SYSTEM AND METHOD FOR ANIMAL IDENTIFICATION USING IRIS IMAGES

• US 20090208064A1
• Filed: 02/13/2009
• Published: 08/20/2009
• Est. Priority Date: 02/14/2008
• Status: Active Grant

• Current Assignee: Iristrac LLC
• Original Assignee: The International Performance Registry, LLC
• Inventors: Cambier, James L.
• Granted Patent: US 8,189,879 B2
• Time in Patent Office: Days
• US Class Current: 382/110
• CPC Class Codes: G06V 40/193 Preprocessing; Feature extraction

Abstract

A system for animal identification includes: an image capture apparatus for obtaining an image of an eye of an animal including a pupil region and an iris region; and a template generation apparatus. The template generation apparatus is for: extracting a set of pixel data from the image, the set of pixel data representing an upper region of interest of the iris region above the pupil region and a lower region of interest of the iris region below the pupil region, the upper region of interest and the lower region of interest have parallel side boundaries that are spaced apart a distance that is substantially independent of a degree of dilation of the pupil region; and transforming the set of pixel data representing the upper region of interest and the lower region of interest into a template of the upper region of interest and the lower region of interest.

Claims

1. A method for animal identification, comprising:

obtaining an image of an eye of an animal including a pupil region and an iris region;

extracting a set of pixel data from the image, the set of pixel data representing an upper region of interest of the iris region above the pupil region and a lower region of interest of the iris region below the pupil region, the upper region of interest and the lower region of interest having parallel side boundaries that are spaced apart a distance that is substantially independent of a degree of dilation of the pupil region; and

transforming the set of pixel data representing the upper region of interest and the lower region of interest into a template of the upper region of interest and the lower region of interest, the template for use in identifying the animal.

2. The method of claim 1, wherein extracting a set of pixel data from the image includes:

determining a pupil aspect ratio of a minor axis of the pupil region to a major axis of the pupil region;

determining a major axis length at a condition of minimum pupil dilation (MAL_min) using the pupil aspect ratio; and

determining a location of the parallel side boundaries of the upper region of interest and the lower region of interest using the MAL_min.

3. The method of claim 2, wherein the image of the eye includes a pupil having an irregular boundary, and wherein determining the pupil aspect ratio further includes:

determining the irregular boundary of the pupil;

constructing a convex hull of the irregular boundary of the pupil; and

using the convex hull of the irregular boundary of the pupil as a boundary of the pupil region in all subsequent operations.

4. The method of claim 3, wherein determining the irregular boundary of the pupil includes:

detecting a specular reflection region in an interior of the pupil by identifying pixels within a central sub-image of the image of the eye having an intensity greater than a specular reflection threshold value;

constructing a histogram of pixel intensity values in an area immediately surrounding the specular reflection region;

determining a pupil threshold value using the histogram of pixel intensity values in the area immediately surrounding the specular reflection region; and

differentiating between the pupil and surrounding areas of the pupil using the pupil threshold value to determine the irregular boundary of the pupil.

5. The method of claim 4, further comprising performing successive image morphological closing operations on an area within the irregular boundary of the pupil until an Euler number representation of the pupil region is equal to one to eliminate internal holes in the pupil region.

6. The method of claim 5, wherein determining the pupil aspect ratio of the minor axis of the pupil region to the major axis of the pupil region includes performing principle component analysis on points lying within the convex hull of the irregular boundary of the pupil to identify the major axis of the pupil region and the minor axis of the pupil region.

7. The method of claim 6, further comprising rotating the image of the eye of the animal so that the major axis becomes horizontal.

8. The method of claim 7,

wherein MAL_min is calculated as

MAL_min=MAL*(1+P₁+P₂*PAR²+P₃*PAR³)

wherein MAL is a measured major axis length, PAR is a measured pupil aspect ratio, and parameters P₁, P₂, and P₃ are predetermined by fitting a curve to empirical values of MAL and PAR for several animal eyes imaged at various degrees of dilation;

wherein a ROI width is a width of the upper region of interest and the lower region of interest calculated as;

ROI width=C₁* MAL_min

wherein C₁ is predetermined by empirical analysis of the several animal eyes imaged at various degrees of dilation;

wherein a ROI height is a height of the upper region of interest and the lower region of interest calculated as;

ROI height=C₂*MAL_min−

Minor Axis Length/2

wherein C₂ is predetermined by empirical analysis of the several animal eyes imaged at various degrees of dilation; and

wherein the upper region of interest and the lower region of interest extend from the convex hull of the irregular boundary of the pupil outwardly a distance equal to the ROI height and have a width extending a distance equal to the ROI width positioned as a function of a position of the major axis of the pupil region, MAL, and PAR.

9. The method of claim 1, wherein transforming the set of pixel data comprises:

processing the set of pixel data from the image using log-Gabor wavelets in a complex transformation to produce phase information for each pixel of the set of pixel data; and

encoding the phase information for each pixel of the set of pixel data as a two bit value for each pixel to produce a template that provides a compact representation of iris texture features.

10. The method of claim 1, further comprising:

locating an upper eyelid in the upper region of interest and a lower eyelid in the lower region of interest; and

encoding a mask image of the upper region of interest and the lower region of interest, the mask image for use in masking the upper eyelid and the lower eyelid from the upper region of interest and the lower region of interest.

11. The method of claim 10, further comprising:

locating eyelashes in the upper region of interest;

creating an eyelash mask image of the eyelashes in the upper region of interest; and

combining the eyelash mask image with the mask image of the upper region of interest to produce a combined mask image for use in masking the eyelashes from the upper region of interest.

12. The method of claim 10, further comprising:

generating a plurality of comparison templates and a corresponding plurality of comparison mask images by;

applying a plurality of scale factors applied to the upper region of interest and the lower region of interest to generate a plurality of re-sampled upper regions of interest and lower regions of interest; and

generating a comparison template and a comparison mask image from each of the plurality of re-sampled upper regions of interest and lower regions of interest;

determining a match score value for each comparison template by comparing each comparison template and each comparison mask image to a previously enrolled template and a previously enrolled mask image of an eye of a previously identified animal; and

determining whether the eye of the animal matches the eye of the previously identified animal by comparing each match score value to a match threshold value.

13. The method of claim 10, comprising:

generating a plurality of comparison templates and a corresponding plurality of comparison mask images by re-sampling the template and to the mask image using a plurality of scale factors;

determining a match score value for each of the plurality of comparison templates by comparing each of the plurality of comparison templates and the corresponding plurality of comparison mask images to a previously enrolled template and a previously enrolled mask image of an eye of a previously identified animal; and

determining whether the eye of the animal matches the eye of the previously identified animal by comparing each match score values to a match threshold value.

14. A system for animal identification, comprising:

an image capture apparatus for obtaining an image of an eye of an animal including a pupil region and an iris region; and

a template generation apparatus for;

extracting a set of pixel data from the image, the set of pixel data representing an upper region of interest of the iris region above the pupil region and a lower region of interest of the iris region below the pupil region, the upper region of interest and the lower region of interest having parallel side boundaries that are spaced apart a distance that is substantially independent of a degree of dilation of the pupil region; and

transforming the set of pixel data representing the upper region of interest and the lower region of interest into a template of the upper region of interest and the lower region of interest, the template for use in identifying the animal.

15. The system of claim 14, wherein extracting the set of pixel data from the image by the template generation apparatus includes:

determining a pupil aspect ratio of a minor axis of the pupil region to a major axis of the pupil region;

determining a major axis length at a condition of minimum pupil dilation (MAL_min) using the pupil aspect ratio; and

determining a location of the parallel side boundaries of the upper region of interest and the lower region of interest using the MAL_min.

16. The system of claim 15, wherein the image of the eye includes a pupil having an irregular boundary, and wherein determining the pupil aspect ratio by the template generation apparatus further includes:

determining the irregular boundary of the pupil;

constructing a convex hull of the irregular boundary of the pupil; and

using the convex hull of the irregular boundary of the pupil as a boundary of the pupil region in all subsequent operations.

17. The system of claim 14, wherein the template generation apparatus is further for:

locating an upper eyelid in the upper region of interest and a lower eyelid in the lower region of interest; and

encoding a mask image of the upper region of interest and the lower region of interest, the mask image for use in masking the upper eyelid and the lower eyelid from the upper region of interest and the lower region of interest.

18. The system of claim 17, further comprising an enrollment database in communication with the template generation apparatus for:

receiving from the template generation apparatus the template of the upper region of interest and the lower region of interest, and the mask image of the upper region of interest and the lower region of interest; and

storing the template of the upper region of interest and the lower region of interest, and the mask image of the upper region of interest and the lower region of interest along with information that includes an identifier for the animal.

19. The system of claim 18, further comprising a variable-scale matching apparatus in communication with the template generation apparatus and the enrollment database for:

generating a plurality of comparison templates and a corresponding plurality of comparison mask images by;

applying a plurality of scale factors applied to the upper region of interest and the lower region of interest to generate a plurality of re-sampled upper regions of interest and lower regions of interest; and

generating a comparison template and a comparison mask image from each of the plurality of re-sampled upper regions of interest and lower regions of interest;

determining a match score value for each comparison template by comparing each comparison template and each comparison mask image to a previously enrolled template and a previously enrolled mask image of an eye of a previously identified animal retrieved from the enrollment database; and

determining whether the eye of the animal matches the eye of the previously identified animal by comparing each match score value to a match threshold value.

20. The system of claim 18, further comprising a variable-scale matching apparatus in communication with the template generation apparatus and the enrollment database for:

generating a plurality of comparison templates and a corresponding plurality of comparison mask images by re-sampling the template and to the mask image using a plurality of scale factors;

determining a match score value for each of the plurality of comparison templates by comparing each of the plurality of comparison templates and the corresponding plurality of comparison mask images to a previously enrolled template and a previously enrolled mask image of an eye of a previously identified animal retrieved from the enrollment database; and

determining whether the eye of the animal matches the eye of the previously identified animal by comparing each match score values to a match threshold value.

21. The system of claim 14, further comprising an infrared illumination source illuminating the eye of the animal with infrared light, and an image acquisition device in communication with the image capture apparatus, the image acquisition device comprising:

an optical lens, an infrared bandpass filter, and a high resolution digital camera in optical communication with the eye of the animal for acquiring an image of the eye of the animal; and

a graphical display for providing a live video image to an operator to facilitate positioning of the image acquisition device relative to the eye of the animal.