Infrared face detection and recognition system

• US 20060104488A1
• Filed: 11/12/2004
• Published: 05/18/2006
• Est. Priority Date: 11/12/2004
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Bazakos, Michael E., Morellas, Vassilios, Ma, Yungian, Johnson, Andrew
• Granted Patent: US 7,469,060 B2
• Time in Patent Office: Days
• US Class Current: 382/118
• CPC Class Codes: G06V 40/166 using acquisition arrangements
• CPC Class Codes: H04N 23/611 where the recognised objects include parts of the human body

Abstract

A face detection and recognition system having several arrays imaging a scene at different bands of the infrared spectrum. The system may use weighted subtracting and thresholding to distinguish human skin in a sensed image. A feature selector may locate a face in the image. The face may be framed or the image cropped with a frame or border to incorporate essentially only the face. The border may be superimposed on an image direct from an imaging array. A sub-image containing the face may be extracted from within the border and compared with a database of face information to attain recognition of the face. A level of recognition of the face may be established. Infrared lighting may be used as needed to illuminate the scene.

Claims

1. A detection system comprising:

a first array of detector pixels having a first field of view;

a second array of detector pixels having the first field of view;

a subtractor connectable to each pixel pair of the first and second arrays; and

a comparator connected to the subtractor; and

wherein;

the first sensing array has a sensitivity to light in a lower portion of a reflective infrared spectrum;

the second sensing array has a sensitivity to light in an upper portion of a reflective infrared spectrum;

a pixel pair includes a first pixel at a location of the first array of detector pixels of the first array which a one-to-one correlation with a second pixel of the pixel pair at a corresponding location of the second array of detector pixels;

the subtractor may detect the differences of magnitudes of output signals from the first and second pixels of each pixel pair and output a magnitude difference signal for each pixel pair; and

a comparator may compare the magnitude difference signals to a certain threshold value and select the pixel pairs having magnitude difference values on one side of the threshold value.

2. The system of claim 1, wherein:

the second pixels of each of the pixel pairs having a magnitude difference value on one side of the threshold value exhibit detected low reflectance areas in the field of view; and

the low reflectance areas may indicate skin of a person.

3. The system of claim 2, further comprising:

a face feature selector;

a framer; and

an extractor; and

wherein;

an image from of the second array of detector pixels may reveal the skin of a person in the field of view;

the feature selector may locate features of a face in the shin;

the framer may place a frame around the face in the image;

pixels of the frame around the face may be matched with pixels at corresponding locations in an image from the first array of detector pixels; and

the extractor may extract a sub-image containing the face within the frame in the image from the first array of detector pixels.

4. The system of claim 3, further comprising:

a face recognition system; and

wherein;

the face recognition system may match the sub-image of the face with another sub-image of an identified face.

5. The system of claim 4, further comprising:

an evaluator; and

wherein the evaluator may indicate a confidence level of recognition of the face within the frame.

6. The system of claim 5, further comprising a database of sub-images of identified faces, connected to the face recognition system.

7. A human detection system comprising:

a multi-band camera having first and second outputs;

a subtractor connected to the first and second outputs;

a thresholder connected to the subtractor;

a feature selector connected to the thresholder;

a framer connected to the thresholder;

a pixel integrator connected to the framer and the first output;

an extractor connected to the integrator;

a recognition system connected to the extractor; and

8. The system of claim 7, further comprising an evaluator connected to the recognition system.

9. The system of claim 8, wherein:

the first output of the multi-band camera has a bandwidth in the lower reflective region of the infrared spectrum; and

the second output of the multi-band camera has a bandwidth in the upper reflective region of the infrared spectrum.

10. The system of claim 9, wherein the subtractor may remove pixels from an image which are not representative of skin.

11. The system of claim 10, wherein:

the feature selector seeks out features of a face in the image;

the feature selector may retain only the face in the image; and

the framer may output a framed image of the face.

12. The system of claim 11, wherein:

the pixel integrator may match the pixels of the framed image with an image from the first output of the multi-band camera; and

the integrator may superimpose the frame on the image from the first output of the multi-band camera.

13. The system of claim 12, wherein the extractor may extract a portion of the image within the frame.

14. The system of claim 13, wherein:

the portion of the image may be sent to a recognition system; and

the recognition system may perform a match of the portion of the image with a similar image from a data system.

15. The system of claim 14, wherein the evaluator may provide a confidence level of recognition of the match.

16. The system of claim 15, further comprising a light source for providing infrared lighting as needed.

17. A method for detecting a face in a scene, comprising:

capturing a first image the scene with a first sensing array sensitive to a first band of reflective infrared light;

capturing a second image of the scene with a second sensing array sensitive to a second band of reflective infrared light; and

wherein;

the first image comprises m×

n pixels;

the second image comprises m×

n pixels;

the first and second images have the same field of view of the scene;

each pixel P_(i,j)1 of the first image is located at an i_th row and a j_th column;

each pixel P_(i,j)2 of the second image is located at an i_th row and a j_th column;

1≦

i≦

m;

1≦

j≦

m;

each pixel P_(i,j)1 of the first image has a counterpart pixel P_(i,j)2 in the second image in a one-to-one mapping of the first image to the second image;

a difference intensity pixel Δ

I_P(i,j) may be determined from an intensity of each pixel I_P(i,j)1 of the first image and each counterpart pixel I_P(i,j) of the second image, for all of the pixels of the first and second images;

each difference intensity pixel Δ

I_P(i,j) has an intensity value that may be measured against an intensity threshold value and has its value nullified if it does not meet a certain range of intensity values relative to the intensity threshold value, and becomes a thresholded difference intensity pixel TΔ

I_P(i,j);

each thresholded difference intensity pixel together with each other thresholded difference intensity pixel forms a third image having a one-to-one mapping of pixels to the first and second images; and

features of a face are sought, and if found, a face is detected in the third image and a frame is scribed as frame pixels in the third image around the face.

18. The method of claim 17, further comprising:

transferring the frame pixels from the third image to the first image on a one-to-one mapping basis; and

extracting the pixels within the frame in the first image to form a fourth image having the face extracted from the first image.

19. The method of claim 18, further comprising:

comparing the face in the fourth image with faces from a database; and

matching the face in the fourth image with a face from the data base.

20. The method of claim 19, further comprising:

evaluating the matching of the face in the fourth image with the face from the database; and

determining a confidence level of recognition of the face in the fourth image relative to the face from the database.

21. The method of claim 20, wherein:

the first band of reflective infrared light is approximately between 800 and 1400 nanometers; and

the second band of reflective infrared light is approximately between 1400 and 2200 nanometers.

22. The method of claim 21, further comprising shining infrared light on the scene as needed for obtaining adequate first and second images of the scene.

23. Means for detecting faces, comprising:

means for obtaining a first image with a first field of view of a scene at a lower reflective infrared band of light;

means for obtaining a second image with the first field of view of the scene at an upper reflective infrared band of light;

means for subtractively combining the first and second images into a third image;

means for thresholding the pixels of the third image to result in a fourth image having only the pixels meeting a set threshold value;

means for finding features of a face in the fourth image;

means for inserting a frame around the face based on the features; and

means for transferring the frame to be around the face in the first image.

24. The means of claim 23, further comprising:

means for extracting a sub-image of the face within the frame; and

means for possibly at least approximately matching the face with at least one other face in a database.

25. The means of claim 24, further comprising:

means for evaluating the matching the face with the at least one other face in the database; and

means for determining a confidence level of recognition of the face within the frame.

26. The means of claim 25, further comprising means for providing infrared lighting on the scene as needed for obtaining the first and second images.

27. The means of claim 26, wherein:

the lower reflective infrared band of light is approximately between 800 and 1400 nanometers; and

the upper reflective infrared band of light approximately between 1400 and 2200 nanometers.