(19)
(11)EP 1 862 111 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
04.03.2015 Bulletin 2015/10

(21)Application number: 07075405.6

(22)Date of filing:  29.05.2007
(51)Int. Cl.: 
A61B 3/113  (2006.01)
A61B 5/18  (2006.01)

(54)

Eye monitoring method with glare spot shifting

Augenüberwachungsverfahren mit Blendspot-Wechsel

Procédé de surveillance de l'oeil de commutation du point de reflet


(84)Designated Contracting States:
AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LI LT LU LV MC MT NL PL PT RO SE SI SK TR

(30)Priority: 01.06.2006 US 444841

(43)Date of publication of application:
05.12.2007 Bulletin 2007/49

(73)Proprietor: Delphi Technologies, Inc.
Troy, MI 48007 (US)

(72)Inventors:
  • Hammoud, Riad I.
    Kokomo, IN 46902 (US)
  • Harbach, Andrew P.
    Kokomo, IN 46902 (US)

(74)Representative: Robert, Vincent 
Delphi France SAS Bât. le Raspail - ZAC Paris Nord 2 22, avenue des Nations CS 65059 Villepinte
95972 Roissy CDG Cedex
95972 Roissy CDG Cedex (FR)


(56)References cited: : 
WO-A1-99/27844
US-A1- 2002 181 774
US-A- 5 598 145
US-A1- 2005 270 486
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description

    TECHNICAL FIELD



    [0001] The present invention relates to monitoring a human's eyes in a video image, and more particularly to a method for producing images of the eye that are not occluded by eyeglass glare.

    BACKGROUND OF THE INVENTION



    [0002] Vision systems frequently entail detecting and tracking a subject's eyes in an image generated by a video camera. In the motor vehicle environment, for example, a camera can be used to generate an image of the driver's face, and portions of the image corresponding to the driver's eyes can be analyzed to assess drive gaze or drowsiness. See, for example, the U.S. Patent Nos. 5,795,306; 5,878,156; 5,926,251; 6,097,295; 6,130,617; 6,243,015; 6,304,187; and 6,571,002.

    [0003] Due to variations in ambient lighting, the vision system typically includes a bank of infrared lamps that are lit during the image capture interval of the camera to actively illuminate the driver's face. While such active lighting ensures that the driver's face will be sufficiently illuminated to enable the camera to produce a high quality image, it can also introduce glare that occludes the eye when the driver is wearing eyeglasses. Such eyeglass glare is troublesome because it can interfere with the operation of the vision system's eye detection and tracking algorithms. It may be possible to remove eyeglass glare from an image, but this typically adds a significant amount of image processing, which may be impractical in a system that already is burdened with complex image processing routines. Accordingly, what is needed is a way of producing high quality eye images that are not occluded by eyeglass glare.

    [0004] Due to variations in ambient lighting, a vision system will typically include a bank of infrared lamps that arc lit during the image capture interval of a video camera to actively illuminate the driver's face. US-A-5598145, for example discloses a way of using different illumination sources, one of which is offset from the optical axis of the video camera, to detect the subject's eye; the two light sources are alternately activated to form two images that are then differenced to remove the influence of the illumination sources and to exaggerate the brightness difference between the eye and the other parts of the subject's face. US 2002/0181774 A1 addresses the problem of eyeglass glare by illuminating the subject with plural illuminators arranged in a predetermined shape so that if the subject is wearing glasses, the reflections off the surface of the glasses will be arranged in the same predetermined shape; as a result, the peculiarly arranged reflections can be identified in the image and ignored to effectively remove eyeglass glare from an image. US 2005/0270486 A1 describes a method of aligning and fusing a succession of images to create an enhanced image for processing. WO99/27844 discloses a method in accordance with the preamble of claim 1.

    SUMMARY OF THE INVENTION



    [0005] The present invention is directed to a novel method for producing a stream of video images of an actively illuminated human eye, where glare due to eyeglass reflection is shifted in a way that allows accurate and efficient eye detection and tracking. First and second sources of active illumination are physically staggered, and preferably disposed above and below a video imaging device. The first and second sources alternately illuminate the subject in successive image capture intervals of the imaging device to produce a stream of video images in which eyeglass glare, if present, shifts from one image to the next. The eye detection and tracking routines are designed to ignore images in which the eye is occluded by eyeglass glare so that the glare does not interfere with the performance of such routines.

    BRIEF DESCRIPTION OF THE DRAWINGS



    [0006] 

    FIG. 1 depicts a diagram of a vehicle equipped with an eye monitoring apparatus according to the present invention;

    FIG. 2 is a block diagram of the eye monitoring apparatus of FIG. 1, including upper and lower illumination sources, a video imaging device and a microprocessor-based digital signal processor (DSP) for carrying out eye detection and tracking routines;

    FIG. 3 is a flow diagram representative of an executive routine carried out by the DSP of FIG. 2 for controlling the upper and lower illumination sources and the imaging device;

    FIG. 4 is a flow diagram representative of an executive routine carried out by the DSP of FIG. 2 for processing the acquired images according to first embodiment of the present invention;

    FIG. 5 is a flow diagram representative of an executive routine carried out by the DSP of FIG. 2 for processing the acquired images according to second embodiment of the present invention; and

    FIG. 6 is a flow diagram representative of an executive routine carried out by the DSP of FIG. 2 for processing the acquired images according to third embodiment of the present invention.


    DESCRIPTION OF THE PREFERRED EMBODIMENT



    [0007] The method of the present invention is disclosed in the context of a system that monitors a driver of a motor vehicle. However, it will be recognized that the method of this invention is equally applicable to other vision systems that monitor a human eye, whether vehicular or non-vehicular.

    [0008] Referring to the drawings, and particularly to FIG. 1, the reference numeral 10 generally designates a motor vehicle equipped with an eye monitoring apparatus 12 according to the present invention. In the illustration of FIG. 1, the apparatus 12 is mounted in the passenger compartment 14 forward of the driver 16 in a location that affords an unobstructed view of the driver's face 18 when the driver 16 is reposed on the seat 20, taking into account differences in driver height and orientation. In general, the eye monitoring apparatus 12 actively illuminates the driver's face 18 and produces a stream of video images that include the driver's eyes 22. The images are processed to locate the driver's eyes 22 and to track the eye locations from one image to the next. The state of the eyes 22 can be characterized for various purposes such as detecting driver drowsiness and/or distraction, or even driver gaze.

    [0009] In the illustration of FIG. 1, the driver 16 is wearing eyeglasses 24, which in general may include sunglasses, goggles, or even a face shield. The eyeglasses 24 introduce the potential for glare in the images produced by eye monitoring apparatus 12 due to reflected active illumination that occludes one or both of the driver's eyes 22. While conventional eye monitoring systems are frustrated by eye-occluding glare, the eye monitoring apparatus 12 of the present invention is utilizes a glare shifting technique to enable effective eye detection and tracking in spite of the eyeglass glare.

    [0010] Referring to the block diagram of FIG. 2, the eye monitoring apparatus 12 includes upper and lower infrared (IR) active illumination devices 28 and 30, a solid-state imaging device 32 focused on the driver's face 18, and a vision processor 34. In the illustrated embodiment, the apparatus 12 provides eye state information to a remote host processor 36, and the host processor 36 selectively activates one or more counter-measure devices or systems 38 such as an alarm or a braking system if it is determined that the driver's lack of alertness or attention may possibly compromise vehicle safety. The active illumination devices 28 and 30 are individually activated by the vision processor 34 via I/O interface 46, and each comprises an array of infrared light emitting diodes as indicated. The vision processor 34 comprises conventional components, including a frame grabber 40 for acquiring video images from imaging device 32, a non-volatile memory 42 for storing various signal processing routines, and a digital signal processor (DSP) 44 for selectively executing the routines stored in memory 42 processing the video images acquired by frame grabber 40. The DSP 44 outputs various control signals to illumination device 30 and imaging device 32 via interface 46, and communicates with host processor 36 via interface 48.

    [0011] The upper and lower active illumination device 28 and 30 are oppositely staggered about imaging device 32 in the vertical direction as indicated in FIG. 2, and are alternately activated during successive image capture intervals of imaging device 32. Due to the proximity of the imaging device 32 to the active illumination devices 28 and 30, the active illumination can reflect off the driver's eyeglasses 24 in a way that creates a glare spot (i.e., a grouping or blob of saturated pixels) in the images produced by imaging device 32. However, the location of the glare spot in the image shifts depending on which active illumination device is lit. For example, if the image produced when the upper active illumination device 28 is lit results in a glare spot that occludes one or both of the driver's eyes 22, the glare spot will be shifted to a non-occluding location in the next image which is produced with driver illumination provided by the lower active illumination device 30.

    [0012] In general, the active illumination devices 28 and 30 must be physically separated or staggered to achieve the desired glare spot shifting, and the separation distance is preferably on the order of 100mm or greater. While the active illumination devices 28 and 30 may be staggered horizontally, vertically, or both horizontally and vertically, vertical staggering is preferred for at least two reasons. First, normal eyeglass curvature is such that the amount of glare shift for a given separation between the active illumination devices 28 and 30 occurs when they are vertically staggered. And second, vertical staggering of the active illumination devices 28 and 30 results in vertical shifting of the glare spot, which is the most effective way to shift the spot away from a feature such as an eye that is dominated by horizontal geometry. Also, it is preferred to oppositely stagger the active illumination devices 28 and 30 about the imaging device 32 as shown in FIG. 2 in order to maximize the separation distance for a given package size of eye monitoring apparatus 12.

    [0013] The signal processing routines residing in the vision processor memory 42 include an eye detection routine 50, an eye tracking routine 52, and an eye analysis routine 54. In general, the routine 50 identifies the regions of a video image that correspond to the driver's eyes 22, the routine 52 tracks the eye location from one video image to the next, and the routine 54 characterizes the state of the driver's eyes (open vs. closed, for example). The eye detection routine and eye tracking routine 50 and 52, as well as the analysis routine 54 and the routines executed by host processor 36 for using the eye state information, may comprise any of a number of known processing techniques. As explained below, however, the eye detection and tracking routines 50 and 52 are capable of detecting and tracking the driver's eyes 22 based on every other image, and ignoring the intermediate images in cases where eye-occluding glare occurs.

    [0014] The flow diagram of FIG. 3 illustrates a coordinated control of active illumination devices 28 and 30 and imaging device 32 by DSP 44. The blocks 60-66 are repeatedly executed as shown to assign the images captured by imaging device 32 to one of two channels, designated as Channel_A and Channel_B. The blocks 60-62 illuminate the driver 16 with just the upper illumination device 28, and then capture the resulting image and assign it to Channel_A. The blocks 64-66 then illuminate the driver 16 with just the lower illumination device 30, capture the resulting image and assign it to Channel_B. Thus, Channel_A contains a stream of images where the driver 16 is actively illuminated by upper illumination device 28, and Channnel_B contains a stream of images where the driver 16 is actively illuminated by lower illumination device 30.

    [0015] The flow diagrams of FIGS. 4-6 illustrate three possible techniques for processing the Channel_A and Channel_B images developed when DSP 44 executes the flow diagram of FIG. 3.

    [0016] Referring to FIG. 4, the first processing technique individually applies the detection and tracking routines 50 and 52 to the Channel_A and Channel_B images, as indicated at blocks 70 and 72. If the routines are not successful with the Channel_A images or the Channel_B images, blocks 70-72 are re-executed. If the routines are successful with the images of at least one of the channels, the block 76 is executed to run the eye analysis routine 54.

    [0017] Referring to FIG. 5, the second processing technique substitutes pixel data from the images of one channel into the images of the other channel to create a series of glare-free images for analysis. The block 80 is first executed to detect and bound a glare spot, if present, in an image assigned to Channel_A (or alternately, Channel_B). The glare spot can be identified, for example, by filtering the eye portion of the image with a morphological bottom-hat filter. The contrast of a local region including the identified spot can be enhanced by histogram equalization. A morphological bottom-hat filter can be applied to the contrast-enhanced region to extract the spot, and the boundary of the glare spot can be defined as the area of overlap between the two filter outputs. Block 82 fetches the corresponding pixel data from a time-adjacent Channel_B image, and substitutes that data into the Channel_A image. The resulting Channel_A image should be substantially glare-free because the glare spot, if present, will be in a different area of the Channel_B image. The block 84 applies the detection and tracking routines to the modified Channel_A images (after defining a search window in the unmodified image based on the previous frame), and the block 86 determines if the routines were successful. If not, the blocks 80-84 are repeated; if so, the block 76 is executed to run the eye analysis routine 54.

    [0018] Referring to FIG. 6, the third processing technique applies the processing technique of FIG. 5 to the images of both Channel_A and Channnel_B, and then consolidates the modified images for analysis. The blocks 90-92 form a modified Channel_A image by substituting pixels from the Channel_B image for the glare spot pixels of the Channel_A image. Conversely, the blocks 94-96 form a modified Channel_B image by substituting pixels from the Channel_A image for the glare spot pixels of the Channel_B image. The block 98 consolidates the modified Channel_A and Channel_B images to form a succession of glare-free images at the full frame update rate of imaging device 32, thirty frames per second for example. The block 100 applies the detection and tracking routines 50 and 52 to the consolidated stream of images. If block 102 determines that the detection and tracking routines were unsuccessful, the blocks 90-100 are repeated; otherwise, the block 104 is executed to run the eye analysis routine 54.

    [0019] In summary, the present invention provides a way of reliably detecting and tracking an actively illuminated eye in a series of digital images that are subject to eyeglass glare that occludes the eye. While the invention has been described with respect to the illustrated embodiments, it is recognized that numerous modifications and variations in addition to those mentioned herein will occur to those skilled in the art. For example, system may include more than two sets of active illumination devices, and so on.


    Claims

    1. A method of monitoring an eye (22) of a human subject (16) to determine an attention of the subject (16) based on a succession of digital images captured by an imaging device (32) focused on the subject (16), where a first image of the subject is captured while the subject is illuminated by a first illumination source (28) disposed in a first location (60, 62) and a second image of the subject is captured while the subject is illuminated by a second illumination source (30) disposed in a second location displaced from said first location (64, 66),
    identifying an eye portion of the first image and determining if the identified eye portion includes a glare spot due (80/90) to a reflection of said first illumination source off eyeglasses (24) worn by the subject;
    characterized by the steps of:

    if the identified eye portion includes such a glare spot, identifying a set of pixels that make up the glare spot (80/90), and replacing the identified set of pixels in the first image with a set of pixels obtained from the same portion of the second image (82/92) to produce a first modified image that does not include a glare spot due to a reflection of said first illumination source; and

    processing said first modified image (84) to detect and track the subject's eye, and thereby determine the attention of the subj ect (16).


     
    2. The method of claim 1, where said first location is above said imaging device (32) and said second location is below said imaging device (32).
     
    3. The method of claim 1, where said second location is vertically displaced from said first location.
     
    4. The method of claim 1, including the steps of:

    identifying an eye portion of the second image and determining if the identified eye portion includes a glare spot (94) due to a reflection of said second illumination source off eyeglasses worn by the subject;

    if the identified eye portion of the second image includes such a glare spot, identifying a set of pixels that comprise said glare spot (94), and replacing the identified set of pixels in the second image with a set of pixels obtained from the same portion of the first image (96) to produce a second modified image that does not include a glare spot due to a reflection of said second illumination source;

    consolidating the first modified image and the second modified images (98); and

    processing the consolidated images to detect and track the subject's eye (100).


     


    Ansprüche

    1. Ein Verfahren zum Überwachen eines Auges (22) einer menschlichen Person (16), um eine Aufmerksamkeit der Person (16) zu bestimmen basierend auf einer Aufeinanderfolge von digitalen Bildern, die von einer Abbildungsvorrichtung (32) aufgenommen werden, die auf die Person (16) fokussiert ist, wobei ein erstes Bild der Person aufgenommen wird, während die Person durch eine erste Beleuchtungsquelle (28) illuminiert wird, die an einer ersten Position (60, 62) angeordnet ist, und ein zweites Bild der Person aufgenommen wird, während die Person durch eine zweite Beleuchtungsquelle (30) illuminiert wird, die an einer zweiten Position versetzt von der ersten Position (64, 66) angeordnet ist,
    Identifizieren eines Augenabschnitts des ersten Bilds und Bestimmen, ob der identifizierte Augenabschnitt einen Blendpunkt aufgrund (80/90) einer Reflexion der ersten Beleuchtungsquelle auf Brillengläsern (24) umfasst, die von der Person getragen werden;
    gekennzeichnet durch die Schritte:

    wenn der identifizierte Augenabschnitt einen derartigen Blendpunkt umfasst, Identifizieren eines Satzes von Pixeln, die den Blendpunkt (80/90) darstellen, und Ersetzen des identifizierten Satzes von Pixeln in dem ersten Bild mit einem Satz von Pixeln, der aus dem selben Abschnitt des zweiten Bilds erlangt wird (82/92), um ein erstes modifiziertes Bild zu erzeugen, das keinen Blendpunkt aufgrund einer Reflexion der ersten Beleuchtungsquelle umfasst; und

    Verarbeiten des ersten modifizierten Bilds (84), um das Auge der Person zu erfassen und zu verfolgen, und dadurch Bestimmen der Aufmerksamkeit des Person (16).


     
    2. Das Verfahren gemäß Anspruch 1, wobei die erste Position über der Abbildungsvorrichtung (32) ist und die zweite Position unter der Abbildungsvorrichtung (32) ist.
     
    3. Das Verfahren gemäß Anspruch 1, wobei die zweite Position vertikal versetzt von der ersten Position ist.
     
    4. Das Verfahren gemäß Anspruch 1, das die Schritte umfasst:

    Identifizieren eines Augenabschnitts des zweiten Bilds und Bestimmen, ob der identifizierte Augenabschnitt einen Blendpunkt (94) aufgrund einer Reflexion der zweiten Beleuchtungsquelle auf Brillengläsern umfasst, die von der Person getragen werden;

    wenn der identifizierte Augenabschnitt des zweiten Bilds einen derartigen Blendpunkt umfasst, Identifizieren eines Satzes von Pixeln, die den Blendpunkt (94) aufweisen, und Ersetzen des identifizierten Satzes von Pixeln in dem zweiten Bild mit einem Satz von Pixeln, der aus dem selben Abschnitt des ersten Bilds erlangt wird (96), um ein zweites modifiziertes Bild zu erzeugen, das keinen Blendpunkt aufgrund einer Reflexion der zweiten Beleuchtungsquelle umfasst;

    Konsolidieren des ersten modifizierten Bilds und des zweiten modifizierten Bilds (98); und

    Verarbeiten der konsolidierten Bilder, um das Auge der Person zu erfassen und zu verfolgen (100).


     


    Revendications

    1. Procédé pour surveiller un oeil (22) d'un sujet humain (16) pour déterminer une attention du sujet (16) en se basant sur une succession d'images numériques capturées par un dispositif d'imagerie (32) focalisé sur le sujet (16), dans lequel une première image du sujet est capturée alors que le sujet est illuminé par une première source d'illumination (28) disposée à un premier emplacement (60, 62) et une seconde image du sujet est capturée alors que le sujet est illuminé par une seconde source d'illumination (30) disposée à un second emplacement déplacé par rapport au premier emplacement (64, 66),
    on identifie une portion d'oeil de la première image et on détermine si la portion d'oeil identifiée inclut une tache brillante (80/90) due à une réflexion de ladite première source d'illumination en dehors de lunettes (24) portées par le sujet ;
    caractérisé par les étapes consistant à
    si la portion d'oeil identifiée inclut une telle tache brillante, identifier un groupe de pixels qui constituent la tache brillante (80/90) et remplacer le groupe identifié de pixels dans la première image par un groupe de pixels obtenus de la même portion de la seconde image (82/92) pour produire une première image modifiée qui n'inclut pas de tache lumineuse due à une réflexion de ladite première source d'illumination ; et
    traiter ladite première image modifiée (84) pour détecter et suivre l'oeil du sujet, et déterminer ainsi l'attention du sujet (16).
     
    2. Procédé selon la revendication 1, dans lequel ledit premier emplacement est au-dessus dudit dispositif d'imagerie (32) et ledit second emplacement est au-dessous dudit dispositif d'imagerie (32).
     
    3. Procédé selon la revendication 1, dans lequel ledit second emplacement est déplacé verticalement par rapport audit premier emplacement.
     
    4. Procédé selon la revendication 1, incluant les étapes consistant à :

    identifier une portion d'oeil de la seconde image et déterminer si la portion d'oeil identifiée inclut une tache brillante (94) due à une réflexion de ladite seconde source d'illumination en dehors de lunettes portées par le sujet ;

    si la portion d'oeil identifiée de la seconde image inclut une telle tache brillante, identifier un groupe de pixels qui constituent ladite tache brillante (94), et remplacer le groupe de pixels identifiés dans la seconde image par un groupe de pixels obtenus de la même portion de la première image (96) pour produire une seconde image modifiée qui n'inclut pas de tache brillante due à une réflexion de ladite seconde source d'illumination ;

    consolider la première image modifiée et la seconde image modifiée (98) ; et

    traiter les images consolidées pour détecter et suivre l'oeil du sujet (100).


     




    Drawing


















    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description