System and method for detecting an eye

• US 7,689,008 B2
• Filed: 06/10/2005
• Issued: 03/30/2010
• Est. Priority Date: 06/10/2005
• Status: Active Grant

• Current Assignee: Aptiv Technologies AG (Aptiv PLC)
• Original Assignee: Delphi Technologies, Inc. (Aptiv PLC)
• Inventors: Hammoud, Riad I., Wilhelm, Andrew L.
• Primary Examiner(s): Bali, Vikkram
• Assistant Examiner(s): Zeilberger, Daniel
• Application Number:
  US11/150,684
• Publication Number:
  US 20070014431A1
• Time in Patent Office: 1,754 Days
• Field of Search: 382/103, 382/115, 382/117, 382/118
• US Class Current: 382/117
• CPC Class Codes: G06V 40/19 Sensors therefor

Abstract

A system and method are provided for detecting one or both eyes of the driver of a vehicle. The system includes a video imaging camera oriented to generate images of the driver'"'"'s face. The system also includes a video processor for processing the images generated with the video imaging camera. Filters are applied to each of potential eye candidates to determine which candidates represent an eye, weighted values are assigned to the filtered eye candidates, and an eye is detected based on the weighting of each eye candidate. According to one aspect, different size patches of potential eye candidates are processed and compared to models.

Claims

1. A system for detecting an eye, said system comprising:

a video imaging camera oriented to generate images of a face of a subject; and

a video processor for processing the images generated with the video imaging camera, said video processor comprising a salient feature detector for detecting potential eye candidates and filters for filtering the potential eye candidates based on appearance, spatial-geometrical constraint, and physiological skin property, said filters comprising a pattern recognition filter for filtering the potential eye candidates based on appearance, wherein the processor compares the potential eye candidates to one or more appearance-based models and classifies the potential eye candidates based on the comparison, said video processor further comprising weighting logic for weighting each potential candidate based on the appearance, spatial-geometrical constraint, and physiological skin property and generating a weighted value for each potential eye candidate, said video processor further comprising detecting logic for detecting an eye based on the weighted value of each eye candidate, wherein the filters comprise a spatial-geometrical constraint filter for weighting the potential eye candidates based on one or more spatial-geometrical constraints which compare horizontal eye candidates and compare vertical eye candidates.

2. The system as defined in claim 1, wherein said detecting logic detects the eye based on the weighted value of each eye candidate by comparing the weighted values of each of the eye candidates and detecting the eye based on the maximum weighted value.

3. The system as defined in claim 1, wherein said filters comprise a physiological skin property filter for weighting the potential eye candidates based on one or more physiological skin properties for comparing physiological properties of eye candidates to physiological properties of a skin region.

4. The system as defined in claim 1, wherein the video processor further includes a multi-resolution filter that compares a first patch of an eye candidate having a first size to one or more first models based on image recognition, and further compares a second patch of the eye candidate having a different second size to one or more second models based on pattern recognition.

5. The system as defined in claim 1, wherein the video imaging camera is located on a vehicle to monitor the eye of the driver of the vehicle.

12. The system as defined in claim 4, wherein the first patch and the second patch of the image are both generally centered on an area of an eye candidate.

13. The system as define din claim 1, wherein the one or more appearance-based models comprises eye and non-eye models.

14. The system as defined in claim 13, wherein the processor computes a distance from the potential eye candidate to each eye and non-eye model and compares the distances in an eigen space to classify the potential eye candidates.

6. A method of detecting an eye, said method comprising the steps of:

storing one or more models in memory;

arranging a video imaging camera to generate images of a face of a subject;

generating an image of the face using the video imaging camera;

detecting salient features in the image as potential eye candidates with a video processor;

filtering the potential eye candidates based on appearance with the video processor, wherein the filtering comprises comparing the potential eye candidates to the one or more appearance-based models and classifying the potential eye candidates based on the comparison;

assigning a weighted value to each eye candidate based on the appearance with the video processor;

filtering the potential eye candidates based on one or more spatial-geometrical constraints comprising comparing pairs of horizontal eye candidates and comparing pairs of vertical eye candidates with the video processor;

assigning a weighted value to each eye candidate based on the spatial-geometrical constraints with the video processor;

filtering the potential eye candidates based on one or more physiological skin properties with the video processor;

assigning a weighted value to each eye candidate based on the physiological skin properties with the video processor; and

detecting an eye based on the weighted values of the eye candidates with the video processor.

7. The method as defined in claim 6, wherein the step of detecting an eye comprises comparing the weighted values of each of the eye candidates and detecting the eye based on the maximum weighted value.

8. The method as defined in claim 6, wherein the step of filtering the potential eye candidates based on appearance comprises applying a pattern recognition filter to the generated image.

9. The method as defined in claim 6, wherein the step of filtering the potential eye candidates based on physiological skin properties comprises comparing brightness of eye candidates to brightness of one or more skin regions.

10. The method as defined in claim 6 further comprising the step of comparing a first patch of the eye candidate having a first size to a first model based on pattern recognition and generating a first output, comparing a second patch of the eye candidate having a different second size to a second model based on pattern recognition and generating a second output, and assigning a weighted value to each eye candidate based on the comparison of the first and second patches.

11. The method as defined in claim 6, wherein said step of arranging a video camera comprises arranging the video camera in a vehicle to monitor the eye of the driver of the vehicle.

15. The method as defined in claim 6, wherein the one or more appearance-based models comprise eye and non-eye models.

16. The method as defined in claim 15, wherein the method computes a distance from the potential eye candidate to each eye and non-eye model and compares the distances in an eigen space to classify the potential eye candidates.