Image analysis – Histogram processing – With pattern recognition or classification
Reexamination Certificate
1998-04-17
2001-01-23
Boudreau, Leo H. (Department: 2721)
Image analysis
Histogram processing
With pattern recognition or classification
C382S118000, C382S218000, C382S298000
Reexamination Certificate
active
06178259
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to object recognition systems; and more particularly, it relates to methods of compensating for pixel histogram distortion which can occur in such systems.
One particular use for the present invention is in object recognition systems which either grant or deny a person access to a secure area based on whether or not an image from a camera of the person's face is recognized by the system. Also, the present invention can be used in object recognition systems which employ a robotic arm to select or reject a component based on whether or not an image from a camera of the component is recognized by the system.
In all of the above applications, a key task which needs to be performed as part of the recognition process is that two images must be correlated. One of these images is the image, from a camera, of an unknown object which the system is trying to recognize; and the other image is of a known object which is stored in the system as a reference. If the correlation between the two images exceeds a predetermined threshold value, then the unknown object is recognized.
However before the two images are correlated, it is highly desirable to adjust the magnification of the image of the unknown object such that it is closely matches the magnification of the reference. This is because as the difference in magnification of the unknown object and the reference increases, their correlation decreases; and that will cause recognition errors.
In the case where a person is seeking access to a secure area, the magnification of that person's image from the camera will increase as the person stands closer to the camera, and vice-versa. Such variations in magnification will occur in a work environment because it is not practical for a person to stand exactly the same distance from the camera each time access is sought.
Before any adjustment in magnification is made to the image of the unknown object, the current magnification of that object must be detected. This can be attempted by sensing the width of the unknown object in the image from the camera and comparing it to the width of the reference.
However, the present inventor has found that under certain conditions, a histogram of the pixels in the image from the camera can be distorted in a particular fashion which causes errors to occur when the width of the unknown object is sensed. If that happens, the magnification of the image from the camera will be adjusted incorrectly, and thus a recognition error will occur.
Accordingly, a primary object of the present invention is to provide an object recognition system in which the above problem is overcome.
BRIEF SUMMARY OF THE INVENTION
With the present invention, the above problem is overcome as follows. Initially, a first image of the object that is to be recognized is received from a camera. Then, a second image is generated from the first image, by scaling the pixel magnitudes in the first image such that a histogram of the pixels in the second image has a larger standard deviation than a histogram of the pixels in the first image. Next, the width of the unknown object is detected as the width of an area in the second image where the pixel magnitudes exceed a certain minimum. Then, a third image is generated from the second image, by adjusting the magnification of the second image based on the detected width; and the unknown object is identified if a correlation between the third image and a reference exceeds a threshold.
One particular embodiment of the present invention recognizes the face of a person who is seeking access to a secure area, and an infrared camera is used to generate an image of that person. However, as the person's face becomes hotter, then a histogram of the pixels in the facial image from the infrared camera will become shifted to the high side of the pixel temperature spectrum and the standard deviation of the pixels will decrease. Such facial heating can occur, for example, if the person recently was exercising or was outside in hot weather.
As the standard deviation of the pixels from the camera decrease, a point is eventually reached where the width of the person's face cannot be accurately detected. This occurs because as the standard deviation decreases, any difference in the magnitude of the pixels in the person's hair and the person's skin tends to decrease. Thus, long hair on the sides of the becomes indistinguishable from skin on the face.
REFERENCES:
patent: 4651199 (1987-03-01), Alkofer
patent: 4959870 (1990-09-01), Tachikawa
patent: 5163094 (1992-11-01), Prokoski et al.
patent: 5497430 (1996-03-01), Sadovnik et al.
patent: 5693940 (1997-12-01), Botti et al.
patent: 5748802 (1998-05-01), Winkelman
patent: 5822453 (1998-10-01), Lee et al.
patent: 5835616 (1998-10-01), Lobo et al.
patent: 5991429 (1999-11-01), Coffin et al.
Adornato Rocco L.
Boudreau Leo H.
Fassbender Charles J.
Sherali Ishrat
Starr Mark T.
LandOfFree
Method of compensating for pixel histogram distortion in an... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of compensating for pixel histogram distortion in an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of compensating for pixel histogram distortion in an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2533576