (b).
When a black dot on the right hair or a black dot on the right side of the eyeglass frame is detected again by scanning in the X direction, the number is counted in step S324,
It is determined whether or not black dots have been counted five times in a row (step S325). Here, when the black dot is counted five times in a row, X = XX2 is stored in the memory as the X coordinate XX2 on the right side of the window as shown in FIG. When X1 is stored in steps S316 to S320, the left end of the right frame of the glasses is stored as XX1, and this XX1
XX2 is stored based on.

If the black dots are not counted five times in a row in step S325, it means that the occupant is facing sideways or the face is oblique, and the white dots are cleared in step S327 and the process proceeds to step S310. Transition. At this time, an occupant can be warned to correct his / her posture by voice or the like.

In this way, when the left X coordinate X1 and the right X coordinate XX2 (or XX1) of the window are detected, the process proceeds to step S331, where the left and right X coordinates X1, XX2 (or
XX1) is stored in the vertical direction as HABAY,
In step S332, an X coordinate for dividing the window into left and right eyes is obtained (see FIG. 8). This division is performed by the following equation, for example, in a case where no glasses are worn. That is, X-axis center = XC = X1 + {(XX2-X1) / 2} left X coordinate of left eye window = X1 right X coordinate of left eye window = X2 = XC-25 left X coordinate of right eye window = XX1 = XC + 25 Right X coordinate of right eye window = XX2 Next, in steps S333 and S335,
The process proceeds to WINSUB for obtaining the Y coordinate of each of the left and right windows (step S334).

As described above, the X coordinates X1, X2, XX1, XX of the left and right windows
2 is detected, step S in the flowchart of FIG.
Then, the process proceeds to step S4, where the eye candidate point is detected within the width of the window in the X direction, and it is determined whether the forehead is detected (step S).
5) Determine one of the patterns 1 to 10 shown in FIGS. 10A and 10B (steps S6 and S7).

FIG. 6 shows a detailed flowchart (WINSUB processing) of steps S4 to S7.

This process is for determining the coordinates of the window in the Y direction, and is performed for each of the left and right eyes.

First, in steps S401 to S403, for example, one row of the left window, that is, from the left X coordinate X1 to the right X coordinate X2 is scanned, and the number E of black dots is counted. Next, it is determined whether or not the counted number E of black dots is equal to or larger than the maximum number Emax (the maximum number in the previous processing) (step S404), and the count value of the current processing is equal to or larger than the maximum number Emax. In the case of, the number E is the maximum number
It is stored as Emax (step S405). Thereafter, the counter E is cleared (step S421), and it is determined whether the scanning has been performed up to 300 dots in the Y direction, that is, whether the scanning has been performed up to the cheek (step S422). For example, the above steps of Emax storage are repeated, and in the case of the pattern 7 in FIG. 10 (b), Emax of both eyebrows and eyes is stored.

Although 300 dots is a fixed value in step S422, the number of dots may be smaller than this, or an appropriate number may be automatically determined by detecting the position of the head end.

If the number E of black dots is less than the maximum number Emax in step S404, the process proceeds to step S406, and the number E of black dots is reduced by 80% or more with respect to the maximum number Emax, that is, 80% or more is white. It is determined whether it is a dot, and if 80% or more is a white dot, it is determined that any of the forehead, the space between the eyebrows and the eyes, or the cheek has been detected, and the process proceeds from step S406 to S407, where the Y coordinate ZR at that time is stored. I do.

Then, in step S408, it is determined whether or not there is one candidate point. If there is one candidate point, it is determined whether or not the amount has been detected five times (step S409). If this determination is YES, it indicates that scanning in the Y direction has reached from the head to the forehead, so the candidate point flag is turned off (step S41).
0), the process proceeds to step S421 to count the number of black dots E
Clear Therefore, the head is not detected as a candidate point.

When the scanning in the Y direction is between the eyebrows and the eye and the candidate point is one of the eyebrows, the process proceeds from step S408 to S409 and proceeds to step S409.
Is NO, and the flow shifts to step S421. If the scan in the Y direction is cheek and there are already a plurality of candidate points, the candidate point flag remains ON (step S408). As described above, it is possible to distinguish whether or not the candidate point has been detected in steps S406 to S410, and whether the detected point is in the scanning of the forehead, between the eyebrows and the eyes, or in the cheek.

In step S406, when the white dot is 80% or less,
It is determined whether the Y coordinate ZR has been stored (step S41).
1) If the Y coordinate ZR is stored, the number E of black dots is 1
It is determined whether the number has increased by 0 or more (step S412). If the ZR is not stored in step S411, the Y coordinate is, for example, at the head, and the process returns from steps S421 and S422 to S401.

If ZR is stored and the number E of black dots has increased by 10 or more in step S412, it is determined that the portion is a portion with many black dots, such as eyebrows, eyes, and eyeglass frames, which may have an eyeball area. Then, in step S413, it is determined whether or not the candidate point flag is ON. If it is ON, it is determined whether or not the number of candidate points is one (step S414). Here, if there is one candidate point, the parameter F of the size of the black area represented by the candidate point is increased in step S416. If this parameter F is very large, the eyes and eyebrows are recognized as one part due to a shadow or the like, and it is determined that there are substantially two candidate points (see the pattern in FIG. 10 (a)). 4).

If the candidate point flag is OFF in step S413, the candidate point flag is turned ON (step S417).
In S418, the number of times of candidate point detection is counted up, and it is determined whether or not the number of times of candidate point detection count is two or more (step S419). If the number of candidate point detection counts is not two or more, the coordinates of the Y axis at that time are stored as the reference coordinates YYS for setting the window in step S420.

After the processes of steps S416 and S420 are performed in this manner, the process proceeds to step S421 to clear the black dot counter E, and then proceeds to step S422.

In step S422, it is determined whether or not the Y axis has been scanned from 0 to 300 dots. When scanning of the Y coordinate has been completed to 300 dots, the process proceeds to step S423, where the position of the HABAY previously stored in the SETWIN process is set to 220. It is determined whether or not: If the position of the HABAY is 220 or less, it is determined that the forehead has been detected (step S424), and it is determined in step S425 whether the number of candidate point detection counts is 2 or more or the parameter F is 70 or more. Is done.

If the number of candidate point detection counts is two or more, or if the parameter F is 70 or more, the pattern is one of the patterns 4, 7, 8, and 10 shown in FIG. 10, and the setting means A shown in the table of FIG. The Y coordinate on the window is YYS + 30, the Y on the window is
The coordinates are set to YYS + 80 (step S426).

In step S425, if the number of candidate point detection counts is 2 or less or the parameter F is 70 or less, it is determined in step S427 whether or not the number of candidate point detection counts is 1 and the number of candidate point detection counts is 1 Is pattern 3 and
The setting means B sets the Y coordinate on the window to YYS-20,
Set the Y coordinate at the bottom of the window to YYS + 30 (Step S
428).

If there is no candidate point detection count in step S427, detection is impossible (step S429), and EXIT is performed (step S430). In this case, it is determined that the driver is facing down. Therefore, a predetermined warning or the like may be given.

Next, if the position of the HABAY is not 220 or less in step S423, it is determined that the forehead has not been detected (step S423), and it is determined in step S432 whether or not the number of candidate point detection counts is two or more. Is done.

If the number of candidate point detection counts is two or more, the pattern 9 is determined.
YYS-20 and the Y coordinate under the window are set to YYS + 30 (step S428).

If the number of candidate point detection counts is two or less in step S432, it is determined in step S433 whether the number of candidate point detection counts is one. If the number of candidate point detection counts is one, the candidate is detected. It is determined whether the point is a frame of glasses (step S434). Here, if the candidate point is a spectacle frame as in pattern 6, the process proceeds to step S435. In step S435, a dot is searched upward from the standard coordinates YYS, and the position YYS2 at which the black dot is detected is stored. The setting means D sets the Y coordinate on the window to YYS2.
−35, the Y coordinate under the window is set to YYS2 + 15 (step S436).

If it is determined in step S434 that the frame is not an eyeglass frame, the pattern is pattern 5, and the setting means E sets the Y coordinate on the window to YYS-20 and sets the Y coordinate under the window to YYS-20.
It is set to +30 (step S428).

In step S433, if there is no candidate point detection count, it is pattern 1 and the process proceeds to step S437. In step S437, it is determined whether the candidate point is an eye or an eyeglass frame. Using the ZR coordinates stored in step S407, the setting means F sets the Y coordinate on the window to YYS-20 and the Y coordinate under the window to YYS + 30 (step S428).

If the candidate point is a spectacle frame in step S437, it is pattern 2 and the process moves to step S439. Step S4
At 39, a dot is searched upward from the ZR coordinates, and the position YYS2 at which the black dot is detected is stored. Then, the setting means G sets the Y coordinate on the window to YYS2-35 and the Y coordinate under the window to YYS + 15 (step S436).

Next, FIG. 12 shows a processing flowchart for determining whether the candidate point is an eye or an eye gun frame in steps S434 and S437, which is performed to determine the setting means D, E, F, and G.

This processing is executed when the amount cannot be detected and the number of candidate points is one or less.

First, in steps S100 to S300, a point that becomes a black dot upward from a point 15 dots above the point a where the candidate point changed from a black dot to a white dot is searched for, and a step S4
At 00, it is determined whether or not the count b until the dot becomes a black dot is 20 times or more. Here, when the count b is 20 or less, as shown in FIG. 13, the candidate point is determined to be an eye (step S500), and when the count b is 20 or more, , The candidate point is determined to be a spectacle frame (step S600).

Then, the reference coordinates YYS2 when the candidate point is an eyeglass frame is calculated by the following equation depending on whether there is one candidate point or not. That is, when there is one candidate point, YYS2 = YYS-15-b, and when there is no candidate point, YYS2 = ZR-15-b.

The window for each eye is set by performing the above processing for both the left and right eyes.

Returning to FIG. 10 and FIG. 11, when the pattern 1 is used, since there is no space between the forehead and the eyebrow and the eye, a window is set by the setting means F from the point a (= ZR) which is a black dot break.

The pattern 2 is the same as the pattern 1, except that the spectacles are worn, so that 15 + b dots above the point a (= ZR) are set as the reference coordinates YYS2, and the setting means G sets the window.

In pattern 3, since the eyebrows are thin, the candidate point is point a (= YYS)
Only, and the window is set by the setting means B.

Pattern 4 is a case in which a black portion of the eye, eyebrow, and eyeglass frame is detected as one shadow. Since the inside of the eyeglasses is black, the candidate point is only point a (= YYS).
The window is set by the setting means A from the value b.

Since the pattern 5 has no forehead exposure and is detected only between the eyebrows and the eyes, the candidate point is one point a (= YYS), and the setting means E sets a window.

Pattern 6 is a case where the forehead is hidden and the black part of the eyes and eyebrows is detected as one in the shade, and it is not possible to detect between the eyebrows and the eyes, and to detect the white part under the eyes, The candidate point is point a, but since the user wears glasses, a window is set by the setting means D using YYS2 as reference coordinates.

In the pattern 7, the candidate points are two points of the point a (= YYS) and the point b in an ordinary form, and the setting means A sets the window based on the point a.

Pattern 8 is a case in which the black part of the eye and eyebrow is detected as one shadow, and the forehead is detected but the space between the eyebrow and eye cannot be detected by the spectacle frame. There are two points, and a window is set by the setting means A.

In the pattern 9, the forehead is hidden and the space between the eyebrow and the eye and the space between the eye and the lower part of the spectacle frame are detected, and the candidate points become the points a and b, and the setting unit C sets a window.

The pattern 10 detects the forehead, the space between the eyebrows and the eyes, and the space between the eyes and the lower part of the spectacle frame. The candidate points are three points a, b and c, and the setting means A sets a window.

Then, in step S8 of FIG. 3, it is determined whether a window has been set for both eyes. When a window has been set for both eyes, the process proceeds to step S9. In addition, it is also possible to perform processing for only one eye.

In step S9, a process of detecting an iris center in the window set above from the binarized image J (x, y) is performed. Since the iris of the eye at this time is generally observed as a dark circular region, the presence or absence of the driver can be determined by detecting this circular region and recognizing the region area.

In this embodiment, since the area is set by detecting the face width, even if the back shadow in the face image is black (nighttime or the color of the headrest), it can be reliably detected.

The details of the processing in step S9 will be described with reference to the flowchart of FIG. 4 which is the main part of the embodiment of the present invention.

First, in step S800, iris candidate points are detected in the window.

 This detection will be described first.

FIG. 15 is an explanatory diagram showing the principle of detecting iris candidate points.

Now, a circle having a radius R centered on an arbitrary point (x, y) in the window is set, and four rectangles are set radially around the point (x, y). The rectangle is set so as to extend by P pixels inside and outside the circle. Then, a difference δ between the sum of the brightness values of the rectangular white portions outside the circle and the sum of the brightness values of the rectangular hatching portions inside the circle is determined.

This is performed from Rmin to Rmax at the above-mentioned arbitrary point (x, y), and a predetermined value of the difference δ is obtained as Δ. Next, the point (x
₊₁ , y) is performed and a predetermined value Δ is stored. Such a calculation is performed centering on all pixel points in the window, and a predetermined value Δ as an iris candidate point is output.

This means that when an iris is present in the window, the iris is detected as a circular figure with lower brightness than other areas,
It is based on the principle that the difference δ found at the center of the iris takes the maximum value.

FIG. 16 shows a flowchart of the above processing.

First, in step S901, counters x and y for scanning the window are reset to 1. The size of the window set here is M dots in the x direction and N dots in the y direction. Next, in step S902, it is determined whether the point J (x, y) of the center coordinates of iris detection is black. If black, the process proceeds to step S903, where the detection radius R is set to Rmin. Subsequently, in steps S904 and S905, Δ and p
Reset.

Next, in steps S906 to S908, it is specifically detected as a black circular area surrounded by white parts. That is, in step S906, the first four terms J (x + R + p,
y), J (x, y−R−p), J (x−R−p, y), J (x, y + R +)
p) represents the brightness at the right, bottom, left, and top positions, respectively, at a radius R + p away from the center coordinates (x, y), and the last four terms J (x + R−p−1, y), J (x , y−R + p + 1), J (x−
R + p + 1, y) and J (x, y + R-p-1) are the right, bottom, left, and left R- (p + 1) away from the center coordinates (x, y), respectively.
The brightness of the upper position is shown.

Then, in step S907, p is incremented by 1 and changed to p−1, and steps S906 to S908 are repeatedly executed, and the brightness value sum of the rectangular hollow portion in FIG. 15 at the radius Rmin (the first half of the equation of step S906: 4) Term sum) and the sum of the brightness values of the rectangular hatched portion (sum of the last four terms of the expression in step S906)
Is obtained as a predetermined value Δ (step S910).
Next, in step S911, the process proceeds to step S904 again with the radius Rmin + 1, and by repeatedly executing steps S906 to S908, the lightness difference δ of the rectangular region when the radius is Rmin + 1 is obtained.

If the brightness difference δ is larger than Δ calculated for the first radius Rmin, the brightness difference δ is set to a predetermined value Δ. Such an operation is repeatedly performed up to the radius Rmax, and a predetermined brightness difference Δ for an arbitrary point (x, y) is obtained (step S912).

This is because the radius of the iris to be detected varies depending on the distance between the individual or the camera and the occupant, and therefore, a zone (Rmin to Rmax) at the detection radius is provided.

Hereinafter, this processing is performed over the entire window from x to 1 to M and y to 1 to N. Δ obtained by such processing is the brightness difference calculated for the iris candidate points.

With the above processing, the processing of step S800 in FIG. 4 is completed, and several regions of the candidate point group are obtained. Next, labeling processing is performed (step S801). As a result, for example, as shown in FIG. 18, the iris candidate points which are dark areas are numbered from to, and the respective center points (Xi, Yi) are stored as representative coordinate points of each area. Next, in step S802, the upper, lower, left and right 5% zones of the detection area (0.05 × lw and 0.05 in FIG. 18)
Xlh) excludes candidate points having representative point coordinates. Then, only the candidate points whose area Si is larger than the predetermined value SS are left as candidate points. This is because it can be clearly determined that the candidate point is not an iris if the candidate point is smaller than the predetermined area value SS (steps S803 to S807). Finally, in step S808, the iris center position is determined by the weighted average of the candidate points by the following equation.

In step S808, the iris center position is determined using the weighted average. Alternatively, an arithmetic average or the like may be used. In general, the iris is often a dark area having the largest area among the candidate points. Therefore, instead of steps S803 to S808, the representative point of the candidate point having the largest area among the candidate points may be set as the iris center position. This is possible, and is effective when the processing speed needs to be increased. The point is that at least the area of the dark area is detected in the set area, the candidate points are weighted by statistical processing according to the size of the detected area, and the iris center position is detected.

Finally, after the iris center leaving is obtained in this way, the state of the occupant, such as falling asleep or looking aside, is determined by the drowsiness aside determination circuit 15 in FIG. That is, the steps in FIG.
The process proceeds to S10 and a determination is made. That is, since the predetermined value Δ of the iris candidate point as the final output is greatly different between when the eye is opened and when the eye is closed, it is possible to determine whether the eye is open or closed by subjecting this to a predetermined threshold value processing. That is, the brightness difference Δ in the window calculated in step S9
Is thresholded, and it is determined that the eyes are open when the brightness difference Δ ≧ Th (threshold), and the eyes are closed when the brightness difference Δ ≦ Th.

Even when blinking, the eye may be determined to be open in the above iris detection processing. Therefore, if it is determined that the driver is dozing in one iris detection processing, there is a high possibility of a false alarm. The iris detection process is repeatedly performed a plurality of times, and it is determined that the user is dozing when the eyes are continuously recognized for a predetermined number of times or more.

For example, as shown in FIG. 18, when an image is input at each time interval indicated by a black dot, and as a result of the iris detection processing, the number of times that the eye is determined to be closed for three consecutive times, it is determined that the driver is dozing off It makes a determination and outputs a dozing determination signal as shown in FIG. When the alarm output circuit 17 receives the dozing determination signal, it issues an alarm to alert the driver.

When it is determined that only one eye is closed, it is considered that one eye is not actually closed and one eye is off the original screen due to looking aside. Therefore, when it is determined that one eye is closed three consecutive times, as in the case of the drowsiness determination, it is determined that the eyes are aside.

In the WINSUB subroutine shown in FIG. 5, EXIT may occur as an undetectable error. In this case, for example, when EXIT is performed three times in a row, only the head is imaged because the driver has turned down. Is considered to be the result of
In such a case, it is possible to determine that the user is dozing and issue an alarm.

(Effects of the Invention) As is clear from the above, according to the configuration of the first aspect of the present invention, at least the area of the dark region can be detected in the region where the iris portion is detected, and the area of the detected dark region is reduced. Since the candidate points are weighted by statistical processing according to the size and the iris center position is specified, the iris center position can be accurately detected.

According to the invention of claim 2, it is possible to accurately determine whether the driver's eyes are open or closed, whether the driver is facing the front, and the like.

[Brief description of the drawings]

FIG. 1 is a block diagram of the present invention, FIG. 2 is a block diagram according to an embodiment of the present invention, FIGS. 3 and 4 are flowcharts based on the block diagram of FIG. 2, and FIG. Flowchart, FIG. 6 is a flowchart of the WINSUB process, FIGS. 7 to 11 are explanatory diagrams relating to the process of this embodiment, FIG. 12 is a flowchart of the eyeglass frame determination process, FIG.
FIG. 14 is an explanatory diagram relating to the processing, FIG. 15 is an explanatory diagram of the principle of the iris detection process, FIG. 16 is a flowchart of the iris detection process, FIG. 17 is an explanatory diagram for determining dozing, and FIG. It is explanatory drawing of the characteristic process of an example. CL1 image input means CL2 binarization means CL3 area setting means CL4 iris center detection means CL5 determination means

Claims (2)

(57) [Claims] 1. An image input means for inputting a face image including both eyes of a driver, a binarization means for binarizing an input image sent from the image input means, and a binarization by the binarization means. Area setting means for setting an area for detecting an iris part in a coded image, and detecting at least the area of a dark area in the area set by the area setting means and statistically determining candidate points in accordance with the size of the detected area. An iris position detecting device, comprising: iris center detecting means for weighting by processing to detect an iris center position. 2. An image input means for inputting a face image including both eyes of a driver, a binarization means for binarizing an input image transmitted from the image input means, and a binarization by the binarization means. Area setting means for setting an area for detecting an iris part in a coded image, and detecting at least the area of a dark area in the area set by the area setting means and statistically determining candidate points in accordance with the size of the detected area. Iris center detecting means for performing weighting by processing to detect an iris center position, and determining means for judging a driver state based on a detection result by the iris center detecting means; apparatus.